Zen gets even with Broadwell-E in Blender

C

ChosenUno

[H]ard|Gawd
Joined
Aug 26, 2011
Messages
1,318
http://seekingalpha.com/article/400...utperforms-1000-intel-broadwell-e-coreminus-7

Rather iffy statement from AMD, but I think they must be quite confident to be able to state that directly.

In a live demonstration AMD matched their upcoming desktop processor "Summit Ridge" against a similarly configured Intel (NASDAQ:INTC) Core I-7 processor (most likely a Core i7-6900K).

Both units were running at a clock speed of 3MH and both utilized 8 cores and 16 threads. Although no details were released about the actual results AMD did say:

"During the event, AMD demonstrated an 8-core, 16-thread "Summit Ridge" desktop processor (featuring AMD's "Zen" core) outperforming a similarly configured 8-core, 16-thread Intel "Broadwell-E" processor when running the multi-threaded Blender rendering software with both CPUs set to the same clock speed."
Click to expand...

Zen's looking more and more appealing each day.
 
Be careful and keep the enthusiasm reined in. It would be great if AMD could trump Intel but highly unlikely. If anything it does show some solid improvement and something to build from.
 
JustReason said:
Be careful and keep the enthusiasm reined in. It would be great if AMD could trump Intel but highly unlikely. If anything it does show some solid improvement and something to build from.
Click to expand...

Actually this might be somewhat of a best case for AMD , but I doubt that performance across the board will be better for Zen then for Broadwell. Now hoping that they can manage some decent clock speeds on 4c8t and 8c16t models
 
I'll be happily swap my 2500k for one of these even if they are on par with it clock for clock. It's not like Intel's latest and greatest today is any faster anyway.

Recently I was shopping for a minor htpc cpu upgrade. Had an idea of changing the pretty nice G3258 for a quad. Stopped on a suitable quad xeon and was about to click buy when it hit me spec wise the xeon is the exact silicon x2 for x4 the cost. Fuck that.
 
In a benchmark in our lab, we were able to show a product that isn't out, has no samples, doesn't exist outside of our labs and just got delayed until some time in 2017 beating a product which has been out since Q2 of this year.

Yay for low price, good price/perf budget solutions??? Maybe?
 
CCLMD said:
In a benchmark in our lab, we were able to show a product that isn't out, has no samples, doesn't exist outside of our labs and just got delayed until some time in 2017 beating a product which has been out since Q2 of this year.

Yay for low price, good price/perf budget solutions??? Maybe?
Click to expand...
This product was shown in private in public down the street from where Intel was, it was shown to reviewers it was testes live and they were there e.g Ryan Shrout, Hot Hardware etc.
 
Not to take away from Shintai or lolfail9001's observation (as it is highly suspect to do a stock offering at this moment in time in the release cycle), but I am wondering, does AMD have any debts that they must refinance or retire which are coming due before the end of 1H 2017?

I ask because I know AMD has basically negligible cash reserves, and would not have the cash on hand to pay a bond note if it were to come due prior to end of 1H-2017. Of course end of 1H-2017 assumes Zen's hard launch is early in Q1-2017 and that it is the smashing success we all wish it would be resulting in them selling like hotcakes and pulling maximum revenue into AMD. So if someone who has been reading AMD's GAAP numbers recently has any insight supporting my hopeful scenario, I would be glad to hear it.
 
If Zen performs well enough and they can muster a decent APU out of it, I hope someone like Microsoft takes over AMD. It would be great for future consoles and maybe even CPU development. If we're really lucky, they'll retire ATi to console duty and then partner w/NVIDIA for AMD chipsets again.
 
5150Joker said:
If Zen performs well enough and they can muster a decent APU out of it, I hope someone like Microsoft takes over AMD. It would be great for future consoles and maybe even CPU development. If we're really lucky, they'll retire ATi to console duty and then partner w/NVIDIA for AMD chipsets again.
Click to expand...

As I've said on many occasions now, the CPU portion of AMD has minimal to zero value. The only thing remotely worth anything are the patents related to x86-64. The GPU portion of the company is attractive however, and I could see a number of companies having interest into purchasing that unit.

Not to take away from Shintai or lolfail9001's observation (as it is highly suspect to do a stock offering at this moment in time in the release cycle), but I am wondering, does AMD have any debts that they must refinance or retire which are coming due before the end of 1H 2017?
Click to expand...

The next *major* debt payments occur in 2019, where AMD somehow needs to come up with $600 Million in cash, plus an additional $400 Million the following year. Aside from that, there's the general day to day operating expenses AMD has to struggle with, given their lack of cash reserves.

