That 3DMark seems under half what it should be, I was getting better than that on Socket A. Unless FX5950 really does suck that bad on DX8, I'd have been running a Gf4200 at somewhere near 4600 clocks.

Edit: dang "nature" particularly slaughtered it, I was in the 30s with a GF3

BitWrangler wrote on 2021-11-22, 04:13:

That 3DMark seems under half what it should be, I was getting better than that on Socket A. Unless FX5950 really does suck that bad on DX8, I'd have been running a Gf4200 at somewhere near 4600 clocks.

Edit: dang "nature" particularly slaughtered it, I was in the 30s with a GF3

Yeah I noticed that too, NFS games also struggle. The FX series is supposed to be the DX8 king. I'll have to look into that.

VDNKh wrote on 2021-11-22, 03:51:

I still can't get CD audio to work. VxD or WDM drivers both produce no sound, not even with CD digital audio that WDM drivers offer. Windows will detect an audio CD and play it, but no sound is made. The analog audio cable I know is good after I tested it with a multimeter. So now I'm wondering if it's an issue with the card or with the DVD drive I have.

Yes, it could be the drive itself.

See this thread: When did optical drives stop support cd audio despite still having the connector for it? How to avoid buying one?

BitWrangler wrote on 2021-11-22, 04:13:

That 3DMark seems under half what it should be, I was getting better than that on Socket A. Unless FX5950 really does suck that bad on DX8, I'd have been running a Gf4200 at somewhere near 4600 clocks.

Edit: dang "nature" particularly slaughtered it, I was in the 30s with a GF3

Well, in a contemporary benchmark, the FX 5950 Ultra reached ~12000 3D marks in 3DMark2001 on a 2.4 GHz P4 HT (Win2K Pro, Detonator 51.16) , which is not really that far from the results shown here.
So there's certainly some room for improvement given the faster CPU, but the overall results aren't that bad either.

Edit: Link (in German): https://www.computerbase.de/2003-12/test-gefo … 5900-roundup/7/

gex85 wrote on 2021-11-22, 09:10:

Well, in a contemporary benchmark, the FX 5950 Ultra reached ~12000 3D marks in 3DMark2001 on a 2.4 GHz P4 HT (Win2K Pro, Detona […]
Show full quote

BitWrangler wrote on 2021-11-22, 04:13:

That 3DMark seems under half what it should be, I was getting better than that on Socket A. Unless FX5950 really does suck that bad on DX8, I'd have been running a Gf4200 at somewhere near 4600 clocks.

Edit: dang "nature" particularly slaughtered it, I was in the 30s with a GF3

Well, in a contemporary benchmark, the FX 5950 Ultra reached ~12000 3D marks in 3DMark2001 on a 2.4 GHz P4 HT (Win2K Pro, Detonator 51.16) , which is not really that far from the results shown here.
So there's certainly some room for improvement given the faster CPU, but the overall results aren't that bad either.

Edit: Link (in German): https://www.computerbase.de/2003-12/test-gefo … 5900-roundup/7/

It's probably a combo of the old drivers and this VIA memory controller, which is just inferior to Intel's. Tightening my timings from whatever Auto set it to, to 3-3-3 1T boosted my score over 600 points.

gex85 wrote on 2021-11-22, 09:10:

Well, in a contemporary benchmark, the FX 5950 Ultra reached ~12000 3D marks in 3DMark2001 on a 2.4 GHz P4 HT (Win2K Pro, Detona […]
Show full quote

BitWrangler wrote on 2021-11-22, 04:13:

That 3DMark seems under half what it should be, I was getting better than that on Socket A. Unless FX5950 really does suck that bad on DX8, I'd have been running a Gf4200 at somewhere near 4600 clocks.

Edit: dang "nature" particularly slaughtered it, I was in the 30s with a GF3

Well, in a contemporary benchmark, the FX 5950 Ultra reached ~12000 3D marks in 3DMark2001 on a 2.4 GHz P4 HT (Win2K Pro, Detonator 51.16) , which is not really that far from the results shown here.
So there's certainly some room for improvement given the faster CPU, but the overall results aren't that bad either.

Edit: Link (in German): https://www.computerbase.de/2003-12/test-gefo … 5900-roundup/7/

New benchmark and tightened the memory timings (3-3-2-5 tRFC: 15T CR :1T 4-way interleave)+ overclocked the FSB to 288 (highest it could go before P95 failed while keeping the tight timings), memory is now at 576MHz and CPU is at 3.6GHz. MemTest86 passes with no errors. Still finding how high this 5950 can clock to, but it passed 3DMark 01 with 520MHz core and 1015MHz VRAM.

This bench seems to fit better with what you'd expect out 5950. If I had some better RAM I could probably get 3-2-2-5 8-way interleave, but I'm just chasing maybe a few percent performance gain at that point.

