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

[HurleyBird]

Not at all, Blender is FP while Handbrake use Integer code..

Is it more integer heavy? FX-8350 performs really poorly in it, which I'd expect from floating point heavy code.

 

[Maverick177]

Petition to have 2016 is the best year ever. 95% my next CPU will be Zen.

 

[Phynaz]

How would anyone know? Footnotes?

I don't see that AMD has published the footnotes yet.

 

[swilli89]

You still can, the market won't respond to this until tomorrow at the soonest. Then you can ride it up as the company continues to regain market share.

Not true in the least. Markets have major volume traded both post and pre market. There has been a significant run-up this past 10 days already, mostly due to anticipation of this event. As it stands AMD's stock price already carries all of the hype and speculation. Remember AMD has not yet posted profits for a few years and by all other metrics is an awful investment. The only ones that invested this past year were the savvy ones that knew Zen and Vega were in the pipeline and what they represent in terms of AMD's financial future.

If you want some action on AMD, my tip is this. Take July 2017 call options at a strike price of $14. Good luck investing!

 

[sirmo]

How could an Intel platform be identical to an AMD platform?

You asked me what the rest of the platform was.. the rest of the platform is identical.. memory.. SSD. Isn't that what you were asking, I am confused.

 

[Abwx]

Is it more integer heavy? FX-8350 performs really poorly in it, which I'd expect from floating point heavy code.

In Blender the FX is handicapped by the code low IPC, one thread is not enough to fully load a core, that s why there s 55% gain with Intel s HT, on Handbrake the FX perform very well, better than the 2600K FI because of higher Integer throughput, in X264 Intel got an advantage recently thanks to AVX256, that is new instructions that are also supported by Zen..

 

[sirmo]

Zen is looking better than I had hoped. The TDP is truly intriguing.

 

[blublub]

Does anyone have a 6900K to test the Blender files running 6900k at fixed 3.2Ghz and then again with boost up to 3.5Ghz to seel if Ryzen was going again a maxed out 3.5Ghz 6900K or a mixed bag between 3.2-3.5Ghz

 

[F-Rex]

It seems 14lpp is not as Bad as expected

It seems 2. 8ghz is not base frequency and ipc is not between Sandy and Ivy...

 

[LTC8K6]

It seems 14lpp is not as Bad as expected

It seems 2. 8ghz is not base frequency and ipc is not between Sandy and Ivy...

Well, the bad ipc info came from AMD.

 

[TheELF]

It seems 14lpp is not as Bad as expected

It seems 2. 8ghz is not base frequency and ipc is not between Sandy and Ivy...

There was not one single thing they showed that would allow you to come to this conclusion.
No explanation of how their SMT works and no test for single core performance.

 

[TheELF]

How would anyone know? Footnotes?

You would "see" * like 250-300FPS converting speeds

*not that you could actually see it,it was so small.

 

[blublub]

You would "see" * like 250-300FPS converting speeds

*not that you could actually see it,it was so small.

Intel Quick-Sync-Video is not available on the 6900k

 

[LTC8K6]

There was not one single thing they showed that would allow you to come to this conclusion.
No explanation of how their SMT works and no test for single core performance.

They did indicate that the IPC goal, 40% up on XV, had been exceeded.

 

[rtsurfer]

Blender files are available here

http://www.amd.com/en-us/innovations/new-horizon

 

[The Stilt]

I wonder why AMD displayed the power consumption of the whole system, instead of just the CPUs. The information can easily be accessed on both systems and AMD has and since there is no full FIVR in Zeppelin.

 

[.vodka]

piednoel claims QSV was not enabled in the cpu demos. Does that specifically affect just handbrake or also blender?

QSV is a fixed function hardware video encoder that started out back in 2011 with Sandy Bridge and gained on speed and features with time.

Did he really throw quicksync (unavailable in HEDT, no iGPU) around at a video encoding test showcasing MT brute force/performance between two competing products?

This was Piednoel throughout the show:

And it smells like denial.

---------------------

They're putting out the Blender demo files. AMD means it! Now we need the footnotes.

I'm pretty sure some will benchmark these and watch it through a profiler so we can get a closer look to what's been shown today...

 

[coercitiv]

piednoel claims QSV was not enabled in the cpu demos. Does that specifically affect just handbrake or also blender?

Piednoël is having a bit of "at first they ignore you, then they laugh at you" moment. I would expect an Intel representative, while voicing personal opinions on Twitter, to show a minimum of respect for the competition. His childish reactions are unbecoming, and reflect badly on the company as well.

Did he really throw quicksync (unavailable in HEDT, no iGPU) around at a video encoding test showcasing MT brute force/performance between two competing products?

He thought they were pitching Zen vs. 6700K.

 

[Doom2pro]

I wonder why AMD displayed the power consumption of the whole system, instead of just the CPUs. The information can easily be accessed on both systems and AMD has and since there is no full FIVR in Zeppelin.

Is it possible they are just trying to obfuscate?

 

[jihe]

Whoohoo! Truly incredible. The Ryzen PC also slightly beat the 6900K in Battlefield 1. Lisa had a huge smile on her face when she mentioned that the 6900K costs $1,100. I can't wait to see Zen's price.

Lol it will be $1200 a pop, no wonder she's smiling.

 

[witeken]

Zen 8-core at worse process node has, according to AMD, lower powe consumption ("TDP"). Uh, we'll see.

Inb4, it's really funny that in the comments we now see people worshipping the TDP numbers, while in the past I've seen people bashing them because, they say, AMD measures TDP at worst case power consumption, while, again they claim, Intel uses a more opportunistic measure.

The value of the TDP name is just as much as the value of the node name: none.

I also remember that it isn't the first time AMD has done this trickery. I remember how well the Jaguar performance per TDP was supposed to be... we know how well those chips did.

 

[rainy]

It seems 14lpp is not as Bad as expected

It seems 2. 8ghz is not base frequency and ipc is not between Sandy and Ivy...

You shouldn't be so sarcastic, even without it, a couple of members of this forum could have tomorrow a quite serious headache.

 

[HurleyBird]

I wonder why AMD displayed the power consumption of the whole system, instead of just the CPUs. The information can easily be accessed on both systems and AMD has and since there is no full FIVR in Zeppelin.

Well, I believe Zen has more chipset features integrated into the processor, so total system consumption is both going to look better for them, and be more accurate.

 

[krumme]

Piednoël is having a bit of "at first they ignore you, then they laugh at you" moment. I would expect an Intel representative, while voicing personal opinions on Twitter, to show a minimum of respect for the competition. His childish reactions are unbecoming, and reflect badly on the company as well.


He thought they were pitching Zen vs. 6700K.

I am pretty sure he will be told to shut up.
You dont actually want someone sounding like a qualcomm fixed function marketeer when you are the worlds no 1 cpu maker. Sad.

 

[AtenRa]

Nice, 95W TDP vs 140W TDP , both at 14nm and RYZEN is faster at least in those two scenarios.

Those two tests make me think that their SMT is working as it should and per Core throughput is way higher vs CMT XCV designs.

One more thing left to know, perf/$