AMD 8-core ZEN Packs a Whallop with Multithreaded Performance

[silentbogo]

Lol. Not this again....

This is "AMD Engineering Sample". We don't even know how many cores or even CPUs was in this hypothetical test machine. Heck, it could've been a new 16-core opteron, or that dual-socket server board with, and I quote "up to 32 cores"...

http://www.anandtech.com/show/10581...-naples-32cores-dual-socket-platforms-q2-2017

It's too early to cheer, guys. Kill the hype before hype kills you.

 

[Chaitanya]

http://www.guru3d.com/news-story/new-amd-engineering-sample-zen-processors-get-higher-clocks.html

Right now, the 8 core zen is suposed to be clocked at 3.3Ghz all core turbo, and 3,6 single core. I' m going to speculate and say that in multuthread it may be as fast/slightly, slightly slower than a :

Intel® Core™ i7-5960X Processor Extreme Edition
(20M Cache, up to 3.50 GHz)

It may not be the intel annihilator that we wanted it to be, but that's not bad. excavator couldn't even touch a 6 core i7.

That might be lower clocked to keep the TDP target at 95W while Intel is a 140W chip.

 

[dyonoctis]

Lol. Not this again....

This is "AMD Engineering Sample". We don't even know how many cores or even CPUs was in this hypothetical test machine. Heck, it could've been a new 16-core opteron, or that dual-socket server board with, and I quote "up to 32 cores"...

http://www.anandtech.com/show/10581...-naples-32cores-dual-socket-platforms-q2-2017

It's too early to cheer, guys. Kill the hype before hype kills you.

If this is a 16 core/32 thread zen cpu, then zen is scaling like dead horse shit. In another blender test a 3Ghz 8 core zen was sligthly faster than a i7 6900k (8 cores) clocked at 3Ghz. Now you are telling me there is a chance that a 16 zen core is barely as good than a 10 core xeon. That's twice the cores count for amd, and only 2 more cores for intel. Unless this 16 cores is clocked at 1Ghz, I can't see how the scalling can be that bad.

I'm not expecting Amd to flat out beat intel in every range, but from the benchmarck that amd showed us right now, it's not going to be buldozer 2.0 were an "8 cores" was only as good as an intel 4 cores.

 

[xorbe]

No core count, no clock rate, it's not even worth postulating.

 

[alucasa]

I wonder... why Blender? It scales well to core count.

Is AMD indirectly saying their single thread performance is like dirt?

 

[cdawall]

I wonder... why Blender? It scales well to core count.

Is AMD indirectly saying their single thread performance is like dirt?

Who knows maybe because they could. He'll amd could be trolling people and this is a 16c at 1.4ghz. No one knows.

 

[Nosada]

I play Blender more than any other games. You can spend hours with Blender while accomplishing absolutely nothing.

That DOES sound like how I play most games.

 

[Fierce Guppy]

It out performs my i7-5960X. Nice...

 

[notb]

Maybe so. But this Zen has 2 cores less. So i'd say they nailed it with multithread. I'm more concerned about that single core performance.

But what's the point of such a comparison?
If a 10-core CPU is better in multicore tasks than an 8-core chip, it is simply faster. There is no point in looking at average core performance., because there isn't that much software that couln't use the extra 2 cores (and if it exists, that's not what we call proper programming...).

 

[TheinsanegamerN]

I wonder... why Blender? It scales well to core count.

Is AMD indirectly saying their single thread performance is like dirt?

Wouldnt this be saying the OPPOSITE?

In order for a 8 core zen to outperform a 10 core xeon at multithreaded benchmarks, assuming the core clock is similar, the zen cores would need to be more powerful then the i cores at the same speed, which would indicate better single threaded performance.

 

[Evo85]

This thread needs more information .

And popcorn. . .

 

[64K]

This thread needs more information .

And popcorn. . .

Still awaiting benchmarks from respectable sites like this one.

 

[alucasa]

/Objective mode on
Just so everyone knows, the original source of this news is from http://blenchmark.com/

It is an addon benchmark tool for Blender. And the result was posted on August 25th of 2016 on Windows Server R2 OS. As Blender isn't that popular among
pc enthusiast, no PC pc enthusiast sites noticed this until yesterday.

