VOGONS


First post, by someperson42

User metadata
Rank Newbie
Rank
Newbie

I'm looking for a GPU for a 440BX machine with the bus running at 133 MHz. Infamously, this configuration causes the AGP bus to be overclocked to 89 MHz, and the GeForce MX440 I have right now crashes when using Direct3D in this configuration. Can anyone recommend a better card for this configuration? Thank you!

Reply 2 of 27, by shamino

User metadata
Rank Oldbie
Rank
Oldbie

The Geforce4 MX cards are among those I would have suggested, so I'm surprised you're having trouble with an MX440.
In general I think cards in the NVidia Geforce2/3/4 range have been known to work well at 89MHz AGP.

Cards I've used that way are Geforce2 MX, Geforce3, and Ti4200.
My Radeon 9800 Pro did not work with 89MHz.

Reply 3 of 27, by darry

User metadata
Rank l33t
Rank
l33t

The assumption being made here is that the crashes are due to the video card because Direct3D happens to be in use when said crashes occur. However, the crashes may have a number of potential other causes

a) Is the RAM able to run at 133MHz
b) Is the board/CPU even stable running at 133MHz FSB with no video card load ?
c) Is the board used one of those that that does not apply a /4 divider to FSB to derive the PCI clock when FSB=133MHz and thus ends up running the PCI bus at 44MHz in that situation ?

Memtest and Prime95 can be used to test a) and b) respectively . Hopefully, the board is not affected by c) .

Reply 5 of 27, by someperson42

User metadata
Rank Newbie
Rank
Newbie
darry wrote on 2021-05-21, 00:52:
The assumption being made here is that the crashes are due to the video card because Direct3D happens to be in use when said cra […]
Show full quote

The assumption being made here is that the crashes are due to the video card because Direct3D happens to be in use when said crashes occur. However, the crashes may have a number of potential other causes

a) Is the RAM able to run at 133MHz
b) Is the board/CPU even stable running at 133MHz FSB with no video card load ?
c) Is the board used one of those that that does not apply a /4 divider to FSB to derive the PCI clock when FSB=133MHz and thus ends up running the PCI bus at 44MHz in that situation ?

Memtest and Prime95 can be used to test a) and b) respectively . Hopefully, the board is not affected by c) .

These are good questions, and I took the time to check all this just to make sure:

a) I'm using PC133 RAM, so I figured this would be fine, but just to be safe, I let memtest86 run for a couple of passes. No issues with that.
b) The CPU I'm using is a 1 GHz Pentium III designed for a 133 MHz bus, but just to test the board stability, I let Prime95 run for about 4 hours. No issues.
c) I don't think so. The board I'm using is an AOpen AX6BC. I don't see any settings for the PCI clock on this system, but I did confirm that AGP is correctly set to 2/3.

Given all this, I'm pretty sure the GPU is the problem, especially since invoking 3D acceleration seems to be the only way I can reproduce the issue, and the system crashes almost immediately when I do.

Doornkaat wrote on 2021-05-21, 08:54:

Some boards offer a setting to slightly increase the AGP signalling voltage. If your board has this option try it and see wether stability increases. Good luck!😃

I don't see an option for this.

Reply 8 of 27, by mothergoose729

User metadata
Rank Oldbie
Rank
Oldbie
someperson42 wrote on 2021-05-21, 18:10:
mothergoose729 wrote on 2021-05-21, 18:06:

Sanity check, does the card work if you set the bus speed to 100mhz?

Yes, it works fine with the bus set to 100 MHz.

If it were me, I might just run my processor at a lower bus speed and call it a day.

I don't have any personal experience with boards that run at this AGP clock. Maybe someone else does. It would be a shame if you went out and bought a pricier FX card and it still didn't work.

Reply 9 of 27, by darry

User metadata
Rank l33t
Rank
l33t

Thank you for extra info.

Based on that, my guess is AGP/video card related as well.

I have an Asus P3B-F with 89MHz AGP base clock and a Geforce FX 5900 XT running fine, for what it's worth .

In your case, it is going to be hard to isolate the issue without either trying your video card in another 89MHz system or trying another video card, unfortunately.

Then again, if 89MHz AGP ends up not being a workable option for your board, getting a modded Tualatin chip and a cheap coppermine slotket (assuming your boards VRMs are up to it) might make 100MHz FSB more palatable . That would definitely help get more mileage out of any Geforce FX you might have ended up buying for testing . Admittedly, all of this has a cost, but it could be worth it .

