add arrow-down arrow-left arrow-right arrow-up authorcheckmark clipboard combo comment delete discord dots drag-handle dropdown-arrow errorfacebook history inbox instagram issuelink lock markup-bbcode markup-html markup-pcpp markup-cyclingbuilder markup-plain-text markup-reddit menu pin radio-button save search settings share star-empty star-full star-half switch successtag twitch twitter user warningwattage weight youtube

$5500 budget dual Xeon ECC RAM with dual GPUs and dual 4K monitors for Linux and must be silent

FitEyes
  • 53 months ago

A few days ago I posted about a similar build, but I was focused on a single i7 and non-ECC RAM. From the comments I learned that a dual Xeon ECC RAM build will be more suitable. Therefore, I am asking a new question for advice on this build.

Here is the initial outline of a parts list:

  • 2x Intel Xeon E5-26XX V3 Processor
  • quiet CPU coolers
  • dual socket LGA 2011-v3 motherboard
    • ATX form factor
    • Intel C612 PCH
    • quad channel memory support
    • passive cooling
    • 2x Intel Gigabit LAN
    • DDR4 2133 support / at least 128 GB RAM supported (I'm only going to install 64 GB now)
    • consumer-type BIOS preferred (I think)
    • support for 2x SLI or 2x Crossfire (I'll probably use Nvidia cards for now)
    • enough room for both GPU's to run cool
    • good PCIe lane design (e.g. using M.2 doesn't prevent SLI, etc.)
    • lots of USB 2.0 and 3.0 ports
    • lots of fan connectors
    • prefer onboard USB 3.1, but can live with a card, if necessary
    • no onboard audio or video needed (but it will probably be there anyway)
    • do not need RAID, do not need SAS, do not need more than 6 SATA ports
    • need support for DDR4-2133 RAM, but not more than 256 GB. I'll start with 64 GB.
    • overclocking support not needed
    • want really good quality capacitors that will last a long time
    • prefer M.2 port onboard that will work with an enterprise SSD (see drive below)
    • if not, maybe I can use something like: Lycom M.2 PCIe SSD to PCIe 3.0 x4 adapter
  • SAMSUNG XP941 M.2 256GB PCI-Express 2.0 x4 MLC Enterprise Solid State Drive (or similar)
    • I hope for Linux boot support from this drive, but can work around it if not.
  • Samsung 850 PRO 2TB 2.5" Solid State Drive
  • 2x Western Digital Red 5TB 3.5" 5400RPM Internal Hard Drive (or similar quiet drives)
  • 2x EVGA GeForce GTX 980 4GB Superclocked ACX 2.0 Video Card (2-Way SLI)
    • or
  • 2x MSI GeForce GTX 980 4GB Video Card (2-Way SLI) (or similar)
    • GPU choice based on whichever is quietest
  • A case similar to: Fractal Design Define R5 (Black) ATX Mid Tower Case
    • needs room for 2 GPU's and 2 large CPU coolers, 2 internal HDD's & 1 SSD
    • should not be more than 19" x 9" x 18" (max). Can be smaller, as long as it runs cool.
  • quiet case fans
  • Asus DRW-24B1ST/BLK/B/AS DVD/CD Writer (or similar)
  • EVGA SuperNOVA 1000 P2 1000W 80+ Platinum Certified Fully-Modular ATX Power Supply (or similar)

Below are some of the things I plan to do with this build, but I'm sure I'll find more purposes for it and extract every ounce of capabilities the rig has.

NFS file server on a small LAN (currently 3 people total, including me, but it could grow)

MariaDB (mysql) database server on the same small LAN

3-D CAD work and rendering

Working with very large spreadsheets (where single core clock speed has a large impact on performance than multiple cores)

Working simultaneously with many open applications (often dozens of CPU and RAM hungry apps) and many hundreds of browser tabs (necessary research work) where multiple CPU cores and lots of RAM will be very helpful. This is often the most demanding use-case based on past experience.

Extensive Photo/Video editing (Gimp, Inkscape, etc.)

Very extensive audio editing with Ardour. USB Audio interface is a Native Instruments Komplete Audio 6 (for now).

Power consumption is not a big issue.

My budget is flexible.

I do not plan to overclock.

No gaming. None.

I do not want a watercooled rig. I want my rig (hardware) to be relatively maintenance-free.

It absolutely must be silent!

It must be compatible with current Linux kernels (e.g., 4.2 or 4.3.3). In fact, I would like to run Kubuntu 14.04 LTS, kernel 3.19, until April when I'll switch to 16.04.

I want to drive two of these monitors: Amazon.com: Philips BDM4065UC 40" Class 4K Monitor UHD 3840 x2160 Resolution DisplayPort, Mini DisplayPort, HDMI, MHL-HDMI http://www.amazon.com/dp/B00SCX78JS/

The monitor choice is driven by the use-case of very large spreadsheets. Clear text rendering and lots of screen real estate is most important.

The monitors take away $1454 from my budget, leaving me $3346 to $3500 for everything else. I have a little flexibility. For example, if some part is a much better value, I will make the wiser choice and go slightly above my budget.

I believe that the video cards must be GTX 970 (minimum) or probably the GTX 980's. I'm going to use the Nvidia proprietary drivers under Linux. I prefer the GPU cards to have multiple DisplayPort connectors. (DP 1.3 and HDMI 2.0 would be nice, but I'm probably dreaming. Even if such cards exist, I probably would have driver issues under current versions of Linux, especially kernel 3.19 in Kubuntu 14.04. I think the 980 is the max card I should consider, but correct me if I'm wrong.)

Comments

  • 53 months ago
  • 1 point

Give me a bit and I can help.

  • 53 months ago
  • 1 point

How about this?

PCPartPicker part list / Price breakdown by merchant

Type Item Price
CPU Intel Xeon E5-2630 V3 2.4GHz 8-Core Processor $628.99 @ SuperBiiz
CPU Intel Xeon E5-2630 V3 2.4GHz 8-Core Processor $628.99 @ SuperBiiz
CPU Cooler Noctua NH-D14 SE2011 CPU Cooler $86.99 @ Mac Mall
CPU Cooler Noctua NH-D14 SE2011 CPU Cooler $86.99 @ Mac Mall
Motherboard Asus Z10PE-D16 WS SSI EEB Dual-CPU LGA2011-3 Motherboard $495.99 @ SuperBiiz
Memory Kingston 64GB (4 x 16GB) Registered DDR4-2133 Memory $437.99 @ SuperBiiz
Storage Samsung XP941 Series 256GB M.2-2280 Solid State Drive $205.98 @ Newegg
Storage Western Digital Red 5TB 3.5" 5400RPM Internal Hard Drive $199.99 @ Amazon
Storage Western Digital Red 5TB 3.5" 5400RPM Internal Hard Drive $199.99 @ Amazon
Video Card EVGA GeForce GTX 980 4GB ACX 2.0 Video Card (2-Way SLI) $503.99 @ SuperBiiz
Video Card EVGA GeForce GTX 980 4GB ACX 2.0 Video Card (2-Way SLI) $503.99 @ SuperBiiz
Case Phanteks Enthoo Pro ATX Full Tower Case $99.99 @ Amazon
Power Supply EVGA SuperNOVA 1000 P2 1000W 80+ Platinum Certified Fully-Modular ATX Power Supply $186.99 @ SuperBiiz
Optical Drive Asus DRW-24B1ST/BLK/B/AS DVD/CD Writer $15.88 @ OutletPC
Case Fan Noctua NF-S12B redux-700 33.5 CFM 120mm Fan $13.85 @ Amazon
Monitor Philips BDM4065UC 60Hz 40.0" Monitor $786.99 @ Best Buy
Monitor Philips BDM4065UC 60Hz 40.0" Monitor $786.99 @ Best Buy
Prices include shipping, taxes, rebates, and discounts
Total $5870.57
Generated by PCPartPicker 2016-01-04 02:59 EST-0500
  • 53 months ago
  • 1 point

Thank you! I like that motherboard choice (Asus Z10PE-D16 WS SSI EEB Dual-CPU LGA2011-3), based on what I have read. But it is an E-ATX size board. That will probably require a larger case than I wanted to use.

The case you recommended looks very nice (and has great reviews), but the dimensions are large:

Dimensions (H x W x D)
21.06" x 9.25" x 21.65"

That's almost 3 inches too tall for where this PC needs to fit.

Is there any other case you know that would accommodate this build and not be more than about 18" tall and 20" long? (The Antec P100 and the Antec Solo II have suitable dimensions, but they are designed for ATX motherboards.)

  • 53 months ago
  • 1 point

Let me look, most full towers are large.

  • 53 months ago
  • 1 point

I should have probably brought this up in our previous correspondence in the other thread, but here in this thread again I'm seeing no specific application mentioned that is going to make use of such a high end GPU, let alone multiple of them. I'm not sure that you understand the purpose of SLI and crossfire technologies. This leads me to believe that you may not have a very good understanding of your actual GPU requirements.

SLI/Crossfire are technologies that allow multiple physical GPU's to be treated as a single logical GPU, so that they can work together on the same real-time render operation. This really only applies to real-time viewports, and I am aware of no real-time viewports in creativity/design applications that have any support for SLI/Crossfire as it is simply not useful/necessary for this. In linux, SLI/CFX support is practically non-existent anywhere.

Software that uses GPU's for openCL or CUDA, can often scale to multiple physical GPU's if the workload can be easily split into separate buckets. Export renderers commonly do this, like for example, Cycles render (Blender) can assign render work to multiple GPU's, but this is NOT SLI/CROSSFIRE. The GPU's are NOT logically combined into "1" GPU for these operations. There is no need to, in fact, doing so would actually cause a reduction in performance. Assign separate work to each GPU makes more since than assigning a single bucket of work for a pair of GPU's to share.


What specific CAD software? How big are the projects?

What specific video editing software? What resolution, frame-rate, and format is the footage you are working with?

What specific rendering/raytracing software?

You can run 4 X 4K@60hz monitors from a GTX950 (3 X DP + 1 X HDMI 2.0). Might be pushing the limits of the 2GB VRAM depending on application VRAM usage but it is technically possible.

A GTX960 4GB would provide ample overhead for a 2 X 4K monitor configuration for most productivity software in linux, and even handle CAD viewports at high resolution pretty darn well as long as the viewport doesn't have really advanced modern visual quality features. (unlikely in any CAD software for linux).


