AMD Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

[ShintaiDK]

For 100 samples with a simulated 6900K scores MUCH better than the Zen and 6900K tested. 27 seconds vs 34 second.

 

[itsmydamnation]

2400Mhz.

A 3930K gets 34 seconds with the 100% setting.

Shame AMD couldn't even come clean on something they told the public to test. Some things never change.

Actually in no point in the demo can we see what the samples are set to ( i just watched it) can we confirm that the other pics are for the same demo and the room looks completely different. 32minute mark to see for your self.

Yes something never change like how much of a .............

@ 200 sample my 3770k @ 4.3 is 1.47.31
now that would line upto the demo scores would it not.........

 

[superstition]

Late 2013 Macbook Pro. i7-4960HQ (2.6 base, 3.8 turbo). OS X 10.9.5. Blender 2.78a, 100 samples

37.73

 

[ShintaiDK]

@ 200 sample my 3770k @ 4.3 is 1.47.31
now that would line upto the demo scores would it not.........

Not even remotely close. With 200 samples and a 5960X/6900K/6950X you get 45-50 seconds. It requires a 4.25Ghz 6950X to get to AMDs 6900K stock number if 200 samples.

Yes something never change like how much of a .............

...con artist AMDs marketing is? Correct. Why change a good habit after 10 years of non stop.

 

[superstition]

FX 8370E was a 32nm SOI part at 95W

It doesn't hit 95 watts even when turboing, though, afaik. So, it seems that the 95W number is only relevant if it's overclocked with APM enabled.

 

[itsmydamnation]

Not even remotely close. With 200 samples and a 5960X/6900K/6950X you get 45-50 seconds. It requires a 4.25Ghz 6950X to get to AMDs 6900K stock number if 200 samples.

So your saying NO IPC gain from IVB to Broadwell? because 107/4300*3500/2 = 43

 

[ShintaiDK]

So your saying NO IPC gain from IVB to Broadwell? because 107/4300*3500/2 = 43

You cant compare like that. My 6700K stock at a lower clock than yours gets 77 seconds.

Dont tell me my 6700K got over 40% higher IPC than your 3770K

 

[majord]

Actually in no point in the demo can we see what the samples are set to ( i just watched it) can we confirm that the other pics are for the same demo and the room looks completely different. 32minute mark to see for your self.

good point... Looks like they are two seperate demo's with different settings,

The press only Demo , where the 100 samples screenshots are from (and possibly showing v2.77), both machines complete in around 25s , not 35 .

https://youtu.be/7yxSFmEOkrA?t=1m8s

 

[ShintaiDK]

good point... Looks like they are two seperate demo's with different settings,

The press only Demo , where the 100 samples screenshots are from (and possibly showing v2.77), both machines complete in around 25s , not 35 .

If you time the LIVE stream you get 34.56seconds or so. Its certainly not 25 seconds.

 

[majord]

If you time the LIVE stream you get 34.56seconds or so. Its certainly not 25 seconds.

Yeah, I know,but i'm talking about the video I linked, which is the only one actually showing version and sample setting (100). Time the run in it, and you'll see it's around 25

 

[superstition]

Yeah, I know,but i'm talking about the video I linked, which is the only one actually showing version and sample setting (100). Time the run in it, and you'll see it's around 25

That makes more sense, then, because a 47 watt mobile i7 from 2013 shouldn't perform only a little worse than an unreleased 16 thread desktop CPU.

 

[itsmydamnation]

If you time the LIVE stream you get 34.56seconds or so. Its certainly not 25 seconds.

and show us the live demo settings.

 

[ShintaiDK]

Yeah, I know,but i'm talking about the video I linked, which is the only one actually showing version and sample setting (100). Time the run in it, and you'll see it's around 25

That one just raises further questions. Including the power usage.

AMD told people to test it at home to directly compare. Download Blender 2.78a, download the file and render it.

AMD needs to come clean. Either what is required or why the results doesn't match at all.

 

[Atari2600]

Ah, the usual story.

There is *quite* never enough information to be sure your doing a like-for-like comparison.

 

[ShintaiDK]

and show us the live demo settings.

That's not my job. AMD told people to replicate it themselves. That's why you can download the file and the specific version on Blender from amd.com.

If there is any more instructions to it than that, then AMD have to tell people so.

 

[cytg111]

Aaand spinning out of control again, not by applied curiosity but that other thing that people here do.

 

[bjt2]

It doesn't hit 95 watts even when turboing, though, afaik. So, it seems that the 95W number is only relevant if it's overclocked with APM enabled.

Good to know... So if the AMD statements (same energy/cycle per core) hold true, we will see more than 3.4GHz base (as already advertised)...

 

[itsmydamnation]

That one just raises further questions. Including the power usage.

AMD told people to test it at home to directly compare. Download Blender 2.78a, download the file and render it.

AMD needs to come clean. Either what is required or why the results doesn't match at all.

Or how about you prove AMD CEO was lying in public......

200 samples gives my IVB 1:47.31 that get pretty close in linear extrapolation as said before, blender is known the scale well with both cores and clocks, its pretty linear ( see http://blenchmark.com/cpu-benchmarks for example) have we considered OS?

my windows 7 FX 6300 (3.6) takes 4:34 (21600mhz) (5:46 @ 2800/16800mhz)
my windows 10 IVB 1:47 (17200mhz)

So does IVB really have 321% more performance per clock over piledriver?
maybe we should all install the benchmark from http://blenchmark.com/cpu-benchmarks and compare to get a point of reference?

 

[CatMerc]

good point... Looks like they are two seperate demo's with different settings,

The press only Demo , where the 100 samples screenshots are from (and possibly showing v2.77), both machines complete in around 25s , not 35 .

https://youtu.be/7yxSFmEOkrA?t=1m8s

You're correct, I missed that the press demo finished in 25 seconds.

 

[HurleyBird]

Shame AMD couldn't even come clean on something they told the public to test. Some things never change.

It sounds like Kyle from [H] is going to get in touch with some guys from AMD and try to sort things out. That said, I doubt in this case AMD was wilfully deceptive, seeing as they wanted people to replicate the test in the first place. If they really wanted to deceive people you wouldn't think they'd give the files out in the first place.

Either way, I think we'll know the full story, one way or the other, in the next couple of days. It's really premature to start attributing bad motive, and I can see you having some egg on your face soon.

 

[inf64]

Usual suspect out of control yet again...
Yes AMD rigged the demo live on purpose and CEO lied on live stream to thousands who watched. Makes perfect sense!

 

[Atari2600]

This is why you don't do this kinda sh!t live and on the hop.

AMD could probably do with releasing just a 2 or 3 page paper on their website detailing the configurations. They also have to be concrete in consistency. No different runs for different groups - too easy to lead to confusion.


I'm hopeful that Zen performance is good given their bullishness, but mindful I've been bitten before, I'm not setting my expectations much above a Sandy Bridge.

 

[Redentor]

At least AMD live demo was better than Intel live demo: https://www.youtube.com/watch?v=EBa_fUC_QaA

 

[zinfamous]

It's been so long since we had a choice in CPUs.. this feels like early 2000s all over again.

Early 2000's you say?!

 

[Atari2600]

At least AMD live demo was better than Intel live demo: https://www.youtube.com/watch?v=EBa_fUC_QaA

Oooofft!

Never seen that before - shocking.