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

What are the specs that make a CPU better?

JoshGer

24 months ago

I'm currently looking at an AMD FX-8320 and a Ryzen 5 1400. I see the only thing different is that the FX-8320 has 4 more cores, the Ryzen 5 1400 has worse(?) l2 cache, and better(?) l1 data cache. I'd obviously know that the Ryzen is way better than the FX, but how? Do the different cache specs make all the difference? (Not experienced with cache values)

Comments

  • 24 months ago
  • 4 points

A Ryzen core has far more execution resources on a more efficient instruction pipeline than a PileDriver module. As a result, a Ryzen core can achieve higher execution throughput at lower clock speeds than a piledriver module.

In fact, a Ryzen 5 1400 at base clock speeds (3.2-3.45GHZ) and sipping only ~50W max, can achieve higher execution throughput in nearly ANY workload, than an FX-9590 at 4.7-5.0GHZ running up to ~220W.


Core count, clock speeds, and cache sizes are all actually very superficial specifications for a CPU. I would not advise using ANY of these measures to compare CPU's except those within the same architecture family.

  • 24 months ago
  • 1 point

Thank you for the information.

  • 24 months ago
  • 1 point

So if those numbers are so superficial, how come theyre the only numbers listed on sites like PCPP and Newegg? Why not more important numbers like execution resources and IPC?

  • 24 months ago
  • 3 points

Why is "watts per channel" and "distortion into a dummy load" the most prevalent specification for the amplification in consumer Receivers? (When distortion characteristics into reactive loads of a real speaker would be far more eye opening)

Why is peak horse power the most prevalent engine specification for cars? (when a graph of torque and power over the RPM range of the engine would be far more meaningful).

Why are Megapixels the most prevalent specification for consumer cameras? (true resolving power of most consumer cameras/phones with 16MP+ sensors is actually only a few MP in ideal conditions, often less than a MP at higher than base ISO).

Why is contrast ratio a major selling point of televisions and monitors when manufactures aren't even using the same standards to measure it?


Consumers are interested in the specifications that are easy to conceptualize and compare, regardless of whether they are meaningful. The marketing departments of the world know this.

  • 24 months ago
  • 1 point

Eh, most people are happier just drinking the Kool-Aid anyway.

The best way to compare a CPU is just with synthetic and real-world benchmarks in situations that you expect to use your computer. Does core count, clockspeed, or architecture matter if your computer meets your performance expectations?

  • 24 months ago
  • 1 point

Eh, most people are happier just drinking the Kool-Aid anyway.

Ive got some of that purple drank right now. If I put it on my CPU instead of thermal paste can I get better overclocks?

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

Thanks so much!

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

Sorry to correct you but AMD is simultaneous threading which makes one core operate as two so four cores with simultaneous threading is actually like eight or so says the studying material I read states

  • 24 months ago
  • 3 points

To clarify this a bit....

A modern CPU core has execution resources spread across many execution pipelines. More heavily used instructions are actually executable on multiple of these pipelines. When a single thread is scheduled on these modern, wide core designs, there are frequently idle execution resources as it is rare to have a workload on a single thread that can make use of all these execution pipes at the same time.

Hyperthreading is just Intel's name for Simultaneously MultiThreading. Same technology. SMT leverages a dual threaded front end that can fetch, decode, and schedule work from 2 threads at the same time on the back-end execution pipelines. By running the second thread on the same back-end, the would-be idle resources can be put to work.

SMT does double the number of "threads" presented to the system, however, SMT does not double the execution throughput of a core. It opportunistically maximizes execution throughput by making use of underutilized execution pipes. The uplift in execution throughput varies depending on the sort of workloads being mixed on the core, but can be anywhere from ~5% - 50%.

  • 24 months ago
  • 1 point

Thanks for that break down I'm in the beginning stages of learning all the tech knowledge so I gave you a point for your analysis

  • 24 months ago
  • 1 point

Also simultaneous threading is more energy efficient than non threading cores

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

I'm a beginner to currently building mine I am going with the ryzen 5 1400

[comment deleted by staff]

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