AMD Zen Performance Preview

Shintai said:
You can download and run the test yourself.
http://www.amd.com/en-us/innovations/new-horizon

My 6700K uses ~65W running the bench and not ~90W. ;)
Click to expand...

65w total system power?, also, how long does it take for you to run it?, i seem to remember that the test was a Watt+Time test.

Still, this is closer to Intel than Bulldozer was. But yeah i can see the point that you all are saying, let's wait and see.
 
pendragon1 said:
Click to expand...

Wait wait wait a second.... Did I hear that right? 3:40...

3.4ghz Zen, Turbo DISABLED vs. a 3.2ghz i7 6900k with Turbo ENABLED to 3.7ghz
and the Zen edged it in one test and tied it in another while using 15 watts less? The 6900k is a Broadwell core 5th gen Core series. Everything I've heard is that performance estimates were Ivy Bridge (3rd gen) or Haswell (4th gen) at most. Either AMD pulled a fast one or we are missing part of the story here.
 
Revdarian said:
65w total system power?, also, how long does it take for you to run it?, i seem to remember that the test was a Watt+Time test.

Still, this is closer to Intel than Bulldozer was. But yeah i can see the point that you all are saying, let's wait and see.
Click to expand...

77 seconds. Nobody have managed to get even remotely close to AMD numbers. Nor with 5960X, 6900K or 6950X.

If I had to guess they use a custom version of Blender and with an IPC load that's so low its down to clocks only.
 
The thing is the blender result isn't a benchmark is it, it's a demo. It's not a standardised test , any 3D rendered can be tweaked to a near infinite levels, passes, ray reflection limits etc. I used to work in animation industry and tweaking settings can change image quality and render time exponentially. We can only assume both machines were setup the same in terms of render settings, but unless AMD tell us what those settings are just having the render files is useless in trying to produce comparable results on our home systems.
 
One thing that has me worried is the comparison test are both for workloads that scale near perfectly with clock and core count. Rendering and video recompression are to paraphrase are ridiculously parallel work loads. Realtime gaming isn't and counts more on other things that just brute force, cache efficiency, bandwidth, branch prediction will be tested far more in realtime gaming. Still if they are even close to intel it has to be the greatest comeback in chip design in years.
 
For me price is going to be the biggest factor. If AMD could manage a CPU that is 8 cores and 16 threads that is in the $400 range and it can compete with a CPU like the 6900k then I'll be back on the AMD bandwagon real quick.
 
qbanb8582 said:
For me price is going to be the biggest factor. If AMD could manage a CPU that is 8 cores and 16 threads that is in the $400 range and it can compete with a CPU like the 6900k then I'll be back on the AMD bandwagon real quick.
Click to expand...

If they can make a CPU that beats the 6900k and sell it for $400, they're idiots. One, they'd probably be losing money on every chip at that price. Two, Intel proved you can sell the chips for twice that price and people will still buy them.

I could see them undercutting Intel's price, but not by that much. If they do have some way to make that complex of a CPU and sell it at a profit for $400, expect Intel to drop prices pretty quick to compete. Either way, if this thing really can perform as AMD would have you believe, we're looking at some good news for consumers.
 
CaptNumbNutz said:
Wait wait wait a second.... Did I hear that right? 3:40...

3.4ghz Zen, Turbo DISABLED vs. a 3.2ghz i7 6900k with Turbo ENABLED to 3.7ghz
and the Zen edged it in one test and tied it in another while using 15 watts less? The 6900k is a Broadwell core 5th gen Core series. Everything I've heard is that performance estimates were Ivy Bridge (3rd gen) or Haswell (4th gen) at most. Either AMD pulled a fast one or we are missing part of the story here.
Click to expand...

Yep, and the results are highly contested.

No one has been able to replicate them, using the blender file downloaded from their site and the version of blender they say to use.
 
6900k in the UK is around £1100 anything around £500-600 would be an absolute steal, and IMHO make people like myself with an old i7-2600k seriously consider upgrading where it really hadn't been on my agenda.
 
Zarathustra[H] said:
Yep, and the results are highly contested.

No one has been able to replicate them, using the blender file downloaded from their site and the version of blender they say to use.
Click to expand...

