Leaked AMD Ryzen Benchmarks?

We'll know for sure once the NDA is lifted. Also, a comment on the Chinese website suggests that Sandra is primarily an AVX bench.
 
Pieter3dnow said:
Isn't that the benchmark that never gets recompiled unless something new from Intel floats to the surface ?
Click to expand...
Pretty much. I can't remember when SiSoft Sandra ever represented actual performance beyond 1 manufacturer. Even in the Athlon 64 days, a Intel Prescott/Northwood would trounce the shit out of an AMD chip. However, it rarely translated verbatim into actual programs/workloads. That's when reviewers started to really make the distinction between synthetic and actual workloads.
 
kac77 said:
Pretty much. I can't remember when SiSoft Sandra ever represented actual performance beyond 1 manufacturer. Even in the Athlon 64 days, a Intel Prescott/Northwood would trounce the shit out of an AMD chip. However, it rarely translated verbatim into actual programs/workloads. That's when reviewers started to really make the distinction between synthetic and actual workloads.
Click to expand...
yup! I distinctly remember this. an amd system would kick the shit out of the intel system in everything. then you run Sandra and youre like WTF how is the intel now doubling the amd?! stopped using it years and years ago for that reason.
 
Everyone, lets remember the name of the site is Canard PC.


The definition of Canard:


ca·nard
kəˈnär(d)/
noun
  1. 1.
    an unfounded rumor or story.
    "the old canard that LA is a cultural wasteland"
  2. 2.
    a small winglike projection attached to an aircraft forward of the main wing to provide extra stability or control, sometimes replacing the tail.
I don';t think they are talking about the PC's small forward wing...
 
New leak of AOTS. Already seems to be deleted from the AOTS benchmark site.
https://www.ptt.cc/bbs/PC_Shopping/M.1485365573.A.E03.html
http://www.ashesofthesingularity.co...-details/409f0980-b72c-4d49-b6b8-2c4847c6bb5d

Ryzen 3.6Ghz base, 4.0Ghz turbo 8 cores, 16 threads.
upload_2017-2-2_13-3-56.png


To compare with a few CPUs:
upload_2017-2-2_13-5-13.png

upload_2017-2-2_13-5-24.png

upload_2017-2-2_13-5-43.png
 
Last edited:
Shintai said:
New leak of AOTS. Already seems to be deleted from the AOTS benchmark site.
https://www.ptt.cc/bbs/PC_Shopping/M.1485365573.A.E03.html
http://www.ashesofthesingularity.co...-details/409f0980-b72c-4d49-b6b8-2c4847c6bb5d

Ryzen 3.6Ghz base, 4.0Ghz turbo 8 cores, 16 threads.
View attachment 16066

To compare with a few CPUs:
View attachment 16067
View attachment 16068
View attachment 16069
Click to expand...

Two of the sample images show 12 million terrain shaders, the other uses 16 million. The Ryzen sample used 16 million. These are not quite apples-to-apples. (It seems like a 33% increase in terrain shaders has an impact, looking at the i7700k vs i6700k framerates.)
 
1) if someone calls themselves AMDFanboy, it is a legit troll
2) I have seen less fake girls online than this benchmark. Photoshoped

clickbait
 
0x0 resolutions seems like the cupcake forgot to add the 4K he was running at lol. You get incompetence amongst trolls now.
 
Still faked or severely scewed and nobody has confirmed a real working 3.6-4 Ryzen part let alone AOTS knowing about it.

as pointed above the tests are not even balanced. Ryzen is run on much higher terrain settings which impacts results a lot.
 
OrangeKhrush said:
Still faked or severely scewed and nobody has confirmed a real working 3.6-4 Ryzen part let alone AOTS knowing about it.

as pointed above the tests are not even balanced. Ryzen is run on much higher terrain settings which impacts results a lot.
Click to expand...

The 3.6Ghz base, 4Ghz turbo part is known isn't it. F4 stepping.

Also the 7700K runs with the exact same settings tho with a 1080 instead of a Titan X.
 
The 7700K has high clocks to compensate. It was probably removed for a reason, authenticity the likely.
 
OrangeKhrush said:
The 7700K has high clocks to compensate. It was probably removed for a reason, authenticity the likely.
Click to expand...

Dude, this is why AMD might not have shown us a true single core IPC test against Intel, its probably still between Ivy bridge- Haswell levels, Hand brake and Blender are heavy multithreaded applications where AMD might be a bit better for them.

No it was probably removed because it shouldn't have been posted up ;)

We have seen this from AOTS benchmarks before, for Polaris, they were real, and were taken down because AMD probably asked them to take them down.

This is a hell of a lot better then your rambling of extrapolating figures that just can't be done lol.
 
Shintai said:
The 3.6Ghz base, 4Ghz turbo part is known isn't it. F4 stepping.

