3dMark05, Should it Look Good?

G

gamesgeek

n00b
Joined
Mar 26, 2005
Messages
17
Last night I got a new machine up-and-running w/ a 3200+ AMD 64 and a GeForce 6800 GT PCI-E 256MB card.

I downloaded 3DMark05 just out of curiousity, and ran it. It seemed like it was running really slow, like ~20FPS, and eventually was slowing down in parts to < 5FPS. I got worried that maybe my GPU was overheating so I killed the Benchmark early, but it gave me a "score" of 4500+.

So I'm curious, is that the expected behavior of 3DMark? Is the score just a false reading sinc eI terminated the Benchmark early?
 
sounds about right, though you should finish the benchmark anyway. if the card overheats when benching it, you've got serious airflow problems to deal with ;)
 
He stopped it id way there fore lowering his score. And do the extra 4 pipes really make a difference? i only got 3981 with mine. :(
 
the card wasn't overheating, 2005 is just a hardbenchmark for most systems to run. If it was overheating you'd be seeing a whitechecker board effect all over the screen, strange colored lines and every thing would be blurring together (atleast that what happens to me when my card overheats). And the reason it got down to 5fps and lower was because it was running the cpu part of the test. By lowering the res it puts more of the stress on the cpu instead of the video card, the lowered framerates are perfectly normal.
 
Sometimes my SLI setup would drop below 30 FPS... it's just a really stressing benching test. You gotta give the cards something taxing or else it's not really a good indicator of peformance. Don't worry about it. I wasn't aware that it gave a score even though the test stops midway? I've stopped mine before and it told me not enough tests were run to give a score.
 
The benchmark tries to predict the performance of games yet to come, like games based on the UE3.0 engine.
So yes, you get horrible framerates on anything less than a 6800U SLI setup :)
3DMark has always been like that, only the most high end available at the time of release would get framerates of more than 30 fps, generally.
3DMark03 is the one that predicted the games of 'today', like Far Cry, Half-Life 2, Doom3.
And your card will probably blaze through that one with ease, exactly like it will blaze through those games with ease.
Enjoy your new system.
 
ray4389 said:
He stopped it id way there fore lowering his score. And do the extra 4 pipes really make a difference? i only got 3981 with mine. :(
Click to expand...

My system is similar to yours and I get 4657, so I think the extra pipes help a good amount. Although going from the 67.03 drivers to 71.84, my score went from 4305 to 4657.
 
Baker_God said:
And the reason it got down to 5fps and lower was because it was running the cpu part of the test. By lowering the res it puts more of the stress on the cpu instead of the video card, the lowered framerates are perfectly normal.
Click to expand...

While you are right that reason for the sub 5fps's was because it was running the CPU tests, you are wrong about how the test works. The "CPU" tests actually runs in what is traditionally called "software" mode. This means the CPU does all the work the videocard normally does while the videocard basically just passes the CPU's results on to the monitor.

If the tests simple reduced resolution then the game would run FASTER (unless it was already CPU bound, which would defeat the purpose of 3dmark05 anyway, so they intentionally design it not to be CPU bound for a long time), not slower. It would also look better than it does in the CPU test currently, but worse than in the normal test.
 
arentol said:
While you are right that reason for the sub 5fps's was because it was running the CPU tests, you are wrong about how the test works. The "CPU" tests actually runs in what is traditionally called "software" mode. This means the CPU does all the work the videocard normally does while the videocard basically just passes the CPU's results on to the monitor.
Click to expand...

I think he just means that they go to a low resolution to minimize the effect of the videocard on the CPU score. In which case he is right.
The CPU tests only use software vertex processing by the way, the rest is still done by the videocard. If you'd actually run it entirely in software, it'd take a few minutes to render a single frame of 3DMark05.

For more information on the 3DMark05 tests, see http://futuremark.com/products/3dmark05/?tests
 
3dmark05, on my 6800NU looks a little better than 3dmark03 did on my Radeon 9000. It's a killer.
 
You must log in or register to reply here.
Back
Top