Leaked AMD Ryzen Benchmarks?

[lolfail9001]

I don't see how a GPU can in anyway be related to how Ryzen operates, it was hitting 4Ghz on stock, Polaris was very different and GPU's in general are far mor sensitive to throttling or putting themselves into imposed V-Sync.

I talk about Polaris power consumption being underrated compared to what it really had. You know, same 15W of difference Polaris had from it's claimed TDP and it's a thermal runaway in tight case.

 

[CSI_PC]

Yeah, the irony if this does not fail to escape me.
The point being SPEC is one of benches AMD still loses, though i do not know if AMD included libquantum in their presentation.

But quite ironically 1800X matches 6900K (with its boost fully enabled) on 1 thread in Cinebench R15, a tool that is designed around the Intel compiler, with both around the same frequency as an extra 100MHz single thread is going to make marginal difference against the Intel at 4GHz.

Cheers

Edit:
Just for Shintai edited to add R15 as he thought I was talking about Cinebench, though my posts on this and context has always been about AMD event and Cinebench R15 scores.

 

[Shintai]

But quite ironically 1800X matches 6900K (with its boost fully enabled) on 1 thread in Cinebench, a tool that is designed around the Intel compiler, with both around the same frequency as an extra 100MHz single thread is going to make marginal difference against the Intel at 4GHz.

Cheers

The Cinebench benchmark isn't the same thing as the Cinebench product that uses AVX for example unlike the benchmark. And it only uses ICC because it also gives the best performance for AMD. The speedup for AMD is actually greater than for Intel using ICC vs MSVC.

 

[CSI_PC]

The Cinebench benchmark isn't the same thing as the Cinebench product that uses AVX for example unlike the benchmark. And it only uses ICC because it also gives the best performance for AMD. The speedup for AMD is actually greater than for Intel using ICC vs MSVC.

Not talking about AVX, Cinebench R15 does not even support AVX unless that changed recently.
There is no way around it, Cinebench benchmark is developed and biased towards Intel's design-compiler even though Maxon states it also benefits AMD, but that is not the same as AMD also have their own specific implementation for Ryzen.

And my context continues with this thread, it is about the R15 Cinebench benchmark not anything else, probably makes sense to delete your post as it is totally out of context (I thought it was pretty clear I have been talking about the AMD event scores for last several pages but you jumped on it anyway just because I said Cinebench even though context was the event) and I will this one.
Thanks

 

[Luka]

As i said so many times a problem i had with AMD testing against Intel is that they have the worst possible cooler on Intel platform causing CPU to stay on lower clock speed as much as possible, also they disabled quad memory channel, disable 2x16x speed for SLI Titan cards they used. In my mind that is just f. bullshit. AMD Ryzen CPU is great and comes with the great price but let's stop talking about bullshit saying that it stomps Intel, it is way faster than Intel counterparts, it kills it in the performance...sorry for bursting your bubbles but that's not the case.

 

[lolfail9001]

As i said so many times a problem i had with AMD testing against Intel is that they have the worst possible cooler on Intel platform causing CPU to stay on lower clock speed as much as possible, also they disabled quad memory channel, disable 2x16x speed for SLI Titan cards they used. In my mind that is just f. bullshit. AMD Ryzen CPU is great and comes with the great price but let's stop talking about bullshit saying that it stomps Intel, it is way faster than Intel counterparts, it kills it in the performance...sorry for bursting your bubbles but that's not the case.

The problem you have with AMD testing is that you bought an 2666v3, that's about it.

But quite ironically 1800X matches 6900K (with its boost fully enabled) on 1 thread in Cinebench R15, a tool that is designed around the Intel compiler, with both around the same frequency as an extra 100MHz single thread is going to make marginal difference against the Intel at 4GHz.

1800X was probably at 4Ghz there, actually. Point notwithstanding, ICC is the best AMD compiler even for con cores. I wonder how it turns out with gcc and in particular with llvm.

 

[Luka]

The problem you have with AMD testing is that you bought an 2666v3, that's about it.

1800X was probably at 4Ghz there, actually. Point notwithstanding, ICC is the best AMD compiler even for con cores. I wonder how it turns out with gcc and in particular with llvm.