Anyway, here's the build:

PCPartPicker part list / Price breakdown by merchant

Type Item Price
CPU Intel Xeon E5-2680 V3 2.5GHz 12-Core OEM/Tray Processor $1639.99 @ SuperBiiz
CPU Cooler Noctua NH-D9DX i4 3U 46.4 CFM CPU Cooler $59.99 @ Amazon
Motherboard Supermicro MBD-X10DAL-I-O ATX Dual-CPU LGA2011-3 Motherboard $301.98 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $249.99 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $249.99 @ Newegg
Storage Samsung 950 PRO 256GB M.2-2280 Solid State Drive $189.99 @ Amazon
Storage Sandisk Ultra II 960GB 2.5" Solid State Drive $244.00 @ Amazon
Storage Seagate Archive 8TB 3.5" 5900RPM Internal Hard Drive $228.98 @ PCM
Video Card Gigabyte GeForce GTX 980 Ti 6GB WINDFORCE 3X Video Card $599.99 @ Amazon
Case Antec P100 ATX Mid Tower Case $92.98 @ Newegg
Power Supply FSP Group 850W 80+ Platinum Certified Fully-Modular ATX Power Supply $154.99 @ Amazon
Optical Drive Asus DRW-24B1ST/BLK/B/AS DVD/CD Writer $15.99 @ Newegg
Other Lycom DT-120 M.2 PCIe to PCIe 3.0 x4 Adapter $29.00
Other GIGABYTE USB 3.1 Add-on Card Model GC-USB3.1 $24.00
Prices include shipping, taxes, rebates, and discounts
Total $4081.86
Generated by PCPartPicker 2016-01-04 03:01 EST-0500

Add another CPU, more RAM, more drives, etc, as needed....

PCPartPicker part list / Price breakdown by merchant

Type Item Price
CPU Intel Xeon E5-2680 V3 2.5GHz 12-Core OEM/Tray Processor $1639.99 @ SuperBiiz
CPU Intel Xeon E5-2680 V3 2.5GHz 12-Core OEM/Tray Processor $1639.99 @ SuperBiiz
CPU Cooler Noctua NH-D9DX i4 3U 46.4 CFM CPU Cooler $59.99 @ Amazon
CPU Cooler Noctua NH-D9DX i4 3U 46.4 CFM CPU Cooler $59.99 @ Amazon
Motherboard Supermicro MBD-X10DAL-I-O ATX Dual-CPU LGA2011-3 Motherboard $301.98 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $249.99 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $249.99 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $249.99 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $249.99 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $249.99 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $249.99 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $249.99 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $249.99 @ Newegg
Storage Samsung 950 PRO 256GB M.2-2280 Solid State Drive $189.99 @ Amazon
Storage Sandisk Ultra II 960GB 2.5" Solid State Drive $244.00 @ Amazon
Storage Samsung 850 EVO 2TB 2.5" Solid State Drive $697.99 @ Adorama
Storage Samsung 850 EVO 2TB 2.5" Solid State Drive $697.99 @ Adorama
Storage Seagate Archive 8TB 3.5" 5900RPM Internal Hard Drive $228.98 @ PCM
Storage Seagate Archive 8TB 3.5" 5900RPM Internal Hard Drive $228.98 @ PCM
Storage Seagate Archive 8TB 3.5" 5900RPM Internal Hard Drive $228.98 @ PCM
Storage Seagate Archive 8TB 3.5" 5900RPM Internal Hard Drive $228.98 @ PCM
Video Card Gigabyte GeForce GTX 980 Ti 6GB WINDFORCE 3X Video Card $599.99 @ Amazon
Compute Card Gigabyte GeForce GTX 980 Ti 6GB WINDFORCE 3X Video Card $600.00
Case Antec P100 ATX Mid Tower Case $92.98 @ Newegg
Power Supply FSP Group 1200W 80+ Platinum Certified Fully-Modular ATX Power Supply $249.99 @ Amazon
Optical Drive Asus DRW-24B1ST/BLK/B/AS DVD/CD Writer $15.99 @ Newegg
Other Lycom DT-120 M.2 PCIe to PCIe 3.0 x4 Adapter $29.00
Other GIGABYTE USB 3.1 Add-on Card Model GC-USB3.1 $24.00
Prices include shipping, taxes, rebates, and discounts
Total $10059.70
Generated by PCPartPicker 2016-01-04 03:07 EST-0500

If you do believe you have use of multiple GPU's, opt for a 1KW or larger PSU for this build, an example of a nice 1.2KW unit is shown in this "upgraded" build, obviously, it would make sense to buy that up front if you have use for it, otherwise, an 850W would be great.

Technically speaking, a 2 X E5-2680 V3 +GTX980Ti build with a few drives could run on a PSU as small as ~650W. By selecting something ultra high efficiency with lots of headroom, it will run quieter, thus, the 850W selection for the single GPU, 1KW+ for 2 GPU.

  • 53 months ago
  • 1 point

You said:

"A GTX960 4GB would provide ample overhead for a 2 X 4K monitor configuration for most productivity software in linux, and even handle CAD viewports at high resolution pretty darn well as long as the viewport doesn't have really advanced modern visual quality features. "

That sounds like my situation. You have educated me that I do not need SLI/CROSSFIRE. Furthermore, based on your feedback, it sounds like I actually do not need dual video cards.

You also said, "This leads me to believe that you may not have a very good understanding of your actual GPU requirements. " Apparently true!

I'm not gaming and I'm not doing any real-time rendering. I'm not using any software that has "really advanced modern visual quality features. " I'm not even editing 4K video yet, but I will do that with this machine. I'm not sure which software I'll use for that, but it will probably be Blender, although Lightworks is under consideration.

Phoronix says that the GTX 980 is the best GPU for Linux gamers. (See links below.) I'm not gaming, but this is relevant in terms of knowing that the drivers for that card work well. (Unfortunately, the good driver is proprietary.)

Based on your recommendation that a "GTX960 4GB would provide ample overhead for a 2 X 4K monitor configuration for most productivity software in linux" it sounds like a single 970 or 980 is a good choice.

However, given all this information, do you think an AMD card using the open source driver would be a good choice?

References:

NVIDIA GeForce GTX 980: The Best GPU For Linux Gamers Review - Phoronix http://www.phoronix.com/scan.php?page=article&item=nvidia_geforce_gtx980&num=1

NVIDIA GeForce GTX 980 Ti: Simply The Best For Linux Gamers Review - Phoronix http://www.phoronix.com/scan.php?page=article&item=nvidia-gtx-980ti&num=1

  • 53 months ago
  • 1 point

Yea in my experience the Nvidia propriatary driver is currently the best linux video driver available, though my Wife's computer runs linux with an AMD APU with AMD's proprietary driver and that still seems to be superior to the open source driver for openGL and translated directX (wine) performance. It's been awhile since I have tried the open source AMD driver on her machine (I don't like to go fussing with stuff if it isn't broke on her computer).

One thing I noticed there... take a look at the GTX960 performance in counter strike at 4K resolution in that link you have there. That's a real-time viewport of fairly decent visual quality running over 90FPS at 4K on a $200 GPU. This is very telling.

If you told me you were planning to use Davinci Resolve to work with 4K video, I would advise a GTX980Ti or Titan X, or a pair of them. If you were planning to do big complex models in blender and export high resolution video sequences, I would advise a pair GTX980Ti's.

If you're editing at 1080P in blender or lightworks, I think you'd be well served by a GTX960 4GB. I don't think anything stronger is going to be of much use but if it makes you feel better go for the GTX970/980.

  • 53 months ago
  • 1 point

What will make me feel better is to find the quietest option among all those choices you mentioned! :=)

I believe the 960 is a bit quieter than the 980 under load. I also found an article that used a 970 in a silent PC build with good results.

Zotac GTX 970 AMP! Extreme Core Noise Review | silentpcreview.com http://www.silentpcreview.com/Zotac_GTX_970_Extreme_Core

silentpcreview.com also gave an editors's choice to the Asus GTX 980 Ti Strix OC.

Do you happen to know whether the Asus Strix line is quieter than the Gigabyte series you included in the build?

BTW, I would like to have the option in the future to add a 2nd GPU should I feel the need for it later.

  • 53 months ago
  • 1 point

http://us.hardware.info/reviews/5867/10/15x-nvidia-geforce-gtx-970-and-980-review-what-card-to-pick-cooling-noise-levels-and-power-use

I believe the 960 is a bit quieter than the 980 under load.

Most Maxwell architecture GPU's actually run pretty quiet. The only exceptions being a few that are really tuned out with radical power envelopes.

Do you happen to know whether the Asus Strix line is quieter than the Gigabyte series you included in the build?

Actually I haven't been able to find a review of the non-G1 (standard windforce 980Ti) so I can't say with absolute certainty. I can however say, that the fans on the windforce and G1 are the same. The G1 editions of Maxwell GPU's from Gigabyte have higher clock speeds and REALLY high power envelopes set in firmware, so they are known for running louder than the competition especially when torture tested with furmark or other high-saturation GPGPU workloads. The non-G1 edition (standard Windforce model) is usually tuned to a much more conservative power envelope limit (closer to reference design). With the reduced power envelope, I would expect the windforce fan setup to keep thermals in check at noise levels that are competitive with other GTX980Ti's that run similar power envelope limits.

These considerations may be moot if you decide to scale back on the GPU to something less overkill.


BTW, I would like to have the option in the future to add a 2nd GPU should I feel the need for it later.

There are many ways to do this. I'll give you some ideas based on the X10DAL....

Consider configuring the machine with 1 GPU as the actual "video" card, (used to drive your monitors and draw your user interface), and the other as a dedicated compute card for export rendering (when the need arises), this way, when you are rendering or computing with the GPU, the responsiveness and smoothness of your desktop environment isn't impacted.

In a case with 7 expansion slots, you'll be force to install the 2 GPU's next to eachother for this build. If you use 2 longer GPU's, I would install the compute card in the lower slot, so that it's air intake is less obstructed, this will help reduce noise. Consider using a GTX960 as your "video" card (which will be able to run cool/quiet even when packed up against another GPU), and a more powerful GTX980/980Ti/TitanX as your compute card. Alternatively, choose an ITX length card (to use as the "video" card) to place below the compute card. The short length of the ITX card will obstruct the compute cards airflow far less.

If you did select a case with 8+ slots, the GPU's could be installed with a large gap between them (better airflow), and I would advise selecting an ITX length (or near there) GTX960/970 as your "video" card to place in the lowest slot on the motherboard, and put the compute card up in the closest 16X slot to the CPU's. The short GPU will minimize, or possibly eliminate the need for low profile right angle adapters/cables for things like USB/SATA connections.

  • 53 months ago
  • 1 point

Thanks for the link. That's a very relevant article... exactly the kind of info I wanted to know.

I like your ideas. Since I'm not using SLI, I see the advantages in choosing two different GPU's, as you suggest.

So far I think the Antec P100 case you recommended looks like the best option for me. If you have alternative suggestions, I'm all ears. However, the P100's dimensions make it a good fit for the space this PC must reside in.

  • 53 months ago
  • 1 point

Thank you for explaining this to me as well. I was just following OP's instructions as I don't know much about CAD. I always love your write ups.

  • 53 months ago
  • 1 point

Storage:

The Seagate Archive 8TB 3.5" 5900RPM Internal Hard Drive is a great recommendation. I was not aware of these drives. I'll use 1 or more of these along with a more traditional HDD and the two SSD's.

