AMD ZEN Engineering Sample AOS - Further Performance Analysis
Click here to post a comment for AMD ZEN Engineering Sample AOS - Further Performance Analysis on our message forum
Ryu5uzaku
I find interesting the fact that through all of those the Zen keeps the same fps for the cpu pretty much. And performing best on extreme. Will be fun to see how this turns out in near future.
Really good analysis. Gives good insight in to the benchmark itself in my opinion also.
Chillin
Interesting perspective, but like you said these results are invalid since the source is questionable and is based on an early engineering sample (with unknown changes).
But it does bring back to mind the major problem, market. Who exactly is AMD targeting these for? Intel and IBM have the server market sealed up pretty tight from what I see, and the consumer desktop market is shrinking and has long ago become a commodity item, which means low margins there. I'm really curious.
DARKSF
The Zen ES is keeping consistent results on different quality settings which means there is somewhere a bottleneck that is not the CPU itself and all the crazines is pointless.Any way for me the AoS is a bad tool for comparing any CPUs it has better CPU clock scaling than number of threads.
Correct me if i'm wrong but if you put some fast DDR3 and put HT link to 2200 or higher on the FX8370 if i'm right you should see around 10-15% boost in the FPS at the same CPU clock and bigger difference between Normal and Heavy like the dive the i7 is doing which is a perfect platform that is not constrained by the system bus or the ram bus.
vbetts
Moderator
Unless it's changed and I could be wrong, I don't believe upping the HT link on an AMD system improves performance that much, if at all. But don't quote me on that. 😀
DARKSF
Well on a K10 series (Phenom II) 2000HT to 2200HT equals 10%-15% FPS boost in most game titles (World of Tanks 15% atleast before update 9.15) and believe me this means the system bus is a big bottleneck.
BiggieShady
Was Intel at 3.2 GHz with all cores loaded? I ask because I don't know if Zen ES had max turbo at 3.2 GHz for single core and much less with all cores loaded.
DeskStar
It sure would be nice to build both of my daughters their monster budget build here soon, but I am going to need the true facts of these processors to make a serious judgement. Really looking forward to some factual numbers here soon!.!
vbetts
Moderator
I had a K10 x3 435 and I never noticed anything by upping the HT link...Just caused more heat and power draw.
BoMbY
Yes, it looks like there was an FPS limiter, or something like that, running for the ZEN sample tests.
Hilbert Hagedoorn
Administrator
If there was an FPS limiter then the results would be flat out all the same across the different workloads. This does look more like the CPU plafond in combo with a bit of difference due to the GPU maxing out. Neither is is vsync as the nummer passes 60Hz.
But I hope you are right and I am not 🙂
BoMbY
It would be very interesting to see if a FRTC setting to 60 FPS would've any influence on the CPU results of the i7 - maybe you could check this if you have the system still up and running?
schmidtbag
I'd say I'm slightly disappointed by these results. And yeah, I know these should be taken with a grain of salt and there's a lot of other things we don't know (such as scheduling optimization, thread usage, and wattage) but I feel like Excavator on DDR4 (so, IIRC, what Bristol Ridge will be) would offer similar performance.
Anybody know how many threads this game actually takes advantage of?
I believe you're correct. Actually back in the day before unlocked multipliers were a common thing, people would actually lower the HT link in order to achieve higher overclocks.
Chaython
After years of saying Zen is the one that will save AMD; now I have proof that I deserve to laugh in their faces!
__hollywood|meo
good test boss. thanks for the interesting perspective & quick results.
when i got my 1090t i did benches. clocking up HT only resulted in instability with zero performance gains. the northbridge clock was what was bandwidth starved - increasing that to about 3ghz resulted in nearly linear fps gains, dropping off around ~2.8ghz. maybe hes talking about vishera chips?
anyway thats neither here nor there. im not interested in seeing how much information this zen cpu can cram through the memory bus (yet). wat i want to know is how these suckers are going to clock when they finish poking & prodding the silicon & go to retail
four.
schmidtbag
So.... you think an experimental CPU that will likely be worth roughly $300 losing to a $1000 CPU in a piece of software that isn't optimized for more than 8 threads is laughable?
IMG source:
http://www.pcworld.com/article/3039552/hardware/tested-how-many-cpu-cores-you-really-need-for-directx-12-gaming.html
Noisiv
uh...
https://abload.de/img/dx12_cpu_ashes_of_the0auzo.png
Yogi
Excuse my ignorance, but what is the CPU framerate in this bench and how does it relate to the rendered frame rate?
Denial
To be fair, isn't the problem with AMD CPU's currently that they are only good in heavily multi-threaded scenarios, which often doesn't occur in general purpose computing?
I think the initial results look promising, but AMD has to get it's single thread performance up or it's not going to stack up well against Intel's lineup. So hopefully clockspeed goes up between now and launch.
__hollywood|meo
boss said the benchmark only utilizes four threads in this very article. is the bench different from the game? i would assume so, otherwise theres been a mistake.
Reardan
The CPU isn't experimental, it's a retail part.
The Zen processor will be priced around 300 dollars, and the Intel around 1000, because of their relative performance. This Zen will be a 300ish dollar part because its performance won't command any more.
You keep editing your damn comment