Reply 10 of 27, by bloodem

User metadata
Rank Oldbie
Rank
Oldbie

I’ve never seen a GeForce 2+ card that doesn’t work with an out-of-spec AGP bus speed. There might be some cheapo models out there with this issue, though…

I’ve successfully tested A LOT of cards such as GeForce 2 GTS/Pro/Ti, GeForce 3 (Ti200/500), GeForce 4 Mx440/460 and GeForce 4Ti 4200/4400/4600 on 440BX motherboards running at FSB133, and never have I had an issue that was caused specifically by the video card. The issues are usually related to the motherboard itself (some fail to run properly at FSB133, but work very well at FSB124), or some other peripheral.

Based on your tests, it does seem (at first) that the video card is to blame, but it’s still not 100% conclusive. I would do some more tests with the bare minimum (just the motherboard, CPU, RAM, video card, HDD) on a fresh installation wih driver version 30.82 (one of the fastest and most stable for GeForce 4 MX/Ti cards).

2 x PGA132 / 5 x Socket 3 / 4 x Socket 7 / 6 x Super Socket 7 / 5 x Slot 1 / 3 x Slot A
5 x Socket 370 / 5 x Socket A / 1 x Socket 478 / 2 x Socket 754 / 3 x Socket 939 / 4 x LGA775 / 1 x LGA1155
Current rig: Ryzen 5 3600X
Backup rig: Core i7 7700k

Reply 11 of 27, by someperson42

User metadata
Rank Newbie
Rank
Newbie
darry wrote on 2021-05-21, 20:06:

In your case, it is going to be hard to isolate the issue without either trying your video card in another 89MHz system or trying another video card, unfortunately.

Unfortunately, I don't have another such system on hand, or even another AGP card, so I'm not sure I have much of a choice but to try buying one.

bloodem wrote on 2021-05-21, 20:22:

I’ve never seen a GeForce 2+ card that doesn’t work with an out-of-spec AGP bus speed. There might be some cheapo models out there with this issue, though…

I’ve successfully tested A LOT of cards such as GeForce 2 GTS/Pro/Ti, GeForce 3 (Ti200/500), GeForce 4 Mx440/460 and GeForce 4Ti 4200/4400/4600 on 440BX motherboards running at FSB133, and never have I had an issue that was caused specifically by the video card. The issues are usually related to the motherboard itself (some fail to run properly at FSB133, but work very well at FSB124), or some other peripheral.

Based on your tests, it does seem (at first) that the video card is to blame, but it’s still not 100% conclusive. I would do some more tests with the bare minimum (just the motherboard, CPU, RAM, video card, HDD) on a fresh installation wih driver version 30.82 (one of the fastest and most stable for GeForce 4 MX/Ti cards).

Interesting. I've attached pictures of the exact card I'm using. What would you recommend trying to test it?

Attachments

Reply 12 of 27, by The Serpent Rider

User metadata
Rank l33t
Rank
l33t

I’ve never seen a GeForce 2+ card that doesn’t work with an out-of-spec AGP bus speed

Ha! Try 1:1 FSB/AGP multiplier on 440BX.

Get up, come on get down with the sickness
Open up your hate, and let it flow into me

Reply 13 of 27, by bloodem

User metadata
Rank Oldbie
Rank
Oldbie
The Serpent Rider wrote on 2021-05-22, 06:49:

Ha! Try 1:1 FSB/AGP multiplier on 440BX.

Hehe! Well, you know what I mean 😀 I'm referring to the typical (lowest) 2/3 AGP divider, which results in a ~ 89 MHz frequency 😀

2 x PGA132 / 5 x Socket 3 / 4 x Socket 7 / 6 x Super Socket 7 / 5 x Slot 1 / 3 x Slot A
5 x Socket 370 / 5 x Socket A / 1 x Socket 478 / 2 x Socket 754 / 3 x Socket 939 / 4 x LGA775 / 1 x LGA1155
Current rig: Ryzen 5 3600X
Backup rig: Core i7 7700k

Reply 14 of 27, by pete8475

User metadata
Rank Member
Rank
Member

I've tried a number of cards in my CUBX-E machine, off the top of my head:

5900XT - perfect
6800XT - crashes
TI4200 - perfect
440MX - crashes
Voodoo 3 - perfect

I'm sure I've used others but that's what I can remember for sure right now.