When it comes to SSD's, you recommended the Samsung 950 PRO 256GB M.2-2280 Solid State Drive. I read an article touting this one: SAMSUNG XP941 M.2 256GB PCI-Express 2.0 x4 MLC Enterprise Solid State Drive

Would you mind commenting briefly on the differences between those choices? I plan to use this to mount /boot and /root. (/home will be mounted on the 2TB SSD and audio, video, etc. will go on the HDD's.)

Speaking of the 2TB SSD, you recommended this one: Samsung 850 EVO 2TB 2.5" Solid State Drive I read that the PRO version was the best choice: Samsung 850 PRO 2TB 2.5" Solid State Drive Care to comment?

  • 53 months ago
  • 1 point

The 950 Pro is an NVME protocol PCIE SSD.

The XP941 is an AHCI protocol PCIE SSD.

AHCI protocol is the same as is used for SATA hard drives and SSDs,

The AHCI protocol is a bottleneck for upward performance scaling of SSD's. NVME replaces AHCI to solve this bottleneck.

The 950 Pro is far faster, and, actually costs less.

http://www.anandtech.com/show/7843/testing-sata-express-with-asus/4


Speaking of the 2TB SSD, you recommended this one: Samsung 850 EVO 2TB 2.5" Solid State Drive I read that the PRO version was the best choice: Samsung 850 PRO 2TB 2.5" Solid State Drive Care to comment?

In the 2TB size, both the EVO and PRO versions have the exact same high-end MHX 3 core controller with DDR3 cache. The difference between the drives is TLC vs MLC NAND. The EVO TLC NAND version has a write endurance rating of 150TB (~100GB/day for 4 years), the PRO MLC NAND version has a write endurance of 300TB (~200GB/day for 4 years).

Performance wise they are so close it doesn't matter (the AHCI protocol effectively bottlenecks both drives to the same real world performance, which is actually the case with most largish SSD's on this protocol these days). To be quite frank, when buying 1TB or larger AHCI SSD's, there is so much interleaving opportunity for the controller that there is very little performance discrepancy between a high end drive in this class like the 850 Pro and a budget friendly drive like a Mushkin Reactor (which in the 1TB size, routinely goes on sale for ~$250).

Speaking of which... the reactor is back on sale for $250 right now.... If it were my money, I'd have a pair of them in RAID0 for $500 instead. The Silicon Motion SM2246EN paired with 16nm Micron MLC NAND is proven quality/reliable guts that I would happily put in any high end single client workstation. You'll get as much or more performance from a pair of these in RAID0 for a /home drive than you will from a 2TB EVO or PRO IMO. The write endurance of a pair of these combined would also be on par with the 2TB 850 Pro.

  • 53 months ago
  • 1 point

That is very timely advice on the SSD's! I just ordered two of the Mushkin Reactor 1 TB drives while they are on sale. (First purchase for this build!)

I will also go with the 950 Pro. Your explanations are very helpful. I like to learn as I go and have at least a basic understanding of why one choice is better than another. Building a PC is always a good learning experience.

  • 53 months ago
  • 1 point

Amazon is listing a "version 2" of the

Seagate Archive HDD v2 ST8000AS0002 8TB 5900 RPM 128MB Cache SATA 6.0Gb/s 3.5" Internal Hard Drive Bare Drive. It costs $40 more. Is v2 worth $40?

http://www.amazon.com/dp/B00UZHA3H8/

  • 53 months ago
  • 1 point

Newegg's listing shown here on PCPP is also for a "V2" drive.

Never noticed this distinction. Not sure if there even is a distinction.

http://www.seagate.com/www-content/product-content/hdd-fam/seagate-archive-hdd/en-us/docs/archive-hdd-dS1834-3-1411us.pdf

Order it from newegg for $235 ($6 more than the lowest shipped price on PCPP) to get one sold with "V2" in the listing if it makes you feel better. I can't find any information about whether there really is a "V2" of this specific drive, or if this designation is having to do with something else.

http://www.newegg.com/Product/Product.aspx?Item=N82E16822178748

  • 53 months ago
  • 1 point

I believe you are right -- no distinction. The "version 2" drive and the non-version-2 drive both have serial number ST8000AS0002. I think the 2 at the end might be the indicator of "version 2" but apparently all the current offerings are for the same item. I ordered the less expensive one.

  • 53 months ago
  • 1 point

Cases:

The Antec P100 ATX Mid Tower Case looks like the perfect size case for me. Thanks for that recommendation. I assume you are confident everything will fit in that case. Will it also work if I end up using 2 GPU's later?

The dimensions of the P100 are: H x W x D)
19.00" x 8.70" x 20.60"

If the P100 will work for this build, what about this one?
Antec Sonata Series SOLO II Black Aluminum / Steel ATX Mid Tower Computer Case - Newegg.com http://www.newegg.com/Product/Product.aspx?Item=N82E16811129177&cm_re=antec_solo_ii-_-11-129-177-_-Product

The dimensions of the Solo II are: H x W x D)
18.50" x 8.10" x 17.30"

That's just slightly smaller. The Solo II was recommended as a very good quiet case. Do you think this build will fit in the Solo II?

Both of these cases are several years old. However, I get the feeling these Antec cases will be a very good match for my needs. I appreciate the recommendation.

  • 53 months ago
  • 1 point

The Antec P100 ATX Mid Tower Case looks like the perfect size case for me. Thanks for that recommendation. I assume you are confident everything will fit in that case. Will it also work if I end up using 2 GPU's later?

Yes I am pretty confident your build will fit comfortably in this case. 2 GPU's should work fine. With the included PCIE cards that will fill or cover all slots ;)


what about this one? Antec Sonata Series SOLO II...... Do you think this build will fit in the Solo II?

I think that would be a pretty tight squeeze there for the motherboard against the 3.5" drive cage. May or may not fit. The P100 has the space for the 10" deep motherboard and (IMO) a far superior drive-cage configuration for such a build. I know you're saying you only expect a few drives... I thought the same thing when I bought my Fractal Design Core 3500's for my workstations (got them on sale for $55 each), they only have 4 X 3.5/2.5" drive sleds, which I never expected to fill, and then... woops, they are all full now.

  • 53 months ago
  • 1 point

I would like to do some thermal imaging on my system to help me position the fans, etc. Do you have any advice or thoughts?

Heat seeker: Meet the thermal-imaging camera you can afford - CNET http://www.cnet.com/news/heat-seaker-thermal-imaging-camera-for-the-masses/

Blogs:

Computer Case Cooling Thermal Image - Thermal Imaging Blog http://thermalimaging-blog.com/123-computer-case-cooling-thermal-image/

Puget Systems Thermal Imaging https://www.pugetsystems.com/thermal.php

  • 53 months ago
  • 1 point

I would like to do some thermal imaging on my system to help me position the fans, etc. Do you have any advice or thoughts?

Never done it. To be honest I probably wouldn't bother unless I were designing a system specifically intended for overclocking at very high caliber. (competition overclocking? or perhaps, a certified-overclocked rig as a workstation).

I would observe temperatures reported by the CPU's, motherboard sensors, GPU, and Hard drives, and work from there. Move fans around, add and delete fans as required until you're happy with temps.

Quiet computing REQUIRES that you embrace thermal margins rather than fight them. Use temperature differentials to your advantage. Fighting for uber-low temps everywhere will cost you decibels.

  • 53 months ago
  • 1 point

My intention for thermal imaging was not to try from uber-low temps, but to use it to help me keep the system quiet. I got the idea partly from here:

Puget Systems Thermal Imaging https://www.pugetsystems.com/thermal.php

https://www.youtube.com/watch?feature=player_detailpage&v=Y9m48H6fKHI#t=658

If you were to watch the entire video, they have a discussion about positioning of fans (and other things) to achieve the quietest system that doesn't overheat.

  • 53 months ago
  • 1 point

wow dude.. You could seriously think about a PC-ADVISER STARTUP.. Might make you a Millionnaire in a year's time.. I'm really serious!! Meanwhile ppl are damn plain lucky to get such professional advice for free..

  • 53 months ago
  • 1 point

Revised Full Build:

I'm going to start with just one CPU, but this is what the full build with 2 CPU's will look like... actually, I expect to eventually add a 2nd GPU. (And this list doesn't include a few smaller items such as case fans.)

Comments are appreciated!

PCPartPicker part list / Price breakdown by merchant

Type Item Price
CPU Intel Xeon E5-2680 V3 2.5GHz 12-Core OEM/Tray Processor $1639.99 @ SuperBiiz
CPU Intel Xeon E5-2680 V3 2.5GHz 12-Core OEM/Tray Processor $1639.99 @ SuperBiiz
CPU Cooler Noctua NH-D9DX i4 3U 46.4 CFM CPU Cooler $59.99 @ Amazon
CPU Cooler Noctua NH-D9DX i4 3U 46.4 CFM CPU Cooler $59.99 @ Amazon
Motherboard Supermicro MBD-X10DAL-I-O ATX Dual-CPU LGA2011-3 Motherboard $301.98 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $244.99 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $244.99 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $249.99 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $249.99 @ Newegg
Storage Samsung 950 PRO 256GB M.2-2280 Solid State Drive $189.99 @ Amazon
Storage Mushkin Reactor 1TB 2.5" Solid State Drive $249.99 @ Newegg
Storage Mushkin Reactor 1TB 2.5" Solid State Drive $249.99 @ Newegg
Storage Western Digital Red 6TB 3.5" 5400RPM Internal Hard Drive $249.00 @ Adorama
Storage Seagate Archive 8TB 3.5" 5900RPM Internal Hard Drive $228.98 @ PCM
Video Card Asus GeForce GTX 960 4GB Video Card $204.99 @ Micro Center
Case Antec P100 ATX Mid Tower Case $92.98 @ Newegg
Power Supply FSP Group 1200W 80+ Platinum Certified Fully-Modular ATX Power Supply $239.99 @ Newegg
Optical Drive Asus DRW-24B1ST/BLK/B/AS DVD/CD Writer $15.99 @ Newegg
Other Lycom DT-120 M.2 PCIe to PCIe 3.0 x4 Adapter $29.00
Other GIGABYTE USB 3.1 Add-on Card Model GC-USB3.1 $24.00
Prices include shipping, taxes, rebates, and discounts
Total (before mail-in rebates) $6491.80
Mail-in rebates -$25.00
Total $6466.80
Generated by PCPartPicker 2016-01-05 02:36 EST-0500
  • 53 months ago
  • 1 point

ALLAN_M_SYSTEMS is the last word.. Period!!

  • 53 months ago
  • 1 point

Yes, indeed! I can tell Allan_M_Systems speaks from a lot of knowledge and experience.

  • 53 months ago
  • 1 point

I notice you have 2 different types of memory in this build (2 dual ranked DIMMs, 2 quad ranked DIMM's). It would be best to use all the same exact DIMM's....

