An interesting thing I noticed on my Asus K8V SE: as expected the Via SATA ports do not work with my SSD, but luckily it has a Promise 378 controller and that works fine with it. But as measured with ATTO, the drive speed is on the low side (IIRC it was 80 MB/s read, and about 40 write). Out of curiosity I tried the drive on one of the IDE ports with a SATA adapter, and the read/write speeds were identical. This is with Win98 so that may not be the case with another OS

shevalier wrote on 2025-10-01, 14:24:

by the way A weird and rare mut from Asrock K8A780LM https://www.asrock.com/mb/photo/K8A780LM(M1).png AMD760/710 + socket754 + […]
Show full quote

by the way
A weird and rare mut from Asrock
K8A780LM

AMD760/710 + socket754
+ problem-free SATA2
+ PCIe v2
+There are no issues with Hypertransport/PCIe/PCi frequencies.

Overall, the chipset is from a completely different era – AM3(+)

Apologies if I have missed this somewhere in the thread, but it would be awesome to get some benchmark comparisons between the Radeon HD3000 graphics on the 760G and some more common mid-range or even higher end graphics cards from the Socket 754 era. For PCI-E GPUs, the X300 (and SE), X550XT, X600 (and SE), X1300 would probably be the most interesting comparisons. If you have a somewhat similar AGP board it'd be cool to compare performance to the Radeon 9550\9600XL (64bit), 9600 Pro, Geforce 4 Ti, FX 5600 (or similar), 6200, 6600 GT, etc.

That board is such a weird thing. The chipset is from the AM2 era (some boards still used it for AM3+, but the chipset itself is from 2009), so the IGP is unusually strong compared to the speed of the CPUs it supports. Of course, DDR memory will severely limit the performance compared to later 760G boards that used DDR2 or DDR3. I built an HTPC back in ~2009 with an Athlon X2 5050e on an AMD 780G chipset board (Radeon HD3200) and just as an experiment I was able to run Crysis on it at low resolution (640x480 or 800x600 I think, 🤣) and it was totally playable. At the time this was crazy, since IGPs tended to be totally useless for the latest games. The HD3000 is noticeably slower, but it's still probably sufficient for running games from 2000-2004.

Also, the model number of the board is so funny and typically ASRock. It seems to indicate that it is an AMD 780G chipset, but it is only the 760G. Would have been cool if they'd made a 780G or even 785G board for Socket 754 as well to really push the IGP performance. 😁

I am sorry to dissapoint you but all the platforms I ran benchmarks on have Abit motherboards and are era correct: so no board with SIS chipset, no weird hybrids such as that Asrock.

Ozzuneoj wrote on 2025-10-09, 16:36:

Apologies if I have missed this somewhere in the thread, but it would be awesome to get some benchmark comparisons between the R […]
Show full quote

shevalier wrote on 2025-10-01, 14:24:

by the way A weird and rare mut from Asrock K8A780LM https://www.asrock.com/mb/photo/K8A780LM(M1).png AMD760/710 + socket754 + […]
Show full quote

by the way
A weird and rare mut from Asrock
K8A780LM

AMD760/710 + socket754
+ problem-free SATA2
+ PCIe v2
+There are no issues with Hypertransport/PCIe/PCi frequencies.

Overall, the chipset is from a completely different era – AM3(+)

Apologies if I have missed this somewhere in the thread, but it would be awesome to get some benchmark comparisons between the Radeon HD3000 graphics on the 760G and some more common mid-range or even higher end graphics cards from the Socket 754 era. For PCI-E GPUs, the X300 (and SE), X550XT, X600 (and SE), X1300 would probably be the most interesting comparisons. If you have a somewhat similar AGP board it'd be cool to compare performance to the Radeon 9550\9600XL (64bit), 9600 Pro, Geforce 4 Ti, FX 5600 (or similar), 6200, 6600 GT, etc.

That board is such a weird thing. The chipset is from the AM2 era (some boards still used it for AM3+, but the chipset itself is from 2009), so the IGP is unusually strong compared to the speed of the CPUs it supports. Of course, DDR memory will severely limit the performance compared to later 760G boards that used DDR2 or DDR3. I built an HTPC back in ~2009 with an Athlon X2 5050e on an AMD 780G chipset board (Radeon HD3200) and just as an experiment I was able to run Crysis on it at low resolution (640x480 or 800x600 I think, 🤣) and it was totally playable. At the time this was crazy, since IGPs tended to be totally useless for the latest games. The HD3000 is noticeably slower, but it's still probably sufficient for running games from 2000-2004.

