Thank you for the detailed reply, mockingbird. As per the 3C509B, I'm using the PCI version which has great overall compatibility and low CPU usage, it is easy to find online at reasonable price.

One issue I've seen recently, when installing Windows XP, is that the initial file copy would freeze at around 30%; a complete hang of the system requiring a hard reboot. I identified the source of the problem as the PSU, essentially the CD-ROM drive would start spinning faster and (I suppose) taking some power from the 5V rail while the CPU is busy decompressing the files, also requiring lots of power. The CPU voltage was at 1.4V initially; increasing the voltage to 1.5V causes installation to freeze even earlier, decreasing the voltage to 1.3V resulted in the installation succeeding. The PSU can provide up to 24A on the 5V rail, unfortunately I don't know where to buy a beefier PSU. Can you provide details about your PSU?

About the soldering pin modification, could you post a picture of the modification, and after it is done, can you still adjust the multiplier in the BIOS or is it fixed? Losing the ability to change the multiplier would be problematic for the purpose of playing some speed-sensitive DOS game that require L1 to be disabled + CPU speed set to lowest. In your case though, I suppose it's less of a concern since you have a separate DOS build 😀

Regarding the JMicron JM20330 SATA to IDE, be aware that it has caused DMA issues for me when used with the stock VT82C686B IDE controller. I'd be interested to know if you can get UDMA100 to work reliably in 98/XP or Linux with this adapter coupled with the Ultra 133 TX2.

Indeed an interesting combo would be the Voodoo 2 + GeForce 2 MX. This would get excellent compatibility with a good range of late 90's DOS/Windows 98 games. Also the Voodoo 2 can ofter be found at lower prices than the Voodoo 3 PCI.

An alternative to the ND-4570A / ND-4571A would be the Lite-on LH-20A1H, it is also CDBQ compatible. By the way, I realize you have already bought the parts but I recommend you check out this thread which also documents a slightly more powerful all-in-1 DOS/98/XP build: MSI MS-7253 / K9VGM-V (KM890, VIA8237A) , Athlon 64 X2 5050e, Radeon X800 XL . An advantage is the possibility of dynamically adjusting the CPU speed down to 386 SX levels.

Finally, I recently bought the WP32 McCake and I would absolutely recommend it. MT-32 emulation is perfect and you can switch dynamically to Fluidsynth mode with some very decent SC-55 soundfonts available (see here: Re: WP32 McCake : MT32 compatible waveblaster board). If your preference is better SC-55 emulation the the X2GS is also really good.

cde wrote on 2021-12-31, 13:56:

Thank you for the detailed reply, mockingbird. As per the 3C509B, I'm using the PCI version which has great overall compatibility and low CPU usage, it is easy to find online at reasonable price.

Ah, ok, thanks for that.

The PSU can provide up to 24A on the 5V rail, unfortunately I don't know where to buy a beefier PSU. Can you provide details about your PSU?

My PSU criteria for ATX builds are simple. 1) Japanese capacitors or equivalent and 2) 5VSB standby circuit that uses an IC, and not a two-transistor design. If your PSU has the old two transistor design, your cooking the innards while the computer is off. You can see this for yourself if you remove the PSU from the case, lay it on its back with the fan pointing up, have it connected to the wall and switched on in the back and connected to the motherboard so that the motherboard is drawing the 5V standby voltage, then leave it overnight and come back the next morning. Put your hand on the fan exhaust (the fan is not spinning of course - because the PSU is in standby), and feel the heat coming out of the thing... Unacceptable, and unsafe in the long run. The overwhelming majority of modern PSUs have a standby circuit... Even inexpensive PSUs.

Which brings us to the next point: use a modern PSU. I understand you want a 5V rail with high ratings, but unless you're running 10,000 RPM SCSI drives or something, it should be more than enough. Personally, I favour the old FSP PN/B units or THN... It's a late 2000s PSU with the old design, but later revisions have the standby circuit. They need re-capping before they can be put to use.

About the soldering pin modification, could you post a picture of the modification

and after it is done, can you still adjust the multiplier in the BIOS or is it fixed?

I am looking into the part early on in your post where you have trouble doing that in software... I believe I can accomplish that and mean to test that for you... It may have to do with Abit disabling a certain feature in the latest BIOS revision. We both know that in theory, XP-M should respond to SETMUL commands. Truth be told, you can still adjust the multiplier, but you're adjusting what are now only high multipliers, so that's still useless to you, because you're not going to go as low down there in Mhz as you personally desire.

