3DMark 11 out on 30th November

[mk-ultra]

i think the point of not allowing physX for competitive OCing in Vantage is that it's an nvidia feature only more than anything

this time it looks like both vendors will benefit from it so idk what to think.
i'm too for going forward but on the other hand, zanabar is right, if enabled it's not a CPU test anymore it's a CPU+gpu test, so cpu subscores won't be comparable unless you have the same GPU too so it makes comparison nearly impossible which is the main point about a benchmark.

[FM_Jarnis]

Many of your questions are answered in the product pages.

http://preorder.3dmark.com/benchmarks/

There are actually *six* tests. Four graphics tests that use two sets of assets from deep sea and high temple demo scenes, each is balanced differently to stress different GPU features of DX11. Then there are two additional tests - a CPU physics test ("the test with the crappy graphics") and a combined test that has it all - Rigid body physics simulation on a moderate number of objects (CPU), Soft body physics using DirectCompute and the Bullet physics library running on the GPU and tessellation & lighting effects - like you would have in a game running on DX11 and using all these at once.

Graphics Test 4 and the combined test are expected to push your hardware the most and all tests in combination are designed to poke each aspect to get a fair 3DMark score that isn't biased based on one specific DX11 feature and/or bottleneck. Yes, the team has done their homework and found some interesting things about DX11 along the way...

And of course in default run you first get two demo scenes with audio that are designed to just look damn cool and run well, showing off your hardware. They don't affect the score and can be disabled if all you want is to bench and get numbers.

[STEvil]

the cpu test is for the cpu, when u run it on the gpu its a cheat and not showing what its intended to or atleast that was the rules with the last one. if its a cpu test it should test the cpu though as when u let a gpu go unloaded and do some kind of cuda or OCL to the whole gpu its not a representative test of what it would do while gaming. now for a feature test or not by default thats fine but making it do it by default just makes it so people cheat inadvertently by not knowing better.

Running the vantage test on GPU with intent to cheat is a cheat, yes. Running it on GPU with intent to compare to other GPU's is fine. Given the limited testing ability of vantage its a little pointless for sure, though.

[FM_Jarnis]

Running the vantage test on GPU with intent to cheat is a cheat, yes. Running it on GPU with intent to compare to other GPU's is fine. Given the limited testing ability of vantage its a little pointless for sure, though.

Correct. You can run Vantage with GPU PhysX, but at that point the score is usable only for comparing against other systems using GPU PhysX.

To get a score that is comparable to any other DX10 hardware, disable PPU is the box to tick (or disable GPU PhysX in NVIDIA drivers).

And yes, 3DMark 11 uses only DX11 API functions (including DirectCompute) so none of this is an issue with it.

[mAJORD]

Actually looks reasonable,

Hope this is a "turning point" for 3Dmark, back to its roots of providing a glimpse into upcoming gaming visuals, not already dated looking crap like 06 and Vantage were

*remembers the good ol days of 99 - 03*

[trinibwoy]

Yes, the team has done their homework and found some interesting things about DX11 along the way...

I didn't see any mention of it on the product page but have you guys included the targeted feature tests for texturing, shading etc like you've done in the past or is it just the game and physics tests now?

[Razrback16]

I miss the days when it was free.

[Johnny87au]

this looks promising, cant wait to try this!

[FM_Jarnis]

I miss the days when it was free.

?

3DMark 11 has free Basic Edition with no "trial key" limitations or anything like that. It'll be pretty much as 3DMark 06 was.

[saaya]

sounds good, but even if the numbers are useful to analyze the gaming performance of a system, its too bad that the benchmark gets those numbers running game atypical animations...
since vantage 3dmark looks more like a 3dsmax or rendermonkey benchmark than a GAME benchmark... everybody loves 2k1 cause it emulates common games and game scenarios... fps, racing, action/mmorpg... its a shame that newer versions dont do this anymore... thats what made 3dmark and madonion/fm famous and successful...

[Manicdan]

sounds good, but even if the numbers are useful to analyze the gaming performance of a system, its too bad that the benchmark gets those numbers running game atypical animations...
since vantage 3dmark looks more like a 3dsmax or rendermonkey benchmark than a GAME benchmark... everybody loves 2k1 cause it emulates common games and game scenarios... fps, racing, action/mmorpg... its a shame that newer versions dont do this anymore... thats what made 3dmark and madonion/fm famous and successful...

i agree the transition has moved from ingame action to cutscenes, there was some action in vantage, but not like 01/03. i dont believe there is anything actually wrong with doing a scene rather than a cut of a ingame timedemo. what i think is wrong is how much we dont really need 60fps with a slow moving panning render. if you get 10fps you can kinda see whats going on just fine in 06 or vantage, while in 01/03 you would be lost. and with more options for running the game, the need for 60fps is easily available by running a lower setting. so if you multi expensive gpu solutions, run on extreme. or if your a midrange guy, run entry, since in real life you would want your 60fps, these "games" in 3dmark make it look good enough because it appears to be smooth.

[zalbard]

Free edition sounds great!
These [paid versions] should also be bundled with MSI cards, apparently, so even better!