Something is seriously wrong here. With this system you should be able to hit 20-25k in 3dmark01 easily. Perhaps even more. For example check my score with FX5800U with similar system https://hwbot.org/submission/2294872_havli_3d … tra_28937_marks

There are some tweaks decreasing image quality in place of course. But on the other hand at the time I didn't know the BSEL mod trick and the board didn't perform optimally. So with FX 5900, anything under 25000 is bad.

Do you have the chipset drivers instaled properly - especially the AGP driver? This might have huge impact on the performance.

havli wrote on 2021-11-23, 13:45:

Something is seriously wrong here. With this system you should be able to hit 20-25k in 3dmark01 easily. Perhaps even more. For example check my score with FX5800U with similar system https://hwbot.org/submission/2294872_havli_3d … tra_28937_marks

There are some tweaks decreasing image quality in place of course. But on the other hand at the time I didn't know the BSEL mod trick and the board didn't perform optimally. So with FX 5900, anything under 25000 is bad.

Do you have the chipset drivers instaled properly - especially the AGP driver? This might have huge impact on the performance.

I don't have any special settings in DX set, it's all defaults.

I installed the 4in1 v443 driver from VIA's site on a clean install. 8x, Fast write, and sidebanding are enabled. Is there a specific way to install it or a different driver I should use? Could it be because I locked the AGP/PCI bus clocks? Or enabled the VIA chipset compatibility option in Rivatuner? You got double or triple the FPS in some of those tests!

I think version 4.43 is way too old for PT880 chipset. In fact I am surprised it even installs without complaining (it is few years older than the chipset itself). Try something more recent. Like 5.14 or 5.24 for example. https://www.philscomputerlab.com/via-chipset-drivers.html

Then you should see some difference.

Yeah, those 3DMark 2001 scores are way too low. I should recheck on my system that has 98se and a x6800 running at about 3.3Ghz which is where the FX5950U stops becoming CPU limited in 3DMark 2001. Pretty sure I was getting a fair bit over 25k.

If I remember correctly I need to mess with the pencil mod or actually do a proper mod to get the fsb stable where I originally had it.... bleh.

Last time I was testing with it, 3DMark 2001 was acting extremely strange no matter what card I tried in it.

havli wrote on 2021-11-23, 15:42:

I think version 4.43 is way too old for PT880 chipset. In fact I am surprised it even installs without complaining (it is few years older than the chipset itself). Try something more recent. Like 5.14 or 5.24 for example. https://www.philscomputerlab.com/via-chipset-drivers.html

Then you should see some difference.

Why yes I did see a difference, it's SLOWER!

I built this PC a little less than a year ago and the v443 driver worked great, there was always 1 programmable interrupt control that never got a driver but it didn't seem to break anything.

I went into the BIOS and synced AGP/PCI bus to CPU speed, that boosted my score into the 13,000 range.

Then I did a clean install of 98SE and used the v524A driver from VIA's website. At first I thought the installer was broken because it gave an error "The wizard was interrupted before VIA HyperionPro Setup Wizard could be completely installed." But after rebooting it found all the drivers. My memory is hazy but I think that's why I opted for the v443 driver back when I set this up in Christmas 2020. ​Fortunately, the AGP and IDE drivers (they fail too but I don't need them it seems) installed without failing like that. Then I installed the 45.23 Nvidia drivers and DX9 August 2006 package, and now my score is back down into the 12,400 range. On a whim I installed the 56.64 Nvidia driver and that alone boosted my score to 14,455.

But upon closer inspection I think I see the issue here.

havli wrote on 2021-11-23, 13:45:

For example check my score with FX5800U with similar system https://hwbot.org/submission/2294872_havli_3d … tra_28937_marks

This benchmark that havli posted was on XP which is multi-threaded, and they have a dual core processor. No doubt that having an extra processor is boosting the CPU limited score, and the graphics driver is probably way better optimized for XP.

gex85 wrote on 2021-11-22, 09:10:

Edit: Link (in German): https://www.computerbase.de/2003-12/test-gefo … 5900-roundup/7/

By contrast the benchmarks here were run in Windows 2k on a single core with early hyper-threading, which I'm not sure Win2k knows how to handle, all while running newer drivers made for XP.

So in summery, I don't think I can get much higher than 13,000 (on 45.23) while in 98SE. For a single threaded DOS based OS running drivers meant to be on XP, I think this is about as good as it gets. If I run the Nature benchmark in 1280x1024 AA2, the FPS average is almost identical (28FPS vs 26FPS). This is definitely a CPU bottle neck. I am going to keep the newer chipset driver around even though it seems slower, maybe it'll surprise me. In the future I may get a second SSD and dual boot to XP as well. If havli wants to run 3DMark01 in 98SE to prove me wrong however, I welcome it.

Well, it is just weird. Installing XP on your rig just with the basic software (drivers + directX + 3dmark) to find out whether it is SW or HW related problem souds like a good idea.

