AMD CPU speculation... and expert conjecture

[juanrga]

More BF_H scores

 

[8350rocks]

More BF_H scores

So, basically...I read that graph and think..."Hmm...even the lowly FX-4100 can play that game so well that a 60Hz monitor would require v-sync."

In other words...what is the bloody point of comparing benchmarks that most people will never see the difference in frame rates on...?

I mean, honestly, the FX 9590 pushes 150 fps minimum who needs more than that?

EDIT: I would also like to add...if you have to run a GTX980 SLI config to play at 1080p...you need to jump to 4K at that point.

 

[juanrga]

^^ Still waiting your explanation to your claim that Zen has AVX-512 support listed. The code only mentions AVX and AVX2:

+ { "CPU_ZNVER1_FLAGS", + "Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|Cpu SYSC ALL|CpuRdtscp|Cpu387|Cpu687|CpuFISTTP|CpuNop|CpuMM X|CpuSSE|CpuSSE2|CpuSSE3|CpuSSE4a|CpuABM|CpuLM|Cpu FMA|CpuFMA4|CpuBMI|CpuF16C|CpuCX16|CpuClflush|CpuS SSE3|CpuSVME|CpuSSE4_1|CpuSSE4_2|CpuAES|CpuAVX|Cpu PCLMUL|CpuLZCNT|CpuPRFCHW|CpuXsave|CpuXsaveopt|Cpu FSGSBase|CpuAVX2|CpuMovbe|CpuBMI2|CpuRdRnd|CpuADX| CpuRdSeed|CpuSMAP|CpuSHA|CpuXSAVEC|CpuXSAVES|CpuCl flushOpt|CpuCLZERO" },

 

[-Fran-]

^^ Still waiting your explanation to your claim that Zen has AVX-512 support listed. The code only mentions AVX and AVX2:

+ { "CPU_ZNVER1_FLAGS", + "Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|Cpu SYSC ALL|CpuRdtscp|Cpu387|Cpu687|CpuFISTTP|CpuNop|CpuMM X|CpuSSE|CpuSSE2|CpuSSE3|CpuSSE4a|CpuABM|CpuLM|Cpu FMA|CpuFMA4|CpuBMI|CpuF16C|CpuCX16|CpuClflush|CpuS SSE3|CpuSVME|CpuSSE4_1|CpuSSE4_2|CpuAES|CpuAVX|Cpu PCLMUL|CpuLZCNT|CpuPRFCHW|CpuXsave|CpuXsaveopt|Cpu FSGSBase|CpuAVX2|CpuMovbe|CpuBMI2|CpuRdRnd|CpuADX| CpuRdSeed|CpuSMAP|CpuSHA|CpuXSAVEC|CpuXSAVES|CpuCl flushOpt|CpuCLZERO" },

Not to say you're wrong, but that's an early push for GCC's headers. So it's not a final spec thing. It can be added on a later date. Since Zen has not been officially announced (spec wise), a lot of things can change in that code push.

Cheers!

 

[cdrkf]

And how many cycles does Haswell take to complete an ADD instruction? How many cycles does Pilediver take? What about MOV? Lets get fancy and talk about a MUL as well. Each instruction has a certain number or promised cycles, giving the throughout put per core of each instruction. But you're right; it was indeed a design decision.

You aren't looking at the whole picture. The actual instruction execution is only part of what goes into IPC. Cache latencies matter. Resource sharing matters. Even memory access times matter to some degree. And that's where the BD arch starts to fall behind.

When you look at the size and capabilities of a haswell i3 core (which critically includes ht), then compare to amd its comparable in most respects to 1 module. Fx 4xxx is directly comparable to an i3, its only poor marketing on amds part that brands this a quad core part.

To be fair, CMT is a relatively aggressive form of SMT, and far superior performance wise to HTT. Which brings us back to the "What makes a CPU core" argument? As far as Windows is concerned, both HTT enabled i3s and the FX-4xxx have four cores, even if HTT is treated differently by the OS.

Also out of interest, the haswell core is so wide that in games the haswell i3 (dual core + ht) is keeping pace with the the i5 with a small clock advantage (highest i3 vs entry i5). Does that mean the i5 is suddenly a bad design?

No, just that the i3 (at least HTT enabled variants) are fast enough to avoid a CPU bottleneck.

Exactly! The problem is though that *some people* are using these things to 'comapre ipc' and you can use this fact against the i5 in the same way as used against the FX.

"Dual core i3 at 3.6 ghz is 75% faster than quad core i5 in gaming" purley on the premis that you ignore HTT and treat the i3 as 2 cores vs 4.

CMT was AMD's attempt at answering HTT. What this tells me is this: CMT works quite well in terms of multi thread performance, as we already know in single thread the i3 is a long way in front, yet when it's 4 threads vs 4 threads they offer similar performance (ergo CMT is scaling better than HTT). The problem for AMD is that Intel's approach is somewhat more elegant, as it delivers both good single and multi thread performance.