Reply 15 of 27, by someperson42

User metadata
Rank Newbie
Rank
Newbie

Since making this post, I've obtained a Ti4200 card. This definitely runs a lot more stable with the 133 MHz bus than the MX440 did. However, there are still issues. Most games actually work just fine, but there are still cases where I run into crashes. Oddly, the most reliable way I've found to reproduce them is to preview the "Underwater" screensaver that comes with Windows 98. I have no idea why, but in this configuration, this crashes the system almost every time. It works fine if I set the FSB to 100 MHz.

This seems to leave me with a few options:

  • Try to find an even more stable GPU for this configuration. I'm not convinced that's possible though.
  • Try to find an SL4BR or SL4KL CPU, which are Slot 1 CPUs intended for a 100 MHz bus. These CPUs seem to be rare though.
  • Get an SL5QV CPU, which seems far more common than the aforementioned Slot 1 CPUs, and try to set it up with a slotket. There are a bunch of unknowns here for me, like which slotkets would work, if they would fit in the retention clip already on my board, and I'm sure I'd have to get some kind of cooler.
  • Replace the board with something more suitable for the CPU I already have. Options? I'd definitely want something with an ISA slot or 2.
  • Live with the CPU underclocked to 750 MHz.

Honestly, it kind of sucks, but I'm leaning towards just living with it.

Reply 16 of 27, by darry

User metadata
Rank l33t
Rank
l33t
someperson42 wrote on 2021-06-10, 23:29:
Since making this post, I've obtained a Ti4200 card. This definitely runs a lot more stable with the 133 MHz bus than the MX440 […]
Show full quote

Since making this post, I've obtained a Ti4200 card. This definitely runs a lot more stable with the 133 MHz bus than the MX440 did. However, there are still issues. Most games actually work just fine, but there are still cases where I run into crashes. Oddly, the most reliable way I've found to reproduce them is to preview the "Underwater" screensaver that comes with Windows 98. I have no idea why, but in this configuration, this crashes the system almost every time. It works fine if I set the FSB to 100 MHz.

This seems to leave me with a few options:

  • Try to find an even more stable GPU for this configuration. I'm not convinced that's possible though.
  • Try to find an SL4BR or SL4KL CPU, which are Slot 1 CPUs intended for a 100 MHz bus. These CPUs seem to be rare though.
  • Get an SL5QV CPU, which seems far more common than the aforementioned Slot 1 CPUs, and try to set it up with a slotket. There are a bunch of unknowns here for me, like which slotkets would work, if they would fit in the retention clip already on my board, and I'm sure I'd have to get some kind of cooler.
  • Replace the board with something more suitable for the CPU I already have. Options? I'd definitely want something with an ISA slot or 2.
  • Live with the CPU underclocked to 750 MHz.

Honestly, it kind of sucks, but I'm leaning towards just living with it.

Another potential option is to get a Coppermine compatible slotket (if you are not already using one) and get one of the modified CPUs described below .

darry wrote on 2021-06-03, 04:36:
If you have a socket 370 board that supports Coppermine CPUs and has flexible VRMs, there is a guy in Korea selling pre-modified […]
Show full quote
BitWrangler wrote on 2021-06-03, 03:45:

Oh man I wish I had access to a plug and play tualatin board, even just for testing CPUs. Got a 1.1A that I don't know if it works, and there's a mighty tempting 1.2 on offer locally. Less locally there's a guy with a poorly described bunch of slotket stuff, but IDK if he will ship and whether there's anything in there that will help me.

If you have a socket 370 board that supports Coppermine CPUs and has flexible VRMs, there is a guy in Korea selling pre-modified Tualatin CPUs (interposer PCB designed by the guy is used) that will work in many Coppermine capable boards .
Prices start at 50ish $US . If you go to the auction site that shall not be names and search for Tualatin Korea , you should find what I am referring to . A list of compatible boards is included in the listings .

If you do take that path. Make sure that

a) your board's VRMs circuit supports sending the proper (low by coppermine standards) voltage to the CPU
b) you provide additional cooling to the VRM circuit on your board as feeding a lower Tualatin-compatible voltage to the CPU will generate more heat than the board designers likely expected when they designed the board for Coppermine voltage requirements .

