VOGONS


Reply 20 of 38, by murrayman

User metadata
Rank Member
Rank
Member

Been playing around with the above recommendations, and things just seem to be getting stranger.

Right now, I'm on another clean install of 98SE with the V5 5500. Decided to keep things simple for now and just go with DirectX 7.0a, Raziel 64's Evolution Driver v1.01.16, and the latest 3dfx Tools -- no updates yet, including the security updates. Everything appears correct under Info in Tools as before, and settings are such to allow use of both GPUs. Installed and ran 3DMark 99 MAX at default settings again, and got close to the same score as before, this time in the 4200s. Swapped out to the V3 1000, made sure the Raziel 64 drivers were still being utilized (just told Windows to install the drive again just to make sure), and ran the benchmark. Approx. 100 3DMarks higher yet again, in the 4300s.

Although these results are basically the same as before, what's strange are the temperature readings -- or at least what can be felt by hand. The V3 uses the same small black heatsink as is found stock on the V5, but with no fan. The chip gets hot as you'd expect. On the V5, both heatsink fans are running, but only one heatsink is getting even remotely warm when in use. Decided to try a game, Descent 3, using Glide at 1024x768 with maxed details. The V3 ran just fine (50fps+) on indoor portions and with particle effects (e.g. smoke), but sluggish on the outdoor parts (<10fps), though it was clearly due to a lack of RAM, not the GPU, as it only struggled when trying to render the farthest distances or several objects at once. The V5 on the other hand did just fine in these parts (50fps+), but performed right on par or slightly worse than the V3 everywhere else.

It seriously seems like the V5 is running on just one chip, but all the settings are correct in Tools. It's running at AGP 1x, voltages are correct, everything I can think of to check based on what's recommended here and on the 3dfx Help Page checks out fine. Any thoughts?

P3B-F 1.04, PIII 1k, 512MB PC133, GF DDR 32MB + DM3DII 12MB SLI, SB0100
P3B-F 1.03, PIII 700, 384MB PC100, V5 AGP, SB0160
CP 5170, PII 350, 256MB PC100, Rage LT 2MB, ESS 1869
PB M S610, PMMX 233, 128MB EDO66, DM3D 4MB, Aztech

Reply 21 of 38, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++

Did you install the VIA chipset drivers?

What power supply are you using?

Is your RAM running at PC133 or PC100?

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

Reply 22 of 38, by murrayman

User metadata
Rank Member
Rank
Member
cyclone3d wrote:

Did you install the VIA chipset drivers?

What power supply are you using?

Is your RAM running at PC133 or PC100?

Haven't installed any chipset drivers, I'll look into that next.

Power supply came with the rig; it's an AOpen FSP250-60GTW rated at 250W.

RAM is running at PC133; down to the one stick of 512mb now, running stable. CPU is also at 933mhz stable; settings are correct in the BIOS and CPUID confirms everything's good.

P3B-F 1.04, PIII 1k, 512MB PC133, GF DDR 32MB + DM3DII 12MB SLI, SB0100
P3B-F 1.03, PIII 700, 384MB PC100, V5 AGP, SB0160
CP 5170, PII 350, 256MB PC100, Rage LT 2MB, ESS 1869
PB M S610, PMMX 233, 128MB EDO66, DM3D 4MB, Aztech

Reply 23 of 38, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++

Try chipset drivers first. If that doesn't fix it, it could very well be that the power supply is not putting out enough power. That is a really old power supply after all.

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

Reply 24 of 38, by murrayman

User metadata
Rank Member
Rank
Member
cyclone3d wrote:

Try chipset drivers first. If that doesn't fix it, it could very well be that the power supply is not putting out enough power. That is a really old power supply after all.

VIA 4in1 chipset drivers appropriate for the Apollo Pro series installed with no issues: no difference in benchmarks or game performance.

Is there a way I could reliably (and safely) check to see if the PSU is putting out enough juice? I'm assuming since wattage rather than voltage would be the issue, that might not be possible. And if not, are there any credible new-stock PSUs being manufactured for old spec motherboards like this?

P3B-F 1.04, PIII 1k, 512MB PC133, GF DDR 32MB + DM3DII 12MB SLI, SB0100
P3B-F 1.03, PIII 700, 384MB PC100, V5 AGP, SB0160
CP 5170, PII 350, 256MB PC100, Rage LT 2MB, ESS 1869
PB M S610, PMMX 233, 128MB EDO66, DM3D 4MB, Aztech

Reply 25 of 38, by mwdmeyer

User metadata
Rank Oldbie
Rank
Oldbie

Sounds like one of the VSA chips might be dead. Can you disable SLI and see if the performance drops?

Vogons Wiki - http://vogonswiki.com

Reply 26 of 38, by murrayman

User metadata
Rank Member
Rank
Member
mwdmeyer wrote:

Sounds like one of the VSA chips might be dead. Can you disable SLI and see if the performance drops?

Just ran a benchmark on single chip mode, and sure enough, the performance was the exact same. 😢 Is this a sure-fire way of determining the second VSA is dead though?

