AMD Zen Performance Preview

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
49,713
Were you able to reproduce their results on your 6950x with two cores disabled at 150 samples?

Because I re-ran the benchmark with 150 samples, and I still have some serious doubts.
Well, from what i see on screen it looks like within a second from live demo.
I am damn near dead on. The results from my timer using the live stream are ~35.75 seconds.
 

Gideon

2[H]4U
Joined
Apr 13, 2006
Messages
2,670
Glad to hear you replicated their result Kyle, makes it far less likely that anything was being hidden or distorted. Now the question is will this result carry over into other tasks.
 

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
49,713
Huh. Then my results I posted in the other thread make absolutely no sense. (45.04s with the 150 sample file on my 3930k @ 4.8)

Blender must have some sort of funky core scaling issues going on for the results to wind up as they do.
Well, you are trying to match results from a different CPU architecture with less cores. I am not sure what you are trying to accomplish with that hardware in relation to what AMD showed us this week.
 

Teenyman45

2[H]4U
Joined
Nov 29, 2010
Messages
2,403
Now watch AMD finally come out with a part almost as good as Intel's but completely screw up pricing it like they did with the Fury and 7900 series gpus.
 

sirmonkey1985

[H]ard|DCer of the Month - July 2010
Joined
Sep 13, 2008
Messages
22,016
AMD doesn't have to outright beat Intel in performance to start winning again (would be a plus but very unlikely), it is more important for them improve their margins again and get back into the black to fund future R&D.
agree that they don't need to outright win, if they can get close with a competitive price this will definitely give people more options. i really do hope it's able to be competitive because i'd like to go back to supporting AMD products.
 

ebduncan

[H]ard|Gawd
Joined
Feb 1, 2008
Messages
2,007
I will say I am bit more excited for Ryzen, or Zen what ever they want to call it.

They come straight out and say 3.4ghz base speed which I think took everyone by surprise. Esp when people were predicting them to be low 3ghz max speed parts. Now that is all left to see is how high can go, and of course a much wider benchmark suite.

Glad I've held out on a upgrade for so long, appears it was worth waiting for.
 

SighTurtle

[H]ard|Gawd
Joined
Jul 29, 2016
Messages
1,412
wow went quiet in here all of a sudden! whered all the tin-foil hats go? still think amd was lying and giving out false info, faking tests and hiding altered settings? geez....
Its not too far to suspect AMD is lying...history informs the future. But, its good everything is resolved, now time for launch and hopefully a HardOCP review soon after.
 

Palladium@SG

Limp Gawd
Joined
Feb 8, 2015
Messages
284
wow went quiet in here all of a sudden! whered all the tin-foil hats go? still think amd was lying and giving out false info, faking tests and hiding altered settings? geez....
Because the last time Core 2 was destroying A64 in everything Intel was sending pre-release samples to everybody to test, instead of just leaking numbers for just one specific app?
 

razor1

[H]F Junkie
Joined
Jul 14, 2005
Messages
10,155
wow went quiet in here all of a sudden! whered all the tin-foil hats go? still think amd was lying and giving out false info, faking tests and hiding altered settings? geez....
Its one application, we should really wait for other tests, added to this the last two time AMD had something good, K8 and A64, they were chest pounding a month before release. Now having said that, AMD keeping quiet can still be very good, because we saw that chest pounding with their GPU's and CPU's (phenom) were just disappointments (we kinda can throw BD in there too, because it was an AMD employee that blew it out of the water even though it wasn't done by official channels), so it could just be an apprehension of what the backlash was or they have learned a lesson.
 

trick0502

Supreme [H]ardness
Joined
Apr 17, 2006
Messages
5,251
Because the last time Core 2 was destroying A64 in everything Intel was sending pre-release samples to everybody to test, instead of just leaking numbers for just one specific app?
If zen was destroying intel, I'm sure they send out samples to everyone too. What we're looking at is a part that is on par with current intel offerings.
 

