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

[Sweepr]

Lol...

I posted the numbers, incredible the denial...

Indeed it is (not me). Average CPU FPS using RX 480, same version of the benchmark @ 1080p High:

i5-4690K (3.5 GHz): 63.4 FPS
i5-6500 (3.2 GHz): 59.2 FPS

Invisible 60%, few % difference at best. And, as already explained to you, that 6700K is likely OCed (already changed the OP).

 

[Abwx]

Invisible 60%, few % difference at best. And, as already explained to you, that 6700K is likely OCed (already changed the OP).

There s several 6700K, there s one that do even higher numbers but they show all a considerable CPU framerate advantage against anything HW, so they are all overclocked but no HW is overclocked, isnt it..?.

http://www.ashesofthesingularity.co...-details/01da06c4-060e-4db6-af28-ac8c8a5774a3

http://www.ashesofthesingularity.co...-details/ab1ac9c9-2b76-445a-9dbc-492fa0f4d8ec

http://www.ashesofthesingularity.co...-details/994bc960-5894-4b91-87db-7514a0573be0

And a 4790 /
http://www.ashesofthesingularity.co...-details/a3510396-f9a5-447f-83d7-748b55c62c5f

http://www.ashesofthesingularity.co...-details/510424cc-c89a-45da-8c85-8eaae9da6d6e

A few % differences in CPU framerate you said..?.
And nothing is overclocked, set apart the 6700Ks of course..


A 4820K :

http://www.ashesofthesingularity.co...-details/5f2b7b89-40ca-4597-9af8-3d74f9d6ed5d

 

[Sweepr]

There s several 6700K, there s one that do even higher numbers but they show all a considerable CPU framerate advantage against anything HW, so they are all overclocked but no HW is overclocked, isnt it..?.

Yes, which is why your comparison is completely flawed. No way to determine the clocks on these 'K' CPUs here - scores ranging from ~90-110 FPS from the links provided. Also if faster memory increases the scores - that could explain part of the difference here (high-speed DDR4 on OCed Skylake-S systems). Still nowhere near your imaginary 60%, easily found 4790Ks above 80 FPS average CPU FPS. My comparison (locked CPUs) is more apples to apples than the random results you chose.

 

[Abwx]

easily found 4790Ks above 80 FPS average CPU FPS..

With a GTX950 that is likely not even capable of rendering all the DX12 effects and is using surely a DX11 path as do all Nvidia card of this gen for this game, you are just aggravating your case even further and as usual using confusion as a smoke screen to hide even more confusion...

And you call this apple to apple comparison..?.

 

[Sweepr]

With a GTX950 that is likely not even capable of rendering all the DX12 effects and is using surely a DX11 path as do all Nvidia card of this gen for this game, you are just aggravating your case even further and as usual using confusion as a smoke screen to hide even more confusion...

And you call this apple to apple comparison..?.

Right, because the GPU makes such a big difference to the CPU framerate. The excuses never end.

4790K with RX 480 scoring 87.7 FPS.

Still searching for the 60% gap between Haswell and Skylake you claimed.

 

[Abwx]

Right, because the GPU makes such a big difference to the CPU framerate. .

Yes because the CPU framerate is the estimation made by AOTS of how much FPS would be rendered if the GPU had unlimited size, so the GPU and the exact DX path matter..

Anyway you are just doing excuses to try justifying your picked, and as demonstrated useless, data to show AMD on as bad light as possible, that s indeed why you created this thread and it is proved by any of your post about anything AMD.

 

[MajinCry]

Aye, Abwx is on the money 'ere. The "CPU Framerate" was stated to be a guesstimate if the CPU had some not-bottlenecked-anywhere graphics card tied tae it. 'Tis not even a synthetic benchmark, so it's just stupid and/or dishonest ta portray it as some sort of decent indicator.

'Tis the other framerate scores that matter, and aye. GPUs aren't created equal when handling draw calls. For example, we ken that NVidia has a software scheduler for Maxwell 'n' Pascal (defo' fer Maxie, not too sure about Pascie, actually), and that AMD's GCN graphics cards have hardware schedulers. And it's not as if it's the same damn one; architecture changes 'n' all that.

And that ain't even taking into driver differences between brands, architectures and individual cards.

Get the CPUs, disable turbo 'n' set them all tae the slowest chip's clocks, slap three 980 ti's on each of 'em and send off the results. If we ain't got that, the comparisons are 'tarded at best.

 

[Sweepr]

Anyway you are just doing excuses to try justifying your picked, and as demonstrated useless, data to show AMD on as bad light as possible, that s indeed why you created this thread and it is proved by any of your post about anything AMD.

The exact same data you used to 'predict' Summit Ridge will 'largely match' a 8C/16T 5960X, despite being dual-channel DDR4? You change your opinion faster than the goalposts.

 

[Abwx]

The exact same data you used to 'predict' Summit Ridge will 'largely match' a 8C/16T 5960X, despite being dual-channel DDR4? You change your opinion faster than the goalposts.

Of course, we know that Guru3D 5960X is at a fixed 3.2GHz while it s unlikely that Zen is significantly higher than 2.8GHz given that 4 cores at least are fully used in single thread each, besides the very CPU frame rate point that it has limited RAM speed since it use DDR4 and is not much faster than the FX wich use DDR3.

We know from Aten Ra s test that 33% higher RAM speed increase the GPU framerate by 8%, add to this that Guru3D s CPU has more than 10% frequency advantage, and do the maths :

Given what we saw from the i5/i7 comparison it s likely that a 4C Zen will perform like the 8C, perhaps even better due to frequency.

 

[.vodka]

Too many unknown variables to even start considering where Zen stands, as is. I wonder when someone will leak a proper CPU benchmark of Zen, this is a waste of time.

