3D Mark 03 out!!!!

Next time new drivers from NV and ATI comes out, witness "dramatic overall 3D-rendering performance increase" (illustrated by 500+ 3dmarks points by optimizing specificly for this particular "benchmark")... That's my prediction... ;)

*G*
 
Got it. 8)

so fresh windows install that my Seismic Edge needs drivers before 3DMark install..

results will follow.
 
ACK! Holy Slide-Show, bat-man!

3dmarkscore.jpg


PIII-700 Mhz, 192 MB Ram, Radeon 8500 64MB, Win XP.

In game 2 and game 3, (and most of the feature tests), was UNDER 5 FPS a majority of the time. I may have one other criticism of 3D Mark'03...too many of the tests are too intensive. ;)

This is billed as a DX8 and DX9 benchmark, and I certainly expect some tests to be such poor performers on my rig...but not virtually all of them.

I was expecting at least one of the DX8 game tests to run about 30 FPS on my system. Oh well, now we know who paid-off Future Mark...BOTH ATI and Nvidia. ;) Talk about feeling of inadequacy and wanting to upgrade. :oops:

I realize that this test needs to look to the future and what cards will be tested with it over the next 2 years perhaps....but man, It just surprises me that one of the better DX8 cards does so poorly in an absolute frame rate sense for both DX8 tests:

Here's Rage3D's scores with a Radeon 8500:

http://www.rage3d.com/reviews/Other/3dmark03/?page=4

6 FPS on a Radeon 8500 with a PIV 2.4 Ghz. It just seems unreasonably low. Again, 1 DX8 test I could see, I think both DX tests with that behavior is too drastic.
 
1004 3D Marks

Athlon XP 2000+, 512MB DDR333, ASUS A7V8X
RadeOn 8500 64MB

It's bloody slow, hopefully, John Carmack will write a much better engine that will run at more than 4fps...
 
3Dmark03 is really more a DX9 test.

It uses DX8 capabilities, but that makes sense as a lot of DX9 titles will still be using Dx8 shaders in games to come simply for a larger userbase.

From that standpoint i think its a pretty good benchmark... It won't ever reflect direct gaming performance, but the aggregate performance from hundreds of thousands of cards will give a good general hierarchy. And it give the benchmark a lot of headroom for benching cards over the next couple years. The Shader limits in the game tests is one of the few ways to help make the bench more card-limited vs. system limited.
 
worm[Futuremark said:
]You all wanted a more GPU/VPU intensive benchmark, didn't you? ;)
Click to expand...

I haven't seen it yet, probably won't till next week as every server is lined up like a Mcdonalds drive through.
 
Joe DeFuria said:
3D Mark 2001 is a DX7 / DX8 test. TNT-2 is DX6.

3D Mark 2003 is a DX8 / DX9 test. Radeon 8500 is DX8.
Click to expand...

Hmm...I think the difference between DX 7/DX 6 cards capability to handle the 3DMark 2001 workload is markedly less than that of DX9/DX8 cards.
With the dramatic increases of shader functionality/speed, the results seem reasonable to me.

I do tend to agree with the 64MB/128MB issue mentioned before though...but only think that should apply to synthetic tests (and doesn't it already?)...I think it is a reasonable barometer of the future to have 64MB card performance suffer due to insufficient memory storage (it was mainstream DX 7 games and ones with minimal DX 8 functionality that ran well with 64MB AFAIK).
 
Absolutely zero of the mirrors work for me. Boy are they getting hammered....

I just bit the bullet and waiting to download off FilePlanet :?
 
You all wanted a more GPU/VPU intensive benchmark, didn't you?
Click to expand...

Actually, to be precise, we wanted a benchmark that "scales" with GPU/VPU moreso than CPUs. ;) (Yes, I know the two concepts are related! ) Note that scaling well doesn't mean by defintion, must runs at 5 FPS on one card (8500), and at 15 FPS on another (9700). it could be 20 FPS and 60 FPS.

I am just stating my preference that ONE of the DX8 tests should not be such a poor performer from an absolute frame rate perspective.
 
You must log in or register to reply here.
Back
Top