VOGONS

3dmark2001se: athlon mp vs p4

Topic actions

First post, by ratfink

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

Here's a result that puzzled me.

System 1 [now defunct]:

dual athlon 1600mp
wildcat vp990
windows 2000
santa cruz
2gb ecc ram

with vertical sync on: 3895
vertical sync off: 6793

System 2:

p4 2.8ghz with HT
wildcat vp870
windows 2000
m-audio card
256mb non-ecc ram

with vertical sync on: 3916
vertical sync off: 7356

The VP990 is slightly slower as I recall from my own testing, than the 870.
Non-ecc ram should be faster - they are both ddr266 iirc.
I assume sound cards are irrelevant to the test.
3dmark2001se is not multi-threaded I assume, but the view seems to be that it's cpu-dependent.

I was expecting a 2.8ghz p4 to be a lot faster than an athlon 1600mp, but the improvement is less than 10%.

The athlon might be running background processes on one cpu, and the benchmark on the other, but that seems unlikely to explain this. Insufficient ram seems a bit unlikely, wouldn't 3dmark crash?

edit:
turning HT off in the bios gives me 7810, which is better though still only an overall change of 6800->7800 for a near-doubling of cpu rated speed

Or is it just that for these cpu/gpu ranges, the gpu is having proportionately more effect - so maybe 60-70% of the score is due to the gpu [say 5000 in this case] and the remainder is down to the cpu [changing from 1800 to 2800, which is more believable]

Reply 1 of 3, by obobskivich

Posted on 
User metadata
Rank l33t
Rank
l33t

The Wildcats do not score very well in 3DMark to begin with (especially 01 IME), so that's probably part of the issue. 3D01 is *very* CPU dependent as well (example: 5900XT w/P4 2.0GHz scores around 7000, with 3.2GHz Extreme scores around 17,000; both have 1GB of RAM, FX 5900XT @ stock, same HDD, etc). On modern systems 3D01 is more of a test of CPU performance than anything else. The AthlonXP series tends to do very well in 3D01 as well, a 2.8GHz P4 isn't going to be a lot better off as a result. You should also have more RAM in the P4 system for this kind of testing (or pull RAM out of the MP system); to make it more fair. I'm not sure if it will or won't make a substantial difference though (I forget if 256MB is a limiter for 3D01; I've never known it to improve (much) with 1GB->2GB but I don't remember where the "break" is).

Per 3DLabs the VP990 is faster on the core than the 870, but not by much.

Reply 2 of 3, by ratfink

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

Ok thanks. I've tried it with 768mb and get 7860 compared to 7810, so probably no real change. Unfortunately I can't find any more 512mb or greater DDR and this board only takes 2 sticks.

Looking back at my old posts I see I am repeating things I've done already though: the following thread actually relates to the same board and gets the same results, I just never compared to the dual athlon before. Must be getting old...

Wildcat VP 3dmark2001SE results - some oddities

That other thread by the way shows my results where a vp990 was actually slower than a vp870, and for some test slower than a vp760.

I'm going to get my slippers and have a kip under the newspaper till nurse comes...

Reply 3 of 3, by obobskivich

Posted on 
User metadata
Rank l33t
Rank
l33t

That's weird that the 990 scores lower; according to 3DL it has higher throughput. For example see here:
http://www.3dlabs.com/legacy/Datasheets/wcvp990.pdf where it shows 225M Vertices/S

But then compared to something lower down (price-wise) in the model range, like the 760:
http://www.3dlabs.com/legacy/Datasheets/wcvp760.pdf which only shows 165M Vertices/S

I only have an 880Pro though, so I don't know how it stacks against other 3DL cards.

Go to top of page Go to top of page
Back to General Old Hardware
Close window
Close window