The FX parts aren't that bad tbh, especially when they were released (PD didn't look terrible against Sandy / Ivy). The problem for them now is they are several generations behind. It's like comparing first gen i7 parts to the latest stuff, its still fast enough to be usable most of the time but it doesn't win any records

 

[gamerk316]

Personally, I think that once or twice a year, review sites should do one MEGA CPU/GPU performance review, going back to the C2D (E8600 OC) and PII X4 days, just to see how things like the Q9650 and i7 920 hold up.

 

[logainofhades]

That would be nice

 

[etayorius]

More BF_H scores

So, basically...I read that graph and think..."Hmm...even the lowly FX-4100 can play that game so well that a 60Hz monitor would require v-sync."

In other words...what is the bloody point of comparing benchmarks that most people will never see the difference in frame rates on...?

I mean, honestly, the FX 9590 pushes 150 fps minimum who needs more than that?

EDIT: I would also like to add...if you have to run a GTX980 SLI config to play at 1080p...you need to jump to 4K at that point.

I would like to have the extra performance, the more you got in the present the longer it will last in the future.

 

[con635]

Personally, I think that once or twice a year, review sites should do one MEGA CPU/GPU performance review, going back to the C2D (E8600 OC) and PII X4 days, just to see how things like the Q9650 and i7 920 hold up.

I've been crying out for this and same comparisons with the new apis.

 

[de5_Roy]

I've been crying out for this and same comparisons with the new apis.

you can always cry after the apis actually launch and they get debugged more post-launch.

 

[gamerk316]

http://techreport.com/news/28017/sec-filing-says-samsung-is-fabbing-chips-for-nvidia

Since NVIDIA is having chips fabbed at Samsung, so Samsung must obviously be ready to purchase them, right? [/sarcasm]

 

[logainofhades]

Well of course they are.

 

[jdwii]

Many things come to mind LCD always blur that is correct but the point you are trying to make about dgpu using lots of power well why should it use more while performing the same as their competitor? Why get a CPU that offers inconsistent performance for the same price as a CPU that doesn't? Also haswell gets pretty hot people are crazy if they think otherwise my 8350 on my cooler master 212+ ran cooler.

 

[etayorius]

http://wccftech.com/amd-allegedly-merge-samsung/

If true, say goodbye to x86 on AMD...

 

[-Fran-]

http://wccftech.com/amd-allegedly-merge-samsung/

If true, say goodbye to x86 on AMD...

Oh Boy if true, lol...

One bit that makes it sound more realistic: "they could structure the buyout in a way that doesn't incur in a violation of the agreement". I don't know how that could work out, but Legal things is like black magic; we engies can not understand it with logic and science.

In any case, not the first rumor, not the last either.

Much News. So Buyout. Very Salt. Wow.

Cheers!

 

[gamerk316]

Again, I could somewhat see Samsung being interested in the old ATI unit, but that's about it.

But yeah, the sharks are starting to circle. The way it's written, it sounds more like a hostile takeover bid then a merger.

 

[COLGeek]

Greetings, thread regulars. After discussion, the Moderation Team has decided to close this thread. While there have been many enlightening discussions throughout the course of this discussion, there has also been a large degree of unacceptable behavior (and subsequent sanctions).

I would suggest that one of you start a new thread regarding AMD products. However, I would caution you to not repeat the conduct often found in this thread. I would further suggest ceasing and desisting on many of the tired, cyclical arguments that this thread is full of. Really, the whole IPC discussion and the relative merits of benchmarks has been taken the "beating a dead horse" analogy to sub-atomic levels.

I would further ask that any new thread be one where civility and respect toward one another be at its very core. No discussion between enlightened members should ever result in a flame war and outright attacks on individuals. Facts and sources go much farther than names, jabs, insults, and other trollish behavior.

Once again, thanks for expanding the minds of many, but it is time to move on from this thread.

Have fun and good luck!

 

[juanrga]

I suggest the title "AMD's Future Chips: News, Rumours & Reviews" for the new thread. I adapted it from the title of the Intel thread.

 

[COLGeek]

Folks, due to technical difficulties (possibly due to the sheer size of the thread) we are having trouble closing this thread. Please don't post any new content. juanrga's suggestion for a new thread seems reasonable. You, sir, have the honors of starting such a thread if you so desire.

Thanks.

 

[de5_Roy]

will this do, colgeek?

"AMD CPUs, SOCs rumors and speculations temp. thread"
http://www.tomshardware.com/forum/id-2585271/amd-cpus-socs-rumors-speculations-temp-thread.html
the current thread is full, so amd cpus related discussion and speculations have moved to the above thread.

 

[COLGeek]

Please stop posting anything here. Thanks.