Kaisersoze wrote on 2022-02-22, 13:43:

Good news guys,
i've changed the psu and now the geforce ti 4600 is back in action ✌️.
Obviously i checked it even after i installed the v2 sli and the awe 64 gold, but so far everything seems ok...Finally i'd say.

Excellent news! Have fun with your new build 😀

Excellent!! This is going to be one hell of a fun gaming system for sure.

Now i would complete my build with a barton mobile one, for both my t-bred's multiplier seemingly being locked so i can't move down to 5x100 mhz when i need it, and for wattage matters...anyway now i am very relieved

EDIT: do you think an epox ep 8kta3+PRO with a may 2003 bios rev could play well with a barton xp-m 2800+ (AXMJ2800FHQ4C)?

There's something strange going on with my ABIT KT7A v1.1.

Athlon XP-M, multiplier x12.5, and voltage default 1,35V, I can get this thing stable maybe just 133FSB (1666MHz). But, if I raise CPU voltage up to 1,5V then maybe 136FSB is stable. Raising northbridge (and memory) voltage to 3,4V or more makes things only worse. It's possible to run memtest whole day without errors, and maybe 3DMark but FarCry crashes after few minutes on memory error.

BUT, if I put in non-mobile but unlocked Barton and pin-mod the CPU to use 15x multiplier (and voltage of 1,65V) then I can get FSB up to 142MHz or even more. (same thing with this CPU with x12.5 multiplier, board is not stable)

Months of struggling and testing I have came to conclusion that the more CPU eats current, the better for getting this board stable.

And yes, all the caps are replaced and the situation is the same before and after.

System is:
-Abit KT7A v1.1
-3x 256MB Micron ECC SDRAM
-Radeon X800Pro AGP
-146GB WD Raptor on PCI-SATA-controller
-Gravis Ultrasound
-Audigy2
-ESS Solo-1
-3COM ethernet

dj_pirtu wrote on 2022-02-23, 08:45:

There's something strange going on with my ABIT KT7A v1.1. […]
Show full quote

There's something strange going on with my ABIT KT7A v1.1.

Athlon XP-M, multiplier x12.5, and voltage default 1,35V, I can get this thing stable maybe just 133FSB (1666MHz). But, if I raise CPU voltage up to 1,5V then maybe 136FSB is stable. Raising northbridge (and memory) voltage to 3,4V or more makes things only worse. It's possible to run memtest whole day without errors, and maybe 3DMark but FarCry crashes after few minutes on memory error.

BUT, if I put in non-mobile but unlocked Barton and pin-mod the CPU to use 15x multiplier (and voltage of 1,65V) then I can get FSB up to 142MHz or even more. (same thing with this CPU with x12.5 multiplier, board is not stable)

Months of struggling and testing I have came to conclusion that the more CPU eats current, the better for getting this board stable.

And yes, all the caps are replaced and the situation is the same before and after.

System is:
-Abit KT7A v1.1
-3x 256MB Micron ECC SDRAM
-Radeon X800Pro AGP
-146GB WD Raptor on PCI-SATA-controller
-Gravis Ultrasound
-Audigy2
-ESS Solo-1
-3COM ethernet

For sure kt133a's are among the most enigmatic environment. You can get an AIO system capable to work on multi-OS in order to cover up to 10-15 years of gaming legacy at least ( and it's no coincidence that kt133a's are very very rare and difficult to retrieve today ), BUT on the other hand you get "in homage" also A LLLLLOT of hard&software issues............

dj_pirtu wrote on 2022-02-23, 08:45:

There's something strange going on with my ABIT KT7A v1.1.
<snip>
And yes, all the caps are replaced and the situation is the same before and after.

Can you provide more details on the re-cap job you did? What did you use to replace the old caps?

I just used standard electrolytic capacitors from local shop, nothing special. Bought a tin sucking machine from banggood.com which helped alot.