That's why I've been so down on AMD: Zen is the absolute FINAL chance, and even then, I don't know if it will be profitable enough by simply undercutting Intel to raise the necessary cash to keep AMD in operation past 2019. $600 Million in cash is hard to raise on just product sales...
 
gamerk2 said:
As I've said on many occasions now, the CPU portion of AMD has minimal to zero value
Click to expand...

Depends whether or not a company like MS could negotiate to keep the x86 patents. If they could, it would be quite valuable to them for console production and it would royally screw Sony over since they'd have to go groveling to Intel/NVIDIA for parts and get charged through the nose.
 
5150Joker said:
Depends whether or not a company like MS could negotiate to keep the x86 patents. If they could, it would be quite valuable to them for console production and it would royally screw Sony over since they'd have to go groveling to Intel/NVIDIA for parts and get charged through the nose.
Click to expand...

Sony would just go ARM+Nvidia. Nothing lost. It would require the usage of big cores to make a change in favour of x86. The only reason they use x86 now have most likely to do with the fact that 64bit ARM wasn't ready at the time.

gamerk2 said:
The GPU portion of the company is attractive however, and I could see a number of companies having interest into purchasing that unit.
Click to expand...

Is it?

The only attractive part of the company is its console contracts that accounts for a majority of all revenue. But loyalty there is a file not found.
 
5150Joker said:
That would be fine if MS was making a handheld, not so much for a top end console. ARM would get destroyed by a Zen CPU.
Click to expand...

They would have to use Zen first. And performance? We have to see. Because its far from top performance from what is leaked.

Hewlett-Packard 500-c60 - Geekbench Browser

A 2Ghz Jaguar is around the same as a 1.45Ghz Zen in GB4 in ST.

ARM to x86 compares are off. But its not exactly stellar performance to beat.

And do I have to mention the inefficient GPU? ;)
 
Last edited:
Shintai said:
Sony would just go ARM+Nvidia. Nothing lost. It would require the usage of big cores to make a change in favour of x86. The only reason they use x86 now have most likely to do with the fact that 64bit ARM wasn't ready at the time.
Click to expand...

Wrong. The CPUs were chosen based on performance/thermal/ease of use constraints. ARM didn't meet the necessary performance, and PPC based designs didn't meet ease of use (doubly so for MSFT).

It's also worth noting this generation of console CPUs were downgrades over the last generation. Especially in Sony's case, where the Cell had double the maximum throughput of the AMD APU chosen to power the PS4. There's a reason why developers like Ubisoft and CDPR complained about CPU limitations very early on in this consoles lifecycle, and why both are getting very early functional HW upgrades, which is unprecedented for consoles.

And if anything, the performance limitations in current ARM designs would force a return back to POWER based CPUs.

Is it?

The only attractive part of the company is its console contracts that accounts for a majority of all revenue. But loyalty there is a file not found.
Click to expand...

Not really; the return is only about $10 per console sold, and console sales are notoriously front-loaded. Sure, it's a nice chunk of change, but not significant enough to warrant purchasing the company. The yields on the GPU division far exceed those contracts, nevermind the tech/patents that would be gained in such a purchase.

Depends whether or not a company like MS could negotiate to keep the x86 patents. If they could, it would be quite valuable to them for console production and it would royally screw Sony over since they'd have to go groveling to Intel/NVIDIA for parts and get charged through the nose.
Click to expand...

Based on how the contract is structured, the x86 license would be non-transferrable in that case.

MSFT would actually hurt itself majorly even if they managed to keep the license, since they'd be competing against their own OEMs. Remember how pissed they were when MSFT made the Surface tablet? Same deal. Besides, outside of the Xbox, MSFT has basically given up being a HW house at this point.
 
gamerk2 said:
That's why I've been so down on AMD: Zen is the absolute FINAL chance, and even then, I don't know if it will be profitable enough by simply undercutting Intel to raise the necessary cash to keep AMD in operation past 2019. $600 Million in cash is hard to raise on just product sales...
Click to expand...

Indeed, trying to undercut Intel is a risky game to play with Intel given Intel cash position. A war of attrition, AMD will lose.
 
I'm really looking forward to Zen and would love to switch back to an AMD platform at some point.

Having to underclock the competition to even match/slightly outpace it doesn't instill a whole lot of confidence. Fuck.

When it's massaged over a bit further and finally released, I hope ZEN has the capability to give (at least) 4GHz stock-clocked out of the box without needing a window unit air conditioner blowing on a 3x140mm radiator sporting a push-pull config of six 120+ cfm case fans...as the manufacturer recommending cooling solution.
 
Shintai said:
They would have to use Zen first. And performance? We have to see. Because its far from top performance from what is leaked.