What does 2666 v3 have anything to do with? I am saying that AMD testing was not fair. You don't disable things on competition platform to make your look better.

 

[CSI_PC]

1800X was probably at 4Ghz there, actually. Point notwithstanding, ICC is the best AMD compiler even for con cores. I wonder how it turns out with gcc and in particular with llvm.

Yeah, but some would complain if I did not allow for it
It was the "best compiler" according to Maxon for previous AMD CPU designs, not necessarily for Ryzen, I think you agree since Cinebench R11.5 (I think) has been developed and optimised around the Intel compiler, as I mentioned also benefits AMD in the past but not necessarily the same as having ones own development.

However the single core results mean 1800X performance is pretty good and not losing out to Intel 6900K even using their compiler and benchmark optimised for it
Doing a smile because maybe a cheeky comment but hey still valid consideration with everything else being thrown around
Anyway better than the IPC slide IMO that can be made to be good/bad/average.
Cheers

 

[mstaab]

too funny right here:

 

[lolfail9001]

What does 2666 v3 have anything to do with? I am saying that AMD testing was not fair. You don't disable things on competition platform to make your look better.

It was fair, not AMD's fault Intel's suggested 140W TDP solution can't cool 100W (without AVX) CPU to hold it in all core turbo at all times. Besides, AMD did enable TBM3.

Anyway better than the IPC slide IMO that can be made to be good/bad/average.

Yeah.

I now wonder what AMD was thinking with positioning 1700 against 7700k. That sort of comparison will sort of relegate 1700 to "niche CPU" tier when reviews comparing the two are out.

 

[NKD]

Now we are fighting over a better cooler. We all know why that is. At 3ghz it probably 65w

Move on. Stop bitching about a better stock cooler. Atleast it's better than intel.

Man some of you have nothing better to do. How about that stock Intel cooler that everyone throws away lol.

 

[CSI_PC]

It was fair, not AMD's fault Intel's suggested 140W TDP solution can't cool 100W (without AVX) CPU to hold it in all core turbo at all times. Besides, AMD did enable TBM3.

Yeah.

I now wonder what AMD was thinking with positioning 1700 against 7700k. That sort of comparison will sort of relegate 1700 to "niche CPU" tier when reviews comparing the two are out.

Just to add to the Intel cooling debate, it is also worth noting the 6900K multi-threaded R15 Cinebench score was at most 5% off other sites (such as Anandtech) with others being comparable (Tom's Hardware), some of these involved watercooling and the ones I checked did not have the standard Intel cooler.
So it was probably throttling some but still debatable how much, tbh I feel the same with the 1800X and stock cooler (which is not shipping with the 1800X nor the 1700X) on as well but less of an impact.
Cheers

 

[OrangeKhrush]

every day i'm throttling.

I will be honest, Intel stock coolers are terrabad, can't even keep a Pentium cool

 

[legcramp]

Lol trolls be arguing over stock coolers now. They be arguing over why Intel uses crappy TIM on their cpu lids and how it's unfair for AMD to use good one or even solder.

It's only fair if AMD uses peanut butter as TIM on their lids and downgrade the stock coolers otherwise it's unfair I tell you!!!

 

[thesmokingman]

Lol trolls be arguing over stock coolers now. They be arguing over why Intel uses crappy TIM on their cpu lids and how it's unfair for AMD to use good one or even solder.

It's only fair if AMD uses peanut butter as TIM on their lids and downgrade the stock coolers otherwise it's unfair I tell you!!!

Trolls and shills be getting on with each other, the old and the new ones. That's kinda funny.

 

[lolfail9001]

I will be honest, Intel stock coolers are terrabad, can't even keep a Pentium cool

The cooler used by AMD is not actually a stock, it is reference 140W TDP solution (other than Asetek AIO that is actually alright).

Now we are fighting over a better cooler. We all know why that is. At 3ghz it probably 65w

You would not believe how much i hope for that to be the case.

 

[Simplyfun]

Most of the reviews won't even be done with stock coolers I bet. And seriously, is anyone here running an enthusiast chip with a stock cooler? Stand and be counted for so I can write you off my Christmas card list.

The reviews will roll shortly, and that's all that matters. Then at least the trolls and shills will have Hard numbers to bicker over. That's when the true entertainment begins. This is just like the trailers before the movie starts.