Speaking of which, here's the dual ranked versions for less money: http://www.superbiiz.com/detail.php?p=D42132G4S&c

  • 53 months ago
  • 1 point

Thanks for catching that. In actually, I ordered 4 of these:

SAMSUNG 32GB 288-Pin DDR4 SDRAM DDR4 2133 (PC4 17000) Server Memory Model M386A4G40DM0-CPB - Newegg.com http://www.newegg.com/Product/Product.aspx?Item=N82E16820147384

The NewEgg page does not mention rank. I'm not sure which rank this is. I canceled my order until I sort it out. I'll order all the same. But I'm not sure if I should order dual ranked or quad ranked.

Just found this:

Memory Configuration Guide - X10_memory_config_guide.pdf http://www.supermicro.com.tw/support/resources/memory/X10_memory_config_guide.pdf

and

Motherboard manual: http://www.supermicro.com/manuals/motherboard/C600/MNL-1701.pdf

How would you recommend I populate the memory on this motherboard with 2 CPUs and 4 sticks of 32 GB RAM?

The manual says:

2 CPUs & 4 DIMMs CPU1 + CPU2 P1-DIMMA1/P1-DIMMB1, P2-DIMME1/P2-DIMMF1

That looks a little strange. From the manual, it would appear that all the RAM is on one side of the CPU...

  • 53 months ago
  • 1 point

Thanks for catching that. In actually, I ordered 4 of these: SAMSUNG 32GB 288-Pin DDR4 SDRAM DDR4 2133 (PC4 17000) Server Memory Model M386A4G40DM0-CPB , ....I'm not sure which rank this is. I canceled my order until I sort it out.

The DIMM's you ordered/cancelled are quad ranked and they would work fine for this.

But I'm not sure if I should order dual ranked or quad ranked.

In this particular application it wouldn't matter since there are only 4 DIMM slots per memory controller on the X10DAL, which makes it impossible to install more than the maximum number of supported ranks per channel (8). Some C612 chipset boards for the E5 Haswell-EP series have 12 DIMM slots per memory controller (3 slots per channel, 24 slots total for a dual socket board). On those boards, it's important to use DIMM's arranged into fewer ranks when loading them up with crap-tons of memory so as not to exceed the maximum number of supported ranked.

There is a performance benefit from rank interleave, but there is no tangible performance scaling beyond 2 ranks per channel. All 32GB DDR4 DIMM's are going to be arranged into at least 2 ranks, so for a build on the X10DAL with 32GB DIMM's, I would buy the least expensive 32GB Registered ECC DDR4 DIMM's from Crucial/Kingston/Samsung I could get my hands on. (The ones that I linked to there are superbiiz for a little over $200 each would probably be the best deal).

How would you recommend I populate the memory on this motherboard with 2 CPUs and 4 sticks of 32 GB RAM?

Exactly as the motherboard manual says to populate the memory for that configuration.....

The manual says: 2 CPUs & 4 DIMMs CPU1 + CPU2 P1-DIMMA1/P1-DIMMB1, P2-DIMME1/P2-DIMMF1 That looks a little strange. From the manual, it would appear that all the RAM is on one side of the CPU...

I'm not sure I follow what is "Strange" about this. If you look at the board diagram in the quick reference guide or manual, there would be 2 DIMM's populated above CPU1 and 2 DIMM's populated below CPU2.

The memory configuration guide shows a generic diagram, the physical arrangement of the slots can vary from board to board. Note, there is no actual conflict of information; the order of population is based on the slot labeling first and foremost, not the relative physical position of the slot.

The population recommendation is actually ideal for the X10DAL, as it leaves the DIMM slots behind PCIE SLOT6 empty until more than 2 DIMM's per CPU are installed, which makes room for a longer PCIE device there. For your build, this is a non issue, as this is likely where you'll want to install the USB 3.1 card, which is physically very short so wouldn't interfere with the memory anyway.

  • 53 months ago
  • 1 point

Thanks for clearing up those several points.

I plan to order the Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory dual rank memory from the link you provided (4 x $212 each).

You are right. The way the memory slots are populated isn't really strange. It just surprised me at first because I don't have any experience with a dual slot motherboard.

I was just reading this very interesting free ebook:

DDR4 For Dummies, HP Special Edition - DDR4forDUMMIES.pdf http://files.hypervisor.fr/doc/DDR4forDUMMIES.pdf

  • 53 months ago
  • 1 point

Am I correct that the dual rank memory is a better choice than quad ranked, in general? I know you said it doesn't make a difference with this motherboard, but from the reading I did it seems like dual rank is a better choice in general. And the SuperBiiz price for dual rank is lower than the NewEgg price for quad rank, so that would seem to be the clear choice (if I'm understanding all this correctly).

  • 52 months ago
  • 1 point

Is it normal for the fan to be noisy in the "FSP Group 1200W 80+ Platinum Certified Fully-Modular ATX Power Supply"? I have the system running on a test bench and everything is quiet except for the PSU. I have to do something about that...

  • 52 months ago
  • 1 point

It should be very quiet. Like 900RPM all the way up till ~800W....

https://www.techpowerup.com/reviews/FSP/PT-1200FM/6.html

If it's running correctly, it should be one of the quietest PSU's you could use for this build.

  • 52 months ago
  • 1 point

My guess is that the fan in my PSU may be faulty. (Maybe it has a faulty bearing.) I'm deciding whether to take the PSU apart, which will void the warranty, and replace the fan; or send it back and wait for a replacement which will further delay the build.

Do you happen to know how the fan is connected internally? Is it hard wired or does it have a plug like a typical case fan?

I have a Noctua 140 mm fan I could use as the replacement.

  • 52 months ago
  • 1 point

I would not advise taking action to fix the fan yourself on this one. It would need to be a fan with a similar speed vs voltage range. Whomever you bought it from should exchange it.

  • 52 months ago
  • 1 point

I have a few questions about performance of this system. Should I ask here or start a new topic?

My first question is about the performance of the Samsung 950 PRO 256GB M.2-2280 NVMe SSD. I ran this quick benchmark:

Under 1 minute hdparm benchmark a drive http://www.servethehome.com/quick-request-under-1-minute-hdparm-benchmark-a-drive-if-you-can/

sudo hdparm -tT --direct /dev/nvme0n1

Across multiple tests my results were about 1.3 GB/s read speed. I expected at least 2 GB/s and hoped I might see 2.5 GB/s.

What results should I expect on this test? The system is running two 2660 v3's and 64 GB of RAM on each CPU (Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory). It's running Kubuntu 15.10.

Is there a better test I can run?

(I thought about the Phoronix test suite, but I am hoping there's another option.)

Thanks!

  • 52 months ago
  • 1 point

http://www.anandtech.com/show/9702/samsung-950-pro-ssd-review-256gb-512gb/8

They tested the 256GB version at ~1.3GB/s read speed also.

Sounds like your results are right about where they should be.

the 512GB model did ~2GB/s read speed.

Honestly, the sequential bandwidth of an SSD used as a system/software drive is not a particularly useful metric. The bandwidth in random read of smaller file sizes is really where SSD's shine, and really where the 950 sets itself apart from previous generation AHCI SSD's. A 950 pro can load/boot software about twice as fast as high end AHCI drives, and there isn't much difference in performance between the 256GB and 512GB version in these workloads.

  • 52 months ago
  • 1 point

Thanks for that link and the information. I bought the 512 GB model. It sounds like my results are a bit low for that one.

Let me ask... Is there any chance I need to move the PCIe adapter to a different slot? Are there any BIOS settings I should check?

I assume any recent Linux kernel should be fine. Let me know if you think I should check anything else. Thank you.

BTW, here is a very interesting quote from the article:

"The sequential read performance is probably the best showcase of what the PCIe 3.0 x4 interface can do. The 256GB 950 Pro attains over half of the link speed, but the 512GB is again bogged down by something—relatively speaking, since it's still more than twice the speed of SATA and faster than even the Intel SSD 750. It's possible that the 950 Pro isn't faithfully implementing the secure form of the NVMe format command and some lingering fragmentation is preventing the 512GB drive from performing as specified. "

I'm still reading the rest of the article.

  • 52 months ago
  • 1 point

I usually use gnome-disk-utility to do quick-n-easy benchmarking for disks in linux.

When I want to do more in-depth testing, I turn to the phoronix-test-suite and run the benchmarks in a terminal. (the web interface is still sort of buggy IMO).


Lets see.... you went with the X10DAL motherboard right?

Which slot is the adapter/SSD installed in?

  • 52 months ago
  • 1 point

Yes, I went with the X10DAL motherboard. I like it a lot.

I had the adapter/SSD in PCIe slot 1 (closest to CPU's) when I got the 1.3 GB/s read speed from this test:

sudo hdparm -tT --direct /dev/nvme0n1

The next day I moved the adapter/SSD to slot 2 and didn't change anything else. Then I got results of 1.75 GB/s from that same test. I repeated the test multiple times (at both settings).

The only factor, besides PCIe slot, that I can think of is temperature. I believe this drive does get slower when it gets hotter and I did not measure temperature so I don't know if it is a factor. But subjectively, the drive was not hot in either test.

Today I ran the phoronix aio-stress test. The SSD is still in slot 2. The aio-stress test reported 1759 MB/s.

I'm running the phoronix-test-suite in the terminal too. Unfortunately, I am having trouble getting many of the tests to download and install so I can't run all the tests I want to run. I'm running phoronix-test-suite v 6.0 stable.

On another subject I subjectively tested two different video cards today.

I compared the Nvidia GTX 970 card to a Radeon 390X. Both cards are from the Asus Strix line. Using open source drivers, the Radeon card is definitely better performing for everyday tasks.

In fact, with the Nvidia card, the system runs horribly slow. With one Firefox browser window open and a couple dozen tabs and just one terminal window open with nothing running, the performance is worse than my slowest laptop. For example, scrolling in Firefox is super choppy. (I try to scroll down the page and nothing happens, then it suddenly jumps to the bottom.) Another example is right clicking and bringing up a context menu. The highlighted menu item lags way behind normal/slow mouse movements. Its is so bad that I couldn't work on the computer. That's why I pulled out the GTX 970 and tried the Radeon card today. I know the Radeon open source drivers are better. But is there potentially something else going on? Would you expect Firefox to lag that much on a system like this with open source Nvidia drivers?

  • 52 months ago
  • 1 point

Update on the GPU's. I mentioned that I have tested both a GTX 970 and a Radeon 390X with open source drivers. The Radeon was better with open source drivers. But the overall performance of both cards was not acceptable to me.

I intend to test both with proprietary drivers, and I expect the GTX to be better based on what I've read and been told (including here).

However, yesterday I had a chance to test two brand new GTX 970's together. So I proceeded with that test using just two monitors. For that test I have to use Nvidia's propertary drivers.

I followed the usual instructions, which include:

sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt-get update

Then use driver manager to install Nvidia driver.

I ultimately ended up doing more than these simple steps. (E.g, even though this is a new Kubuntu 15.10 installation and no proprietary drivers had been installed previously, I purged nvidia and bumblebee.)

I also tried both driver version 352 and 361 (the recommended driver and latest).