Regarding the JMicron JM20330 SATA to IDE, be aware that it has caused DMA issues for me when used with the stock VT82C686B IDE controller. I'd be interested to know if you can get UDMA100 to work reliably in 98/XP or Linux with this adapter coupled with the Ultra 133 TX2.

Yes, absolutely... The Ultra 133TX2/JMicron combo is rock solid... The later JMicron adapters are perhaps even better than Marvell adapters.

Indeed an interesting combo would be the Voodoo 2 + GeForce 2 MX. This would get excellent compatibility with a good range of late 90's DOS/Windows 98 games. Also the Voodoo 2 can ofter be found at lower prices than the Voodoo 3 PCI.

Yes, the MX is an excellent choice as well, because with the GTS and MX, you can go as low as the 5.xx drivers, which is sometimes necessary for older games.

An alternative to the ND-4570A / ND-4571A would be the Lite-on LH-20A1H

I don't recommend the Lite-ON drives... They're nice on paper but they more often than not need re-capping. And they use odd 8mm height caps. So you're often forced to lay 11.5mm caps flat to accommodate that.

I realize you have already bought the parts but I recommend you check out this thread which also documents a slightly more powerful all-in-1 DOS/98/XP build

I tried to do something similar to you... A P4 Northwood build with a Solo connected to SB-Link... That is still a work in progress, but it's not looking good. The P4 board doesn't seem to respond to PCISET commands, so I can't get the SB-Link working. If I could get that working, I'd ditch everything else and stick the Vortex 2 in there for Windows, along with the Solo for DOS.

Finally, I recently bought the WP32 McCake and I would absolutely recommend it.

I have a working EWS64 XL, it sounds nice and all and has a certain soundfont which resembles the MT32 to the tee, but I might just sell the thing. It's a white elephant. I purchased the ESS wavetable card from our gentleman here on the forum who sells it, and it's good enough for me. Any wavetable is good enough for me, as long as it's replacing FM synth.

Ok, will report back on the SETMUL findings. Cheers for now.

Forget the soldering of the pins. That is not needed.

All you need is a couple single strands of wire, bent in a U-shape and put in the socket in specific places in order to unlock the higher multipliers.

That is how I have mine set up... except I am using the Barton Mobile 2800+

Guides.
https://www.ocinside.de/workshop_en/amd_socket_a_oc/
https://www.ocinside.de/workshop_en/amd_pinmod/

5th multiplier bit
http://www.emboss.co.nz/amdmults/

cyclone3d wrote on 2021-12-31, 18:53:

Forget the soldering of the pins. That is not needed. […]
Show full quote

Forget the soldering of the pins. That is not needed.

All you need is a couple single strands of wire, bent in a U-shape and put in the socket in specific places in order to unlock the higher multipliers.

That is how I have mine set up... except I am using the Barton Mobile 2800+

Guides.
https://www.ocinside.de/workshop_en/amd_socket_a_oc/
https://www.ocinside.de/workshop_en/amd_pinmod/

5th multiplier bit
http://www.emboss.co.nz/amdmults/

That doesn't work on the KT7A 1.1

I tried it.

mockingbird wrote on 2021-12-31, 20:54:

cyclone3d wrote on 2021-12-31, 18:53:

Forget the soldering of the pins. That is not needed. […]
Show full quote

Forget the soldering of the pins. That is not needed.

All you need is a couple single strands of wire, bent in a U-shape and put in the socket in specific places in order to unlock the higher multipliers.

That is how I have mine set up... except I am using the Barton Mobile 2800+

Guides.
https://www.ocinside.de/workshop_en/amd_socket_a_oc/
https://www.ocinside.de/workshop_en/amd_pinmod/

5th multiplier bit
http://www.emboss.co.nz/amdmults/

That doesn't work on the KT7A 1.1

I tried it.

What exact wire mod did you do? For the higher multipliers, you may have to play around with it to get one that will allow you to still boot. That is what the 5th multiplier bit page is about.

mockingbird wrote on 2021-12-31, 18:07:

My PSU criteria for ATX builds are simple. 1) Japanese capacitors or equivalent and 2) 5VSB standby circuit that uses an IC, and not a two-transistor design. If your PSU has the old two transistor design, your cooking the innards while the computer is off. You can see this for yourself if you remove the PSU from the case, lay it on its back with the fan pointing up, have it connected to the wall and switched on in the back and connected to the motherboard so that the motherboard is drawing the 5V standby voltage, then leave it overnight and come back the next morning. Put your hand on the fan exhaust (the fan is not spinning of course - because the PSU is in standby), and feel the heat coming out of the thing... Unacceptable, and unsafe in the long run. The overwhelming majority of modern PSUs have a standby circuit... Even inexpensive PSUs.