I got this machine about perfect today:
Modded mobile-Athlon XP 2400+ (Barton) to get multipliers working up to x18 by cutting one of the L6 bridges. (http://fab51.com/cpu/barton/athlon-e23.html)
Then with WPCREdit 55th register from 89 to 8D so KT133A can change CPU multipliers on-the-fly. Without that it just hangs.
With CrystalCPUID I can change now multipliers from 3x to 18x. With 145FSB x16 is quite spot-on, x16.5 works if voltage is raised a little.

In DOS there's SetMul -software but I still need to make that register change. So I need to use PCISet first with line: pciset 0305 1106 55 00000100 00000100 which does the same trick as WPCREdit in Windows.
And now, SetMul works and I can set the whole 3x to 18x range in DOS on-the-fly.

When FSB is 100MHz and multiplier 3x = 300MHz and I can play Hi-octane in SVGA mode with no problems!

That's awesome! Congratulation 😀 Could you elaborate on which bridge you had to cut to get to 18x? Also, perhaps ping the author of SETMUL to suggest adding the register change you mention.

My CPU is THIS:
13.5x 1800 2250 :C:C: AXMH2400FQQ4C

so counting from L6 text it's from right to left: cut, connected, cut, connected, cut

and I need:
18.0x 2400 3000 :::C:

so I cut the second last (counting from right to left) bridge and got max. multiplier up to x18

dj_pirtu wrote on 2022-02-25, 12:39:

I just used standard electrolytic capacitors from local shop, nothing special. Bought a tin sucking machine from banggood.com which helped alot.

Ah, ok, you should have used something equivalent to the originals... General purpose capacitors are inadequate, maybe that's causing your instability issues.

I got this machine about perfect today:
Then with WPCREdit 55th register from 89 to 8D so KT133A can change CPU multipliers on-the-fly. Without that it just hangs.
And now, SetMul works and I can set the whole 3x to 18x range in DOS on-the-fly.

Brilliant!!! You found the solution for why SETMUL hangs...

Just to be clear for anyone reading, the cutting he did with the bridges is not required for the multiplier changes that was just for achieving a higher multiplier (you can also do this by bridging pins on the underside of the CPU socket)...

So, just pciset 0305 1106 55 00000100 00000100, and then SETMUL after.... Fantastic. This takes KT133A to a whole new level.

This calls for a BIOS mod. I'll have to play around with it and see if I'm able to figure out how to do it (to eliminate the need for PCISET).

I have modded the AXMH2400FQQ4C successfully (CrystalCPUID reports the original multiplier as 18x) however even after changing register 55h from 89 to 8D I still get a system hang when changing the multiplier or voltage in CrystalCPUID (version 2.0). Same behaviour in DOS when changing multiplier after using PCISET. Any idea? (what BIOS version and CPU voltage do you use, btw)?

cde wrote on 2022-02-25, 17:51:

I have modded the AXMH2400FQQ4C successfully (CrystalCPUID reports the original multiplier as 18x) however even after changing register 55h from 89 to 8D I still get a system hang when changing the multiplier or voltage in CrystalCPUID (version 2.0). Same behaviour in DOS when changing multiplier after using PCISET. Any idea? (what BIOS version and CPU voltage do you use, btw)?

I will try to test this early next week to confirm.

My hypothesis is that my modern PSU, which can only output 24A on the 5V rail, is too frail. For example, I sometimes cannot boot with a voltage >= 1.5V which seems anormal. Voltage must be lowered to 1.35V for Windows XP installation to succeed.

mockingbird wrote on 2022-02-25, 15:17:

Ah, ok, you should have used something equivalent to the originals... General purpose capacitors are inadequate, maybe that's c […]
Show full quote

dj_pirtu wrote on 2022-02-25, 12:39:

I just used standard electrolytic capacitors from local shop, nothing special. Bought a tin sucking machine from banggood.com which helped alot.

Ah, ok, you should have used something equivalent to the originals... General purpose capacitors are inadequate, maybe that's causing your instability issues.