However, in every case the system is incredibly unstable and not even usable. It crashes many many times. kwin_x11 has segmentation faults, for example. Depending on the driver and the settings, it either will not boot all the way to the desktop, or it crashes repeatedly if it does get to the desktop.

So far I'm not finding a lot of topics about this online. It seems most people don't have much problem with Nvidia proprietary drivers. But I saw very few posts by people running two Nvidia GPU's in Ubuntu. Is there a forum I should check for more help with this?

For now, I guess my next step is to test single GPU's with proprietary drivers...

  • 52 months ago
  • 1 point

I would advise installing only 1 GTX970, then installing the nvidia proprietary drivers, ideally using a distro specific video driver install tool, and if the distro you're using doesn't have this, then I'd say dump it and move on because package manager installs of video drivers are usually not well implemented and will be incomplete. Most modern/fancy/UI oriented distros require a custom install script. I'm not sure if ubuntu has an automated built-in tool for this.

After you have nvidia's driver installed and working, then add the second GPU. The driver will recognize it and allow you to configure it. I've run as many as 3 nvidia GPU's in linux....

With a pair of modern Nvidia GPU's you can run up to 8 monitors, however, each GPU will have to have it's own Xscreen configured with up to 4 monitors per Xscreen. (this can all be done from the nvidia control panel). If you're not using more than 4 monitors, I would advise using only 1 GPU so that they can all be configured in the same Xscreen.


I do not use ubuntu/debian based distros anymore.

When I first started using linux, I was told to start with ubuntu because it is "easier."

I call bull.

I wasted months, maybe a year or more fighting with ubuntu trying to "learn" linux (this was quite a few years ago now). Instability was common, and I was constantly breaking the install with what should have been routine changes to settings for user interface elements and routine software installs.. Trying to have media codecs, video drivers, and a kernel version all new enough to work with hardware that wasn't from the stone age, in ubuntu, with a few custom desktop environment settings, was basically a nightmare.

All I learned was how to re-install the OS over and over again, and all the little buttons that I couldn't touch else break the install beyond my expertise to recover. Linux Mint was much better, but was still a struggle at times.

I'm over 2 years on a Manjaro XFCE install for my primary desktop. I can't seem to break it no matter what I do, and even when I think I have broke it, it turns out there's usually some very simple way to recover and get it working again with a simple command or a config edit. Point and click install's of new kernel versions, video drivers, and lots of linux software without any conflicts or stability issues is a daily norm. In my experience. YMMV. I also like openSuse and Mageia.

  • 52 months ago
  • 1 point

Thanks for your reply. I posted two updated messages in the forum thread we've been using. In the first post I listed the current build components so you (and everyone) could see exactly the parts I am using now. (I changed a few things, but not much.)

In the second post today I described the main problem I'm having with this build -- GUI performance. I would love to know your thoughts on that issue.

In regard to distros, this is a work computer and I am limited to using Kubuntu 14.04, 15.10 or (soon) 16.04. Personally, I prefer opensuse. I have also wanted to try Arch, but so far I haven't gotten around to it. And I am interested in checking out Manjaro or Mageia, but I have to stick with Kubuntu for this work computer.

Basically, I have worked enough with Kubuntu that I feel comfortable dealing with any reasonable issues. I can usually fix anything I break ;-)

The last few days I have been testing with a single GTX 970 GPU. I'm using a very solid driver installation method. It is based on this PPA: Proprietary GPU Drivers : “Graphics Drivers Team” team https://launchpad.net/~graphics-drivers/+archive/ubuntu/ppa

I do not believe the driver installation is the problem.

  • 52 months ago
  • 1 point

problem I'm having with this build - GUI performance. I would love to know your thoughts on that issue.

In the Nvidia X Server Settings, navigate to PowerMizer and adjust to "Prefer Maximum Performance."

See if that helps any.

this is a work computer and I am limited to using Kubuntu 14.04, 15.10 or (soon) 16.04

I always felt like KDE was sluggish feeling. Maybe that's why I like XFCE. That being said, v-sync supporting window managers in linux often feel a bit sluggish. The XFCE window manager does not properly support this so I switch back and forth from compiz to xfwm4 on an as needed basis. Better interactive performance with xfwm, better video playback with compiz.

I believe this may be partly a consequence of the archaic X11 system. My advise would be to try to run KDE 5.5 / Wayland and see if that doesn't work better. This would be a good excuse to try Kubuntu 16.04 as soon as possible. I've actually been meaning to try Manjaro KDE to see if KDE has improved since last I tried it as a primary desktop environment.

Bottom line here, is that if you're having interactive performance problems on a 20 core haswell rig with any modern GPU with GDDR5 VRAM, you can be darn well sure it's not for lack of hardware. Almost certainly the issue is software related. I have noticed interactive performance varies dramatically with different desktop environments, window managers, and environment settings. Linux has plenty of gremlins.

  • 52 months ago
  • 1 point

I'm upgrading to Kubuntu 16.04 now. :-) I'll let you know how that goes.

Question - it seems to me that with the same OS and equal settings, the i7 with the GTX 960 has better interactive performance than this system. Should that be so?

The test I'm using now is:

  1. open Firefox
  2. open this movie trailer: https://www.youtube.com/watch?v=N0io2w_6vT8 (Both computers have the YouTube setting "The HTML5 player is currently used when possible.")
  3. resize the Firefox window horizontally while the video is playing. I note how far the window resizing lags the mouse cursor at different mouse speeds.
  4. I repeat the test with different system settings.

As I said, at equal settings, the i7 is more responsive. At some settings the difference is very large.

BTW, KDE can be a very fast desktop environment. One reason I like KDE is because everything is configurable. By turning off desktop animation effects and tweaking other settings, KDE can run very well even on a very old and slow computer.

With the right settings*, I can get the i7 system to resize the Firefox window playing the movie trailer with almost no lag behind fast mouse movement. I cannot get this system to respond that well no matter what the settings.

Is that expected?


footnote:
*The most responsive settings include turning off most desktop animations and enabling "Resize Windows" which is an effect that scales windows with a fast texture scale rather than redrawing/updating window contents.

  • 52 months ago
  • 1 point

With a pair of modern Nvidia GPU's you can run up to 8 monitors

I currently run 3 monitors. That's enough for me. In fact, I hope to switch to just 2 40" 4k monitors in the future.

Here's an interesting thread with a picture of a guy's 6-monitor setup on Linux.

AMD Crossfire On Linux? | GamersOnLinux http://www.gamersonlinux.com/forum/threads/amd-crossfire-on-linux.238/

  • 52 months ago
  • 1 point

Yea....

The dirty little secret there is that in order to support more than 2-3 monitors (depending on the specific AMD GPU in question), the additional monitors must be either native displayport, or connected to an active hub that generates legacy digital video for each output. On an AMD GPU with 2 X Displayports + 2 X HDMI/DVI, it is possible to support up to 6 X 1440P displays. Doing so requires the addition of 2 active displayport hubs. Each ~$70. Alternatively such a card would support up to 2 X 4K + 2-3 X 1440P depending on the specific GPU. (some AMD GPU's support 2 legacy digital video streams, some support 3. I think the new Fiji cards might support 4 legacy digital video outputs but I'm not sure).

Nvidia is indeed limited to 4 Monitors per GPU, but a lot of them (even inexpensive cards) support up to 4 X 4K monitors via 3 X DP + HDMI 2.0 right out of the box. And each of those ports supports full legacy digital as well, so can be passively (inexpensively) adapted to DVI/HDMI for lower resolution displays that lack display-port native connectivity.

  • 52 months ago
  • 1 point

Revised Full Build - UPDATED:

I'm still testing, benchmarking and trading parts here and there. This is the most current parts list.

PCPartPicker part list / Price breakdown by merchant

Type Item Price
CPU Intel Xeon E5-2660 V3 2.6GHz 10-Core Processor $1359.99 @ SuperBiiz
CPU Intel Xeon E5-2660 V3 2.6GHz 10-Core Processor $1359.99 @ SuperBiiz
CPU Cooler Noctua NH-D9DX i4 3U 46.4 CFM CPU Cooler $59.99 @ Amazon
CPU Cooler Noctua NH-D9DX i4 3U 46.4 CFM CPU Cooler $59.99 @ Amazon
Motherboard Supermicro MBD-X10DAL-I-O ATX Dual-CPU LGA2011-3 Motherboard $301.98 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $229.99 @ SuperBiiz
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $229.99 @ SuperBiiz
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $229.99 @ SuperBiiz
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $229.99 @ SuperBiiz
Storage Samsung 950 PRO 256GB M.2-2280 Solid State Drive $181.99 @ Best Buy
Storage Mushkin Reactor 1TB 2.5" Solid State Drive $249.99 @ Newegg
Storage Western Digital Red 6TB 3.5" 5400RPM Internal Hard Drive $246.99 @ SuperBiiz
Storage Seagate Archive 8TB 3.5" 5900RPM Internal Hard Drive $222.99 @ Mac Mall
Video Card EVGA GeForce GTX 970 4GB SSC ACX 2.0+ Video Card $324.99 @ Amazon
Case Fractal Design Define R5 w/Window (Black) ATX Mid Tower Case $119.99 @ NCIX US
Power Supply FSP Group 1200W 80+ Platinum Certified Fully-Modular ATX Power Supply $229.99 @ Newegg
Optical Drive Asus DRW-24B1ST/BLK/B/AS DVD/CD Writer $15.99 @ Newegg
Other Lycom DT-120 M.2 PCIe to PCIe 3.0 x4 Adapter $29.00
Other GIGABYTE USB 3.1 Add-on Card Model GC-USB3.1 $24.00
Prices include shipping, taxes, rebates, and discounts
Total (before mail-in rebates) $5717.82
Mail-in rebates -$10.00
Total $5707.82
Generated by PCPartPicker 2016-02-04 16:53 EST-0500

However, I'm experiencing performance issues. In trying to understand the problems, I am exchanging parts. For example, I have tried two other video cards (a Radeon R9 390X and an Asus GTX 970 Strix).

  • 52 months ago
  • 1 point

The main problem with this build:

One of the main reasons I wanted an upgraded system is graphical performance -- not for gaming, but for simple everyday productivity. My current system does a few very annoying things. When I resize a window with the mouse by dragging the side edge of the window (to make it wider or narrower), the window resizing lags significantly behind the mouse -- often by several seconds delay.

Another symptom is selecting a menu item from a drop down menu. I point to the item with the mouse, but the GUI doesn't catch up immediately and I have to pause before I can click. These are symptoms of a GUI that is slow enough to impact my productivity and to frustrate me. I find myself having to do a lot of very common tasks in slow motion so the GUI can keep up with me. Even when selecting text (to copy/paste), I have to be slow and deliberate. Very annoying. (BTW, this is a single CPU Xeon E3-1230 / Asus P8B WS system. But the video card is a ATI Radeon HD6750 driving 3 monitors. It can't handle the requirements well.)

So I wanted this new build to have instant responsiveness to these simple / routine tasks of interacting with the GUI. Unfortunately, the above build is proving to be surprisingly slow in regard to these metrics.