Also the 7700K runs with the exact same settings tho with a 1080 instead of a Titan X.
Click to expand...

Welp, I see you are doing your usual desperation hate on AMD, good luck with that. Have a legit benchmarks that are absolutely known to be an 8c 16t Ryzen cpu and specific clocks and everything is being compared apples to apples?
 
razor1 said:
Dude, this is why AMD might not have shown us a true single core IPC test against Intel, its probably still between Ivy bridge- Haswell levels, Hand brake and Blender are heavy multithreaded applications where AMD might be a bit better for them.

No it was probably removed because it shouldn't have been posted up ;)

We have seen this from AOTS benchmarks before, for Polaris, they were real, and were taken down because AMD probably asked them to take them down.

This is a hell of a lot better then your rambling of extrapolating figures that just can't be done lol.
Click to expand...

Which numbers? All of the estimations had Haswell or close there to numbers.

I said from the beginning that AMD needed to at least match Haswell type performance or thereabouts for the product to be successful, unless AMD called it well and everything becomes multi threaded which inevitably will happen over the next few years. Needless to say the user is titled AMDFanboy, a derogatory term so there is likely extreme skewing. Most benches like this can be scewed by editing the cfg file, games like F1 and the first time I saw it being done was farcry 2 where editing the ingame bench file to a CPU you don't actually have is possible.

The source is just very bad and AOTS removed it because it was not validated.
 
Shintai said:
Are they?



They get removed from AOTS benchmarks due to AMDs request. Its done with CPUs and GPUs before.
Click to expand...


There is plausibility to it, anything less would likely result in winding up and Intel charging $250 for an i3.

I know you want failure so try not beat around the bush and call a spade a spade, you want the results to be low.
 
http://www.ashesofthesingularity.co...-details/a79a8ffb-284b-44d3-b237-31ccb0f2b4df

4790 non k with very similar clock speed to Ryzen scored a 60 CPU FPS on lesser settings than the faked result, I notice that all the K or X CPU's are showing base clock, which means AOTS CPU authentication only shows native speeds not the OC clocks, which means that the other scores may have been posted at overclocked speeds and the AOTS benchmark doesn't show it.
 
OrangeKhrush said:
Which numbers? All of the estimations had Haswell or close there to numbers.

I said from the beginning that AMD needed to at least match Haswell type performance or thereabouts for the product to be successful, unless AMD called it well and everything becomes multi threaded which inevitably will happen over the next few years. Needless to say the user is titled AMDFanboy, a derogatory term so there is likely extreme skewing. Most benches like this can be scewed by editing the cfg file, games like F1 and the first time I saw it being done was farcry 2 where editing the ingame bench file to a CPU you don't actually have is possible.

The source is just very bad and AOTS removed it because it was not validated.
Click to expand...


really? And you think those numbers are valid? Why, because AMD says so on applications that are highly multithreaded?

I have seen no synthetics or breakdown of Ryzen that shows IPC not at all!

The only benchmarks we have seen outside of AMD has been Canard with an ES sample, and in games it seems to be once you factor in clocks of what its releasing at a 6 core Intel in games. That still holds true to what I'm saying. AMD seems to be more efficient with multithreading than Intel but with less IPC than Broadwell/Haswell.

Edit, the reason why the games show this more is because IPC is more important in games, because certain game and graphcis routines can't be distributed across CPU cores since they are dependent on one another and with the older API's as well but to a less degree now.
 
Last edited:
OrangeKhrush said:
http://www.ashesofthesingularity.co...-details/a79a8ffb-284b-44d3-b237-31ccb0f2b4df

4790 non k with very similar clock speed to Ryzen scored a 60 CPU FPS on lesser settings than the faked result, I notice that all the K or X CPU's are showing base clock, which means AOTS CPU authentication only shows native speeds not the OC clocks, which means that the other scores may have been posted at overclocked speeds and the AOTS benchmark doesn't show it.
Click to expand...


Boosts are multipliers of the base clock. So when overclocking you overclock the base clock and boost takes it from there.
 
razor1 said:
really? And you think those numbers are valid? Why, because AMD says so on applications that are highly multithreaded?

I have seen no synthetics or breakdown of Ryzen that shows IPC not at all!

The only benchmarks we have seen outside of AMD has been Canard with an ES sample, and in games it seems to be once you factor in clocks of what its releasing at a 6 core Intel in games. That still holds true to what I'm saying. AMD seems to be more efficient with multithreading than Intel but with less IPC than Broadwell/Haswell.
Click to expand...

I said IPC would be around Haswell or close to it all along, so what exactly are you trying to imply?
 