Also, the model number of the board is so funny and typically ASRock. It seems to indicate that it is an AMD 780G chipset, but it is only the 760G. Would have been cool if they'd made a 780G or even 785G board for Socket 754 as well to really push the IGP performance. 😁

Socket939 got that (ASRock 939A785GMH/128M) and i would like it to be another way around because s754 got many amazing HTPC CPU options, but not the 939 socket. 24w vs 59w is big difference in low power HTPC land and there are no good one for S754 bigger brother.

59w S939
https://www.cpu-world.com/CPUs/K8/AMD-A ... NBOX).html
vs
24w S754
https://www.cpu-world.com/CPUs/K8/AMD-T ... DWOF).html

nd22 wrote on 2025-10-09, 18:04:

I am sorry to dissapoint you but all the platforms I ran benchmarks on have Abit motherboards and are era correct: so no board with SIS chipset, no weird hybrids such as that Asrock.

Oh, sorry, I thought you posted about the board because you had one to use for testing. My mistake.

Also, wasn't looking for SIS chipsets or anything strange like that. Just the AMD 760G chipset was interesting to me.

No worries. 😀

AsRock had more intresting boards, like this Socket754/Socket939/SocketAM2 hybrid,
just these future CPU S939 / AM2 add-in boards are more rare than the motherboards.

https://www.asrock.com/mb/VIA/K8Upgrade-VM800/index.asp
My K8Upgrade-VM800 Rev 1.01 board came with VT8237R Plus 😀

marxveix wrote on 2025-10-09, 21:42:

AsRock had more intresting boards, like this Socket754/Socket939/SocketAM2 hybrid, just these future CPU S939 / AM2 add-in boar […]
Show full quote

AsRock had more intresting boards, like this Socket754/Socket939/SocketAM2 hybrid,
just these future CPU S939 / AM2 add-in boards are more rare than the motherboards.

https://www.asrock.com/mb/VIA/K8Upgrade-VM800/index.asp
My K8Upgrade-VM800 Rev 1.01 board came with VT8237R Plus 😀

Yeah, those boards are really crazy. Hard to believe all the time and energy that must have gone into designing such a strange niche product. I would almost wonder the same thing about any Socket 754 or 939 board because they were so short lived... to produce a board that was meant to span those generations plus AM2... wow. Socket 754 was first available in fall of 2003, 939 was available in summer 2004 and AM2 was available in spring 2006. I wonder how many of those 754\939\AM2 upgradable boards made it to market and how many add-in CPU boards for them were sold.

Also... just to reiterate how unusually powerful the HD3000 IGP is on the AMD 760G Socket 754 board, that K8Upgrade-VM800 has a Via UniChrome Pro IGP which doesn't even have hardware T&L and likely performs somewhere in the TNT2 M64 range... despite being used on boards that run anything from Socket 754 to AM2 processors. Imagine running an FX 62 with that IGP. Oof!

Ozzuneoj wrote on 2025-10-09, 19:34:

Just the AMD 760G chipset was interesting to me.

Approximately the 760G is equivalent to a 2400 non-pro/XT.
https://www.techpowerup.com/gpu-specs/radeon-hd-2400.c2087
For Windows XP this is already a very weak solution, there are no drivers for 98.
On the positive side - 55nM technology, problem-free PCIe v2 and SATA2.
I haven't used this particular motherboard, but I have had extremely positive experiences with the 7 series chipsets.
This is much cooler than any nForce 4.

nd22 wrote on 2025-10-09, 18:04:

I am sorry to dissapoint you but all the platforms I ran benchmarks on have Abit motherboards and are era correct: so no board with SIS chipset, no weird hybrids such as that Asrock.

What a pity that the divine Abit did not make motherboards for the VIA K8T890
https://theretroweb.com/motherboards/s/asus-k8v-xe

And we will never know its strengths and weaknesses.
Although, in any case, this is exactly what an epic "battle" will look like.

