Thursday, November 3rd 2016
AMD 8-core ZEN Packs a Whallop with Multithreaded Performance
AMD's upcoming 8-core "ZEN" processors pack serious multithreaded performance muscle. The company's design focus on empowering the cores, and getting rid of the shared-resource approach to multi-core chips; appears to have paid of big dividends in multithreaded performance, as tested on the Blender benchmark. An 8-core "ZEN" engineering sample was found to be belting out performance rivaling 10-core Intel Xeon E5-2600 V2 series chips, indicating that AMD appears to have made huge gains in per-core performance over its previous generation chips.
The Blender benchmark scores of an alleged AMD ZEN "Summit Ridge" engineering sample were posted by Blender benchmark scores aggregator Blenchmark; and unearthed by this redditor. According to these scores, the "ZEN" sample cruches the Blender benchmark render in 69 seconds, the same time it takes for a 10-core Xeon E5-2650 V2 processor. The ZEN chip is also closely trailing Xeon E5-2600 V4 series chips. AMD is expected to launch its first ZEN "Summit Ridge" 8-core processors in early 2017.
Sources:
Blenchmark, WCCFTech
The Blender benchmark scores of an alleged AMD ZEN "Summit Ridge" engineering sample were posted by Blender benchmark scores aggregator Blenchmark; and unearthed by this redditor. According to these scores, the "ZEN" sample cruches the Blender benchmark render in 69 seconds, the same time it takes for a 10-core Xeon E5-2650 V2 processor. The ZEN chip is also closely trailing Xeon E5-2600 V4 series chips. AMD is expected to launch its first ZEN "Summit Ridge" 8-core processors in early 2017.
116 Comments on AMD 8-core ZEN Packs a Whallop with Multithreaded Performance
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.
Is AMD indirectly saying their single thread performance is like dirt? :p
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...).
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.
And popcorn. . .
Still awaiting benchmarks from respectable sites like this one.
Just so everyone knows, the original source of this news is from 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.
On Zen's alleged performance, it's all hype until it's released or some respectable entity tests it.
It happens weekly lol
and..
Incomplete analysis.
"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."
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
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.
Amd 8 core at ???
My assumption amd is clocked at 3.2ghz+
10 × 2.6 = 26
26 ÷ 8 = 3.25ghz
Sounds in line with expectations. amd doesnt need to beat intel in single thread they just need to match them or be close enough within 15%ipc to compete because the rest they can maneuver between clock speed amd core count.
Also for those wondering why all the benchmarks and hype is around multithreaded its because thats logically where all the ironing out part comes as far as engineering samples go. because its easy to clock one core to max but its definitely more tricky to iron out the thread management and power delivery and fine tuning for the chip as a whole, and thats what decides the total performance of the overall package and what will allow single cores to clock higher when running single thread