No other information is available. No core count. No core frequency. No socket count.
/Object mode off

/Subjective mode on

The chip was likely a server version of Zen arch. Probably 8 or 16 core Zen. Since Blender is one of those applications where large number of low frequency cores rape smaller number of high frequency cores, it could be 16 core 2.0Ghz or 8 core 3.0Ghz

/Subjective mode off

The conclusion? We don't know anything at this point.

 

[cdawall]

I am just letting you all know I am going full troll mode and we are going to see how much attention the clowns on WCCF pay. Time to see what my AMD ES chips can do.

 

[SaltyFish]

This is an advertisement.

Blender is an awesome open source 3D all-in-one suite. Learn it, use it. No need to spend over a grand for other 3D applications.

Granted, it's not accepted as the industry standard, but for hobbyists like me, Blender is Godsend.

I play Blender more than any other games. You can spend hours with Blender while accomplishing absolutely nothing.

It's great for making 3D models and the likes. Now if only they can do something about their Game Engine...

On Zen's alleged performance, it's all hype until it's released or some respectable entity tests it.

 

[TheGuruStud]

I feel as though you could make some nerd reality TV show over this kind of drama...

It happens weekly lol

 

[atomicus]

I'm going to the bank tmrw to remortgage my house, then to the bookmakers and putting the entire amount on the odds of AMD's share price falling the day reviews for ZEN come out. Hype train is starting folks, jump on-board!!

 

[alucasa]

I'm going to the bank tmrw to remortgage my house, then to the bookmakers and putting the entire amount on the odds of AMD's share price falling the day reviews for ZEN come out. Hype train is starting folks, jump on-board!!

And your avatar is perfect for your post.

 

[DeathtoGnomes]

This thread needs more information .

And popcorn. . .

I totally agree! Pass the popcorn!

and..
Incomplete analysis.

 

[nemesis.ie]

2nd paragraph: "the "ZEN" sample cruches the Blender benchmark"

crunches?
crushes?

Several other things need a check over too, e.g. "whallop" instead of wallop. Looks like it was typed up in haste.

 

[alucasa]

Several other things need a check over too, e.g. "whallop" instead of wallop. Looks like it was typed up in haste.

Probably with shaking hands.

"Have got to get this out ASAP....!"

He probably came at the moment he published the article, thinking "I've done it. I've burnt it all; I've become the white ash."

 

[yotano211]

This thread needs more information .

And porn. . .

*fixed

 

[Nuckles56]

Yawn, wake me up when we know that it isn't the 16 or 32 core version being "benchmarked" and there are actual reviews for Zen rather than this BS

EDIT: And I would be selling my stock in AMD now when prices are crazy, this hype will make Zen flop badly if it doesn't perform as well as broadwell at least

 

[Bayonet]

The Xeon E5-2650 V2 is incorrectly quoted in this article as having 10 cores, when it actually has 8 (http://ark.intel.com/products/75269/Intel-Xeon-Processor-E5-2650-v2-20M-Cache-2_60-GHz).
Perhaps the author was confusing it with the V3 variant, or the E5-2660 V2, both of which do have 10 cores?

The Xeon E5-2680 V2 (which performs better than the Zen ES in this benchmark) has 10 cores, so Zen seems to be between an 8 and 10 core Ivy Bridge-EP Xeon, which in multi-thread aware benchmarks (like this one) is where a 6 core i7-6850K would place. This is probably not the conclusion everyone else in this thread was getting by going off that initial mistake in the article.

Edit: This benchmark seems a bit off. The must be some other variable at play, otherwise there is no way a Haswell-EP 2697-V3 (14 cores) could lose to an Ivy Bridge-EP 2690-V2 (10 cores) so badly.

 

[cdawall]

Edit: This benchmark seems a bit off. The must be some other variable at play, otherwise there is no way a Haswell-EP 2697-V3 (14 cores) could lose to an Ivy Bridge-EP 2690 (10 cores) so badly.

They are random user benchmarks with no word on what the user was doing when he ran it. The 2697V3 could have been running other programs at the same time, or even had cores shut off.