P3B-F 1.04, PIII 1k, 512MB PC133, GF DDR 32MB + DM3DII 12MB SLI, SB0100
P3B-F 1.03, PIII 700, 384MB PC100, V5 AGP, SB0160
CP 5170, PII 350, 256MB PC100, Rage LT 2MB, ESS 1869
PB M S610, PMMX 233, 128MB EDO66, DM3D 4MB, Aztech

Reply 27 of 38, by mwdmeyer

User metadata
Rank Oldbie
Rank
Oldbie

Sorry I have no idea but it doesn't sound great. I'm sure there are some more knowledgeable people here that can suggest some other things and/or repairs.

Its pretty old that the card works fine otherwise....

Vogons Wiki - http://vogonswiki.com

Reply 28 of 38, by fitzpatr

User metadata
Rank Member
Rank
Member

[url]https://arstechnica.com/civis/viewtopic.php?t=1066253[\url]

You are using different drivers, but who knows...it could help.

As far as PSUs go...in that configuration, any modern power supply should be ample. 250W was considered mandatory for the V5, AFAIK, so swap in anything newer and more powerful.

MT-32 Old, CM-32L, CM-500, SC-55mkII, SC-88Pro, SC-D70, FB-01, MU2000EX
K6-III+/450/GA-5AX/G400 Max/Voodoo2 SLI/CT1750/MPU-401AT/Audigy 2ZS
486 Build

Reply 29 of 38, by F2bnp

User metadata
Rank l33t
Rank
l33t

Do you have another system you could use to test the card? Also, could you try the latest official reference drivers instead of 3rd party ones?

Also, isn't the Voodoo3 1000 an underclocked Voodoo3 with one TMU disabled and 8MB RAM? For all intents and purposes, that should be maybe slightly faster than 1/4 the performance of a Voodoo5, so even if one of those VSA100 chips is dead (which I doubt), a single one of these should still be quite a bit faster.

Another thing to try is Via 4in1 4.25, anything newer than that usually causes more issues.

Reply 30 of 38, by havli

User metadata
Rank Oldbie
Rank
Oldbie

There is a simple test to see whether both VSA are working - just activate 4xAA in 3dfx tools. If it works, then the problem lies elsewhere, not in the V5 itself.

Also try to run for example Q3A at highest details 1024x768 or 1280x1024. When using the maximum performance mode, your fps should be ~2x higher than with single chip or 2xAA and ~4x higher than with 4xAA.

Weak PSU wouldn't cause low performance, your PC would simply shut down or restart itself.

HW museum.cz - my collection of PC hardware

Reply 32 of 38, by Arctic

User metadata
Rank Oldbie
Rank
Oldbie

I dont know which driver you have installed (didnt backread)

But there is an undocumented bug in the 3dfx tools that the enable and disable functions for vsync are swapped
vsync enabled means disabled and vice versa...

Reply 33 of 38, by murrayman

User metadata
Rank Member
Rank
Member

Started this project off with the Reference 1.04 drivers; they provided the same performance as the Raziel 64 driver I'm currently running, so no impact there. However, I tried 4xAA as recommended by havli, and it worked; ran 3DMark 99 again and, as expected, got roughly half the score in the 2700s. Benchmark ran just fine though.

Downloaded the VIA 4.25a drivers, but haven't installed them just yet. Currently running 4.43 as acquired direct from their site, and since I haven't seen any performance impact from since before I installed them, I'm not inclined to think they're having a negative effect. I can still try though.

I don't own a copy of Q3A (surprise, right?), but I tried 3DMark 99 at 1024x768 32bit and switched between single chip and fastest performance mode. This is the first time I've noticed a difference: at fastest performance, the first game test locks in at around 36 - 37fps. In single chip mode, it locks in at exactly 26fps. Scaled down the resolution back to default 800x600 16bit, and found the performance to be the exact same between the two modes again. Since the fps are around 75ish at best, tried swapping between vsync modes to see if it's a glitch as suggested by Arctic, but saw no measurable difference.

This at least seems to confirm both VSAs are working -- which is great! -- but still haven't been able to identify the bottleneck.

EDIT: I keep forgetting to mention, I have no other system with which to try either the V3 or V5. This is the only computer I have left that has a motherboard with AGP.

P3B-F 1.04, PIII 1k, 512MB PC133, GF DDR 32MB + DM3DII 12MB SLI, SB0100
P3B-F 1.03, PIII 700, 384MB PC100, V5 AGP, SB0160
CP 5170, PII 350, 256MB PC100, Rage LT 2MB, ESS 1869
PB M S610, PMMX 233, 128MB EDO66, DM3D 4MB, Aztech

Reply 34 of 38, by Srandista

User metadata
Rank Oldbie
Rank
Oldbie
murrayman wrote:

I don't own a copy of Q3A (surprise, right?).

Try demo version. You can find it here and you can also compare your result with others in that topic.

