Athlon X4 845 ExcavatorIPC benchmarks

[AtenRa]

Zen should be a big improvement over XV, so it should handle CB much better.

Hard to believe that it won't, and if it doesn't, it's not good.

Lets assume it doesnt perform good in CB(Floating Point perf mostly), why only looking at a single benchmark ???

 

[Hi-Fi Man]

My old Phenom II X4 965 running at 4GHz (2.6GHz uncore) with DDR3 1.6GT/s cl9 RAM scored 4.77 in cinebench R11.5 and 409 in R15. At 3.8GHz it's somewhere in the ~4.5 range in R11.5. XV is unfortunately still behind Phenom II/Core 2 IPC.

The Athlon X4s are interesting chips but it's hard to recommend them because IPC is too low and they end up being outclassed by Pentiums a lot of the time.

 

[Sweepr]

My old Phenom II X4 965 running at 4GHz (2.6GHz uncore) with DDR3 1.6GT/s cl9 RAM scored 4.77 in cinebench R11.5 and 409 in R15. At 3.8GHz it's somewhere in the ~4.5 range in R11.5. XV is unfortunately still behind Phenom II/Core 2 IPC.

The Athlon X4s are interesting chips but it's hard to recommend them because IPC is too low and they end up being outclassed by Pentiums a lot of the time.

Indeed. Quite a shame that the last iteration of the Bulldozer architecture still can't beat 7 year old Phenom II here.

 

[LTC8K6]

Lets assume it doesnt perform good in CB(Floating Point perf mostly), why only looking at a single benchmark ???

Not performing well in popular benchmarks does wonders for sales.

I'm glad you are confident enough to dismiss the possibility.

 

[hrga225]

Zen should be a big improvement over XV, so it should handle CB much better.

Hard to believe that it won't, and if it doesn't, it's not good.

Yeah,CB likes fast SSE and fast cache ,and is more optimized for IA,BD architectures have none.It would not be suprise if Zen is like 80% faster than EXC.

 

[AtenRa]

Not performing well in popular benchmarks does wonders for sales.

I'm glad you are confident enough to dismiss the possibility.

Im not dismissing anything, i dont know how ZEN will perform in Cinebench.

But if it will perform very good in the rest of the benchmarks and only has a low perf in Cinebench. What then ??

 

[LTC8K6]

Im not dismissing anything, i dont know how ZEN will perform in Cinebench.

But if it will perform very good in the rest of the benchmarks and only has a low perf in Cinebench. What then ??

Then it will have low performance in anything similar to CB?

Is this Rocket Science?

 

[AtenRa]

Then it will have low performance in anything similar to CB?

Is this Rocket Science?

What other programs do you know similar to CB ??? I mean similar in architecture bias.

Because there are other 3D content creation applications that BD architecture is performing a lot better vs Intel.

 

[Shehriazad]

I hear it's about equal to a 4.3 Ghz Kaveri chip in quite a few benchmarks....depsite having even less L2 cache and a relatively low boost.


And yes yes...we know it's still gonna be behind Intel in Single Threaded...but if it comes with the 16 threads and the right pricing (now imagine 16 Zen threads for the price of 8 Skylake threads) it might just be enough for many people to buy into it.


It might be Bulldozer all over again(moar cores)...except this time with quite a bit less suckage? Or so I hope.

 

[The Stilt]

What other programs do you know similar to CB ??? I mean similar in architecture bias.

There is no "architecture bias" in Cinebench. Cinebench R11.5 uses MMX, SSE and SSE2 instructions and R15 uses MMX, SSE, SSE2 and SSE3.

AMD CPUs have inferior FPU performance compared to Intel and that's the only reason why they perform bad in Cinebench.

They perform just as bad or even worse in open-source FP heavy application / benchmarks (C-Ray, Rodinia Euler CFD, X265, VP9, etc). In X265 Skylake is around 90% faster than Excavator due the comprehensive AVX2 support. Skylake gets around 30% boost from AVX2 alone, whereas Excavator gets additional performance penalty of 2-5%.

 

[LTC8K6]

I can't see it being BD again. I doubt that very much.

 

[AtenRa]

There is no "architecture bias" in Cinebench. Cinebench R11.5 uses MMX, SSE and SSE2 instructions and R15 uses MMX, SSE, SSE2 and SSE3.

AMD CPUs have inferior FPU performance compared to Intel and that's the only reason why they perform bad in Cinebench.

There are other applications that FX is doing just fine in FP loads and its very competitive even against 8Threaded Intel SKUs up to Ivy.
Like Pov-Ray, 3DS max, Solidworks etc.

Im not saying the FX is better than the Intel in everything but CB is the only benchmark that Intel has such a huge lead in performance.

https://www.youtube.com/watch?v=PzLxCo5qofo

The Intel Vtune Performance Analyzer is heavily optimized for the Intel mArch. You can optimized your application and take 100% of the Intel CPU performance. This is not happening on the AMD CPUs, they need to brute force in order to compete in this benchmark.

They perform just as bad or even worse in open-source FP heavy application / benchmarks (C-Ray, Rodinia Euler CFD, X265, VP9, etc). In X265 Skylake is around 90% faster than Excavator due the comprehensive AVX2 support. Skylake gets around 30% boost from AVX2 alone, whereas Excavator gets additional performance penalty of 2-5%.

If im not mistaken x264 is Integer workload, they use Int AVX not FP. But i could be wrong.

