Futuremark 3DMark 2011 VGA & CPU scaling review

[matose]

Now the NDA of the new 3D Mark is over and I'm sure the results will keep flowing !

But how does the new benchmark scale with different VGA's and CPU's? What's the CPU influence now on the final result? Is the CPU Test from 3DMark 2011 more stressing than the one in Vantage? These are questions that Monstru tried to find an answer to...

Futuremark 3DMark 2011 | lab501 | Translate



Thanks for reading,
matose

[eva2000]

Thanks for the hard work.. saves us doing it

[duploxxx]

conclusion, only really written for 4 cores with 6 cores the real limit or HT has no added value, should try again with 2 cores HT on to see the real result..

[Particle]

Four cores is definitely the end of the large gains for this app. Nice to see it continue to benefit to a small degree beyond that with six-core systems.

[sofos1990]

Oh that's nice man !

Thanks for sharing it with us

Can you check what's the score with different ram timings/freqs., uncore and OSs ?

[Sn0wm@n]

2011 and only 4 cores gives big gains ... lol wat????

[Calmatory]

conclusion, only really written for 4 cores with 6 cores the real limit or HT has no added value, should try again with 2 cores HT on to see the real result..

First read about Amdahl's law and then try to spit out such nonsense please.

[eXceeded]

First read about Amdahl's law and then try to spit out such nonsense please.

Cut your hair you hippy!

Interesting results, look forward to actually being able to run this benchmark!

[Gener_AL (UK)]

Quite a few variables still to consider before spouting things like "its only using 4 cores" / first is use of crossfire and SLI / Then there is type of CPU and different speeds memory/uncore/NB.

Looking at those results I see GPU saturation more then I do CPU limitation.

[Monstru]

When Nvidia will provide a proper SLI working driver for 2k11, SLI test will come too, including CPU scaling. But even if you look at the score formula you can see that the CPU counts less this time.

[safan80]

How well does it scale at 2560x1600?

[M.Beier]

How well does it scale at 2560x1600?

Now why would that be interesting?

E; 600p
P; 720p
X; 1080p

Why wasting time on using none comparable settings?

[Hornet331]

conclusion, only really written for 4 cores with 6 cores the real limit or HT has no added value, should try again with 2 cores HT on to see the real result..

If it would have been written for only 4 threads, you wouldn't have seen any scaling at all past 4 threads.

[hersounds]

four cores
http://3dmark.com/3dm11/22477

vs

(oldie) two cores

http://3dmark.com/3dm11/42266

anyway,not same vga's ...but i still love E8600

[dinos22]

thanks very much for the article, it's very informative

interesting results there in terms of CPU scaling and cores as well as the fact AMD is not getting pretty close to Intel. AMD can resume the MHz race with 4 cores again and still be reasonably competitive next year from the looks of it

would have been nice to see MHz on both Intel and AMD going to say 5.5Ghz with multi card setups but i do undersatnd that cant be done with nvidia hasnt done their job and made drivers for it yet

what we need from intel now is a cheap 4core K SKU for X58 chipset

[saaya]

awesome! i actually like this!
why? its realistic... if you want to predict gaming perf of a system this is spot on... this is exactly how most games scale with additional cores and i agree with fms speculation that this wont change within the next 2 years.

the cpu mhz scaling graph is confusing me though... barely any scaling above 2400mhz?
correct me if im wrong, but the latest games really scale well with cpu clocks above 3g, dont they?
i remember reading that sc2 even scales past 4g... i dont know how respresentative that is, but... im surprised, especially with the physics focus and all, that there is no scaling past 2.4ghz?

wow, has anybody else noticed how a 1090T outperforms a 980X in 2k11 at X?
how?

pardon my ignorance, i havent run 2k11 yet, but whats the difference between GT1 and GT2, and GT3 and GT4?
they look identical on the screenshots

i really like that X includes 4aa, but im disappointed that P doesnt have any aa at all...
if it were up to me i think id have made X 8aa and P 4aa actually... cause i mean SERIOUSLY, we are a decade into the 21st century already, i cant believe people actually play games without aa at all.... and i cant believe ati and nvidia arent spending more efforts and transistors on optimizing aa and image quality and performance...
thats why i think all games and benchmarks should be run with top of the line image quality settings, at least for highend hardware... cause thats the only way ati and nvidia feel the burn to improve AA quality and performance...

it barely scales with higher memory clocks on the vga?
that makes sense... i guess...
the problem isnt feeding the gpu cores but having too few gpu cores and them gnawing on the work like an old dog instead of finishing it quickly

hey monstru or matose, could you test in detail how the physics tests scale?
im curious... it seems they run on the gpu only and dont use the cpu at all?
but i thought they were opencl and should work on both cpus and gpus?

[Tomasis]

saaya, if it adds 8x aa, it would be like 1 fps

just now it makes 15fps on my 6850

[JkS]

.... and i cant believe ati and nvidia arent spending more efforts and transistors on optimizing aa and image quality and performance...

Didn't AMD introduce morphological AA recently?
I've heard rumours that they are introducing another more advance AA for Cayman.

[Gener_AL (UK)]

After a Full read, all I can say is great review.

[zalbard]

Great article, thanks!

When Nvidia will provide a proper SLI working driver for 2k11, SLI test will come too, including CPU scaling. But even if you look at the score formula you can see that the CPU counts less this time.

Looking forward to it! Will we have an update for Cayman, too?

[Monstru]

Yes, once Caymans are out, I will update the charts with 6950, 6970 and GTX 570.

[PatRaceTin]

GPU more effect than CPU

[RPGWiZaRD]

Pretty good that the difference between CPU performance influence is huge in Xtreme vs Entry & Performance so we can bench for both GPU and CPU depending on test used but Vantage already looked pretty much the same in this regard but yea that Xtreme preset is really taxing for today's GPUs. ^^ Doesn't look like CPU performance influence in xtreme preset will change a lot over the next 2 GPU generations at least.

Thx for the investigation.

[Monstru]

Thank you guys, my graph is complete now....but y'all just gonna have to wait a bit more for the full list

[saaya]

nice, nice