3DMark01 definitely is not multithreaded in any meaningful way (single digits of percents gain at most). Also while it might be a bit slower in win98 compared to the XP, the difference is not that big. I don't have any FX score at hand, but I have this GeForce 7950 GT one - https://hwbot.org/submission/3207601_havli_3d … _gt_47887_marks

Also the Nature test is not CPU limited at all under normal circumstances. Definitely not at values around 30 fps.

I just ran 3dmark2001 with my fx 5950 ultra that is in my retro pc that has win98se and a slower cpu than yours and I got 18588 score. So I also think your score is too low for the specs.

Edit: driver version in win98se seems to be ForceWare 71.84

Baoran wrote on 2021-11-28, 06:03:

I just ran 3dmark2001 with my fx 5950 ultra that is in my retro pc that has win98se and a slower cpu than yours and I got 18588 score. So I also think your score is too low for the specs.

3dmark01.jpg

Edit: driver version in win98se seems to be ForceWare 71.84

Yeah I think you and havli are right. I've found several other threads reporting the issue with similar hardware to me. I think it might be the USB drivers, as 2 USB gamepads I tried have extremely lagged and jittery inputs. I'm still troubleshooting it.

Pull out all the USB stuff including USB cards you have and try again.

Just to be blunt to be sure and to confirm. This is part of troubleshooting.

Cheers,

VDNKh wrote on 2021-11-28, 18:36:

Baoran wrote on 2021-11-28, 06:03:

I just ran 3dmark2001 with my fx 5950 ultra that is in my retro pc that has win98se and a slower cpu than yours and I got 18588 score. So I also think your score is too low for the specs.

3dmark01.jpg

Edit: driver version in win98se seems to be ForceWare 71.84

Yeah I think you and havli are right. I've found several other threads reporting the issue with similar hardware to me. I think it might be the USB drivers, as 2 USB gamepads I tried have extremely lagged and jittery inputs. I'm still troubleshooting it.

Mine is the factory overclocked Gigabyte GT version that has 520Mhz core clock and 475Mhz memory clock, but it should not make too much of a difference. I installed Arctic NV silencer cooler before using the card to make sure both core and the memory would be well cooled and card would not be as noisy.

Still at a loss as to what is causing this performance issue. Disabling the USB controller on the motherboard did not help much (+200 points on 3DMark01). Though I fixed my USB gamepad lagging issue by upgrading NUSB from 33E to 36E. I tried the 45.23 based Omega driver, no change there either (Slightly slower in the Nature benchmark). Fiddled with more BIOS settings but no big changes. I ran Sandra 2001 CPU and RAM benchmarks and got these scores, I don't have anything to compare it to though. Is this about what an E6800 should be pushing?

Some good news at least, I installed a "Sabrent Laptop 2.5" to Desktop 3.5" IDE Hard Drive Adapter" (as B&H Photo calls it) so I could use an 80-wire IDE cabled with my IDE SSD. It works perfectly! I have full UDMA-133 speeds now. ATTO benchmark:

Not sure why exactly but I tried running 3dmark01 with 45.23 driver and it dropped score from 18588 to 16293, so it does not seem like 45.23 version is fastest driver at least for 5950 ultra and 3dmark01.

Still no luck on getting this performance issue solved. I re-imaged the computer and tried it with no SB drivers installed, no difference in performance. Made some improvements to the hardware though!

I installed a SATA SSD to install Windows XP on. I'm trying to make it so that I can use 98 with the SATA drive still plugged in. Just plugging it in and leaving it on AHCI mode will give the 'ol

1Your multi-function device (Standard Dual PCI IDE Controller) has some child devices using 32-bit drivers and others using compatibility-mode drivers. This configuration is not supported, so your computer has been halted to prevent corruption.

blue screen message, and breaks CD drive functionality.

Setting the SATA controller to RAID mode causes 98 to hard lock as soon as the splash screen displays. Fortunately, I narrowed this down to EMM386.EXE and RAID not playing nice with each other. If I boot without EMM386, it works. If I boot with the SATA cable unplugged and EMM386 enabled, it works. I'll have to tweak CONFIG.SYS (any tips would be helpful, DOS was before my time) to get it to work properly. With RAID mode working in 98, and the VIA RAID driver installed, I was able to create a DOS partition and format the 128GB SSD to be usable in 98. Even though I don't intended for 98 to ever touch this normally, I wanted to see how well it performed.

I was worried that the RAID controller would cause some bad I/O performance, but I guess that's not an issue! (queue the comments telling me that's half the score I should get) I want to get to get some wrapped IDE cables to clean up the interior a bit but cosmetic purchases are on hold.

I'm working on getting XP installed with the proper RAID driver.

Also updated the BIOS to 2.20 Rev:a with the microcode update found here. Don't see any performance changes here either.

And finally changed the coin cell battery.