They've been replicated, well at least the blender test once someone on Anand zoomed in and saw the sampling size. Something else I read is that the previous release of blender default was 100 and the newest version is 200. It could be a mistake or someone not on the job failing to realize everyone always has their pitchforks at the ready anytime it comes to AMD. I think that's the thing AMD doesn't realize. Everyone wants them to fail, they are the kid everyone loves to kick.

Here's my 6700k lol

awwVWWkl.jpg
 
Sampling size of 100 doesn't fit the results either. Neither does 128 for that matter.
 
their pitchforks at the ready anytime it comes to AMD. I think that's the thing AMD doesn't realize. Everyone wants them to fail, they are the kid everyone loves to kick.
Click to expand...

I don't think this is the case at all. However we do have the right to be skeptical. I mean considering all the lies that AMD's marketing team / JF_AMD posted before the bulldozer release.. Although they did fire many involved in that including John Fruehe.
 
drescherjm said:
I don't think this is the case at all. However we do have the right to be skeptical. I mean considering all the lies that JF_AMD posted before the bulldozer release..
Click to expand...

Bullshit, slightest hiccup and everyone is screaming. It was a nice gesture for the ceo to try, it didn't play out like how they imagined because a simple setting caused a mix up. But that's not how a lot saw it and you know it. The NDA hasn't lifted afaik and when it does, there will be plenty of time to get to the nitty gritty. Ppl are expecting to spend the night when its just flirting at the bar.
 
Well, their Blender time at whatever setting they used for both systems was at 20seconds less, but yeah those point to a massively parallelized test to show such a difference between your 6700's 77seconds and their time.

I will maintain what i said in the past, that noone even thought that this was a possibility. Similarly with the BF1 test were both systems were able to run the game at 4k with around the same performance. Of course that 4k was chosen to showcase a specific cap of up to 60fps, since that is what 4k monitors can handle, so there may be extra leeway into the 6900s.

**modifying the time and explaining that it was less than your time, theirs was at 55-59 no?... i need to rewatch the benchmarks released or just wait until jan... naw just kidding, i will look for them lol.

editting again, this time to quote a paragraph from a forbes article:
"The ultimate performance comparison tests against the 6900K will be when third-party reviewers test production systems in Q1 2017 after Ryzen launches. It’s likely both AMD and Intel will have some tricks up their sleeve between now and then to make themselves look better. AMD hasn’t even shown turbo speeds which will not only boost maximum performance, but could also impact power. This is the most desktop excitement in years and just the notion of Ryzen being even close to Intel’s highest-end processor is impressive as this hasn’t happened in nearly a decade."

That about sums it up. Congrats to AMD, hopefully it works out, as the market could use this.
 
Revdarian said:
Well, their Blender time at whatever setting they used for both systems was at 20seconds less, but yeah those point to a massively parallelized test to show such a difference between your 6700's 77seconds and their time.

I will maintain what i said in the past, that noone even thought that this was a possibility. Similarly with the BF1 test were both systems were able to run the game at 4k with around the same performance. Of course that 4k was chosen to showcase a specific cap of up to 60fps, since that is what 4k monitors can handle, so there may be extra leeway into the 6900s.

**modifying the time and explaining that it was less than your time, theirs was at 55-59 no?... i need to rewatch the benchmarks released or just wait until jan... naw just kidding, i will look for them lol.
Click to expand...


That blender file rendering was not heavy on the CPU's nor is it heavy on parallelization, as bender itself is good with parallelization. The first time AMD showed off with Blender they tried to show IPC, and now you are trying to switch that to parallelization? What they are trying to show is they can match up with Intel's specific chip, without giving us any details about the settings used to get those results.

here is just an example of what I'm talking about

https://www.blenderguru.com/articles/4-easy-ways-to-speed-up-cycles/

http://www.blenderguru.com/articles/13-ways-to-reduce-render-times/

The second one is quite interesting, because all of the settings they are talking about, are specific to CPU work load and cycles, these change depending on the scene and CPU (different institution sets) can have different performance affects.

And the BF1 test is a farce, its heavily GPU limited even on a titan X at those settings so it doesn't show us anything about the CPU performance.
 
They used SLIed Titan X, so not really a farce on that side, more on the 60fps cap side IMHO.