 

[The Stilt]

Pov-Ray is integer, X264 is integer, 7-Zip (LZMA) is integer. That's the reason why 15h can keep up even remotely in these tests.

 

[Abwx]

So if we add 40% to get the best case Zen ipc...?

The 40% is moot in this case when it comes to estimate how it will perform in CB 11.5 ST, a better approach is to assume that a Zen core will have as much or more throughput (in MT = 1C/2T) than a EXV module since they use the same FPU or so.

From here one can estimate the expected IPC improvement in ST, so unless Zen SMT scale at 100% and has the same ST perf as EXV, wich is unlikely, the ST IPC in CB 11.5 will be increased by significantly more than 40%..

 

[Abwx]

Then it will have low performance in anything similar to CB?

Is this Rocket Science?

Is it rocket science to guess that the only CB score that matters is the MT score, or do you think that renderings are done in ST eventualy..?..

 

[LTC8K6]

Is it rocket science to guess that the only CB score that matters is the MT score, or do you think that renderings are done in ST eventualy..?..

Perhaps you should ask the poster who asked the question?

 

[dark zero]

40% better than this would still be slower than Intel's architectures from recent years.

Remember something... It lacks L3 cache and has half of L2 cache. More? Yeah. This is not the full Excavator. Bristol Ridge is the final one.

 

[Dresdenboy]

Pov-Ray is integer, X264 is integer, 7-Zip (LZMA) is integer. That's the reason why 15h can keep up even remotely in these tests.

POV-Ray as a raytracer uses floating point numbers.

 

[Sweepr]

Remember something... It lacks L3 cache

So does Trinity/Richland and Kaveri.

More? Yeah. This is not the full Excavator. Bristol Ridge is the final one.

Bristol Ridge is basically Carrizo with DDR4, there is no magic pixie dust.

 

[myocardia]

Are you incapable of forming complete sentences?

I am at least 90%, if not 99%, sure that English is not csbin's first language. If he's from somewhere in Europe, like most of the people who post around here who are still learning English, English is also not his second language. Forming complete, intelligible sentences is the very last step of learning a language.

Give him time, and I'm willing to bet that he'll become every bit as fluent in our native tongue as the other two people who've posted in this thread who don't speak English natively have become. They are both more fluent in English today than the average Anandtech forum member, including the ones whose native language happens to be English.

 

[Dresdenboy]

I'm impressed by this. CPU-Z bench might be a pure integer workload, this might imply it would have very good desktop multithread performance. But FP is not what this arch good at, same as previous gen(BD,PD,SR)

The bench is procedural landscape generator based on a Perlin noise. It is a mix of integer operations and single precision float operations. It is pure VC++, and I let Visual Studio optimizer do the job. As far as I can see, the x32 compiler uses x86 & x87, and the x64 version uses scalar SIMD (but I need to double check that point). The x64 version is much faster anyway. The result of the computation is a 3D landscape that will be displayed in the next-to-come GPU bench.

More here:
http://forums.anandtech.com/showpost.php?p=37645739&postcount=193

 

[dark zero]

So does Trinity/Richland and Kaveri.



Bristol Ridge is basically Carrizo with DDR4, there is no magic pixie dust.

And without that limiter that was the High Density Library. That explains why the higher clocks in their chips. In few words, another change on their CPU.

 

[myocardia]

I can't see it being BD again. I doubt that very much.

I would guess that it being just another iteration of the construction cores is the one thing we can more or less guarantee ourselves that Zen will not be. No company in existence is that dumb. This is AMD's chance to have a Conroe moment, and while they don't have anywhere near the monetary resources of Intel, it doesn't require much money at all to learn from your past mistakes.

 

[SPBHM]

Yes and yes

Look at the score of A10-7700K 95W TDP 3.8GHz Turbo vs 45/55/65W TDP A8-7600 3.8GHz Turbo.

http://www.hardware.fr/articles/898-3/cpu-consommation-efficacite-energetique.html

1t load = 38W over idle, it can't be a 100W TDP CPU when only loading one thread.

it would perform the same for this with a 65W TDP I think (if the IGP is not used).

it's useless to compare the TDP considering it's (big) IGP vs no IGP I think.

My old Phenom II X4 965 running at 4GHz (2.6GHz uncore) with DDR3 1.6GT/s cl9 RAM scored 4.77 in cinebench R11.5 and 409 in R15. At 3.8GHz it's somewhere in the ~4.5 range in R11.5. XV is unfortunately still behind Phenom II/Core 2 IPC.

The Athlon X4s are interesting chips but it's hard to recommend them because IPC is too low and they end up being outclassed by Pentiums a lot of the time.

your results are very similar to my i5 2310 (3GHz 4t, up to 3.2 1t)

but also something to consider with the AX4s is the lack of l3 cache, which might not hurt them much on Cinebench but will be more damaging in games for example.

 

[Insert_Nickname]

Yes, but reduced clocks. So why would a 6800K owner want to upgrade in the end of the day. Sitting with a faster product.

Speaking as a 6800K owner, I find it sad that AMD still hasn't released an APU that can beat its single thread performance.

Keep in mind the 6800K does have a 600MHz (turbo/non-turbo) frequency advantage. Carrizo would need to have ~15% better IPC to just keep up.

Anybody around here planning to buy an XV to do more tests?

If I can get one at current pre-order pricing I might get one to play around with. Could be a fun little chip.