 

[Luka]

Trolls and shills be getting on with each other, the old and the new ones. That's kinda funny.

No trolls, i am not asking AMD to cripple their platform...as far as i am f. concerned AMD should set up AIO cooler on both CPUs, not to disable quad memory channel, etc...

I have no problem with Intel TIM on their CPU lids as i have i7 5960k running at 4.6Ghz with rather good temperature. Don't use anything special but thermaltake 360 AIO cooler.

Enough of this, can't wait for real independent benchmarks.

 

[OrangeKhrush]

No trolls, i am not asking AMD to cripple their platform...as far as i am f. concerned AMD should set up AIO cooler on both CPUs, not to disable quad memory channel, etc...

I have no problem with Intel TIM on their CPU lids as i have i7 5960k running at 4.6Ghz with rather good temperature. Don't use anything special but thermaltake 360 AIO cooler.

Enough of this, can't wait for real independent benchmarks.

Out the box test, means out the box test.





Cinebench is now no longer a good enough benchmark, because AMD does well at it now. Quad channel, like where other than memory benchmarking has that every mattered to anything. Here is the scary thing, AMD produce more efficient bandwidth on the IMC than intel do, AMD gets 90+% maximum bandwidth possible, while Intel only scraps 75%, it is the secret weapon behind 50% of Ryzen's success, the IMC has come from Starwars new hope to Revenge of the Sith levels of epicness.

 

[lolfail9001]

AMD gets 90+% maximum bandwidth possible, while Intel only scraps 75%

Does not really matter if AMD does that at 1.5 the latency and has to settle on 20-50% lower mem clocks

 

[OrangeKhrush]

Does not really matter if AMD does that at 1.5 the latency and has to settle on 20-50% lower mem clocks

so far hasn't exibited any real world downside.

 

[CSI_PC]

The cooler used by AMD is not actually a stock, it is reference 140W TDP solution (other than Asetek AIO that is actually alright).

You would not believe how much i hope for that to be the case.

I must say looking into it the Wraith is a pretty ok cooler and better than I thought, but was the Wraith cooler spec ever released as I cannot find it?
Best I could find was they bundle it with the 125W FX8370, where one also could semantically debate it was stock as it is put with these.

If it was 140W, then why not bundle it with the 1700X and 1800X where those products seem to be without any coolers looking at the launch info from AMD, not everyone will go for OCing.
Under AMD Ryzen launch list cooler is shown as N/A for included cooler.

But all of this is probably a bit academic with regards how badly scores were influenced, because the Intel score was not far off that of other sites.
Cheers

 

[bobzdar]

I must say looking into it the Wraith is a pretty ok cooler and better than I thought, but was the Wraith cooler spec ever released as I cannot find it?
Best I could find was they bundle it with the 125W FX8370, where one also could semantically debate it was stock as it is put with these.

If it was 140W, then why not bundle it with the 1700X and 1800X where those products seem to be without any coolers looking at the launch info from AMD, not everyone will go for OCing.
Under AMD Ryzen launch list cooler is shown as N/A for included cooler.
Cheers

I think they assumed only enthusiasts would be pre-ordering the X cpus so would toss the stock in the trash anyway. However, given the lack of aftermarket availability at launch, that may have been a mistake. One of the reasons I got the 1700.

 

[Luka]

I think the best CPU to get is Ryzen 1700 because it is also unlocked and will reach 4.0Ghz easily. I don't think that even x1800 will run stable enough pass 4.0Ghz without adding too much voltage. 1700 reminds me of old AMD64 with 1.8Ghz clock speed which was able to run San Diego speed of 2.6+ Ghz for fraction of the cost. I owned one.

 

[CSI_PC]

I think they assumed only enthusiasts would be pre-ordering the X cpus so would toss the stock in the trash anyway. However, given the lack of aftermarket availability at launch, that may have been a mistake. One of the reasons I got the 1700.

Possibly, but it is interesting how many I know buy the K models of Intel and not overclock, just for the higher frequency by default.
Cheers

 

[Magic Hate Ball]

Now we are fighting over a better cooler. We all know why that is. At 3ghz it probably 65w

Move on. Stop bitching about a better stock cooler. Atleast it's better than intel.