And this will depend on the aggressiveness of the BIOS settings and the absence of errors.
This is not the era of Socket A/370 with a memory controller inside the chipset.

PS. Look what a wonderful document I found online.
https://www.aceshardware.com/pdfs/AcesHardware-65000298.pdf

shevalier wrote on Yesterday, 04:05:

Approximately the 760G is equivalent to a 2400 non-pro/XT. https://www.techpowerup.com/gpu-specs/radeon-hd-2400.c2087 For Window […]
Show full quote

Ozzuneoj wrote on 2025-10-09, 19:34:

Just the AMD 760G chipset was interesting to me.

Approximately the 760G is equivalent to a 2400 non-pro/XT.
https://www.techpowerup.com/gpu-specs/radeon-hd-2400.c2087
For Windows XP this is already a very weak solution, there are no drivers for 98.
On the positive side - 55nM technology, problem-free PCIe v2 and SATA2.
I haven't used this particular motherboard, but I have had extremely positive experiences with the 7 series chipsets.
This is much cooler than any nForce 4.

Of course. I would never claim that it's a high performance solution for XP gaming, but since we're talking about Socket 754, I don't think we're really shooting for absolute peak XP gaming performance either.

It would just be interesting to see how the HD3000 compared to GPUs from the time period of Socket 754 processors (2002-2004).

For example, if you were still using a Socket 754 system in 2009, your motherboard dies and you decide to just get that ASrock 760G to replace it (presumably what it was built for)... since you couldn't use your AGP card anymore (Radeon 9600, FX 5600, X600, X1300 etc.), would the HD 3000 IGP be fast enough to match your previous card, despite the single channel DDR1 limitation? Would just be neat to investigate.

Anyway, this isn't too relevant anymore since no one seems to even have this board. If I ever find one I'll surely tinker with it and post my results though! 😀

Ozzuneoj wrote on Yesterday, 05:54:

If I ever find one I'll surely tinker with it and post my results though! 😀

That would be great.
I wonder how smoothly it would work with Windows 98.
Of course, with something like the X600/800 or PCX5700.
If everything works well, it's the ultimate solution.
The Socket 754 multiplier increases the frequency from 800 to 2.4 GHz, or three times. (And this is without even changing the frequency of Hypertransport.)
256/512 MB DDR1 modules are also quite affordable.
It still amazes me that all retro gamers want the K6-3/P3S and not the S754.

s754 is also ATX and you have V12 ATX CPU connector. VIA boards have DDMA so you can use them in DOS.

I hope nd22 will consider focusing to answer questions such as:

• What is the viability of s754 with VIA chipset for Windows 98 and DOS? VIA boards should work fine, including sound in DOS via Sound Blaster Live/Audigy or ESS Solo 1. Covering DOS, Windows 98 and early Windows XP in one platform is interesting.
• Identify BIOS cut off date for 0 AGP memory VIA bug.
• What is the viability of nForce 3/4 chipset for Windows 98?
• What PCIe GPU works best in Windows 98 (as PCI device)?
• How does AGP and PCIe performance compare in s754 using the same GPU?
• What is the most powerful s754 system we can build?
• How does s754 compare to Athlon XP and s939 at the same clock speed and same GPU?

Interestingly, there was even a dual socket 754/939 board https://theretroweb.com/motherboards/s/jetway … gtdual-std-g-ec . It as both socket 754 and 939, although probably they can't both be used simultaneously.

AlexZ wrote on Yesterday, 06:59:

[*] Identify BIOS cut off date for 0 AGP memory VIA bug.

There are two revisions of the K8T800—"A" and "B."

The older chip has a metal heat spreader, with a Vcore of 2.5 V. (=V_ram)
The newer chip is entirely plastic, with a Vcore of 1.5 V. (=V_AGP)

As far as I know, this issue was not detected in the older revision.

shevalier wrote on Yesterday, 07:25:

The older chip has a metal heat spreader, with a Vcore of 2.5 V. (=V_ram)
The newer chip is entirely plastic, with a Vcore of 1.5 V. (=V_AGP)

As far as I know, this issue was not detected in the older revision.

It's been reported that using older BIOS can fix the issue.

AlexZ wrote on Yesterday, 08:05:

shevalier wrote on Yesterday, 07:25:

The older chip has a metal heat spreader, with a Vcore of 2.5 V. (=V_ram)
The newer chip is entirely plastic, with a Vcore of 1.5 V. (=V_AGP)

As far as I know, this issue was not detected in the older revision.

