VOGONS


A very persnickety PIII machine

Topic actions

First post, by murrayman

User metadata
Rank Member
Rank
Member

AOpen 250w PSU, AOpen AX63 Pro, PIII 933 Slot 1, 256mb Toshiba PC133 SDRAM (2x128mb), Voodoo 5 5500 AGP, Sound Blaster Live! 5.1 SB0100, Western Digital WD400JB 40GB IDE HDD.

I’ve been working at various points on this computer since early 2018 after a HDD failure. You can find the bulk of details of everything that has been done up to this post HERE and HERE. I have other threads floating around for some of the below-listed issues as well.

The issues I currently have are some of the same that can be found in the first link:
- I get a very similar score on 3DMark99 MAX between a Voodoo 3 1000G 8MB and two different Voodoo 5 5500s (got the second one this week and sold the first one) despite confirming SLI is enabled and the molex connection is plugged in on the latter two, switching between AGP 1x and 2x in 3dfx Utilities, and switching between Reference 1.04 and a few different Raziel drivers. Sometimes I actually get a significantly lower score with the V5s than the V3. Other users were kind enough to provide their test scores (found in the first link) which show I should be getting a significantly higher score with a V5.
- Running 3DMark99 MAX on the V3 will crash on the 1000 pixel polygons (individual) test about 9 times out of 10, but never on either V5.
- Installing a copy of the VIA 4in1 4.25a chipset drivers (downloaded off VogonsDrivers) through Setup.exe goes from 0% to 100% instantly and without seeming to install anything, and I don’t know how to confirm if the drivers have installed.
- On all three videos cards, Descent 3 main menu’s animated background displays graphical anomalies at times and eventually crashes Main.exe after a few secs / mins (varies at random) using Glide or D3D APIs, but not OpenGL (though the menu background is not animated when using this). Using the same game discs and installation on my Compaq with an ATI Rage LT Pro does not crash in D3D (OpenGL not supported).

The issues I no longer have are the ones found in the second link:
- In early 2018, I had a 20gb WD IDE HDD installed that I had been using in the system for over a decade. It failed after moving from one house to another when I turned on the system for the first time after the move; motor didn’t fire up, and the BIOS could no longer identify the drive. After trying a number of IDE-to-SATA adapters and SATA HDDs with varying kinds of read / write and hang issues, I installed a NOS 2008 40gb WD HDD, and began having issues after installing some games where VoptXP defrag would encounter crosslinked files, allocation table problems, and unknown issues where it would defer to ScanDisk, which could not properly fix the issues. Around this time, I replaced the Creative 52x CD drive that had been in the system since new with an LG 42x DVD drive, and after multiple clean installs of Win98SE and subsequent game / software installs, both VoptXP and ScanDisk now always show good from top to bottom, including surface scan. So I haven’t had one issue with the HDD since replacing the optical drive.

There are a lot of factors at play and a number of parts that have been swapped since the start of the issues two years ago. For the full details, please see both above links. A rough list of the parts are:
- IDE cables (40 pins swapped to 80 pins on primary and secondary)
- Floppy ribbon cable
- AGP video cards
- SDRAM (from two unknown brand 512mb PC133 sticks to two Toshiba 128mb PC133 sticks)
- Removal and reinstall of SB Live! Drive front panel ribbon cable and power lead
- Optical drive
- Floppy drive
- BIOS battery

I’m looking for help with anything that anyone can address. It’s hard to see the forest for the trees now that I’ve been messing with this thing for two years after nearly two decades of problem-free use. Guess we’ve been growing old together 😜

Last edited by murrayman on 2020-01-19, 20:12. Edited 1 time in total.

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 1 of 42, by PARKE

User metadata
Rank Oldbie
Rank
Oldbie

In one of the links you posted the specs for your AOpen AX63 Pro motherboard read that it has the VIA 693A north bridge.
If that is indeed the case(?) then this motherboard must be considerably slower than 440BX, VIA Apollo Pro133 A and i815 motherboards and comparing your 3DMark99 results with results from either of these latter 3 chipsets does not seem very helpful.
With the VIA Apollo Pro (without A) chipset the 933/133 Mhz. cpu can produce a >cpu benchmark< in 3DMark99 of appr. in between 10.000 and 11.000 points but in for example a 440BX board the 933 will hit on 14.000 points. This may well have an effect on the scaling between a V3 and a V5 but I have no direct experience with that aspect.

As afterthought... you describe running 2x512Mb RAM on this motherboard in the past but the VIA Apollo Pro chipset does not support 512Mb RAM sticks - installed software may 'recognize' that 512Mb is installed but it can only address 256Mb (which you can see in Windows system).