OrangeKhrush said:
I said IPC would be around Haswell or close to it all along, so what exactly are you trying to imply?
Click to expand...

its going to be less, that is why AMD isn't showing synthetics that show IPC only or games for that matter. Anything that will cover up deficiencies they will show, anything that shows weaknesses they won't show. The moment anyone says AMD hasn't reached Intel's Core IPC, AMD sales will get hurt. Because IPC has been ingrained into peoples heads as = to performance for the past 10 years or so or more, remember MIPS?

Think about why OEM's aren't where AMD is focusing on for Ryzen? Reason they hinted at is they are interested in the enthusiast market, there are more valid reasons than that which are more likely on the money. I'll list in a few minutes. In the mean time wrap your head around this.

No for the reasons:

1) OEM's don't want Ryzen yet because of left over stock of older AMD products
2) something went wrong with validation and still waiting for the final sample of Ryzen
3) OEM's know the total sales will not increase on AMD's computers due to what I just mentioned

All of these correlate to what AMD stated about their first quarter AMD predictions a slight loss (in the red in Q1)

The enthusiast market doesn't drive the OEM market, they both go hand in hand. We see this with graphics cards, OEM's get their sales above and beyond what enthusiast's buy as in upgrades. And when we are talking about a new platform and CPU, the entire machine gets gutted, OEM's are better options for such upgrades, enthusiasts will do it anyways.
 
Last edited:
razor1 said:
Boosts are multipliers of the base clock. So when overclocking you overclock the base clock and boost takes it from there.
Click to expand...

I went through most of the first page scores then filtered the Crazy 4K scores then went to the only listed non overclockable CPU's and the i7 4790 Devils Canyon Haswell 3.6-4 Ghz scored much lower than the 4790K yet both have the same IPC but one has lower clock speed which means that the AOTS doesn't know what clockspeed the bench was at, most gamers would be running high overclocks so the benches are all scewed. Also 16 million samples at 3.6Ghz is going to net lower yields than an 5+ Ghz 7700K because in all probability it was benched at 5+ or why would that person really be running a 7700K. The lack of authenticity on clocks make it difficult to guage.

If we take a 80 CPU FPS at stock clocks with heavy sampling, that is probably right in line with estimations somewhere in the Haswell/Broadwell area. I don't really recall anyone claiming they would beat a KBL/SKL.
 
OrangeKhrush said:
I went through most of the first page scores then filtered the Crazy 4K scores then went to the only listed non overclockable CPU's and the i7 4790 Devils Canyon Haswell 3.6-4 Ghz scored much lower than the 4790K yet both have the same IPC but one has lower clock speed which means that the AOTS doesn't know what clockspeed the bench was at, most gamers would be running high overclocks so the benches are all scewed. Also 16 million samples at 3.6Ghz is going to net lower yields than an 5+ Ghz 7700K because in all probability it was benched at 5+ or why would that person really be running a 7700K. The lack of authenticity on clocks make it difficult to guage.

If we take a 80 CPU FPS at stock clocks with heavy sampling, that is probably right in line with estimations somewhere in the Haswell/Broadwell area. I don't really recall anyone claiming they would beat a KBL/SKL.
Click to expand...


There is an easier explanation, AOTS is hard to benchmark, its fly by wire, user has no control lol, and the game isn't always the same in different runs.
 
razor1 said:
its going to be less, that is why AMD isn't showing synthetics that show IPC only or games for that matter. Anything that will cover up deficiencies they will show, anything that shows weaknesses they won't show. The moment anyone says AMD hasn't reached Intel's Core IPC, AMD sales will get hurt. Because IPC has been ingrained into peoples heads as = to performance for the past 10 years or so or more, remember MIPS?

Think about why OEM's aren't where AMD is focusing on for Ryzen? Reason they hinted at is they are interested in the enthusiast market, there are more valid reasons than that which are more likely on the money. I'll list in a few minutes. In the mean time wrap your head around this.
Click to expand...

and they will gain anything from showing it how?

Much of MT is determined by single thread performance, weak single thread equates to weak multithread hence a Thuban beating a 8370 in Blender handsomely at lower clocks. Amadahls law states that, MT is still determined by underlying ST performance.
 
OrangeKhrush said:
and they will gain anything from showing it how?

Much of MT is determined by single thread performance, weak single thread equates to weak multithread hence a Thuban beating a 8370 in Blender handsomely at lower clocks. Amadahls law states that, MT is still determined by underlying ST performance.
Click to expand...

Dude have ever done MT code?

Yeah the base of what you stated is true, but that isn't all of it.

http://cseweb.ucsd.edu/~tullsen/ipdps04.pdf

You can make multithreaded code that is better for more core vs, IPC.

What you do, is distribute the instructions over more cores. Most games don't do this very well, because of dependencies, rendering programs on the other hand do because the dependencies are limited because of not using a graphics card.
 