It's been reported that using older BIOS can fix the issue.

I finally remembered where this discussion started.
It said that BIOS rollback doesn't seem to help on revision B.
This post contains photos of different revisions, but bloodem was mistaken in thinking that the one with metal was a new revision.
Re: Not Another Ultimate Windows 98 Build

PS. The amount of heat generated depends on the square of the voltage, not linearly.
Therefore, the older version has a metal heat spreader, like the PM800 (based on the B revision, but with an integrated graphics core).
Revision A produces 2.7 times more heat then B.

shevalier wrote on Yesterday, 04:15:

What a pity that the divine Abit did not make motherboards for the VIA K8T890 https://theretroweb.com/motherboards/s/asus-k8v-xe […]
Show full quote

nd22 wrote on 2025-10-09, 18:04:

I am sorry to dissapoint you but all the platforms I ran benchmarks on have Abit motherboards and are era correct: so no board with SIS chipset, no weird hybrids such as that Asrock.

What a pity that the divine Abit did not make motherboards for the VIA K8T890
https://theretroweb.com/motherboards/s/asus-k8v-xe

And we will never know its strengths and weaknesses.
Although, in any case, this is exactly what an epic "battle" will look like.

And this will depend on the aggressiveness of the BIOS settings and the absence of errors.
This is not the era of Socket A/370 with a memory controller inside the chipset.

PS. Look what a wonderful document I found online.
https://www.aceshardware.com/pdfs/AcesHardware-65000298.pdf

1. Unlike Asus, Abit was a small company focused on enthousiasts with no/very limited OEM presence and did not manufactured boards with every single chipset.
2. I have read the review on hothardware and i do not agree with one of the basic principle they used: testing at low resolution and settings. You will not see that once I put ut the screenshots.
3. I have that document since december 2024 when I gathered all the components and started preliminary tests. I was going to bring it up after I put all the results on the forum.

nd22 wrote on Yesterday, 11:04:

1. Unlike Asus, Abit was a small company focused on enthousiasts with no/very limited OEM presence and did not manufactured boards with every single chipset.

Pf... I hate Asus.
The best motherboards for AMD were made by Epox.
Let's start with the implementation of the BusDisconnect function for Athlon XP.
And I still prefer Soltek, although I don’t have their motherboards at the moment.

If you see different results, the problem is in the BIOS.
The AGP interface was thoroughly tested on the nForce 2/KT600.
The RAM controller was moved to the processor.
The performance difference is solely due to an unoptimized BIOS.

AlexZ wrote on Yesterday, 06:59:

s754 is also ATX and you have V12 ATX CPU connector. VIA boards have DDMA so you can use them in DOS. […]
Show full quote

s754 is also ATX and you have V12 ATX CPU connector. VIA boards have DDMA so you can use them in DOS.

I hope nd22 will consider focusing to answer questions such as:

• What is the viability of s754 with VIA chipset for Windows 98 and DOS? VIA boards should work fine, including sound in DOS via Sound Blaster Live/Audigy or ESS Solo 1. Covering DOS, Windows 98 and early Windows XP in one platform is interesting.
• Identify BIOS cut off date for 0 AGP memory VIA bug.
• What is the viability of nForce 3/4 chipset for Windows 98?
• What PCIe GPU works best in Windows 98 (as PCI device)?
• How does AGP and PCIe performance compare in s754 using the same GPU?
• What is the most powerful s754 system we can build?
• How does s754 compare to Athlon XP and s939 at the same clock speed and same GPU?

Interestingly, there was even a dual socket 754/939 board https://theretroweb.com/motherboards/s/jetway … gtdual-std-g-ec . It as both socket 754 and 939, although probably they can't both be used simultaneously.