I got this machine about perfect today:
Then with WPCREdit 55th register from 89 to 8D so KT133A can change CPU multipliers on-the-fly. Without that it just hangs.
And now, SetMul works and I can set the whole 3x to 18x range in DOS on-the-fly.

Brilliant!!! You found the solution for why SETMUL hangs...

Just to be clear for anyone reading, the cutting he did with the bridges is not required for the multiplier changes that was just for achieving a higher multiplier (you can also do this by bridging pins on the underside of the CPU socket)...

So, just pciset 0305 1106 55 00000100 00000100, and then SETMUL after.... Fantastic. This takes KT133A to a whole new level.

This calls for a BIOS mod. I'll have to play around with it and see if I'm able to figure out how to do it (to eliminate the need for PCISET).

Just a question: is this trick valid only for abit or could i apply it even to my epox ( 8kta3pro+) kt133a machine? cause at this point it's gonna be wiiiiiiild...

cde wrote on 2022-02-25, 18:12:

My hypothesis is that my modern PSU, which can only output 24A on the 5V rail, is too frail. For example, I sometimes cannot boot with a voltage >= 1.5V which seems anormal. Voltage must be lowered to 1.35V for Windows XP installation to succeed.

No, I reject that... 24A on 5V is not enough? What are you running, a RAID 5 array of 15,000RPM SCSI drives?

Kaisersoze wrote on 2022-02-25, 18:24:

Just a question: is this trick valid only for abit or could i apply it even to my epox ( 8kta3pro+) kt133a machine? cause at this point it's gonna be wiiiiiiild...

Ostensibly, yes. Kind of sucks the life out of those wretched AMD K6-2 systems with the irredeemably buggy Aladdin or Apollo chipsets now, doesn't it?

I'm going to test this on both my Azza KT3-AV and my Abit KT7A 1.1 some time in the near future.

Well for 24A x 5V stated, 96W, multiply by 0.8 for on motherboard voltage conversion efficiency and there's about 77W to play with when TDP of mobile barton 2800 is 72W or 45W if you know you've got the super duper extra low wattage one, which you gave up a kidney for, so you'd know about it. The motherboard is also gonna need 20W and the drives another 20W out of the 5V. So I'd say it's coming up short.

BitWrangler wrote on 2022-02-25, 18:57:

Well for 24A x 5V stated, 96W, multiply by 0.8 for on motherboard voltage conversion efficiency and there's about 77W to play with when TDP of mobile barton 2800 is 72W or 45W if you know you've got the super duper extra low wattage one, which you gave up a kidney for, so you'd know about it. The motherboard is also gonna need 20W and the drives another 20W out of the 5V. So I'd say it's coming up short.

Fair point, but the failure mode of overdrawing from the 5V rail on a modern PSU isn't going to be system instability. It's going to be the PSU going into OCP and shutting down, or the output transistor failing completely if the OCP isn't working. And the overwhelming majority of even somewhat decent quality but low-end PSUs out there have at least a partially functioning OCP circuit.

What does cause system instability is ripple current... When the PSU output filter capacitors are not doing their job, or the motherboard CPU VRM or memory VRM capacitors are not suppressing ripple, it need not be said that the PSU with even the most capable 5V rail will still cause instability.

I can't math though 24x5 is 120W 🤣 ... but yeah, not all PSUs are perfect as you into high load area.

But I must say it's little sad that all this tweaking to get the shit out of the Abit KT7A and then you try the same CPU on MSI KT6 Delta (VIA KT600 with DDR).
With 200MHz lower CPU clocks it just destroys KT7A 😁 Performance is just in a whole new level.

But, if you want that ISA slot... And I do because of Gravis Ultrasound. But all the later VIA-chipsets are really good, 100% SB PRO support in DOS with ESS SOLO-1.

I have tried the ESS Solo-1 in a much newer system (MSI MS-7253 / K9VGM-V (KM890, VIA8237A) , Athlon 64 X2 5050e, Radeon X800 XL) that still retains DOS compatibility and can be slowed down to 386 levels. It's a really nice card coupled with the X2GS or WP32 McCake.