ZEN ES Benchmark from french hardware Magazine

[Timmah!]

I really hope AMD's cpu is good. My last AMD cpu purchase was an Athlon64 X2 over 10 years ago, I've bought nothing but intel since. I want a nice 8 core cpu so I really hope this AMD chip lives up to the hype and is released at a great price ($500 - $600).

My last AMD was A64 3200+ with a Venice core back at the end of 2005 i think? I was so happy about it, as it was quite an upgrade over my old Duron 700 rig... only about half year later for the first dual-cores to pop-up and making it obsolete. I had excellent plans to replace it with 4400+ next year, as soon as i was going to have spare money, but (luckily! before i could, Core 2 Duos hit the market. So evetually i got E8400 at the beginning of 2008. Since then, only Intel about for me as well, but i guess thats true for 99 percent of resident posters.

Anyway, its good to see AMD back. Only recently i got X99 rig, so any potential price war wont affect me, but hopefully AMD can keep this up for coming years, so when its time to upgrade again, i wont have to pay over the odds for just one brand with no competition again.

 

[Timmah!]

What core 2 did was historic. Sandy is the closest thing since. Ryzen could be another great moment in cpu history, but it's not like Conroe. It is not faster than anything else. It looks good but it's pricing and overclockablity is still in question.

IMO Core2 to Nehalem was bigger step than Nehalem/Westmere to Sandy. I may be wrong though.

 

[bjt2]

In the last page of this scan https://imgur.com/a/KOXPd , it can be seen that they have an ES of kabylake, running at 2.7GHz... Noone believe that this will be the final clock. But in the case of Zen, early ES of 2.8GHz was the indication of max 3GHz base clocks...

 

[Doom2pro]

In the last page of this scan https://imgur.com/a/KOXPd , it can be seen that they have an ES of kabylake, running at 2.7GHz... Noone believe that this will be the final clock. But in the case of Zen, early ES of 2.8GHz was the indication of max 3GHz base clocks...

Bias, some of it AMD's fault, some of it excessive irrational enthusiasm for a favorite company.

 

[TheELF]

In the last page of this scan https://imgur.com/a/KOXPd , it can be seen that they have an ES of kabylake, running at 2.7GHz... Noone believe that this will be the final clock.

Because we already knew skylake clocks...Zen is completely new so no one can guess where it will end,not even lisa knew 2 weeks ago, she was still like "or" (maybe) (if we get lucky) it will be higher,.

 

[Abwx]

Because we already knew skylake clocks...Zen is completely new so no one can guess where it will end,not even lisa knew 2 weeks ago, she was still like "or" (maybe) (if we get lucky) it will be higher,.

Lol, total non sense, in line with your post history that amount basically to spread FUD.

Of course that she dont know exactly since they are in the stage of deciding the exact segmentation and number of SKUs :

the base clock for the high-end SKU will be at least 3.4+ GHz. The fact that AMD says ‘at least’ dictates that they are still deciding exactly what to do here

http://www.anandtech.com/show/10907...nvme-neural-net-prediction-25-mhz-boost-steps

 

[lolfail9001]

In the last page of this scan https://imgur.com/a/KOXPd , it can be seen that they have an ES of kabylake, running at 2.7GHz... Noone believe that this will be the final clock. But in the case of Zen, early ES of 2.8GHz was the indication of max 3GHz base clocks...

Are you still dreaming of 4Ghz 8 core at 95W TDP ? Seriously, if you want to know, early ES of Skylake-S have 2.2Ghz base clock, so you may bring it up as well. But, newsflash, 6700T has 2.8Ghz base clock. Not that large of a gain in absolute values, is it ?

Of course that she dont know exactly since they are in the stage of deciding the exact segmentation and number of SKUs :

I see they don't plan to release Ryzen before Skylake-E, do they? With Polaris AMD started mentioning exact market target explicitly since the very early samples.

 

[TheELF]

Lol, total non sense, in line with your post history that amount basically to spread FUD.

Of course that she dont know exactly since they are in the stage of deciding the exact segmentation and number of SKUs :



http://www.anandtech.com/show/10907...nvme-neural-net-prediction-25-mhz-boost-steps

Why do you feel the need to counter argue with the exact same things I said?
We knew kaby would have at least skylake clocks,but not even lisa knows what clocks zen will end up with,how does what I say differ from what you say?
They only know it will be 3.4 OR higher because they reached 3.4 JUST NOW.

 

[Abwx]

Why do you feel the need to counter argue with the exact same things I said?
We knew kaby would have at least skylake clocks,but not even lisa knows what clocks zen will end up with,how does what I say differ from what you say?
They only know it will be 3.4 OR higher because they reached 3.4 JUST NOW.

She "dont know" for the public, do you expect her to give clues to the competition on how positioning
their own offerings, and this weeks before Zen is even launched..?.