Reply 2 of 42, by murrayman

User metadata
Rank Member
Rank
Member
PARKE wrote on 2020-01-17, 17:11:
In one of the links you posted the specs for your AOpen AX63 Pro motherboard read that it has the VIA 693A north bridge. If tha […]
Show full quote

In one of the links you posted the specs for your AOpen AX63 Pro motherboard read that it has the VIA 693A north bridge.
If that is indeed the case(?) then this motherboard must be considerably slower than 440BX, VIA Apollo Pro A and i815 motherboards and comparing your 3DMark99 results with results from either of these latter 3 chipsets does not seem very helpful.
With the VIA Apollo Pro (without A) chipset the 933/133 Mhz. cpu can produce a >cpu benchmark< in 3DMark99 of appr. in between 10.000 and 11.000 points but in for example a 440BX board the 933 will hit on 14.000 points. This may well have an effect on the scaling between a V3 and a V5 but I have no direct experience with that aspect.

As afterthought... you describe running 2x512Mb RAM on this motherboard in the past but the VIA Apollo Pro chipset does not support 512Mb RAM sticks - installed software may 'recognize' that 512Mb is installed but it can only address 256Mb (which you can see in Windows system).

Okay, understood! I had no clue since online specs for the board showed it supported 1.25GB and the BIOS recognized it fine when I had them installed. In any case, they’re out of the picture now.

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 4 of 42, by murrayman

User metadata
Rank Member
Rank
Member
PARKE wrote on 2020-01-17, 17:33:

What are your results for the >cpu benchmark< when you run 3DMark99 ?

12,993, so almost 13,000

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 5 of 42, by lost77

User metadata
Rank Member
Rank
Member

Check for a VIA CPU to AGP controller in device manager to see if the driver was installed. You also might want to use the recommended 4.43 driver instead.

Try to benchmark Quake III, both with one and two chips active.

Reply 6 of 42, by sledge

User metadata
Rank Member
Rank
Member
PARKE wrote on 2020-01-17, 17:11:

With the VIA Apollo Pro (without A) chipset the 933/133 Mhz. cpu can produce a >cpu benchmark< in 3DMark99 of appr. in between 10.000 and 11.000 points but in for example a 440BX board the 933 will hit on 14.000 points.

My ECS P6BAP-ME has the same chipset (if I'm not mistaken) and it's close to 13400 CPU 3DMarks. (With some tweaking via WPCRSET and with "Memory Interleave Enabler 0.15" installed).

doshaven.eu / high-voltage.cz

Reply 7 of 42, by murrayman

User metadata
Rank Member
Rank
Member
lost77 wrote on 2020-01-17, 18:06:

Check for a VIA CPU to AGP controller in device manager to see if the driver was installed. You also might want to use the recommended 4.43 driver instead.

Try to benchmark Quake III, both with one and two chips active.

Via CPU to AGP controller is the Microsoft-provided driver that was already installed. Also, I thought 4.43 was commonly known to have issues, so most recommend 4.25a? Unless I’m misremembering?

I’ll bench Q3A in both states when I get a chance this evening!

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 8 of 42, by SirNickity

User metadata
Rank Oldbie
Rank
Oldbie
murrayman wrote on 2020-01-17, 18:23:

I thought 4.43 was commonly known to have issues, so most recommend 4.25a? Unless I’m misremembering?

I'm curious about this too. I've always tried to steer clear of VIA, but I have one now. Is it OS or chipset specific, or should one always use a particular version?

Reply 9 of 42, by murrayman

User metadata
Rank Member
Rank
Member
SirNickity wrote on 2020-01-17, 19:15:
murrayman wrote on 2020-01-17, 18:23:

I thought 4.43 was commonly known to have issues, so most recommend 4.25a? Unless I’m misremembering?

I'm curious about this too. I've always tried to steer clear of VIA, but I have one now. Is it OS or chipset specific, or should one always use a particular version?

IIRC, it’s OS-specific; newer drivers work well on W2k+, but not W9x. I’m pretty sure the older drivers, specifically 4.25a, were considered more reliable for W9x, but I’m also waiting for feedback

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 11 of 42, by jaZz_KCS

User metadata
Rank Oldbie
Rank
Oldbie

I was hopping through the 4 in 1 drivers as well, but all instabilities on VIA chipsets were - in the end, at least for me - caused by other culprits and the chipset drivers were not at fault.
I have - since then - gone by this:

[...]
- For VIA MVP3, MVP4, Apollo 133 - use 4 in 1 4.35 or older. - better AGP compatibility and performance. Don't use any nvidia driver newer then 5x.xx!!!! 6x.xx drivers for example don't mesh well with 4 in 1 drivers and might bsod
- For VIA Apollo PRO, KT133 and KT266 use 4.43
- For KT333, KT400, KT600, KT880 and newer use 4.56 under Windows 98 - you should manually install all devices (CPU to AGP bridge, CPU to PCI bridge, etc) - sometimes they don't install correctly. Newer chipsets will work well under 98 with the newest drivers (Hyperion PRO 5.xx) - again, you might have to manually install the bridges from the device manager. For Win XP use Hyperion PRO 5.23 / 5.24A
[...]

stability may vary and depends on the MB/Chipset combination.

