3dmarks 2003 discrepancy

[rc240sx]

I ran the new 3dmarks03 yesterday and got a score of 3445. I then ran the previous version and got 11235. I am running a p4 2.4 ghz with 512 mb ram, 200gig hd, audigy2, ati radeon 9500 pro and direct x 9. Did anyone else get such a low score with the new 3dmarks?

 

[chizow]

Your score is normal, the decrease is by design. Its supposed to be forward-looking test to gauge the requirements of future games. I wouldn't even bother with it. Stick to with 3DMark2K1 as it'll be a more accurate assessment of how your system performs in games.

Chiz

 

[xSauronx]

what he said, 2001 is for dx8 games, which is far more prevalent than dx9 games, which is what 2003 evaluates performance of

i got a bit over 9100 3d marks in 3dmark 2001
but only 1300+ in 2003

 

[kylebisme]

well 03 is actuly mostly dx8 but then 2k1 was mostly dx7, but in essence it is still just the advancement of technology. don't expect to be geting as high of 3dmark in 03 until your setup is as much newer than the benchmark as your current setup is when compared to 2k1.

 

[rc240sx]

Thanks guys. I was second guessing my decision for buying the 9500 rather than the 9700. But with the 9800 and 9600 debuting who knows. I feel at ease because my UT2k3 runs flawless on a 17 inch monitor crt full antialiasing, 1024x728 and 32 bit color on every level except tokara forest.

 

[aphex]

my 9700 will be here on friday, i'll post a review compared to my ti4600 asap.

 

[optimistic]

Originally posted by: aphex
my 9700 will be here on friday, i'll post a review compared to my ti4600 asap.

Here's my comparison of my PNY ti4200 128MB (280/600) and Sapphire 9500np > 9800 (360/300) with 4x FSAA enabled

 

[nemesismk2]

Originally posted by: chizow
Your score is normal, the decrease is by design. Its supposed to be forward-looking test to gauge the requirements of future games. I wouldn't even bother with it. Stick to with 3DMark2K1 as it'll be a more accurate assessment of how your system performs in games.

Chiz

3dmark 2001 isn't an accurate assessment of anything, get some video cards benchmark them with 3dmark 2001 and then with some real games. You'll find that 3dmark 2001 results are no way a reflection of actual game performance.

An example:-

My Radeon 9000 performs atleast 1000 points higher than my GF4 MX440 with 3dmark 2001, however with UT2003 my GF4 MX440 outperforms the Radeon 9000 and with most other software they are equal.

Where is the accurate assessment of performance with dx8 games?

3dmark 2003 will be just the same, there will be a large difference between 3dmark 2003 results and game performance which imo makes 3dmark 2003 results worthless.

 

[kylebisme]

nemesismk2, ut2003 is prety muhc just a dx7 game.

 

[nemesismk2]

Originally posted by: TheSnowman
nemesismk2, ut2003 is prety muhc just a dx7 game.

True but finding any game which really supports DX8 is hard, maybe in a few years time DX8 will be more supported.

 

[Keysplayr]

Originally posted by: TheSnowman
nemesismk2, ut2003 is prety muhc just a dx7 game.

It IS?!? Wow, I can't believe it. I cant wait to see what a DX9 game is capable of.

Keys

 

[chizow]

Originally posted by: nemesismk2
3dmark 2001 isn't an accurate assessment of anything, get some video cards benchmark them with 3dmark 2001 and then with some real games. You'll find that 3dmark 2001 results are no way a reflection of actual game performance.

Actually, I've found 3dmark2k1 to be very accurate in terms of scaling to GPU, CPU, platform/chipset upgrades, and FSB speeds. It does a good job of reflecting any changes in the northbridge subsystems, which is why it also makes a decent stress test. The only thing it doesn't take into account well is game optimizations for vendor specific code paths.

I've gone through a Voodoo5, GF2 GTS, GF3 Ti200, Radeon 8500 (original), GF4 Ti4200 Turbo, and a 9700pro using 3DMark2k1, and in all instances other than the 8500, my results reflected actual game performance. Same with the various clockspeeds (1GHz through 2.4GHz) as well as changes to FSB speeds and there were distinct differences as well. The only one of the cards that didn't mirror real-world results were the 8500 (at launch), but that was b/c of poor driver support in games. The score did however reflect the potential of the R200 core, which is clearly evident today (with mature drivers and better game support).

An example:-

My Radeon 9000 performs atleast 1000 points higher than my GF4 MX440 with 3dmark 2001, however with UT2003 my GF4 MX440 outperforms the Radeon 9000 and with most other software they are equal.

Probably has to do with UT2K3's dependency on fillrate and bandwidth; I'm not familiar with the 9000 (I typically don't research cards I won't buy), but I'm pretty certain the GF4's high clocks and bandwidth are greater than the 9000. 3DMark2K1 has a single DX8 test (nature) which probably accounts for the 1000 point discrepancy. Let me guess, your 9000 can run nature, while your GF4 MX can't? UT2K3 might also be better optimized for GF products, as that big nVidia "Way its meant to be played" logo probably wasn't just a sign of goodwill.

Where is the accurate assessment of performance with dx8 games?

As someone mentioned already, UT2K3 is still for the most part a DX7 engine, so its no suprise a GF2 on steroids performs well on it.

3dmark 2003 will be just the same, there will be a large difference between 3dmark 2003 results and game performance which imo makes 3dmark 2003 results worthless.

I do agree 3dmark2k3 is a worthless benchmark atm. Its mostly a DX8 benchmark that doesn't take into account system/platform, CPU, FSB changes much at all. Although I still don't agree with the scoring system it uses and its nearly exclusive dependency on the GPU while rendering DX8 and DX9 tests, there's pretty good evidence games will be moving in that direction in the future (2005 probably, heh).

I certainly don't use 3dMark2KX for anything more than a north bridge subsystem stress test and as a comparison to my own platform, but its important to some people, which is why I offered 3DMark2k1 as a better gauge of system performance in current games.

Chiz