As for the KBL sample they got it back in june, not sure that you had any info on exact frequencies at the time but that s anothe debate, the only thing of interest is that it was a A0 sample, just like the Zen sample they tested wich is from the first batches that were used in June for the first Blender demo..

 

[Arachnotronic]

Are you still dreaming of 4Ghz 8 core at 95W TDP ? Seriously, if you want to know, early ES of Skylake-S have 2.2Ghz base clock, so you may bring it up as well. But, newsflash, 6700T has 2.8Ghz base clock. Not that large of a gain in absolute values, is it ?

I see they don't plan to release Ryzen before Skylake-E, do they? With Polaris AMD started mentioning exact market target explicitly since the very early samples.

Ryzen should beat Skylake-X to market in enthusiast. Both Skylake-EP/EX and Zen Opterons should formally launch in Q2 2017 (though Skylake-EP/EX are shipping to select server customers now).

 

[cytg111]

IMO, arguing 3.4 or 3.5 or +- 5% ipc trying to deduce summed benches from a review of an engineering sample in terms of absolute facts is pointles right now.. Remeber the TLB? Bulldozer had thing too fixed with microcode that resultet in lesser perf IIRC.
From now to release they can find another 300MHz turbo headroom only to loose it all to a bug.
Predictions are so volatile right now, getting our panties in a bunch over 5% is lunacy. Wait it out and dont go internet tough-guy. My advice anyway. Happy holidays.

 

[DrMrLordX]

IMO Core2 to Nehalem was bigger step than Nehalem/Westmere to Sandy. I may be wrong though.

I agree. Moving away from the onboard memory controller to the IMC was a big leap for Intel platforms. It gave uniformity of memory controller quality to all Intel CPUs, and it (along with the ring bus) effectively eliminated the FSB bottleneck. It was a big leap forward for Intel's desktop platform.

Remeber the TLB?

Gahhh please AMD don't have another one of those problems.

I also remember them promising that Phenom would run on AM2 motherboards, which often did not work well (if at all). Abit dropped support for my nf-m2 nview, scuttling my plans to run Agena or even Kuma/Rana on that board. Meh. It's a good thing Summit Ridge is making a fresh start on AM4.

 

[Sven_eng]

Bugs aren't a problem when the arch is good, remember SB chipset bug? TLB was extra bad because Phenom I was terrible and this was the final nail.

 

[TheELF]

She "dont know" for the public, do you expect her to give clues to the competition on how positioning
their own offerings, and this weeks before Zen is even launched..?.

Kabylake is already being sold in small numbers,production is done and intel can't do anything more or less then what they can in a few months,whenever zen get's released.
A stronger showing from zen would get them more attention and more willingness for people to wait until they are released.

As for the KBL sample they got it back in june, not sure that you had any info on exact frequencies at the time but that s anothe debate, the only thing of interest is that it was a A0 sample, just like the Zen sample they tested wich is from the first batches that were used in June for the first Blender demo..

We didn't knew exact numbers for final clocks but we knew they wouldn't be lower then sky,we had something to compare them with .Zen is a completely new design we knew almost nothing about,we have nothing to compare them with.

 

[ksec]

From an ES and Turbo turned off, I expect the final version to be at least 3-5% better. Turbo is extremely important in single thread performance which I expect Intel will continue to lead. So while those numbers aren't leapfrogging Intel ( Who ever had thought of that may be dreaming too much), they certainly are very encouraging.

The power usage is also great! Which is really the surpressing factor here. It may not be hard for AMD to beat Intel with a 150W chip compared to 95W. But getting it close and similar power usage! This is also on an inferior node compared to Intel.

I am pretty sure the 14nm GF using are not the latest from Samsung. Samsung 14nm are on generation 3, my guess at best GF is on gen 2. GF next node is 7nm targeting early 2018, and product to market at Mid to Fall 2018. Not sure how AMD would play this out. They could have Zen+ 14nm Gen3 for 2018, and Zen+ 7nm from 2019.

Intel's lousy desktop roadmap means 10nm aren't coming until 2019. And 6 core Coffeelake ( which is pretty much the same as Kabylake, and pretty much the same as Skylake) coming in early 2018. This is about milking the bloody market at its best. And competition is good !

But of course Intel could always change its roadmap. Their roadmap has been very inconsistent for the past 2-3 years. Since they have virtually no competition on both PC and Server market. I guess they have the flexibility to adopt quickly according to market situation rather then rushing it.

On another important point which no one else has mentioned yet, is software optimization. AMD has presented a few benchmarks previously but they were all open source projects. Which means you could tune and compile with optimization that is not Intel only. Now these are games on the market and very likely with no tuning at all. That means there are even more potential performance! And since the gaming development world has moved to using middleware game engine like Unity / Unreal etc, any optimization done in these few game engine will see results in all games ( Assuming they provide you a download update )

I pray AMD don't fxxk it up this time. They have a long history of doing so.