[cold2010]

3DMark 11 settings: eXtreme

http://img201.imageshack.us/img201/1423/36045208.jpg

http://img41.imageshack.us/img41/3050/21107786.jpg

http://img541.imageshack.us/img541/7125/48494198.jpg

http://img44.imageshack.us/img44/2641/40390408.jpg

http://img840.imageshack.us/img840/8040/31478559.jpg

http://img535.imageshack.us/img535/7703/42346701.jpg

http://img140.imageshack.us/img140/742/27314663.jpg

http://img403.imageshack.us/img403/8498/12691303.jpg

http://img228.imageshack.us/img228/2351/95388818.jpg

http://img3.imageshack.us/img3/601/79676003.jpg

http://img14.imageshack.us/img14/7974/86555172.jpg

More Photos

Source

http://www.chiphell.com/thread-141186-1-1.html

[kaktus1907]

thank you cold2010

Test platform

Processor
Intel Core i7-950 3.06GHz (133x23)

Radiator
Thermalright HR-02

Motherboard
ASUS P6X58D Premium X58

Memory
Corsair TR3X6G1600C8D DDR3 1600 2Gx3 (8-8-8-24) 1.65v

Hard drive
Intel X25-M G2 160G SSD

Power
Seasonic X-750

Monitor
DELL UltraSharp 3008WFP 30 "

Chassis
SilverStone RV02-E

System
Microsoft Windows 7 64bit

Driver
ATi Catalyst 10.10
NVIDIA 262.99

Patch
DirectX End-User Runtimes (2010.6)

3DMark 11 eXtreme mode

HD6870___2302

HD6850___2269

GTX 460 1GB___1772



http://www.chiphell.com/thread-141210-1-1.html

[Manicdan]

3DMark 11 eXtreme mode

HD6870___2302

HD6850___2269

GTX 460 1GB___1772



http://www.chiphell.com/thread-141210-1-1.html

something about the 6870 and 6850 being that close in score makes me wonder something is wrong. the specs between the 2 are no where near just 1.5% different, so why are the scores only 1.5% different. unless its cpu bound, but then why is a 460 30% behind

[Iconyu]

something about the 6870 and 6850 being that close in score makes me wonder something is wrong. the specs between the 2 are no where near just 1.5% different, so why are the scores only 1.5% different. unless its cpu bound, but then why is a 460 30% behind

I would have said the only spec they both share that closely would be tessellation, but that wouldn't explain the 460 performance unless Fermi really does dislike combined loads. So second guess would be drivers, I expect both sides will have new drivers released soon to show their cards in the best light.

[trinibwoy]

Nice it leaked earlier than I thought it would.

[cold2010]

3DMark 11 RC

GTX580：X1748
GTX 460：X923
HD5830：X845
HD5870：X1234
HD5870CF：X2350
HD6850：X899
HD6870：X1120

[Manicdan]

ok so the numbers get more confusing

extreme mode on both attempts shows a range of 100% difference. so i guess we can say that we have no idea what the scores will be. or the first run wasnt really extreme, but performance.

edit, or the previous numbers were gpu score, and this is total score

[FM_Jarnis]

I would take all of this with a huge mountain of salt. Just saying...

[Syn.]

I would take all of this with a huge mountain of salt. Just saying...

LOL. Funny how "pinch of salt" has lost all meaning. Its a pinch of salt because it is very tinny amount, that way the saying works. Saying a huge mountain of salt actually means that you should believe what ever it is said because it has a lot of proof behind it.

BOT: Vantage was a hit and miss benchmark for me. They got few things wrong. 11 looks ok so far. Will decide when i give it a run.

[FM_Jarnis]

Figure of speech fail.

My point is: There is no final build yet, so even if those scores are from the benchmark, they may or may not have anything to do with the final numbers... and obviously any builds that exist at this point are under NDA - not that it seems to bother chiphell with, well, anyone

[trinibwoy]

Figure of speech fail.

Nah you got it right. The less reliable/tasty the rumour/food the more salt is required.

[sniper_sung]

Figure of speech fail.

My point is: There is no final build yet, so even if those scores are from the benchmark, they may or may not have anything to do with the final numbers... and obviously any builds that exist at this point are under NDA - not that it seems to bother chiphell with, well, anyone

The owner of chiphell claims that these results are from RC version (instead of beta version in one of his previous post), and should reflect close to RTM results.

[saaya]

i agree the transition has moved from ingame action to cutscenes, there was some action in vantage, but not like 01/03. i dont believe there is anything actually wrong with doing a scene rather than a cut of a ingame timedemo. what i think is wrong is how much we dont really need 60fps with a slow moving panning render. if you get 10fps you can kinda see whats going on just fine in 06 or vantage, while in 01/03 you would be lost. and with more options for running the game, the need for 60fps is easily available by running a lower setting. so if you multi expensive gpu solutions, run on extreme. or if your a midrange guy, run entry, since in real life you would want your 60fps, these "games" in 3dmark make it look good enough because it appears to be smooth.

very good point! in earlier benches high fps really made a difference, in the newer benchmarks from fm thats not the case as the camera is moving slowly... thats actually something that even 2k1 doesnt do very well, everything is 3rd person... a real game benchmark HAS to have some first person view tests imo... i mean just compare the most popular games and check how many are first person and how many third person... yet all "game benchmarks" from fm are third person... i think fm should focus more on really simulating and measuring game performance than building their investor/manufacturer relations like theyve been doing in the past years :/