Hewlett-Packard 500-c60 - Geekbench Browser

A 2Ghz Jaguar is around the same as a 1.45Ghz Zen in GB4 in ST.

ARM to x86 compares are off. But its not exactly stellar performance to beat.

And do I have to mention the inefficient GPU? ;)
Click to expand...

this is exactly why I think that es sample is shit. Think about it. Naples is like a year away from full release. It will be out in second half 2017. At this point I am cautious that these samples may have disabled parts.

I think its totally rubbish to think that if jaugar is performing same at 2ghz then we are getting close to excavator performance probably, and zen is not an improvement at all. I really think everyone is going nuts over these benches. I don't know after 8 cores since its an MCM design so it might lose multi core efficiency. But these benches I am pretty certain are not reflective of zen core. I think the chip might have l3 disabled. Any how people will believe what they believe and I think some people really do wish it fails though. lol. its pretty evident.
 
DejaWiz said:
I'm really looking forward to Zen and would love to switch back to an AMD platform at some point.

Having to underclock the competition to even match/slightly outpace it doesn't instill a whole lot of confidence. Fuck.

When it's massaged over a bit further and finally released, I hope ZEN has the capability to give (at least) 4GHz stock-clocked out of the box without needing a window unit air conditioner blowing on a 3x140mm radiator sporting a push-pull config of six 120+ cfm case fans...as the manufacturer recommending cooling solution.
Click to expand...

intel broadwell E 6900k is clocked at 3.2ghz. AMD does not need 4ghz. They will be fine with 3ghz even as long ipc improvements are significant. However I do expect 4 core zen to be clocked higher. Since 8 core design is geared towards multi threaded app. if they can launch at 3.2 with 3.5 turbo that will be just fine for the 8core zen part.

shit if amd can get even within 20% of broadwell and price their chip at 500 vs 1099 from intel. I say that is one hell of a deal lol.
 
Mav451 said:
Where are you seeing this?
Click to expand...
dont know benchmark says 0, but some people are saying it might be disabled others are saying geekbench does it on all cpus which is sort of not true. It recognizes mine L3 on my 6700k fine. So its a hit or miss. If could be or could not be enabled. Comparing to older amd cpu results the single core looks too low so I think something is off suspect.
 
NKD said:
intel broadwell E 6900k is clocked at 3.2ghz. AMD does not need 4ghz. They will be fine with 3ghz even as long ipc improvements are significant. However I do expect 4 core zen to be clocked higher. Since 8 core design is geared towards multi threaded app. if they can launch at 3.2 with 3.5 turbo that will be just fine for the 8core zen part.

shit if amd can get even within 20% of broadwell and price their chip at 500 vs 1099 from intel. I say that is one hell of a deal lol.
Click to expand...


Yes, ZEN will eventually need 4GHz because Intel sure as hell isn't going to stay static with their core clocks over the next few years.

ZEN better be able to hit 4GHz (or higher) as a stock-clock out of the box at some point in the future in order to keep up with Intel.
 
There's no way Zen is even going to approach 4GHz. The process node AMD chose (or was forced to choose via the WSA) is optimized for < 3GHz clockspeeds, and that partially explains the ES sample at 2.8, and why AMD chose to underclock an Intel chip, rather then OC their own, when doing a head to head comparison. This also explains why Zen got pushed at least one Quarter: AMD is hoping yields improve to the point they can do a clock bump.

I'm predicting a base 3.2 GHz base clock, and a chip that doesn't OC well.
 
The node alone is far from the blame. IPC and high clocks is really hard.

it wouldn't clock much different on 16FF+ for example.

From the leaked info, its 3.2Ghz singleturbo boost, 3.05Ghz all cores tubo and 2.8Ghz base for both the quad and octocore. 65W and 95W respectively.
 
someone should take their broadwell chip and lock it to 3ghz and get us some info on single thread performance.

passmark show the intel broadwell E 6900k at 2182 single thread but I bet that is with it turboing a single core to 4ghz.

core i5 5575r is in the 1850 range but again that is a 2.8/3.3ghz chip.

So I guess ZEN should be some place around 1850 +/- 100

That might also put multi threading in the 17000 range with SMT. 6900k is 17544.
 
Last edited:
gamerk2 said:
There's no way Zen is even going to approach 4GHz. The process node AMD chose (or was forced to choose via the WSA) is optimized for < 3GHz clockspeeds, and that partially explains the ES sample at 2.8, and why AMD chose to underclock an Intel chip, rather then OC their own, when doing a head to head comparison. This also explains why Zen got pushed at least one Quarter: AMD is hoping yields improve to the point they can do a clock bump.