Man some of you have nothing better to do. How about that stock Intel cooler that everyone throws away lol.

I couldn't even use the Intel stock cooler on my i3-6100 that was using a dGPU (so even less heat on the chip).

Seriously pathetic effort by Intel.

 

[GroundZero]

Does not really matter if AMD does that at 1.5 the latency and has to settle on 20-50% lower mem clocks

You still going on about the perfectly normal latency being slow? And the Memory issue is a Bios issue.

We squashed those over at AT yesterday...you were in the middle of it.

That latency was the same you would get on an intel platform with those timings.
DDR 2133 CL16 is ~70ns. On Intel or AMD. There is no Latency problem, and you keep bagging AMD about things you know are wrong should get you banned.

 

[lolfail9001]

You still going on about the perfectly normal latency being slow?

Of course i am, when you turn ~15ns memory latency into ~60-70ns memory+IMC latency, there is an obvious hole for you to fix in the next iteration.

We squashed those over at AT yesterday...you were in the middle of it.

We squashed it in a sense that it probably won't matter, but it's there and it is painfully visible.

That latency was the same you would get on an intel platform with those timings.

http://ranker.sisoftware.net/show_run.php?q=c2ffcdf9dfbedfe2d1e7d5e3c5b78aba9cf99ca191b7c4f9c9&l=en
So, your claim is that 2 more clocks on CAS add 50 nanoseconds. Let me check math real quick, 4 (to give you a breathing room) clocks on 1Ghz clock = 4nanoseconds. Bold claim, Cotton, Bold freaking claim. Nope man, that's like 3rd and maybe even 4th independent verification that there are indeed memory latency weirdness in synthetics. How do they translate into real life and whether they are BIOS or hardware related? I have no clue. But at this point, it's as good as true.

There is no Latency problem, and you keep bagging AMD about things you know are wrong should get you banned.

Keep that to S/A. If you cannot prove your claims with evidence, as you evidently cannot, hugboxes may just be the right place for you.

 

[ole-m]

Of course i am, when you turn ~15ns memory latency into ~60-70ns memory+IMC latency, there is an obvious hole for you to fix in the next iteration.

We squashed it in a sense that it probably won't matter, but it's there and it is painfully visible.

http://ranker.sisoftware.net/show_run.php?q=c2ffcdf9dfbedfe2d1e7d5e3c5b78aba9cf99ca191b7c4f9c9&l=en
So, your claim is that 2 more clocks on CAS add 50 nanoseconds. Let me check math real quick, 4 (to give you a breathing room) clocks on 1Ghz clock = 4nanoseconds. Bold claim, Cotton, Bold freaking claim. Nope man, that's like 3rd and maybe even 4th independent verification that there are indeed memory latency weirdness in synthetics. How do they translate into real life and whether they are BIOS or hardware related? I have no clue. But at this point, it's as good as true.

Keep that to S/A. If you cannot prove your claims with evidence, as you evidently cannot, hugboxes may just be the right place for you.

 

[NKD]

Of course i am, when you turn ~15ns memory latency into ~60-70ns memory+IMC latency, there is an obvious hole for you to fix in the next iteration.

We squashed it in a sense that it probably won't matter, but it's there and it is painfully visible.

http://ranker.sisoftware.net/show_run.php?q=c2ffcdf9dfbedfe2d1e7d5e3c5b78aba9cf99ca191b7c4f9c9&l=en
So, your claim is that 2 more clocks on CAS add 50 nanoseconds. Let me check math real quick, 4 (to give you a breathing room) clocks on 1Ghz clock = 4nanoseconds. Bold claim, Cotton, Bold freaking claim. Nope man, that's like 3rd and maybe even 4th independent verification that there are indeed memory latency weirdness in synthetics. How do they translate into real life and whether they are BIOS or hardware related? I have no clue. But at this point, it's as good as true.

Keep that to S/A. If you cannot prove your claims with evidence, as you evidently cannot, hugboxes may just be the right place for you.

Wait you squashed it that it wont matter, but how is something that doesn't matter so painful? We are talking about electronics here not a broken relationship that sure doesn't matter but still hurts lol.