I ran a number of CPU, disk and other benchmarks and they seemed fine. But the GUI still felt slow. So I have simply resorted to "benchmarking" it by using a Firefox window and resizing it with the mouse. I note how well the window border follows the mouse.

I am comparing the above build to another system I built recently. It is an Intel Core i7-6700K in an Asus Z170A with 32 GB of RAM, a GTX 960 and a Mushkin Reactor SSD. No slouch of a system, but it cost about 1/4 the amount of this one.

When I do my simple window-resizing test on that i7 system, there is zero lag (as far as I can see). The windows resize as fast as I can move the mouse. The GUI is fully responsive.

That video card is a 960. The one in this build is a GTX 970.

Both systems are running the same OS -- Kubuntu 15.10. I built both systems and I installed the OS on both and I did everything the same way I have always done it. They even use the same video driver -- the Nvidia 361 blob from:

Proprietary GPU Drivers : “Graphics Drivers Team” team https://launchpad.net/~graphics-drivers/+archive/ubuntu/ppa

So, should the GUI responsiveness on this build be slower than the one I compared it to? I realize the i7 is clocked much higher. Are my expectations for this build off base? Or is something wrong?

I even turned off all desktop animation effects and set KDE to the settings that would be used for a low powered notebook and I still see significant slowness in the window resizing test. The system I'm comparing against is using normal KDE settings and has zero lag on this test.

I have reinstalled the OS several times and I have carefully installed the Nvidia drivers several times. There is almost no opportunity to install the driver incorrectly using the PPA linked above (afaik). The driver installation has gone smoothly. No problems there.

  • 52 months ago
  • 1 point

Revisiting the issue with your interactive performance. Just spent some more time with firefox open on Manjaro with a few tabs of websites I commonly visit. (like this one).

I immediately noticed significantly higher CPU usage from ordinary browsing and typing within firefox. For example, simply trying to type this response in firefox, pegs a CPU core at 100% and it can't even keep up with my fumbling fingers typing speed. The whole system feels like it is dragging its feet. I then attempted to open Chrome, and do other things on the computer. I opened up a system monitor to see more detail about what was going on. FireFox had 1 thread running at 100% CPU usage while I was doing nothing. It stayed that way for several minutes before subsiding. The laggy/choppy interactive performance continued system wide regardless. As soon as I closed firefox, (Which took it about 10 seconds of pegging a CPU core), the system responsiveness and performance returned to normal.

I ditched FireFox due to its awful performance many years ago. Here I'm using much more powerful hardware today and FireFox is even worse now than it was back then. I'm not even sure how they are still in business... Who would use a browser that brings a modern CPU to it's knees just to type something? That causes the entire desktop environment and all running applications to lag as if they were running on a single core system from 15 years ago?

My advise, would be to write a lengthy letter filled with foul belittling language to the FireFox developers, then stop using it immediately.


Based on current observations, chrome is running better with 100 tabs open on an Athlon II X4+8GB+GT520 rig, than FireFox does with 3 tabs open on an overclocked FX-8350+16GB+GTX650Ti.

  • 52 months ago
  • 1 point

That's really interesting. I will try Chrome as a test and see how that does. Are you using Chrome or Chromium (oss)?

UPDATE on other stuff:

Make sure NUMA is enabled in BIOS

It is.

make sure numa support is enabled in your kernel

I'm not sure... reinstalling the OS now and also not sure how to check it. Reading this:
linux kernel - enabling numa for Intel Core i7 - Unix & Linux Stack Exchange http://unix.stackexchange.com/questions/92302/enabling-numa-for-intel-core-i7

Operating System:

I was not able to get Kubuntu 16.04 to install using any method known to me. The installer has a bug that makes it crash. (It's alpha.) Changing the repos to devel and upgrading that way made the system unbootable so I took that chance to try another distro...

I installed opensuse Tumbleweed (and the proprietary Nvidia drivers) as a test. The interactive performance was largely the same. But I was testing with Firefox (before I read your message).

Now I'm getting ready to try something else (maybe non-KDE).

Next Steps?

Let's say I'm stuck using some slow software (e.g., Firefox). Is there anything else I can do in terms of hardware to help overcome that? (My company is standardized on Firefox because they have a problem with Google, and also on Kubuntu.)

I can turn off all animations in Kubuntu and I can make any configuration changes I wish.

Are there any other bios settings I need to check?

  • 52 months ago
  • 1 point

The system has more than enough power to do all the fancy interactive graphics and compositing in a modern desktop enviroment. You shouldn't have to resort to disabling all the eye candy stuff to make any sort of newish hardware run it smoothly. It should run fast regardless. There's obviously something else amiss here.


I've used both chrome and chomium back and forth over the last several years. Sometimes an update breaks one or the other so I switch for awhile on an as-needed basis. I just use what works and performs. I have no need to placate to specific software.


As far as other things to try. Consider configuring the cpupower config file manually...

/etc/default/cpupower

mc_scheduler=1
smp_scheduler=0
perf_bias=0

https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/7/html/Power_Management_Guide/cpu_power_saving.html

Not sure if it will help any, but something to play with anyway. If you don't notice any difference go ahead and comment these settings out again.

I believe you'll use the numactl command to make more sophisticated policy changes for scheduling, and it should give you some insight into whether the system is indeed running with numa support at the kernel.

  • 52 months ago
  • 1 point

simply trying to type this response in firefox, pegs a CPU core at 100%

That is most likely because of the spelling checker

  • 52 months ago
  • 1 point

yea I'm not sure what the deal is but it CPU load was really bad... I may have to experiment more with that to determine what the deal is. I'm now typing in firefox on a fresh install of Manjaro KDE... Been meaning to try this out for awhile as KDE has come a long way since I tried it last.

For whetever reason, I'm seeing no significant performance issues while typing or browsing here in FireFox on KDE. However, I have to agree that the window resizing with youtube open is not very smooth. Functional, but nowhere near as responsive as XFCE.

Scrolling performance in FireFox is crap compared to Chrome on both desktop enviroments. Fairly choppy. I'm not impressed.

(I'm running this side-by-side on 2 separate machines, XFCE on one, KDE on the other, both are running piledriver architecture chips in the mid-4ghz range, so single threaded performance should still fall slightly behind your haswell-EP's.).

  • 52 months ago
  • 1 point

@Allan_M_Systems, here's an update. You previously asked me about numa. I confirmed it is enabled in bios. Now I have this info too:

dmesg | grep -i numa
[    0.000000] NUMA: Initialized distance table, cnt=2
[    0.000000] NUMA: Node 0 [mem 0x00000000-0x7fffffff] + [mem 0x100000000-0x107fffffff] -> [mem 0x00000000-0x107fffffff]
[    0.000000] mempolicy: Enabling automatic NUMA balancing. Configure with numa_balancing= or the kernel.numa_balancing sysctl
[    1.100894] pci_bus 0000:00: on NUMA node 0
[    1.110746] pci_bus 0000:80: on NUMA node 1

You also asked me a couple times about cpupower. I have now (almost) finished installing Arch Linux now and I just ran cpupower as you suggested. See below.

These are the original settings: (perf-bias: 7 and governor "powersave")

cpupower info
analyzing CPU 0:
perf-bias: 7
cpupower -c 1 info
analyzing CPU 1:
perf-bias: 7

cpupower frequency-info
analyzing CPU 0:
driver: intel_pstate
CPUs which run at the same hardware frequency: 0
CPUs which need to have their frequency coordinated by software: 0
maximum transition latency: 0.97 ms.
hardware limits: 1.20 GHz - 3.00 GHz
available cpufreq governors: performance, powersave
current policy: frequency should be within 1.20 GHz and 3.00 GHz.
                The governor "powersave" may decide which speed to use
                within this range.
current CPU frequency is 1.20 GHz (asserted by call to hardware).
boost state support:
    Supported: yes
    Active: yes

Then I changed the governor to performance and I set the perf-bias to 0 (maximum performance).

With a video playing in VLC full screen, I saw the CPU frequency scale to 2.80 GHz. So it is definitely responding as expected with these aggressive settings. See below.

cpupower frequency-info
analyzing CPU 0:
driver: intel_pstate
CPUs which run at the same hardware frequency: 0
CPUs which need to have their frequency coordinated by software: 0
maximum transition latency: 0.97 ms.
hardware limits: 1.20 GHz - 3.00 GHz
available cpufreq governors: performance, powersave
current policy: frequency should be within 1.20 GHz and 3.00 GHz.
                The governor "performance" may decide which speed to use
                within this range.
current CPU frequency is 2.80 GHz (asserted by call to hardware).
boost state support:
    Supported: yes
    Active: yes

Even with the performance settings, I'm not getting the GUI performance I'm hoping to get. (This is with a GTX 970.) For the type of desktop productivity 2D tasks I'm doing, I wonder if a higher end video card would even make any difference... as you know, I'm not gaming. But apparently, I like to run "slow" software (e.g., Firefox, Libre Office, etc.).

I decided to stick with Firefox. I like it better than Chrome (and in fact, Chrome won't even support my use-case at work). So I simply need the power to run Firefox and KDE. I don't mind turning off some or all of KDE's desktop effects if needed, but you seemed to think that wouldn't be needed.

I have now seen similar GUI performance with Kubuntu 14.04, Kubuntu 15.10, opensuse Tumbleweed, and now Arch Linux.

This Arch install is particularly lean. For example, I don't even have a display manager installed. There is no bloat in this install. I carefully picked every package by hand. I even did some testing in straight Xorg (with a proprietary nvidia driver installed). Then I installed plasma-desktop (not full KDE). This is plasma 5.5, which most people say is really nice. I definitely like it better than Kubuntu 15.10's plasma 5.4.

I almost installed Wayland instead of Xorg, but I decided to stick with something familiar for the moment.

So that's my update for now. I appreciate your continued interest!

  • 52 months ago
  • 1 point

For the type of desktop productivity 2D tasks I'm doing, I wonder if a higher end video card would even make any difference

Nope. The bottleneck is software at this point. Crappy software. I wish I could tell you that wayland is the answer.... but I don't know that it would help. I'm running wayland + KDE 5.5 Plasma on my 2nd machine now. I mean, it seems pretty good most of the time but it's horrifically unstable. Lots of application crashes and I believe it is related to wayland.

  • 51 months ago
  • 1 point

Have not solved main performance problem yet...

Unfortunately, so far, the new system is falling far short of meeting my goal for interactive GUI performance on everyday tasks. I have described the issue in prior posts, but please ask if I need to further clarify the issue. (One simple test: take a Firefox browser window and resize it by dragging the corning with a mouse. Does window resizing significantly lag the cursor?)

I need some advice. Here are the details.

I recently built a Core i7 system and I'm using that system for comparison. Both systems are running Linux. The Core i7 system performs great, but the dual Xeon system is annoyingly slow on common GUI tasks such as resizing windows (all applications). The performance problem is not a Linux problem. It is somehow related to the hardware or hardware configuration (or BIOS) of this build.