OrangeKhrush said:
Also 16 million samples at 3.6Ghz is going to net lower yields than an 5+ Ghz 7700K because in all probability it was benched at 5+ or why would that person really be running a 7700K. The lack of authenticity on clocks make it difficult to guage..
Click to expand...

If you read the [H] review of a retail purchased processor you'd know that they were barely able to get 5.0ghz with HT on the 7700K and had to have crazy voltage and temps to do so. Therefore you aren't going to see a metric shitton of 5ghz 7700Ks out there.
 
razor1 said:
There is an easier explanation, AOTS is hard to benchmark, its fly by wire, user has no control lol, and the game isn't always the same in different runs.
Click to expand...

Which means posting what Shintai posted was useless, if that is the case.
 
ManofGod said:
Which means posting what Shintai posted was useless, if that is the case.
Click to expand...

You aren't going to get exact duplicates, but you get a general understanding where it might fall. Take a margin of error of 10% and it will still end up around a 6 core Intel. That is what Canard showed us, (well less than but that was with clock speeds of the ES being capped, so taking that into consideration it looks to be around a 6 core core Intel).

And this is what I stated a long time ago, Its going to be hard for Zen to keep up with Intel, I can see them cutting their deficient in performance greatly, but catching up to Intel, that is a monumental task. Zen looks to be good as a good foundation for future products, and good for improving current margins, but to gain substantial market share and thus great profits, like they had with Athlon they will need to be able to do the same thing Athlon did, and this is why AMD's forecasting has been well rather tame, because they know this too, although their boasts about "we are back" and potential market share gains have been a bit out-worldly before, those do not match up with what AMD's financial expectations are in the next few quarters, nor does their release schedule of different market segments.

Now you can see where the BS walks and real things happen and for what reasons they happen.
 
Last edited:
razor1 said:
You aren't going to get exact duplicates, but you get a general understanding where it might fall. Take a margin of error of 10% and it will still end up around a 6 core Intel. That is what Canard showed us, (well less than but that was with clock speeds of the ES being capped, so taking that into consideration it looks to be around a 6 core core Intel).

And this is what I stated a long time ago, Its going to be hard for Zen to keep up with Intel, I can see them cutting their deficient in performance greatly, but catching up to Intel, that is a monumental task. Zen looks to be good as a good foundation for future products, and good for improving current margins, but to gain substantial market share and thus great profits, like they had with Athlon they will need to be able to do the same thing Athlon did, and this is why AMD's forecasting has been well rather tame, because they know this too, although their boasts about "we are back" and potential market share gains have been a bit out-worldly before, those do not match up with what AMD's financial expectations are in the next few quarters, nor does their release schedule of different market segments.

Now you can see where the BS walks and real things happen and for what reasons they happen.
Click to expand...
razor1 said:
You aren't going to get exact duplicates, but you get a general understanding where it might fall. Take a margin of error of 10% and it will still end up around a 6 core Intel. That is what Canard showed us, (well less than but that was with clock speeds of the ES being capped, so taking that into consideration it looks to be around a 6 core core Intel).

And this is what I stated a long time ago, Its going to be hard for Zen to keep up with Intel, I can see them cutting their deficient in performance greatly, but catching up to Intel, that is a monumental task. Zen looks to be good as a good foundation for future products, and good for improving current margins, but to gain substantial market share and thus great profits, like they had with Athlon they will need to be able to do the same thing Athlon did, and this is why AMD's forecasting has been well rather tame, because they know this too, although their boasts about "we are back" and potential market share gains have been a bit out-worldly before, those do not match up with what AMD's financial expectations are in the next few quarters, nor does their release schedule of different market segments.

Now you can see where the BS walks and real things happen and for what reasons they happen.
Click to expand...

Horse Manure. The Ryzen 8 core will be at worst less than 5% less than Broadwell 6900K, and at best equal. The highest clock 3.6 GHZ with 4.0 boost stands toe to toe with 6900k. Even in single thread it only 5 to 10% slower, which is insignificant in those poorly designed games like Sky Rim.
 
os2wiz said:
Horse Manure. The Ryzen 8 core will be at worst less than 5% less than Broadwell 6900K, and at best equal. The highest clock 3.6 GHZ with 4.0 boost stands toe to toe with 6900k. Even in single thread it only 5 to 10% slower, which is insignificant in those poorly designed games like Sky Rim.
Click to expand...
Hook me up with that proof, guy!
 
os2wiz said:
Horse Manure. The Ryzen 8 core will be at worst less than 5% less than Broadwell 6900K, and at best equal. The highest clock 3.6 GHZ with 4.0 boost stands toe to toe with 6900k. Even in single thread it only 5 to 10% slower, which is insignificant in those poorly designed games like Sky Rim.
Click to expand...

hows that worked out in the past each time people have said same thing nothing will really be proved till they get the cpu released all the fake leaks dont tell us jack.
 
You must log in or register to reply here.
Back
Top