Nenu

[H]ardened
Joined
Apr 28, 2007
Messages
19,200
wow went quiet in here all of a sudden! whered all the tin-foil hats go? still think amd was lying and giving out false info, faking tests and hiding altered settings? geez....
Can you link us to the last time this wasnt an issue?
 

Palladium@SG

Limp Gawd
Joined
Feb 8, 2015
Messages
284
If zen was destroying intel, I'm sure they send out samples to everyone too. What we're looking at is a part that is on par with current intel offerings.
Matching Intel HEDT per core is already destroying Intel. AMD's current CPUs are in such shambles that they have nothing to lose by previewing samples.
 

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
49,713


Kyle didn't you mention AMD should spell check their slides?
Hmm, I did fix that. It is pulling from a cached version. Lots of security changes lately and this has to do with those. Thanks for pointing this out.
 

ManofGod

[H]F Junkie
Joined
Oct 4, 2007
Messages
11,936
Now watch AMD finally come out with a part almost as good as Intel's but completely screw up pricing it like they did with the Fury and 7900 series gpus.

Fury X and Fury Nano, yes. The 7970 and the 7950 were priced just as they should have been though. I have a feeling their prices will be exactly were we are willing to pay but, we will see. :)
 
  • Like
Reactions: N4CR
like this

MavericK

Zero Cool
Joined
Sep 2, 2004
Messages
29,423
I'm all for hopping back over to AMD CPUs if the performance is there. Intel's latest offerings don't really seem to be much progress over previous generations. Hell, you can still get by with a 1st or 2nd gen i7 for a lot of uses.

I also find it weird that my current processor (4790K) is still essentially the same price as a 6700K.
 

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
49,713
I'm all for hopping back over to AMD CPUs if the performance is there. Intel's latest offerings don't really seem to be much progress over previous generations. Hell, you can still get by with a 1st or 2nd gen i7 for a lot of uses.

I also find it weird that my current processor (4790K) is still essentially the same price as a 6700K.
Well, what all new tech are these new motherboards going to support and how well are those going to work. Big part of the picture....
 

muxr

Limp Gawd
Joined
May 27, 2016
Messages
151
Fury X and Fury Nano, yes. The 7970 and the 7950 were priced just as they should have been though. I have a feeling their prices will be exactly were we are willing to pay but, we will see. :)
You really can't complain about Fiji prices either. It is not like AMD was acting greedy. It was a test run for HBM more than anything and implementing HBM was not cheap. I would be surprised if they made any money on it. They probably lost money, considering how Fury series barely even registers on Steam hardware survey.

Ryzen doesn't have this problem. They are also already selling product on the 14nm, so the yields are probably decent at least. I fully expect we'll see a competitive price.
 

Ultima99

Supreme [H]ardness
Joined
Jul 31, 2004
Messages
4,890
Potentially matching Broadwell-E is a good start. By time this thing launches it will be going up against Skylake-X and Kaby Lake-X so lets hope it can bring the clockspeeds to match those.
 

KingGlade

Limp Gawd
Joined
Dec 15, 2016
Messages
147
I am very excited about these new AMD CPU's and if performance can even come close to broadwell-e at a decent price I'll purchase it just support AMD. Intel needs some competition.
 

thesmokingman

Supreme [H]ardness
Joined
Nov 22, 2008
Messages
5,912
Thank you Kyle, for confirming this. Stupid on AMD, but this should clear up anyone saying this will never match and amd is straight out lying. Like you said they should have paid more attention to the file, lol. WTF amd is as true as it gets.
It's clear they delegate way too much, so much so that the details get lost. They need to hire some continuity people aka script supervisors from Hollywood to maintain at least some appearance of singularity of thought and logistics. Script or continuity supervisors maintain the flow of continuity from scene to scene with regards to props, wardrobe, anything and everything. For ex. if ya link a 100 sample file, you should note it so that ppl are not left scratching their heads, going wtf? All these stupid mistakes only provide fuel for the haters.
 