 

[dark zero]

Ok, after seeing all of this, I expected the following results:
- ST 5% less than Haswell
- MT 5% less than Broadwell

Is not bad, I feel that pricing is the key here despite the haters here.

If they prices the Octacore like the Intel Hexacore, is likely to see a massive drop in prices if Intel wants to win.

However since there are Quads spotted, is likely to see Quads performing much better than the Octacore and putting on Core i7 Vanilla non K levels. And if pricing is around 200 dollars... Intel would be in a problem... Their own ones cost 300.

 

[aigomorla]

All of you guys take the Intel stuff out of this thread NOW.
This ain't a discussion on Intel, its a discussion on Zen.

Sweeper you especially I don't want to read another post about Intel from you in this thread.

Unless its a direct comparison, and I better see Zen numbers next to it, I will not tolerate ANYTHING which is INTEL in this thread.

And if its a direct comparision, you better have credible sources to back it up, or i will assume you are flame baiting / trolling and hand out infractions til no tomorrow.

Moderator Aigo

 

[radieroslaw]

Check this out: http://videocardz.com/63242/amd-zen-stuff

 

[The Stilt]

I think they are going to show Blender running on both Zeppelin and Broadwell-E at identical clockspeed, instead of implying that the clock speeds would be identical. Because they won't be (3.2GHz base, 4.0GHz boost on i7-6900K).

 

[blublub]

I think they are going to show Blender running on both Zeppelin and Broadwell-E at identical clockspeed, instead of implying that the clock speeds would be identical. Because they won't be (3.2GHz base, 4.0GHz boost on i7-6900K).

You mean both running on 3Ghz i.e.? - If true, IPC of ZEN would be pretty good, not Skylake or KabyLake but hey it would be an imense improvement for AMD. If the power consumptionon par or even less than Intels Broadwell ....

 

[cytg111]

I think they are going to show Blender running on both Zeppelin and Broadwell-E at identical clockspeed, ....

That is excatly what the slide states they will do / have done.. identical clocks.

"There was a presentation for the press, you will probably see a video from it in 2 hours. There was also 4k gaming demo as far as I know."
- posted about 30 mins ago ... lets check back in a couple of hours Moar rumors.. yesss!

btw.. isnt Blender one of those workloads that benefit from AVX2 ?

 

[The Stilt]

btw.. isnt Blender one of those workloads that benefit from AVX2 ?

Probably slightly depends on the build and compiler used.

 

[LTC8K6]

That is excatly what the slide states they will do / have done.. identical clocks.

"There was a presentation for the press, you will probably see a video from it in 2 hours. There was also 4k gaming demo as far as I know."
- posted about 30 mins ago ... lets check back in a couple of hours Moar rumors.. yesss!

btw.. isnt Blender one of those workloads that benefit from AVX2 ?

A marketing cynic might say that it's probably one thing Zen happens to do well.

 

[SpaceBeer]

They should compare Zen & Broadwell-E with some GPU in Blender. My R7 260X is faster than i7-4790 :d

Still, it's good to know they will have something to compete with intel in (sub) 250$ range. As I said, I don't expect Zen based CPUs will reach same clocks, but performance will be quite good for end users. At this moment AMD has no competitive product to intels i5-6500 for example, and anyone who needs PC for a little bit more seriuos tasks (gaming, programming, CAD...) has no choice but to go for i5 or i7.

Even if the process is not as good as it should be, I think architecture improvements will be good enough. Though it will probably be harder to reach desired TDP levels. So I hope they will do the same thing as they did with Kaveri – release unlocked CPUs wtih TDP of 95W, configurable to 65/45/(35) W, but also possible to overclock even though TDP might go to 120+ W. In the end, performance (per watt / per $) is important, not clocks (as we know from AM3+/FM2+ offer)

 

[PhonakV30]

I think they are going to show Blender running on both Zeppelin and Broadwell-E at identical clockspeed, instead of implying that the clock speeds would be identical. Because they won't be (3.2GHz base, 4.0GHz boost on i7-6900K).

But damn Beautiful Slide!

Edit : It's Official

http://www.anandtech.com/show/10578...lers-micro-op-cache-memory-hierarchy-revealed

 

[cytg111]

A marketing cynic might say that it's probably one thing Zen happens to do well.

"The power of accurate observation is commonly called cynicism by those who have not got it. "
- George Bernard Shaw

I have given warnings about staying on topic.
I hope this quote was worth it, because it just earned you an infraction.

Moderator Aigo

 

[Abwx]

I think they are going to show Blender running on both Zeppelin and Broadwell-E at identical clockspeed, instead of implying that the clock speeds would be identical. Because they won't be (3.2GHz base, 4.0GHz boost on i7-6900K).

That seems right :

To make the point that it has, AMD put a Summit Ridge engineering sample running at 3GHz up against an eight-core, sixteen-thread Core i7-6900K artificially limited to the same 3GHz speed. AMD ran the same Blender 3D rendering workload on both chips at the same time.

http://techreport.com/review/30540/amd-gives-us-our-first-real-moment-of-zen

Zen is about 3% faster...

Edit : According to Techreport article :

Tantalizingly, AMD says the 3GHz figure isn't the final clock speed it expects production Zen chips to top out at, either.

 

[Bryf50]

I'm not so quick to take marketing slides at face value, but that's pretty impressive. Especially considering Broadwell is Intels latest Xeon architecture and will be for some time still. I hope they can push the clock speeds up a bit more though.

 

[cytg111]

If one were to assume that AMD's SMT implementation is inferior to Intels then IPC gets a notch more..
I also take note that 3.0 <> 2.8

 

[blublub]

Yeah they seem to have upped the clock, so it is likely that it will meet or maybe surpass broadwell-e clock and that at 95w tdp