Reply 17 of 27, by someperson42

User metadata
Rank Newbie
Rank
Newbie
darry wrote on 2021-06-10, 23:50:

Another potential option is to get a Coppermine compatible slotket (if you are not already using one) and get one of the modified CPUs described below .

darry wrote on 2021-06-03, 04:36:
If you have a socket 370 board that supports Coppermine CPUs and has flexible VRMs, there is a guy in Korea selling pre-modified […]
Show full quote
BitWrangler wrote on 2021-06-03, 03:45:

Oh man I wish I had access to a plug and play tualatin board, even just for testing CPUs. Got a 1.1A that I don't know if it works, and there's a mighty tempting 1.2 on offer locally. Less locally there's a guy with a poorly described bunch of slotket stuff, but IDK if he will ship and whether there's anything in there that will help me.

If you have a socket 370 board that supports Coppermine CPUs and has flexible VRMs, there is a guy in Korea selling pre-modified Tualatin CPUs (interposer PCB designed by the guy is used) that will work in many Coppermine capable boards .
Prices start at 50ish $US . If you go to the auction site that shall not be names and search for Tualatin Korea , you should find what I am referring to . A list of compatible boards is included in the listings .

If you do take that path. Make sure that

a) your board's VRMs circuit supports sending the proper (low by coppermine standards) voltage to the CPU
b) you provide additional cooling to the VRM circuit on your board as feeding a lower Tualatin-compatible voltage to the CPU will generate more heat than the board designers likely expected when they designed the board for Coppermine voltage requirements .

Hmm... that's an interesting option, but it also seems kind of risky. I don't know if the AX6BC's VRMs support the proper voltage. I'm also not using a slotket currently (the CPU I have is an SL4BS) and it isn't clear to me what all I would need to use one.

Reply 18 of 27, by darry

User metadata
Rank l33t
Rank
l33t
someperson42 wrote on 2021-06-11, 00:54:
darry wrote on 2021-06-10, 23:50:

Another potential option is to get a Coppermine compatible slotket (if you are not already using one) and get one of the modified CPUs described below .

darry wrote on 2021-06-03, 04:36:
If you have a socket 370 board that supports Coppermine CPUs and has flexible VRMs, there is a guy in Korea selling pre-modified […]
Show full quote

If you have a socket 370 board that supports Coppermine CPUs and has flexible VRMs, there is a guy in Korea selling pre-modified Tualatin CPUs (interposer PCB designed by the guy is used) that will work in many Coppermine capable boards .
Prices start at 50ish $US . If you go to the auction site that shall not be names and search for Tualatin Korea , you should find what I am referring to . A list of compatible boards is included in the listings .

If you do take that path. Make sure that

a) your board's VRMs circuit supports sending the proper (low by coppermine standards) voltage to the CPU
b) you provide additional cooling to the VRM circuit on your board as feeding a lower Tualatin-compatible voltage to the CPU will generate more heat than the board designers likely expected when they designed the board for Coppermine voltage requirements .

Hmm... that's an interesting option, but it also seems kind of risky. I don't know if the AX6BC's VRMs support the proper voltage. I'm also not using a slotket currently (the CPU I have is an SL4BS) and it isn't clear to me what all I would need to use one.

You would not be the first one to try a Tualatin on an AX6BC . Tualatin CPUs are all socketed, which is why you would need a slotket, unless I am misunderstanding what you mean .

See https://web.archive.org/web/20060219190343/ht … fl/slot-t_2.htm

Reply 19 of 27, by someperson42

User metadata
Rank Newbie
Rank
Newbie
darry wrote on 2021-06-11, 01:18:
someperson42 wrote on 2021-06-11, 00:54:
darry wrote on 2021-06-10, 23:50:

Another potential option is to get a Coppermine compatible slotket (if you are not already using one) and get one of the modified CPUs described below .

Hmm... that's an interesting option, but it also seems kind of risky. I don't know if the AX6BC's VRMs support the proper voltage. I'm also not using a slotket currently (the CPU I have is an SL4BS) and it isn't clear to me what all I would need to use one.

You would not be the first one to try a Tualatin on an AX6BC . Tualatin CPUs are all socketed, which is why you would need a slotket, unless I am misunderstanding what you mean .

See https://web.archive.org/web/20060219190343/ht … fl/slot-t_2.htm

I get that I would need a slotket. It is good to hear that others have done it with my board before, but I don't know which particular slotkets would be compatible. The ones I see on eBay don't seem to come with the plastic bracket thing shown in the pictures in that link, so I'm also not sure what I would have to do about that. I also assume I would need some kind of cooling solution.

Also, aren't Tualatins 133 MHz bus CPUs? I'm starting to wonder if this makes any sense.