Socket 775 - ASRock 4CoreDual-VSTA, Pentium E6500K, 4GB RAM, Radeon 9800XT, ESS Solo-1, Win 98/XP
Socket A - Chaintech CT-7AIA, AMD Athlon XP 2400+, 1GB RAM, Radeon 9600XT, ESS ES1869F, Win 98

Reply 36 of 38, by murrayman

User metadata
Rank Member
Rank
Member
Srandista wrote:
murrayman wrote:

I don't own a copy of Q3A (surprise, right?).

Try demo version. You can find it here and you can also compare your result with others in that topic.

Appreciated! Here are the results:

Voodoo 3 1000 = 27.3fps

Voodoo 5 5500:
1) Single Chip Only = 37.3fps
2) Fastest Performance = 58.5fps
3) 2xAA = 37.7fps
4) 4xAA = 16.8fps

The fastest performance results seem on par with some of the results found in that thread for Y2K CPU & GPU period-correct specs, and this is further confirmation that both VSAs are in working order. But am I right in reading the results right, or should I still be doing better than this?

swaaye wrote:

Did anyone ask if there are any missing surface mount doodads? And verify clock speed? Maybe the BIOS has been modified in some way?

CPU clock speed is correct, and the motherboard BIOS settings are correct; no overclocking, FSB @ 133, CPU is confirmed running in BIOS and Win98 at 933Mhz. The V5's BIOS is 1.06 and shows both VSAs running at 166Mhz in 3dfx Tools.

I've heard tale of V5s being prone to missing something on the cards at times, but I've not found anything put simply online when I did a search, just the usual "VSA thermal death" threads. What would I be looking for?

P3B-F 1.04, PIII 1k, 512MB PC133, GF DDR 32MB + DM3DII 12MB SLI, SB0100
P3B-F 1.03, PIII 700, 384MB PC100, V5 AGP, SB0160
CP 5170, PII 350, 256MB PC100, Rage LT 2MB, ESS 1869
PB M S610, PMMX 233, 128MB EDO66, DM3D 4MB, Aztech

Reply 37 of 38, by F2bnp

User metadata
Rank l33t
Rank
l33t

Okay, so this probably sounds like a case of V-Sync? I saw someone mention that it's probably bugged or something, you could try V.Control by KoolSmoky (http://www.3dfxzone.it/koolsmoky/) which will allow you to disable it once and for all and also allow you to tweak other parameters of the VSA100 chips. It's a really cool utility, I use it all the time.

As far as the drivers go, when trying to figure out who the culprit is, it's best to use the latest reference drivers along with DirectX 7.0a, which are known to be the most stable. I also suggested older versions of 4in1, because 4.43 was usually causing issues with older chipsets, even massive performance drops at times (you have to remember that Via Apollo Pro 133 is among the earlier AGP chipsets from Via). Your mileage may vary!
There was another thread by retrogames100 (where did that crazy guy disappear to huh? 🤣 ), where he found out that for Apollo Pro 133, 4.25 was producing the best results for example.

Reply 38 of 38, by murrayman

User metadata
Rank Member
Rank
Member
F2bnp wrote:

Okay, so this probably sounds like a case of V-Sync? I saw someone mention that it's probably bugged or something, you could try V.Control by KoolSmoky (http://www.3dfxzone.it/koolsmoky/) which will allow you to disable it once and for all and also allow you to tweak other parameters of the VSA100 chips. It's a really cool utility, I use it all the time.

As far as the drivers go, when trying to figure out who the culprit is, it's best to use the latest reference drivers along with DirectX 7.0a, which are known to be the most stable. I also suggested older versions of 4in1, because 4.43 was usually causing issues with older chipsets, even massive performance drops at times (you have to remember that Via Apollo Pro 133 is among the earlier AGP chipsets from Via). Your mileage may vary!
There was another thread by retrogames100 (where did that crazy guy disappear to huh? 🤣 ), where he found out that for Apollo Pro 133, 4.25 was producing the best results for example.

Alright, rolled back to the Reference drivers, removed 3dfx Tools, installed V.Control, and nixed vsync off both OpenGL/Glide and D3D. Definitely love the available options in V.Control! Anyway, ran both 3DMark 99 and Q3A again: no difference. I was expecting perhaps a little more headroom on 3DMark since the fps kept capping out at 75, which indeed is my monitor's refresh rate, but no such luck.

I "installed" the VIA 4.25a chipset drivers, but I'm not sure if they successfully overwrote the newer drivers. Tried looking in device manager, but there was no discernible difference; the dates were the same, etc. Tried using Everest to confirm the driver IDs to no avail. How can I confirm the installation was successful / the correct drivers are being used? In any case, ran both 3DMark and Q3A again: no difference.

P3B-F 1.04, PIII 1k, 512MB PC133, GF DDR 32MB + DM3DII 12MB SLI, SB0100
P3B-F 1.03, PIII 700, 384MB PC100, V5 AGP, SB0160
CP 5170, PII 350, 256MB PC100, Rage LT 2MB, ESS 1869
PB M S610, PMMX 233, 128MB EDO66, DM3D 4MB, Aztech