 

[juanrga]

1700X beating 7700K in single-thread.

http://valid.x86.fr/bench/rjmzdu/1

Probably OC'd, but still.

(2) is also the reason for the high scores in some Passmark subtests.

 

[GroundZero]

Of course i am, when you turn ~15ns memory latency into ~60-70ns memory+IMC latency, there is an obvious hole for you to fix in the next iteration.

You are mistaking timings with total DDR4 latency. You just don't get it. Am I going to have to go over it all again here because you forget yesterday when you fall asleep?

Here look at some values for memory latency(That's total latency BTW). That was the figure that has you up in arms. It is generally 40-70ns FOR ALL DDR3 & DDR4!!
https://www.google.com/search?q=max...VixBxIQ_AUIBygC&dpr=1.82#imgrc=5xKeSNv1f9ncdM:

 

[juanrga]

But quite ironically 1800X matches 6900K (with its boost fully enabled) on 1 thread in Cinebench R15, a tool that is designed around the Intel compiler, with both around the same frequency as an extra 100MHz single thread is going to make marginal difference against the Intel at 4GHz.

Using the Intel compiler doesn't mean anything per se. AMD uses the Intel compiler in its own SPEC submission because it extracts the maximum performance from AMD hardware compared to other compilers as GCC.

Also The Stilt said something about Cinebench being a best-case for RyZen performance or something like that. Anyone remember his exact words?

 

[ole-m]

(2) is also the reason for the high scores in some Passmark subtests.

if it's designed for 10 mb L3 and you feed it 20mb L3 and it performs better.. should one disregard it?
just a question about progress in cache size that intel have really rocked for years
if some app is made using same algorythm and it greatly benefits from larger cache, it would mean it performs much better? no? but in cpu-z's defence it would not give a overall performance number that the application probably wants.
Anyhow, it just tells us it's not a great datapoint

 

[Schmave]

There are so many haters and shills posting in the last few pages of this thread. Why don't you wait a couple of days to see how the real reviews turn out.

Seriously, I don't understand why some of you want AMD to fail. It's as if you like being Intel's bitch. Or maybe it's to justify an overpriced Intel purchase.

 

[NKD]

sample overclocked, bla bla bla. 7700k is running at fuckin 4.2ghz stock, weren't you all bitching about AMD needs clock speeds lol. If its not one thing its other. I haven't build an amd rig for years and I don't understand some of you wanting amd fail so hard and will some how gain something by making shit up. I am glad amd has something competitive, Intel can suck it with their outrageous pricing.

 

[Magic Hate Ball]

sample overclocked, bla bla bla. 7700k is running at fuckin 4.2ghz stock, weren't you all bitching about AMD needs clock speeds lol. If its not one thing its other. I haven't build an amd rig for years and I don't understand some of you wanting amd fail so hard and will some how gain something by making shit up. I am glad amd has something competitive, Intel can suck it with their outrageous pricing.

Shouldn't single thread 7700K boost to 4.5ghz stock?

 

[thehoj]

I think the best CPU to get is Ryzen 1700 because it is also unlocked and will reach 4.0Ghz easily. I don't think that even x1800 will run stable enough pass 4.0Ghz without adding too much voltage. 1700 reminds me of old AMD64 with 1.8Ghz clock speed which was able to run San Diego speed of 2.6+ Ghz for fraction of the cost. I owned one.

That's what I'm hoping for.. I had an Opteron 165 (1.8Ghz) which I believe I was able to overclock to around 2.8Ghz or just under. I was using one of those jet engine sounding coolers though, with a Delta 60mm (7000rpm) fan I think... No fan curves, just full throttle all the time Obscenely loud.

 

[NKD]

Shouldn't single thread 7700K boost to 4.5ghz stock?

yea but no point beating my head against trolls. lol. I am just waiting a few more days.

 

[Magic Hate Ball]

That's what I'm hoping for.. I had an Opteron 165 (1.8Ghz) which I believe I was able to overclock to around 2.8Ghz or just under. I was using one of those jet engine sounding coolers though, with a Delta 60mm (7000rpm) fan I think... No fan curves, just full throttle all the time Obscenely loud.

That's when you put a hole through a wall to run your cables to a desk in the adjacent room