VOGONS


First post, by gex85

User metadata
Rank Member
Rank
Member

Hi guys,
I am currently putting together a new P3-S Tualatin system.
Mainboard is a Fujitsu Siemens D1219 (i815E chipset), 512 MB RAM, Windows 2000.

Problem: Terrible 3D performance with different video cards. I have other systems with similar configurations so I can compare.

GeForce 4 Ti4200 --> around 3600 3DMarks, should be between 8000 and 9000.
Radeon 9600 Pro --> around 4200 3DMarks, should be able to hit 10000.

I'm running Win2k with SP4 and DirectX 9.0c installed, default BIOS settings, tried different driver versions for both cards, but to no avail.

Any ideas why there's such a hefty performance hit? Seems I'm not able to figure it out myself...

My retro computers

Reply 2 of 12, by gex85

User metadata
Rank Member
Rank
Member

Good hint. I checked CPU and Memory scores with PCMark2002.
CPU Score ~2100 (should be in the 4400 range)
Mem Score ~1350 (should be around 2700)

So CPU and memory performance is also about half of what it should be.

Edit: Super PI v1.1 1M even worse:
3m 55s (should be around 1m 45s)

My retro computers

Reply 4 of 12, by LubieCipy

User metadata
Rank Newbie
Rank
Newbie

I get around 8150 points in that benchmark on windows 98SE. Used detonators 44.03 and VIA 694T mainbord. Can you run 3DMark'99 and tell me how many CPU points you get? I noticed a large variance in points between tests on different chipsets for tualatin

Reply 5 of 12, by gex85

User metadata
Rank Member
Rank
Member
kixs wrote on 2021-11-11, 11:28:

What FSB is the system running at?

It's humming along nicely at 133 * 10.5 = 1400 MHz. Verified with WCPUID and Everest.

LubieCipy wrote on 2021-11-11, 11:40:

I get around 8150 points in that benchmark on windows 98SE. Used detonators 44.03 and VIA 694T mainbord. Can you run 3DMark'99 and tell me how many CPU points you get? I noticed a large variance in points between tests on different chipsets for tualatin

I just tried to run it but it complains that I am not on DirectX 6.1...

My next steps would be:
1. Swap out the CPU for a different one. I still have some known good P3-S 1400 in stock.
2. Try some different memory sticks
3. Install a fresh copy of Windows 98 on a different disk and benchmark again

If none of this helps, I don't know what to do. I haven't seen a problem like this ever before.

My retro computers

Reply 6 of 12, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++

Are you using something USB? Perhaps a USB PCI card? Those old PCI USB cards can kill performance.

Something else that is hammering the interrupt controller could cause it as well.

Yamaha modified setupds and drivers
Yamaha XG repository
YMF7x4 Guide
Aopen AW744L II SB-LINK

Reply 7 of 12, by gex85

User metadata
Rank Member
Rank
Member

No additional USB controller card, but for the HDD I am using a Promise SATA 300 TX2plus controller. Haven't had any issues with these in the past, though. But I might try hooking up the disk to the IDE controller using a simple SATA to IDE converter and see if it makes any difference.

My retro computers

Reply 8 of 12, by Macca70

User metadata
Rank Newbie
Rank
Newbie

Win2k is a pita for drivers, have you tried this configuration in 98 ?
Wouldn't mind betting it's fine there.
Chipset drivers installed ?

Build 1
Win98se
Dell Dimension XPS B866r
1.0ghz PIII
512mb 800mhz Rdram
Asus GF2 Ti 64mb

Build 2
XP System
Lian Li PC65 USB
Athlon 64 x2 4800+
Asus A8N SLI Premium
2gb Corsair XMS 2
Asus 6600GT 256mb

Reply 9 of 12, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

Looks like CPU is constantly busy with some background processes. Could be USB flash stick.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 10 of 12, by gex85

User metadata
Rank Member
Rank
Member

I have now ruled out Win2k, the Promise SATA controller and the Audigy2 sound card as possible problem sources.
I removed both cards and booted the computer using a Live-CD of Win XP (Hirens Boot CD) and ran SuperPI, which gave the same lousy results (just under 4 Minutes for calculating 1M).
Will try some other RAM sticks and a different CPU later.

Edit: Not the RAM either. Same results with completely different sticks.

Edit: Seems it really was the CPU. I swapped it for a known-good Tualatin 1400 MHz and now the results are just slightly below my expectations (7940 3D marks in 3DMark2001 with the GF4 Ti4200, SuperPI 1M = 1m 52s, PCMark2002 CPU score a bit low, but memory score OK).
I will now do some back-and-forth swapping of the CPUs to verify the first one is really defective. However, this comes as a bit of a surprise, I haven't seen such a behavior ever before. Maybe some of the CPU cache is damaged, causing this dramatic slowdown.

Edit: Swapped the first CPU back in, ran some benchmarks - everything's fine. Maybe one of the pins didn't make good contact in the socket...

My retro computers

Reply 11 of 12, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++

Yeah... the cache could be flaky and the error checking just happens to be able to fix it but it is resulting in a huge slowdown.

Yamaha modified setupds and drivers
Yamaha XG repository
YMF7x4 Guide
Aopen AW744L II SB-LINK

Reply 12 of 12, by PentAmd

User metadata
Rank Newbie
Rank
Newbie

Hello!

I have also a Fujitsu D1219 as Gex85 has.
If I put a SL5QL 1266MHz Tualatin cpu into the mobo, it boots.
I have also 2pcs 1400Mhz (one of them is SL6BY), but the mobo is not booting with them.
Gex85, what CPU did you use, what was the cpu code?

I have the latest bios, see top of the picture.

Edit: The Board has the version Nr. D1219-D21 GS 2
According to the Fujitsu's website, this D21 does not support tualatin at all. But how is that possible that some tualatin works. (see the picture)

Attachments