Reply 12 of 42, by murrayman

User metadata
Rank Member
Rank
Member
lost77 wrote on 2020-01-17, 19:42:

I've gone with the recommendations in this thread:

VIA 4-in-1 driver experiences

Never had any issues.

jaZz_KCS wrote on 2020-01-17, 19:43:
I was hopping through the 4 in 1 drivers as well, but all instabilities on VIA chipsets were - in the end, at least for me - cau […]
Show full quote

I was hopping through the 4 in 1 drivers as well, but all instabilities on VIA chipsets were - in the end, at least for me - caused by other culprits and the chipset drivers were not at fault.
I have - since then - gone by this:

[...]
- For VIA MVP3, MVP4, Apollo 133 - use 4 in 1 4.35 or older. - better AGP compatibility and performance. Don't use any nvidia driver newer then 5x.xx!!!! 6x.xx drivers for example don't mesh well with 4 in 1 drivers and might bsod
- For VIA Apollo PRO, KT133 and KT266 use 4.43
- For KT333, KT400, KT600, KT880 and newer use 4.56 under Windows 98 - you should manually install all devices (CPU to AGP bridge, CPU to PCI bridge, etc) - sometimes they don't install correctly. Newer chipsets will work well under 98 with the newest drivers (Hyperion PRO 5.xx) - again, you might have to manually install the bridges from the device manager. For Win XP use Hyperion PRO 5.23 / 5.24A
[...]

stability may vary and depends on the MB/Chipset combination.

Okay, got it. Little confusion: Since I have an Apollo Pro133, should I use 4.35/older or 4.43? I see Apollo 133 and Apollo Pro listed as separate entities, but not Apollo Pro133.

Also confirmed - at least by AOpen's website, whether correct or no - that chipset is VIA 693A, and that the mobo supports 1.5GB of RAM. Both bits of info found here. Everest Home 2.20 confirms the northbridge as VT82C693A Apollo Pro133.

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 13 of 42, by murrayman

User metadata
Rank Member
Rank
Member

Tried running the same battery of tests havli suggested in 2018 using Q3A Demo 1.11 at 1280x1024 w/ highest texture detail.

SLI = 31.6 seconds, 42.7 fps
Single-chip = 72.2 seconds, 18.6 fps

So both VSAs are working just fine on this second V5.

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 14 of 42, by lost77

User metadata
Rank Member
Rank
Member

So it seems it working correctly. My guess is that Vsync is still on.

Setting to "Default off" or similar settings wont work, it needs to be forced. This is an issue specific to 3DMark99, later versions will turn it off if they are allowed. For older cards I usually use Rivatuner of Powerstrip to do that but I don't know if they support the Voodoo 5 properly (or at all).

Reply 15 of 42, by murrayman

User metadata
Rank Member
Rank
Member
lost77 wrote on 2020-01-17, 23:34:

So it seems it working correctly. My guess is that Vsync is still on.

Setting to "Default off" or similar settings wont work, it needs to be forced. This is an issue specific to 3DMark99, later versions will turn it off if they are allowed. For older cards I usually use Rivatuner of Powerstrip to do that but I don't know if they support the Voodoo 5 properly (or at all).

This was theorized and tested back in 2018 as well. V.Control is the way to force vsync off with a V5. Tried it with this card: 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

Reply 16 of 42, by SirNickity

User metadata
Rank Oldbie
Rank
Oldbie
jaZz_KCS wrote on 2020-01-17, 19:43:

I was hopping through the 4 in 1 drivers as well, but all instabilities on VIA chipsets were - in the end, at least for me - caused by other culprits and the chipset drivers were not at fault.
I have - since then - gone by this:

If bees had knees, you would be they. Thanks for this!

Reply 17 of 42, by PARKE

User metadata
Rank Oldbie
Rank
Oldbie

murrayman wrote:
Okay, got it. Little confusion: Since I have an Apollo Pro133, should I use 4.35/older or 4.43? I see Apollo 133 and Apollo Pro listed as separate entities, but not Apollo Pro133.
>>
>>
The 'official' denominations are [Apollo Pro 133] for the VIA 693A chipset
and [Apollo Pro 133A] for the VIA 694x chipset

Your 3DMark99 cpu score looks nice for the VIA 693A. The VIA 693A system that I have here scores considerably lower - it runs btw on the VIA 4in14.28 chipset driver, a vanilla install without any additional tweaks.
**

murrayman wrote:
Also confirmed - at least by AOpen's website, whether correct or no - that chipset is VIA 693A, and that the mobo supports 1.5GB of RAM. Both bits of info found here. Everest Home 2.20 confirms the northbridge as VT82C693A Apollo Pro133.
>>
>>
There have been Vogons threads on this subject before and it seems that 'supports' does not mean that the 512Mb sticks are actually working as such - only half of the 512Mb is effectively available.
see for example:
Computer only detects half the RAM

Reply 19 of 42, by murrayman

User metadata
Rank Member
Rank
Member
lost77 wrote on 2020-01-18, 19:35:

I still think it might be the vsync issue. Easy enough to test, run 3DMark99 at 640x480 and 60hz. The FPS counter in the first game test should be at 60 most of the time.

Gotcha! Ran test as such, and first game FPS max was 72.

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