1. I have test installed windows me - yes, I know i commited a cardinal sin - on all the platforms but have not done any tests except 3dmark 2001. I know I am going to get a lot of hate for saying it but I believe socket 754 works best with XP. On all AGP platforms ME installed just fine, on NV8 it crashed during setup. I gave up after 1 day.
2. On all abit boards I did not encountered that issue.
3. I do not believe nforce4 is a good candidate for windows 98, at least I expect to get windows installed not to fight it at every step. Original ISO from winworld used without any modifications.
4. I believe that for windows 9X AGP and not PCI-express is the way to go - expecting a lot of hate for this!
5. My tests will answer that.
6. Again wait for the results.

shevalier wrote on Yesterday, 11:19:

Pf... I hate Asus. The best motherboards for AMD were made by Epox. Let's start with the implementation of the BusDisconnect fun […]
Show full quote

nd22 wrote on Yesterday, 11:04:

1. Unlike Asus, Abit was a small company focused on enthousiasts with no/very limited OEM presence and did not manufactured boards with every single chipset.

Pf... I hate Asus.
The best motherboards for AMD were made by Epox.
Let's start with the implementation of the BusDisconnect function for Athlon XP.
And I still prefer Soltek, although I don’t have their motherboards at the moment.

DFI also made great boards, at least with the LanParty series.

I still have my Abit NF7-S 2.0 which has been used in the same PC with the same overclocked XP 1700+ with the same cooler and fan for something like 23 years now. I use it regularly for testing AGP cards still.

I also have an EPoX 9NDA3J, Socket 939 with the Nforce 3 Ultra (AGP) chipset. This was the board that I replaced the NF7-S with when I upgraded back in the day. I originally ran it with an A64 3000+, later used the board in a PC I sold to someone and 10+ years later they asked me for a new PC and gave me this one back, still in 100% working order! Now it has the beta BIOS installed I have it running an X2 4200+.

The board I purchased after the EPoX was a DFI LANParty UT nF4 Ultra-D. Wish I still had that, but it went into another person's system and I'm sure it's long gone. Very likely it ended up with bad caps. So far the Epox and Abit boards are rock solid with no signs of cap issues. 😀

Prior to these boards I had Tyan KX133 (Slot A), ECS K7S5A and a Gigabyte KT333 board. Still have the ECS too, but it needs lots of caps replaced (I did some 12+ years ago, but it needs more now).

Anyway, the EPoX and Abit boards have definitely earned a good name with me. Too bad both companies are no longer around.

Ozzuneoj wrote on Today, 02:25:

DFI also made great boards, at least with the LanParty series. […]
Show full quote

shevalier wrote on Yesterday, 11:19:

Pf... I hate Asus. The best motherboards for AMD were made by Epox. Let's start with the implementation of the BusDisconnect fun […]
Show full quote

nd22 wrote on Yesterday, 11:04:

1. Unlike Asus, Abit was a small company focused on enthousiasts with no/very limited OEM presence and did not manufactured boards with every single chipset.

Pf... I hate Asus.
The best motherboards for AMD were made by Epox.
Let's start with the implementation of the BusDisconnect function for Athlon XP.
And I still prefer Soltek, although I don’t have their motherboards at the moment.

DFI also made great boards, at least with the LanParty series.

I still have my Abit NF7-S 2.0 which has been used in the same PC with the same overclocked XP 1700+ with the same cooler and fan for something like 23 years now. I use it regularly for testing AGP cards still.

I also have an EPoX 9NDA3J, Socket 939 with the Nforce 3 Ultra (AGP) chipset. This was the board that I replaced the NF7-S with when I upgraded back in the day. I originally ran it with an A64 3000+, later used the board in a PC I sold to someone and 10+ years later they asked me for a new PC and gave me this one back, still in 100% working order! Now it has the beta BIOS installed I have it running an X2 4200+.

The board I purchased after the EPoX was a DFI LANParty UT nF4 Ultra-D. Wish I still had that, but it went into another person's system and I'm sure it's long gone. Very likely it ended up with bad caps. So far the Epox and Abit boards are rock solid with no signs of cap issues. 😀

Prior to these boards I had Tyan KX133 (Slot A), ECS K7S5A and a Gigabyte KT333 board. Still have the ECS too, but it needs lots of caps replaced (I did some 12+ years ago, but it needs more now).

Anyway, the EPoX and Abit boards have definitely earned a good name with me. Too bad both companies are no longer around.

Ultra-D and NF7-S are my favorites from back then. The Ultra-D was insane running the memory off the 3volt rail. You'd modify the rail to overvolt to 3.3v with some Winbond bh5 cas 2,2,2,4 1t @ 500mhz... Good times.