Here are links to the two systems I'm comparing:

The Core i7 cost about $1500. It was a very simple build. I put it together in a few hours and put it in service on a co-worker's desk. Works great, no problems.

The dual Xeon machine is intended to be my work system. The parts list is posted above (or see permalink). I saved a bit on this system too by buying the CPU's used on ebay from a reputable seller. The two CPU's cost $704.88 total with shipping (compared to $2750 new). So this build has cost about $3900 so far (some extra fans and a few items are not listed in the parts list above and I have had to return/exchange a few parts which involved some extra fees).

Am I wrong to expect that the dual Xeon system should perform about the same as the i7 system on my simple window resizing test? (It has a better GPU than the i7.)

On the i7, the mouse cursor stays glued to the corner of the window and the window resizes as fast as I care to move the mouse. It is a smooth process.

On the Xeon system, I can easily get the mouse cursor half a screen ahead of the window - there is that much lag. The window resizing is choppy and slow. Even when moving the mouse slowly, it is an annoying experience with lag.

Both computers run the same OS. Initially I tuned every parameter I could think of (at least those I understand) on the Xeon system, but even with max performance settings it lagged the untuned i7 (which is using power-save settings not performance settings).

When I applied jus a few of the tweaks to the OS on the i7, the two systems are not even in the same league. The i7 has no lag in my simple tests that mimic my everyday GUI productivity tasks. The Xeon has tons of lag. You can see the difference without any special measurements. Noticeable lag. Does that sound right?

I need some advice regarding what to try next.

The options I am considering are:

Test 1. replace the dual Xeon E5-2660 V3 2.6GHz 10-Core CPU's with a single E5-2637 V3 3.5GHz Quad Core CPU -- for testing. This might not be a permanent change, but it is intended to answer two questions:

A. How much is single core clock rate a factor in the performance issues I'm seeing. The Core i7 has only one main advantage I can see: higher clock rate. The dual Xeon build is more performant in almost every other way with a faster GPU, faster storage, etc.

B. Given that I bought the CPU's used, and I am having problems, it seems logical to swap out this part and test the result. My testing indicates that the Xeon CPU's are performing properly. There are no errors of any type in any logs I have seen. The CPU's respond to perf-bias and governor settings. Everything seems OK, but I need to take a next troubleshooting step, and this seems a logical choice.

Buying one Xeon E5-2637 V3 3.5GHz Quad Core CPU will cost me $1050. (Will buy at Amazon in case I need to return it.)

Test 2. Another option is to try a whole different mobo, CPU, RAM combination. I can either go with Skylake parts similar to the i7 I'm testing against or I can go with X99 parts. Either way, I can get a CPU, mobo and 32 GB RAM for about $1000, so this option costs nearly the same as option 1.

This will indicate how much the i7 vs Xeon CPU choice infuences the issue I'm seeing. Some people say that a Xeon cannot touch an i7 in certain games. I am not gaming, but I wonder if any of that is relevant to my issue.

Do other people with Xeon workstations see this kind of GUI lag???

Test 3. I can go back through every BIOS setting and see if there is something that magically transforms the Xeon system. I'm skeptical that will happen. After tweaking the CPU governor, the CPU perf-bias, the nvidia PowerMIzer settings, and many OS settings, the problem did not go away. It go a bit better or worse, depending on the settings, but the lag was always there. Pushing performance settings to the max did not get rid of the lag.

Test 4. I can try something suggested here...

Suggestions, please.

  • 51 months ago
  • 1 point

single i7 MUCH faster than dual xeon E5-2650 v3 !!! -- CFD Online Discussion Forums http://www.cfd-online.com/Forums/hardware/144936-single-i7-much-faster-than-dual-xeon-e5-2650-v3.html

I just read this...

"hi there!

I'm posting this info in here so no one throw away they money as I've done.

I did run a FSI (with fluent) analysis in Ansys 15.0.7 under windows 7 64 bits pro with sp1. I did perform the same analysis over a SINGLE i7 3820 and over a DUAL xeon E5 2650 v3 , and the SINGLE i7 is 2 to 3 time much faster !!!!! So guys, be aware, a 1200 Euro computer is 2 to 3 times faster than a 6000 Euro workstation. I really should have check CFD online forums before I bought this computer, I know, I´m guilty, but how in the hell I should imagine this would happen!!! I have tried many bios settings for the new dual xeon, like hyperthreding on and off, overckloking/turbo on and off, numa on and off, QPI auto or fixed, power management on and off, etc etc, and it seems to make small difference in performance. I've been thinking that I should use 8x8 GB RAM instead of 4x16 GB RAM configuration, so I could have the processor at full 4 channel ,but some people said it won't make a big difference.

So please guys, tell me, have I waste my money?? Any idea of what is going on? Is it hardware company fault, OS software or CFD software ? Are they trying to sell products not worth it the price AT ALL compared to others?? Or is it me, and my poor knowledge on the subject? I really hope its my fault, otherwise, guys, be aware and don't spent your money on those products."


Another comment from the same thread:

"I work on either a cluster computing with four nodes with 2 x 8-cores Intel Xeon E5-2650v2 2,6GHz each and on PCs with i7-3770 and i7-4790, and I can tell confirm you that already a single i7-3770 (a little old in this moment) is MUCH faster than two aforementioned xeons... when comparing them with the 4790, the ratio is almost 32 xeon cores to 4 i7-4790 cores, for Matlab, Ansys HFSS and CST. Maybe the xeon processors are made for servers, but for computing they just suck! The features shown in comparison on several sites (for example, http://www.velocitymicro.com/blog/xe...ts-difference/) are all craps.

People, if your scope is computing, don't buy XEON processors!!!! "

The thread also gives me another option to test: memory modules.

Quoting:

Greetings to all!

@acasas: Quote: Originally Posted by acasas

I've been thinking that I should use 8x8 GB RAM instead of 4x16 GB RAM configuration, so I could have the processor at full 4 channel ,but some people said it won't make a big difference. It makes so much of a big difference, that you should sue whomever told you that. The explanation is simple: you have 2 Xeon processors and only 4 RAM modules. This means that technically you only have 2 RAM modules per processor. Technically, what this means, is that you have created a massive bottleneck on your system.

If you look at the specs for the 2 processors and notice the maximum memory bandwidth:

E5-2650 v3: http://ark.intel.com/products/81705/...Cache-2_30-GHz - Max 68 GB/s
i7 3820: http://ark.intel.com/products/63698/...up-to-3_80-GHz - Max 51.2 GB/s

Now, remember that I mentioned that you created a massive bottleneck, by using only 2 RAM modules for each processor E5-2650 v3? This means that instead of having 68 GB/s for each processor, you only have 34 GB/s. Which is almost half of what the bandwidth the i7 3820 has got.

Now, the other bottleneck:

The E5-2650 v3 processor has 10 cores, which will likely operate at around 2.5GHz when all cores are running at full throttle. This means that it has an equivalent potential of 25 GHz processing power per processor, which is more than 3.6*4 = 14.4 GHz of the i7 3820.
But when we overlap the constriction imposed by using only 2 RAM modules, this means that you have 10 cores fighting for access to 2 RAM modules, over 2 memory channels. This means that, with luck, each processor E5-2650 v3 is only being used at 50% capacity, i.e. potentially at 12.5GHz.
But since 10 cores are fighting for the existing RAM, the capacity is likely being dropped down to 30-40%, due to the entropy.

Worst even is if the 4 RAM modules are all allocated only to the first processor E5-2650 v3, and the second processor is nonetheless (somehow) turned on and trying to ask the first processor to share its memory, so that it can do its own share of the calculation as well. Although this shouldn't be possible.

On the other side, the i7 3820 only has 4 cores and also has 4 memory channels, which means that it pretty much gives optimum access to memory for each core, at full bandwidth.

Now, this essentially equates to you having bought a 500000$ super-car and then shooting off 2 of its tyres and expecting that it can still travel at 300 km/h... when it can barely run at 50 km/h, while burning out the rims of the wheels

The solution: take the expensive machine back to the shop where you bought it and ask for replacing the 416GB modules of RAM for 88GB modules.

Want proof, before heading back to the shop? It's simple:

Turn off HyperThreading. It's useless for CFD. Use pure cores only.
Use only 2 cores on each processor for running your case. This equates to roughly 4*2.6GHz = 10.4GHz. And use "cpu affinity binding", do it manually if you have to, via Windows' Task Manager.
This means that the simulation should take roughly 14.4/10.4 = 1.38 times longer to run, i.e., it will take 30 * 1.38 = 41 to 45 minutes to run.

Best regards, Bruno

  • 51 months ago
  • 1 point

42% gain from dual E5-2643 to i7-3930K (x2) -- CFD Online Discussion Forums http://www.cfd-online.com/Forums/hardware/132152-42-gain-dual-e5-2643-i7-3930k-x2.html

"I finally got around to testing this out to answer the whole "2 node cluster vs. dual CPU workstation" question. I used ANSYS CFX v15 for this benchmark, and I made the case very simple if others wanted to compare systems.

CASE: Geometry: 1m x 1m x 5m long duct Mesh: 100 x 100 x 500 "cubes" all 1x1x1cm (5M cells) Flow: Default Water enters @ 10m/s at 300K, goes out other side at 0Pa. Walls are 400K. High Resolution Turbulence and advection Everything else default. Double Precision: ON CPUs: Platform MPI: 8 cores = 4+4 for cluster 20 iterations

Setup 1 is a dual XEON E5-2643 (3.3 GHz) with 128GB of 1600 MHz RAM. Setup 2 is a two node cluster of i7s 3930K @ 4.2GHz, each with 64GB of 2133 MHz RAM. Connected with 20Mbps Infiniband. Maybe in the future I'll check the performance only using a gigabit connection, though I've heard it doesn't matter much with 2 nodes.

Solver Wall Time for dual XEON: 869.96 s Solver Wall Time for two i7s: 612.70 s

I saw the roughly same trends in mechanical benchmarks as well. (30-50% performance increase, and in some benchmarks a single i7 on 4 cores could solve faster than the dual XEON on 8!) "

  • 53 months ago
  • 1 point

I have just a few last minute questions before I order parts.

  1. Which adapter should I get to use the Samsung 950 PRO -Series 512GB PCIe NVMe - M.2 Internal SSD with the Supermicro MBD-X10DAL-I-O Motherboard? One that looks promising is this one:

Addonics Product: M2 PCIe SSD adapter http://www.addonics.com/products/adm2px4.php

But it doesn't say if it works with this type M.2 drive. From what I can see in the pictures, this drive has 5 contacts on the small end, meaning it is has "M" keying. Am I right?