Well, this requirement (5VSB standby) is important if you are connecting the PSU at the wall socket during all the night.

I am using a switchable power distribution unit, and i think that lots of retro friends are using these kind of unit also...
During the last years i have re-capped several older PSUs, also with stronger 5V rail. To me, this feature 5V standby voltage is irrelevant because i will switch off the power distribution unit anyway.

cyclone3d wrote on 2022-01-01, 01:56:

What exact wire mod did you do? For the higher multipliers, you may have to play around with it to get one that will allow you to still boot. That is what the 5th multiplier bit page is about.

Which one didn't I try from that page is the better question. They do not work on the earlier (and possibly later) KT7A boards. If you want to use a Barton with this model, you need to do the full 5 jumper wire mod from the ocinside.de pin mode calculator.

Set the parameters as follows:

PCB View
AMD Barton
8,00 x Systembus (16 x FSB)

Leave VCore and FSB unselected, you don't need to do those because you have full control of that in the BIOS.

You also still have some multiplier control in the BIOS afterwards as well... You need to set it to the correct multiplier setting anyhow to 'sync' with the pin mod you just did on the back of the board.

melbar wrote on 2022-01-01, 15:54:

Well, this requirement (5VSB standby) is important if you are connecting the PSU at the wall socket during all the night.

I am using a switchable power distribution unit, and i think that lots of retro friends are using these kind of unit also...
During the last years i have re-capped several older PSUs, also with stronger 5V rail. To me, this feature 5V standby voltage is irrelevant because i will switch off the power distribution unit anyway.

Re-capping an old PSU is nice, but not always sufficient. Many old models used an unstable cellulose glue which becomes conductive as it ages. FSP was notorious for this in their older models but got better. Lite-ON did this too, so did Besttec, and not to mention Antec's CWT builds (but not their Delta builds). HiPro are better in this regard and are usually easier to repair than others.

You also need to give some attention to the exhaust fan.

Seriously, you can't go wrong with a new PSU.

mockingbird wrote on 2022-01-02, 01:17:

Which one didn't I try from that page is the better question. They do not work on the earlier (and possibly later) KT7A boards. […]
Show full quote

cyclone3d wrote on 2022-01-01, 01:56:

What exact wire mod did you do? For the higher multipliers, you may have to play around with it to get one that will allow you to still boot. That is what the 5th multiplier bit page is about.

Which one didn't I try from that page is the better question. They do not work on the earlier (and possibly later) KT7A boards. If you want to use a Barton with this model, you need to do the full 5 jumper wire mod from the ocinside.de pin mode calculator.

Set the parameters as follows:

PCB View
AMD Barton
8,00 x Systembus (16 x FSB)

Leave VCore and FSB unselected, you don't need to do those because you have full control of that in the BIOS.

You also still have some multiplier control in the BIOS afterwards as well... You need to set it to the correct multiplier setting anyhow to 'sync' with the pin mod you just did on the back of the board.

melbar wrote on 2022-01-01, 15:54:

Well, this requirement (5VSB standby) is important if you are connecting the PSU at the wall socket during all the night.

I am using a switchable power distribution unit, and i think that lots of retro friends are using these kind of unit also...
During the last years i have re-capped several older PSUs, also with stronger 5V rail. To me, this feature 5V standby voltage is irrelevant because i will switch off the power distribution unit anyway.

Re-capping an old PSU is nice, but not always sufficient. Many old models used an unstable cellulose glue which becomes conductive as it ages. FSP was notorious for this in their older models but got better. Lite-ON did this too, so did Besttec, and not to mention Antec's CWT builds (but not their Delta builds). HiPro are better in this regard and are usually easier to repair than others.

You also need to give some attention to the exhaust fan.

Seriously, you can't go wrong with a new PSU.

My KT7A is version 1.00 and it worked fine. Maybe it is because I have the Mobile 2800+ and not eh Mobile 2500+ ?

cyclone3d wrote on 2022-01-02, 02:12:

My KT7A is version 1.00 and it worked fine. Maybe it is because I have the Mobile 2800+ and not eh Mobile 2500+ ?

No sir, I tried it with the desktop 2800+.

Let me be clear, the 5-only pinmod, 5-3, 5-4, and 5-1 did not work.

What kind of numbers are you pulling with the 2800+ AXMH?

mockingbird wrote on 2022-01-02, 06:39:

No sir, I tried it with the desktop 2800+. […]
Show full quote

cyclone3d wrote on 2022-01-02, 02:12:

My KT7A is version 1.00 and it worked fine. Maybe it is because I have the Mobile 2800+ and not eh Mobile 2500+ ?