 

[itsmydamnation]

if AMD do another Zen iteration on 14nm i tihkn we are far more likely to see it on 14 HP then the LPP follow on (forget its name) .

 

[krumme]

What are gf 14 HP plans?

 

[itsmydamnation]

What are gf 14 HP plans?

its IBM's 14nm that GF took on board with the agreement to handle IBM's chip manufacturing going forward.

 

[bjt2]

Are you still dreaming of 4Ghz 8 core at 95W TDP ? Seriously, if you want to know, early ES of Skylake-S have 2.2Ghz base clock, so you may bring it up as well. But, newsflash, 6700T has 2.8Ghz base clock. Not that large of a gain in absolute values, is it ?

I see they don't plan to release Ryzen before Skylake-E, do they? With Polaris AMD started mentioning exact market target explicitly since the very early samples.

It's not a dream. Maybe not with the first batches and not in 95W, but in the future it will be achievable... What was the clock of first 14nm INTEL batches? They went mobile first with low power broadwell, so we may not ever know the difference in clock potential between early and mature 14nm... AMD is going desktop first and demonstrated an overvolted 3.4GHz chip to draw less than 95W... I am confident that first batches will be at least 3.6GHz base at 95W and maybe a black edition 125W with 3.8-3.9 base... Then optimizations will come...

If we stick with AMD promise of same energy/clock per core as XV and we know that XV is more efficient than the FX8370 (4/4.3 @ 125W), we could even have 4GHz base, but with 125W... Time will improve the silicon and we can even have 95W parts with 4GHz...

Remember that Zen has lower FO4 than BW/SKL... Otherwise they couldn't manage to have higher clocks and lower power on an inferior process, with an overvolted ES...

 

[bjt2]

She "dont know" for the public, do you expect her to give clues to the competition on how positioning
their own offerings, and this weeks before Zen is even launched..?.

As for the KBL sample they got it back in june, not sure that you had any info on exact frequencies at the time but that s anothe debate, the only thing of interest is that it was a A0 sample, just like the Zen sample they tested wich is from the first batches that were used in June for the first Blender demo..

Indirectly AMD gave us clue of final clock: they promised same energy/clock/core than XV. I already made calculations that led to the famous 4GHz@95W. Maybe not the first batches, but eventually i think my forecast will come true...

 

[krumme]

its IBM's 14nm that GF took on board with the agreement to handle IBM's chip manufacturing going forward.

And GF 7nm is that a Samsung variant?

 

[BeepBeep2]

And GF 7nm is that a Samsung variant?

GF uses Samsung 14nm LPE/LPP if I remember correctly. (ie. Part of the reason why you saw AMD going Samsung rumors)

The IBM transaction was mostly for engineering talent moving forward and fully owning Fab 8 and Fab 10 in NY. I don't think either was on 14nm yet when the acquisition was announced.

 

[itsmydamnation]

GF uses Samsung 14nm LPE/LPP if I remember correctly. (ie. Part of the reason why you saw AMD going Samsung rumors)

The IBM transaction was mostly for engineering talent moving forward and fully owning Fab 8 and Fab 10 in NY. I don't think either was on 14nm yet when the acquisition was announced.

The guy who gave the power9 hot chips presentation called it the global foundries 14nm HP process (34:34 into this https://www.youtube.com/watch?v=Ln9WKPEHm4w)

 

[BeepBeep2]

The guy who gave the power9 hot chips presentation called it the global foundries 14nm HP process (34:34 into this https://www.youtube.com/watch?v=Ln9WKPEHm4w)

http://www.samsung.com/semiconductor/foundry/process-technology/14nm/
http://www.globalfoundries.com/technology-solutions/leading-edge-technology/14-lpe-lpp
http://globalfoundries.com/docs/def...lobalfoundries-14nm-collaboration---final.pdf
FX-14 ASIC is produced on 14nm LPP (originated from Samsung) which should be the same process that will be used for the Ryzen CPUs

There is some evidence that the 14nm HP process IBM is using for POWER9 is different from the Samsung nodes but GlobalFoundries doesn't seem to reference any process by this name

 

[itsmydamnation]

http://www.samsung.com/semiconductor/foundry/process-technology/14nm/
http://www.globalfoundries.com/technology-solutions/leading-edge-technology/14-lpe-lpp
http://globalfoundries.com/docs/def...lobalfoundries-14nm-collaboration---final.pdf
FX-14 ASIC is produced on 14nm LPP (originated from Samsung) which should be the same process that will be used for the Ryzen CPUs

im not talking about Zen, originally we where talking about a Zen follow on and what process it would use. Which lead to the does GF actually run the 14nm process IBM created, to which i answered with it looks like it because the IBM chip architect calls the process GF 14nm HP.