Still we shall wait for real world benchmarks, and actual prices to go with them.
 
Shintai said:
They used 2.77. And sample size of 150 wont do either it seems. Specially not for the press one.

And of this time the Blender file is unchanged.

Credibility down a notch more.
Click to expand...
For the press one was done at 100 samples and times do look to line-up.

Now, up to test the 150 samples or something.
 
Shintai said:
They used 2.77. And sample size of 150 wont do either it seems. Specially not for the press one.

And of this time the Blender file is unchanged.

Credibility down a notch more.
Click to expand...

*Shrug* They could do nothing that could be credible in your eyes anyways so no lose there.
 
I've been thinking about this... What's the point in using Blender as a benchmark for a gaming or enthusiast audience? Seem we would be more interested in 3dmark or something similar. How about Ashes?

Seems to me if the chip were up to snuff we'd see more than a single benchmark.
 
Spirit_Retro said:
I've been thinking about this... What's the point in using Blender as a benchmark for a gaming or enthusiast audience? Seem we would be more interested in 3dmark or something similar. How about Ashes?

Seems to me if the chip were up to snuff we'd see more than a single benchmark.
Click to expand...

The eis an AOTS bench out before it got pulled and it wasn't as pretty as the Blender. Does that explain it? ;)
 
Spirit_Retro said:
I've been thinking about this... What's the point in using Blender as a benchmark for a gaming or enthusiast audience? Seem we would be more interested in 3dmark or something similar. How about Ashes?

Seems to me if the chip were up to snuff we'd see more than a single benchmark.
Click to expand...

Maybe they should have given us no taste of the cpu before NDA?
 
drescherjm said:
I wish they would have lifted the NDA now. Although we are only a month away.
Click to expand...

Yea, I think everyone wants a crack at it. We're like two months out from release. I can hardly remember the last time we got a concrete metric this far out with such a new comer.
 
drescherjm said:
I wish they would have lifted the NDA now. Although we are only a month away.
Click to expand...

Well, judging from their comments at the New Horizon thing they aren't done yet. They are still tweaking it to come up with the turbo frequencies.

They can't lift the NDA on something they haven't finished yet :p
 
Zarathustra[H] said:
Well, judging from their comments at the New Horizon thing they aren't done yet. They are still tweaking it to come up with the turbo frequencies.

They can't lift the NDA on something they haven't finished yet :p
Click to expand...

They are done, else it wont be ready in time. At this time they should start to fill up distribution centers and inventories to hit a Q1 launch.
 
Shintai said:
They are done, else it wont be ready in time. At this time they should start to fill up distribution centers and inventories to hit a Q1 launch.
Click to expand...

Well, they literally said that they hadnt established the turbo clocks yet at New Horizon.

Maybe January 17th is a soft launch?
 
Zarathustra[H] said:
Well, they literally said that they hadnt established the turbo clocks yet at New Horizon.

Maybe January 17th is a soft launch?
Click to expand...

What she says at a PR event and what is reality is not the same. If you could get her to say the same at a finance call, then sure. But at this time it must be set in stone. And the info just saved for a later announcement.
 
Bandalo said:
Q1, not day 1 of Q1. These might not hit the shelves until late March.
Click to expand...

Yeah, the date that has been out there for some time has been January 17th, but who knows what availability will be like at the official launch.

They may launch, with great fanfare, with a select few review samples out to trusted publications, but have an availability date in the future?
 
Zarathustra[H] said:
Yeah, the date that has been out there for some time has been January 17th, but who knows what availability will be like at the official launch.

They may launch, with great fanfare, with a select few review samples out to trusted publications, but have an availability date in the future?
Click to expand...

The January 17th is based on a rumour. AMD just says Q1. The same source as January 17th also said 3.15-3.3Ghz base clock with 3.5Ghz turbo.
 
Shintai said:
The January 17th is based on a rumour. AMD just says Q1. The same source as January 17th also said 3.15-3.3Ghz base clock with 3.5Ghz turbo.
Click to expand...

Ah, I could have sworn I had seen that quoted as straight from the horse's mouth somewhere, but I can't remember my source, so chances are good I'm wrong.

Too much mixing of rumor and fact to keep straight on this nonsense anymore.
 
You must log in or register to reply here.
Back
Top