No sir, I tried it with the desktop 2800+.

Let me be clear, the 5-only pinmod, 5-3, 5-4, and 5-1 did not work.

What kind of numbers are you pulling with the 2800+ AXMH?

The Mobile chips are unlocked. Desktop are not going to work as far as I know.
It has been a while since I powered up that system, but I am pretty sure it is running at either 2.2 or 2.3Ghz.

I believe some desktop chips, made before a certain week in 2003?, are unlocked. However I never managed to get one; all the chips I got were locked. So it was easier to use the mobile version, which also had the advantage of undervolting better.

mockingbird wrote on 2022-01-02, 01:17:

Which one didn't I try from that page is the better question. They do not work on the earlier (and possibly later) KT7A boards. […]
Show full quote

cyclone3d wrote on 2022-01-01, 01:56:

What exact wire mod did you do? For the higher multipliers, you may have to play around with it to get one that will allow you to still boot. That is what the 5th multiplier bit page is about.

Which one didn't I try from that page is the better question. They do not work on the earlier (and possibly later) KT7A boards. If you want to use a Barton with this model, you need to do the full 5 jumper wire mod from the ocinside.de pin mode calculator.

Set the parameters as follows:

PCB View
AMD Barton
8,00 x Systembus (16 x FSB)

Leave VCore and FSB unselected, you don't need to do those because you have full control of that in the BIOS.

You also still have some multiplier control in the BIOS afterwards as well... You need to set it to the correct multiplier setting anyhow to 'sync' with the pin mod you just did on the back of the board.

melbar wrote on 2022-01-01, 15:54:

Well, this requirement (5VSB standby) is important if you are connecting the PSU at the wall socket during all the night.

I am using a switchable power distribution unit, and i think that lots of retro friends are using these kind of unit also...
During the last years i have re-capped several older PSUs, also with stronger 5V rail. To me, this feature 5V standby voltage is irrelevant because i will switch off the power distribution unit anyway.

Re-capping an old PSU is nice, but not always sufficient. Many old models used an unstable cellulose glue which becomes conductive as it ages. FSP was notorious for this in their older models but got better. Lite-ON did this too, so did Besttec, and not to mention Antec's CWT builds (but not their Delta builds). HiPro are better in this regard and are usually easier to repair than others.

You also need to give some attention to the exhaust fan.

Seriously, you can't go wrong with a new PSU.

I was unaware of this. Could you tell me some more about this and how to recognize this particular type of glue and which models are affected?

cyclone3d wrote on 2022-01-02, 08:06:

The Mobile chips are unlocked. Desktop are not going to work as far as I know.
It has been a while since I powered up that system, but I am pretty sure it is running at either 2.2 or 2.3Ghz.

Week 0338 is right about where they started locking them.

Now whether it would have made a difference using the 5-pin only mod with a mobile - I don't think so. Unlocked mobile and desktop chips should behave the same. The exception is that power-saving feature built into the mobile chips which allows some sort of frequency manipulation in software, which cde couldn't get working, but it is something I am going to test for him. Abit disabled some sort of power-related feature in the last BIOS revision. Maybe that has something to do with it.

...can you burn an Ultimate BootCD and run Prime95 v28 or later on it (under the CPU category in the menu) please? For accurate temperature readings, you need to remove the CPU, and pull up on that blue thermistor in the inner portion of the socket, apply some thermal grease to it, and then re-insert the CPU so that the bottom is touching the greased tip of the thermistor when it's installed. I would love to know how good those 2800+ AXMH chips are. They should be pretty expensive and rare these days.

Tetrium wrote on 2022-01-02, 12:44:

I was unaware of this. Could you tell me some more about this and how to recognize this particular type of glue and which models are affected?

Certainly, but Adrian Black did a pretty good job demonstrating its effects in a very recent video. I recommend you watch that to get a better understanding:

Can I diagnose and repair this small IBM VGA monitor?

This stuff is very time consuming to remove... Sometimes it comes off in blotches, sometime you have to pick away at it like an archaeologist unearthing some rare artifact... It depends on its stage of decay. One thing's for sure, it will eventually short component leads together, or at the very least, act as an (albeit) high-resistance conductor.

I can do the prime95 test but it is not going to match pretty much any other system out there as I am using a Thermaltake Big Typhoon with a better fan and that has also been lapped.

I know I bent the thermistor up to touch the bottom of the CPU.

I actually have a few of the mobile 2800+ CPUs. They pop up every once in a while for a good price.