I'm predicting a base 3.2 GHz base clock, and a chip that doesn't OC well.
Click to expand...

What you said right there is why I've always felt that AMD spinning off their fabs was a bad idea. They will always be at the mercy of a service provider rather than "making it happen" themselves.
 
DejaWiz said:
Yes, ZEN will eventually need 4GHz because Intel sure as hell isn't going to stay static with their core clocks over the next few years.

ZEN better be able to hit 4GHz (or higher) as a stock-clock out of the box at some point in the future in order to keep up with Intel.
Click to expand...

thats the thing. Alot of people here assume amd will never be able to clock it above 3.2 or 3. I call that simply bullshit. As the process improves they will eventually clock it higher as the time goes on. its kind of nonsese for anyone to claim that they will forever be stuck at 3ghz range.

I certainly expect a little higher on the quad cores though. we will see. I think eventual goal will be to shoot for higher clock speeds with new revisions. I am sure they have a 2-3 year plan in place based off zen.
 
Last edited:
NKD said:
thats the thing. Alot of people here assume amd will never be able to clock it above 3.2 or 3. I call that simply bullshit. As the process improves they will eventually clock it higher as the time goes on. its kind of nonsese for anyone to claim that they will forever be stuck at 3ghz range.

I certainly expect a little higher on the quad cores though. we will see. I think eventual goal will be to shoot for higher clock speeds with new revisions. I am sure they have a 2-3 year plan in place based off zen.
Click to expand...

I'm sure they will get to those kinds of clock speeds as well, but the other part of my concern is going to be critical to business/enterprise: power draw and heat output.

If AMD can nail that down with ZEN, then they will have a high-volume segment winner.
 
DejaWiz said:
I'm sure they will get to those kinds of clock speeds as well, but the other part of my concern is going to be critical to business/enterprise: power draw and heat output.

If AMD can nail that down with ZEN, then they will have a high-volume segment winner.
Click to expand...

Ofcourse. I think the main part I am concerned with is performance over excavator. if they can get ipc with 20% of intel I think thats a win looking at how far behind they are. Then only thing left is doing everything they can to to get to the higher clock mark and eventually 4ghz range.
 
Shintai said:
They would have to use Zen first. And performance? We have to see. Because its far from top performance from what is leaked.
Hewlett-Packard 500-c60 - Geekbench BrowserA 2Ghz Jaguar is around the same as a 1.45Ghz Zen in GB4 in ST.ARM to x86 compares are off. But its not exactly stellar performance to beat.
And do I have to mention the inefficient GPU? ;)
Click to expand...

Which would not matter to consoles at all since consoles use a different API which does not revolve around single thread performance. Zen gets even with Broadwell-E in Blender
 
Pieter3dnow said:
Which would not matter to consoles at all since consoles use a different API which does not revolve around single thread performance. Zen gets even with Broadwell-E in Blender
Click to expand...

Consoles depends very much on ST performance. Just because its a console doesn't mean everything that is a challenge for all coding just vanished with the wink of a magic wand. Scaling past 4 cores on the consoles also include quite a penalty due to the ~200 cycles latency between the clusters.

You better pray for a miracle that the Blender test wasn't a canned benchmark, specially after 2 other benchmarks shows it was. Not to mention the new stocks ;)

And AMDs history of lies? Do I even have to mention it?
 
Last edited:
Shintai said:
You better pray for a miracle that the Blender test wasn't a canned benchmark, specially after 2 other benchmarks shows it was. Not to mention the new stocks ;)
Click to expand...

I downloaded Blender, it's not a huge program - 100MB or so IIRC. And then downloaded the "BMW Benchmark" file from the blender site. I haven't tried the "Blenchmark" benchmark - it doesn't appear to be an official Blender release.

I just Open the BMW file and click "Render". It does run all 8C/16T up to 100% for most of the render, backing off to nearly nothing for the last few seconds.

Anyway, running the BMW file is fine and dandy, but there are a shitload of settings it looks like you could tweak to skew your results. Without altering anything, it takes about 2:00 minutes (anywhere from 1:59 to 2:01) to render the BMW file with a 5960X @ 4.6GHz. But there are guys on the Blender forum doing it in half that time (with much less powerful rigs) by changing tiles and stuff that I don't really understand. So no way it could ever be a benchmark, just too many user changeable options that can really alter the time to render a given file.

That, and the fact that the desktops don't appear to be the same on the AMD and Intel rigs in the vid make me question whether the runs were really 100% identical.

I watched the vid again, and didn't notice the differences in desktops I thought I had noticed before, so I retract that part of my statement.
 
Last edited:
You must log in or register to reply here.
Back
Top