Doward

[H]ard|Gawd
Joined
Jul 3, 2004
Messages
1,541
One reason I'm considering Zen vs Skylake-E, is ability to upgrade the CPU across generations, without buying an entire new motherboard. I'd be happy with Broadwell-E levels of performance.
 

noko

Supreme [H]ardness
Joined
Apr 14, 2010
Messages
5,149
One reason I'm considering Zen vs Skylake-E, is ability to upgrade the CPU across generations, without buying an entire new motherboard. I'd be happy with Broadwell-E levels of performance.
Much agree, great having options without having to update the whole system, OS etc. Now I wonder if the dual channel DDR 4 will ever really become a limitation for the Platform over time and what advantage does Intel quad channel memory has if any.
 

hurleybird

Weaksauce
Joined
Jan 14, 2012
Messages
106
Much agree, great having options without having to update the whole system, OS etc. Now I wonder if the dual channel DDR 4 will ever really become a limitation for the Platform over time and what advantage does Intel quad channel memory has if any.
There are probably some really obscure edge cases where the quad channel memory controller makes a tangible difference, although DDR 4 will continue to get faster for awhile. That said, and I'm by no means predicting AMD will do this, but they could always put a stack of HBM on the package as a last level cache if they wanted to maintain platform capability without increasing the number of memory channels in the event they became limited by memory bandwidth in the future.
 

noko

Supreme [H]ardness
Joined
Apr 14, 2010
Messages
5,149
There are probably some really obscure edge cases where the quad channel memory controller makes a tangible difference, although DDR 4 will continue to get faster for awhile. That said, and I'm by no means predicting AMD will do this, but they could always put a stack of HBM on the package as a last level cache if they wanted to maintain platform capability without increasing the number of memory channels in the event they became limited by memory bandwidth in the future.
I wonder if the X370 motherboards will be able to support a type of Interposer for the CPU and HBM memory if that ever becomes available. APU's in particular with the platform.
 

tangoseal

[H]F Junkie
Joined
Dec 18, 2010
Messages
8,577
Getting excited at Zen. Hope it pans out well in your [H] official review. We will know in a few weeks once they hit the shelves.
 

ole-m

Limp Gawd
Joined
Oct 5, 2015
Messages
452
It's for sure faster than my 10 core:
E5-2680 V2 10 core 2.8ghz/ 3.6ghz turbo
40 seconds
 
Joined
Jul 7, 2011
Messages
48
@Kyle :
I'm just wondering if the 25MB of L3 Cache of the 6950X is causing any discrepancies.
The 6900K only has 20MB of L3 cache.

Disabling two cores is fun, but you can't disable the extra 5MB of L3 cache ?
 

aztekk

Limp Gawd
Joined
Nov 4, 2014
Messages
166
@Kyle :
I'm just wondering if the 25MB of L3 Cache of the 6950X is causing any discrepancies.
The 6900K only has 20MB of L3 cache.

Disabling two cores is fun, but you can't disable the extra 5MB of L3 cache ?
You can't but I'm guessing it's probably not gonna make a huge difference since this is not a memory intensive benchmark. Usually rendering tasks do perform better with more cache however, so it would still be helpful if someone with a real i7-6900k could reproduce this result, if only to convince the last remaining skeptics around.
 
Joined
Jul 7, 2011
Messages
48
You can't but I'm guessing it's probably not gonna make a huge difference since this is not a memory intensive benchmark. Usually rendering tasks do perform better with more cache however, so it would still be helpful if someone with a real i7-6900k could reproduce this result, if only to convince the last remaining skeptics around.
Yup, there's nothing better than a real comparison.

The only way to do a correct reproduction of the results, is to use the same hardware+timings in all the tests, where possible.
Using a CPU with 25% extra L3 Cache, invalidates the results of the i7 for me.

As we see, Kyle is almost spot-on with the Zen chip, when the parameters are correct.
 

JustReason

razor1 is my Lover
Joined
Oct 31, 2015
Messages
2,485
wow went quiet in here all of a sudden! whered all the tin-foil hats go? still think amd was lying and giving out false info, faking tests and hiding altered settings? geez....
My good sir, you deserve an award for that. Laughed my rear-end off.
 
Top