Pretty sure the motherboard is the limiting factor on this setup as well.

cyclone3d wrote on 2022-01-02, 17:14:

I can do the prime95 test but it is not going to match pretty much any other system out there as I am using a Thermaltake Big Ty […]
Show full quote

I can do the prime95 test but it is not going to match pretty much any other system out there as I am using a Thermaltake Big Typhoon with a better fan and that has also been lapped.

I know I bent the thermistor up to touch the bottom of the CPU.

I actually have a few of the mobile 2800+ CPUs. They pop up every once in a while for a good price.

Pretty sure the motherboard is the limiting factor on this setup as well.

Indeed. How did you adapt that for Socket370/462?

Maybe you won't mind 'donating' one of yours to me and we can have a contest... I'll put my KT7A up against yours.

mockingbird wrote on 2022-01-02, 18:42:

cyclone3d wrote on 2022-01-02, 17:14:

I can do the prime95 test but it is not going to match pretty much any other system out there as I am using a Thermaltake Big Ty […]
Show full quote

I can do the prime95 test but it is not going to match pretty much any other system out there as I am using a Thermaltake Big Typhoon with a better fan and that has also been lapped.

I know I bent the thermistor up to touch the bottom of the CPU.

I actually have a few of the mobile 2800+ CPUs. They pop up every once in a while for a good price.

Pretty sure the motherboard is the limiting factor on this setup as well.

Indeed. How did you adapt that for Socket370/462?

Maybe you won't mind 'donating' one of yours to me and we can have a contest... I'll put my KT7A up against yours.

Couldn't remember exactly how I did it so I just took a look. I am actually using all stock brackets. The KT7A has through the board holes so it is quite easy to use a cooler like that.
In fact, the Big Typhoon is stock compatible with Athlon XP - https://www.techpowerup.com/review/thermaltake-big-typhoon/

I am also using one of those copper shims that keep the cooler from tipping on the core.

Edit: This biggest problem with the old Thermaltake coolers that used that mounting system was that they didn't come with a way to secure the bolts to the motherboard. Nylon washers and Nylon nuts fixes that right up.
So the bolts come though the back mounting bracket and through the back of the board and then the nylon washers and nylon nuts are put on.

Without that, those coolers move all over the place.

My original "mod" for that fix was to weld the bolts to the rear bracket but that makes for a system where you can't reuse the rear bracket or the bolts if you change the type of system it is being used on.

Oh yeah, I thought my board was v1.0, but it is v1.1

cyclone3d wrote on 2022-01-02, 20:47:

Oh yeah, I thought my board was v1.0, but it is v1.1

Thanks for the cooler information, much appreciated and kudos on your inventiveness and creativity.

I am puzzled then, if your board is revision 1.1 and you had success with the simple mods...

Ok, I'll leave it to future members to come on here and post more info to establish a consensus, in the mean time, they have plenty of info to get up and running if they wish to build such a system.

I forget the exact wire mod I did. I will need to take the CPU out to take a look. Not happening tonight though.

Last weekend I replaced all the caps in my Abit KT7A v1.1 from my main retrorig (did a post to Retro Machines FB-group too) and I was hoping it would overclock a bit more, but I was wrong.

Specs:
-Athlon XP mobile-Barton 1,7GHz (136x12,5)
-768MB RAM
-Radeon 9700PRO
-Silicon Image PCI-SATA with 143GB WD Raptor
-SB Audigy2 with front panel
-ESS Solo-1 for DOS SB Pro support
-Gravis Ultrasound MAX
-3,5" and 5,25" floppy drives
-CD-ROM and DVD-ROM drives (one burner)
-Dual-boots Win98SE and WinXP

I'm not interested in CPU multipliers because memory bandwith is the main bottleneck here. Last stable FSB is 136MHz and no more and I'm getting nuts because I just don't figure out what's the reason.

Let's say I try 140FSB:
I can run memtest whole day, no errors.
3dmark2001SE few times, no problems.
Doom3, no problems.
Command&Conquer Zero Hour, no problems.
Need for Speed Underground, no problems.
But, Far Cry crashes immediately.

Then I loose every setting in BIOS, memory and AGP settings, raise voltage and stuff but Far Cry just won't run. But, if I take FSB down to 136MHz then Far Cry works?!?
I have tried different drivers, graphics cards, memory sticks, with sound, no sound... nothing helps, not even a hint what could be the problem here.

I have overclocked computers since 1993 but this is hard one.

Any ideas or is the problem V1.1 board revision?

And yes, I have faster machines that I can use for Far Cry but still...

I would recommend undervolting more the CPU, down to 1.3V. Personally I get better results through undervolting rather than overvolting (that way the rest of the system has more juice, so to speak).