P4 at 2GHz should be as fast, or faster than, Pentium 3 at 1200-1300 or Athlon 1400 depending on the event (regardless of internal minutiae). Sounds like something is wrong, but exactly what I couldn't say - if you have an 845 I'd say try it out, and also see if there's different driver versions available for the Voodoo2 and if they have any influence on performance.

One question that does come to mind: is the hard-disk/memory appropriately quick for a P4 and not damaged in any way?

I have an i875 SLI setup (Gigabyte GA-8IK1100 with a 2.4ghz cpu) I can bench if it helps

keropi wrote:

I have an i875 SLI setup (Gigabyte GA-8IK1100 with a 2.4ghz cpu) I can bench if it helps

You could?

Wow that would be awesome. Attached Incoming demo that I use.

There is a shortcut with flags -secondary -screenmode -nosound

-secondary just picks the secondary graphics card by default.

You might get negative average fps, so just look at the current fps and let me know what sort of numbers you get...

My Tualatin 1400 gets 116 for single card and 193 for SLI.

Nortwood 2.0 gets 113 for single and 142 for SLI.

My 2.4 gives me negative benchmark numbers, so it's hard to tell.

Other benchmarks to compare are GLQuake, Quake II, Forsaken or Unreal.

I also have a SIS chipset board, might try this...

It sounds like the SLI scaling is the biggest issue with the 2.0; the 1400 Tualatin being ever-so-slightly faster (it may not even be possible to make that claim - what's the margin of error for this benchmark?) doesn't sound all that wrong. But gaining almost nothing with SLI enabled sounds like an issue. How does it do in pure synths like 3DMark?

If you dump the Voodoo2s out of the equation and go with some entirely different card, let's use a TNT2 as a random example, put that in the P4 and try it again in the P3, is the same trend reflected or does it "scale up" with the P4?

Well that system is now off the bench, I'm trying this AOpen board with SIS chipset:

AX45F-4DN

It has bulging caps but the only "different" board I got.

Optical drive is standard IDE, storage is IDE > SATA adapter > 32GB SATA drive + floppy emulator.

The thing is, under W98 the P4 was super quick. I did a benchmark in the 3dmark 99max thread with the same Voodoo 2 SLI. It scored as expected.

So yes, something going on with SLI scaling. But in other games, such as Quake II, a single card result is also quite off:

Tualatin 1400: 126.50

Northwood 2.0: 95.60

Hopefully the SIS board sheds some light on the issue...

just did the test, negative average but worst was 90fps and best 197fps with SLI setup...

keropi wrote:

just did the test, negative average but worst was 90fps and best 197fps with SLI setup...

Hmm that sounds similar. Not enough to beat the average 200 of the Tualatin system.

A 2.4 GHz should smoke the PIII.

Mau1wurf1977 wrote:

A 2.4 GHz should smoke the PIII.

I'm thinking if it can accomplish that feat with a non-3dfx card then the issue is probably driver/hardware related, but if it cannot the issue is probably with how the game itself is interacting with the P4 (poor optimization or utilization or something like that). I know that when I had an AthlonXP still functioning, it would stack-up against my Pentium 4 systems like this in older applications and benchmarks (where SSE2/SSE3 aren't a consideration), but when the P4 could leverage SSE2 (or especially SSE3) it would just blast the AthlonXP. Unfortunately I don't think there's any games or benchmarks that the Voodoo2 can actually run that will also take advantage of SSE2/3.

Aside from testing with a non-3dfx card, do you have a significantly faster (and ideally with 1MB+ cache) Pentium 4 you can test with?

Well the 3.2 GHz Northwood was able to match the 1.4 Tualatin 1.4 in MOST SLI runs at 512 x 384. But not in all...

My SIS chipset board is almost up and running.

In the reviews I've seen the early P4 machines are all faster than the PIII at 640 x 480. So this is quite puzzling...

I also have a 6800GT in that pc, I changed the parameter to "primary" and the result was worse, lower/higher was ~78fps IIRC

Not having much luck with the AOpen board. It chrashes when I try Quake II in 3dfx mode.

Software and Direct3D (Geforce) work fine.

Might try another driver.

Pretty sure it is because of the caps. Shame I don't have any 3300 lying around 🙁

Managed to get a GLQuake SLI run though.

255.9 instead of 229.5 fps. But Tualatin 1.4 still leads with 264.70.

Ok crash was a driver issue.

I'll try a higher FSB CPU next and see what happens.

I give up 😵

Tried 3 boards, same results. The only benchmark that the P4 can pull away is Unreal.

I'm going to kick out the P4 results from my paper but will write about it. Really seems that Voodoo 2 is happiest with Slot and Socket 370 😀

Mau1wurf1977 wrote:

I give up 😵

Tried 3 boards, same results. The only benchmark that the P4 can pull away is Unreal.

I'm going to kick out the P4 results from my paper but will write about it. Really seems that Voodoo 2 is happiest with Slot and Socket 370 😀

Honestly this is interesting, and kind of surprising; thanks for taking the time to flesh it out too! 😀

Do you have 'CPU to PCI post write' or similar options enabled in bios? If so, don't forget to turn them on

Delayed transaction I usually leave on, off it can cause issues for some pci cards.
If its off can't hurt to try that.
Never bothered using V2's in sli mode other then once... I don't like the IQ loss.
This problem exhibited with a single card?

Maybe it's just the games not being optimized for a p4 or something...