  1. which case fans and how many should I order for the Antec P100 case? I'm considering:

2x Noctua 120mm, Anti-Stall Knobs Design,SSO2 Bearing PWM Case Cooling Fan NF-S12A PWM by noctua 3x Noctua 140mm Premium Quiet Quality Fan with AAO Frame Technology (NF-A14 PWM) by noctua

Thanks

  • 53 months ago
  • 1 point

Which adapter should I get to use the Samsung 950 PRO -Series 512GB PCIe NVMe - M.2 Internal SSD with the Supermicro MBD-X10DAL-I-O Motherboard? One that looks promising is this one: Addonics Product: M2 PCIe SSD adapter

I believe that the Lycom DT-120 and Addonics ADM2PX4 are the exact same product. Either one should work fine.

But it doesn't say if it works with this type M.2 drive. From what I can see in the pictures, this drive has 5 contacts on the small end, meaning it is has "M" keying. Am I right?

I'm not sure exactly what you're referring to here, but I think you are over-complicating this. These adapters are nothing more than a physical adapter allowing the smaller PCIE connection used in M.2 to be "wired/mapped" to a traditional PCIE slot. Any M.2 to PCIE adapter should work fine here.

Just do a google search for "950 pro PCIE adapter" and you'll see there are people using these basic inexpensive adapters for these with no problems.


which case fans and how many should I order for the Antec P100 case?

With your initial configuration of a single CPU and GTX960, I predict no need to add any more fans than the 2 that come with the case. I would try them out first before replacing them. Check thermals and noise with those fans to get an idea where where you are, and where you might want to go.

If/When you do decide to add/replace fans (might be necessary with both CPU's and dual GPU's later on), I would just leave the stock 120mm fan in the rear, the front and top of the case both support 2 X 140mm, so filling out the case with 4X ~<900RPM 140mm airflow/noise optimized fans from Noctua/Nanoxia/Phanteks/Corsair/Cougar/BeQuiet/Fractal/SilverStone would be great . There's a lot of room for subjectivity here (there are lots of options capable of running pretty quiet and cooling your components, but may have variations in tonal qualities, or slight performance tradeoffs with different levels of restriction, or different aesthetics, etc). Noctua is a safe bet, but you have options.

Get the machine up and running with the included fans, then, spend some time reading reviews for fans to get an idea of the subtle differences and decide which is best for your tastes.

  • 53 months ago
  • 1 point

OK, thanks! However, I did go ahead and order 2 CPU's (and 2 of the Noctua CPU coolers you recommended). I decided I wanted to know now if everything would fit in the case as expected.

It looks like all that's left at this point is deciding whether to go with dual rank or quad rank memory. I have ordered all the parts except RAM. I plan to order 4x Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory as soon as I learn whether to go with dual rank or quad rank.

  • 53 months ago
  • 2 points

Please when you're done with your build, post it in the 'completed builds' section, would love to 'see' it, never seen a powerful build like this !!! ;)

  • 51 months ago
  • 1 point

Thanks. That guy is pretty funny.

"Linux is the worst operating system, except for all the others." - me

“Democracy is the worst form of government, except for all the others.” - Winston S. Churchill

  • 49 months ago
  • 1 point

Did you ever resolve the issue with your Dual Xeon build? I built a very similar system, X10DAL-i, 2x e-2660 V3, 4x 8gb ram, 256gb Samsung 950 pro M.2 ........ I ran windows 7 and now windows 10 and it's working great. I'll be glad to share info if you wish

  • 49 months ago
  • 1 point

I'm still working on my build. Do your Xeon E5-2660 v3 CPU's support Intel® AES New Instructions? The CPU's are supposed to have that support. But my OS indicates the support is not there. In linux I can run "cat /proc/cpuinfo" to check this.

  • 48 months ago
  • 1 point

yes they do. But no TSX.

  • 48 months ago
  • 1 point

It's a few dollars over, but what about this:

PCPartPicker part list / Price breakdown by merchant

Type Item Price
CPU Intel Xeon E5-2609 V4 1.7GHz 8-Core Processor $301.99 @ SuperBiiz
CPU Intel Xeon E5-2609 V4 1.7GHz 8-Core Processor $301.99 @ SuperBiiz
CPU Cooler Cooler Master Hyper 212 EVO 82.9 CFM Sleeve Bearing CPU Cooler $24.89 @ OutletPC
CPU Cooler Cooler Master Hyper 212 EVO 82.9 CFM Sleeve Bearing CPU Cooler $24.89 @ OutletPC
Motherboard Asus Z10PA-D8 ATX Dual-CPU LGA2011-3 Motherboard $395.99 @ Newegg
Memory Crucial 64GB (2 x 32GB) DDR4-2133 Memory $489.99 @ Adorama
Storage Samsung 850 PRO 2TB 2.5" Solid State Drive $849.99 @ NCIX US
Storage Western Digital Red 5TB 3.5" 5400RPM Internal Hard Drive $192.99 @ NCIX US
Storage Western Digital Red 5TB 3.5" 5400RPM Internal Hard Drive $192.99 @ NCIX US
Video Card MSI GeForce GTX 1080 8GB Founders Edition Video Card $699.99 @ NCIX US
Case Fractal Design Define R4 w/Window (Black Pearl) ATX Mid Tower Case $79.99 @ Newegg
Power Supply EVGA SuperNOVA 1000 P2 1000W 80+ Platinum Certified Fully-Modular ATX Power Supply $165.98 @ Newegg
Optical Drive Asus DRW-24B1ST/BLK/B/AS DVD/CD Writer $16.99 @ SuperBiiz
Case Fan Corsair Air Series SP120 High Performance Edition (2-Pack) 62.7 CFM 120mm Fans $21.99 @ Newegg
Case Fan Corsair Air Series SP120 High Performance Edition (2-Pack) 62.7 CFM 120mm Fans $21.99 @ Newegg
Monitor Philips BDM4065UC 60Hz 40.0" Monitor $871.99 @ Best Buy
Monitor Philips BDM4065UC 60Hz 40.0" Monitor $871.99 @ Best Buy
Prices include shipping, taxes, rebates, and discounts
Total (before mail-in rebates) $5566.62
Mail-in rebates -$40.00
Total $5526.62
Generated by PCPartPicker 2016-06-01 09:23 EDT-0400

I couldn't find the M.2 you were talking about on PCPP.

  • 48 months ago
  • 1 point

Hi. Thanks for the suggestion. However, I already have all the parts -- except the monitors. I ruled out the Philips BDM4065UC 60Hz 40.0" Monitors because they use PWM dimming which produces flicker. I'm looking for a flicker-free 40" 4K monitor with HDMI 2 and DisplayPort 1.2 or 1.3.

Here is the list of parts I purchased:

PCPartPicker part list / Price breakdown by merchant

Type Item Price
CPU Intel Xeon E5-2660 V3 2.6GHz 10-Core Processor $1359.99 @ SuperBiiz
CPU Intel Xeon E5-2660 V3 2.6GHz 10-Core Processor $1359.99 @ SuperBiiz
CPU Cooler Noctua NH-D9DX i4 3U 46.4 CFM CPU Cooler $59.99 @ Newegg
CPU Cooler Noctua NH-D9DX i4 3U 46.4 CFM CPU Cooler $59.99 @ Newegg
Motherboard Supermicro MBD-X10DAL-I-O ATX Dual-CPU LGA2011-3 Motherboard $301.98 @ Newegg
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $192.99 @ SuperBiiz
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $192.99 @ SuperBiiz
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $192.99 @ SuperBiiz
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $192.99 @ SuperBiiz
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $192.99 @ SuperBiiz
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $192.99 @ SuperBiiz
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $192.99 @ SuperBiiz
Memory Samsung 32GB (1 x 32GB) Registered DDR4-2133 Memory $192.99 @ SuperBiiz
Storage Samsung 950 PRO 256GB M.2-2280 Solid State Drive $177.73 @ Newegg
Storage Mushkin Reactor 1TB 2.5" Solid State Drive $219.99 @ Newegg
Storage Western Digital Red 6TB 3.5" 5400RPM Internal Hard Drive $239.99 @ Amazon
Storage Seagate Archive 8TB 3.5" 5900RPM Internal Hard Drive $229.00 @ B&H
Storage Seagate Archive 8TB 3.5" 5900RPM Internal Hard Drive $229.00 @ B&H
Video Card EVGA GeForce GTX 970 4GB SSC ACX 2.0+ Video Card $294.99 @ NCIX US
Case Fractal Design Define R5 Blackout Edition w/ Window ATX Mid Tower Case $109.99 @ Newegg
Power Supply FSP Group 1200W 80+ Platinum Certified Fully-Modular ATX Power Supply $209.99 @ Newegg
Optical Drive Asus DRW-24B1ST/BLK/B/AS DVD/CD Writer $19.99 @ Newegg
Other Addonics ADM2PX4 M2 Pcie Ssd Pcie 3.0 4-lane Accs Adapter Purchased For $23.47
Prices include shipping, taxes, rebates, and discounts
Total (before mail-in rebates) $6475.00
Mail-in rebates -$35.00
Total $6440.00
Generated by PCPartPicker 2016-06-01 13:08 EDT-0400
  • 47 months ago
  • 1 point

update:

A few days ago I posted about a similar build, but I was focused on a single i7 and non-ECC RAM. From the comments I learned that a dual Xeon ECC RAM build will be more suitable.

I think that was a mistake. It seems like I should have gone with an i7

  • 47 months ago
  • 1 point

The other possibility is that my CPUs are somehow defective...

https://pcpartpicker.com/forums/topic/165245-did-i-get-scammed-on-my-cpus

[comment deleted]
[comment deleted by staff]
  • 53 months ago
  • 1 point

Thanks for the recommendations, especially the video card. I don't know enough about the pros / cons of going with a Quadro card, however. I see that it is aimed at my use-case. But many years ago I used Matrox cards designed for CAD and design professionals and I ultimately concluded that the consumer cards were a better choice at the time. That was a long time ago, but ever since then I have not bothered to look at the professional graphics card line ups. Maybe I should... but right now I have no idea whether a Quadro is really a better choice. Do you have any articles I could read that compare both types of cards in professional use cases? Thanks

EDIT: just found this: GeForce vs Quadro as Fast As Possible - YouTube https://www.youtube.com/watch?v=U5Npt1BSF04

Granted, one short video doesn't make me an expert. But this video makes me think I'm still better off with the consumer line of cards (GeForce).

[comment deleted by staff]
  • 53 months ago
  • 1 point

I'm not going to do any gaming at all on this computer. I work on a lot of different tasks, and 3D rendering isn't the main one. Spreadsheets have been the main thing recently, but that will probably change in the next few months. Audio and video editing are another big part of that I'll be doing. So I need a good all-around machine. This computer also functions as the file server and database server for a small group of people. Obviously, that's not the ideal setup, but we're too small to have a dedicated database server right now.

Sort

add arrow-down arrow-left arrow-right arrow-up authorcheckmark clipboard combo comment delete discord dots drag-handle dropdown-arrow errorfacebook history inbox instagram issuelink lock markup-bbcode markup-html markup-pcpp markup-cyclingbuilder markup-plain-text markup-reddit menu pin radio-button save search settings share star-empty star-full star-half switch successtag twitch twitter user warningwattage weight youtube