Archer57 wrote on 2025-08-12, 09:20:

And no devices are detected.

My bad.
It's not "R" that's needed, it's "R plus". VIA is really making me nervous with these chips naming.

It seems to me that now this SATA-1 in the chipset is not needed at all.
VT8237 supports ATA133, on which any SSD via an ATA-SSD adapter will work no slower then SATA-150, and without any problems.

You should be using identical memory sticks otherwise you can run into SPD being read from wrong stick. It cannot set different timings for different sticks so it has to choose one.

I'm not surprised that 4 memory sticks don't want to work. It probably isn't stable at DDR400 and resets to DDR333 automatically. Just use 2 sticks, similar problem to what I encountered with AM2 except it was stable with 4 sticks at DDR2 800 there. Maybe DDR needs slightly more voltage at 400 speed.

Use different SATA cable and/or cap the drive at SATA-1.

See https://www.pcgamingwiki.com/wiki/List_of_gam … t-in_benchmarks

Some of these games have also dedicated downloadable benchmarks that are free and much smaller. Also you can test on games you wouldn't normally play on s939 such as Crisis.

shevalier wrote on 2025-08-12, 10:19:

My bad. It's not "R" that's needed, it's "R plus". VIA is really making me nervous with these chips naming. […]
Show full quote

My bad.
It's not "R" that's needed, it's "R plus". VIA is really making me nervous with these chips naming.

It seems to me that now this SATA-1 in the chipset is not needed at all.
VT8237 supports ATA133, on which any SSD via an ATA-SSD adapter will work no slower then SATA-150, and without any problems.

Yeah, ide-sata works fine and that's what i am using. But that's extra adapter i need to buy, it is a shame i can not just use sata ports which are there.

Speeds... at least on nforce2 my experience was that sata1 is ~20-30% faster than ATA133 for sequential operations. Absolutely not critical, but still faster.

AlexZ wrote on 2025-08-12, 10:46:

You should be using identical memory sticks otherwise you can run into SPD being read from wrong stick. It cannot set different timings for different sticks so it has to choose one.

Yeah, i should. However if i am setting everything manually it should not matter as long as all sticks work at configured settings. The issue with this board i am complaining about is that it ignores my manual settings and does... something. If it is not stable i want it to fail or have memory errors, not try to be "smart". That's how things usually work, even on modern hardware.

AlexZ wrote on 2025-08-12, 10:46:

I'm not surprised that 4 memory sticks don't want to work. It probably isn't stable at DDR400 and resets to DDR333 automatically. Just use 2 sticks, similar problem to what I encountered with AM2 except it was stable with 4 sticks at DDR2 800 there. Maybe DDR needs slightly more voltage at 400 speed.

The same CPU on different board works with 4 sticks though, so this are either issues with board layout or BIOS being difficult. I'll just use 2 sticks obviously, still annoying...

Voltage... 2.6-2.7 does not help, do not want to go further given how... uncommon this CPUs are. This 4800+ is a lot more pricey than even 6400+ windsor...

AlexZ wrote on 2025-08-12, 10:46:

Use different SATA cable and/or cap the drive at SATA-1.

Its a known VIA chipset bug, cable is fine, needs sata1 drive to work. But i want ssd, so have to just use ide-sata adapter...

AlexZ wrote on 2025-08-12, 10:46:

See https://www.pcgamingwiki.com/wiki/List_of_gam … t-in_benchmarks

Some of these games have also dedicated downloadable benchmarks that are free and much smaller. Also you can test on games you wouldn't normally play on s939 such as Crisis.

Thanks, i'll look into this.

Crysis... should probably work on this hardware with something like HD3850, but older cards will show single digit FPS, not sure how useful that is...

Sometimes booting into Windows with 4 sticks may be possible, but memtest still reveals lot of memory errors. They just happen to be at addresses not used initially. This is how I got initially optimistic about Phenom II and DDR2 1066 with 4 sticks but it turned out not to be stable.

Radeon HD 2600 XT is from 2007, so it may be able to play Crisis, perhaps not with max settings. Athlon 64 X2 4800+ should be able to score about 30 fps in dx9 crisis benchmark with a capable GPU. The actual game then runs better as long as nothing is happening. It would be interesting to see what a good AGP card can do in Crysis, including 7600 GT.

Archer57 wrote on 2025-08-12, 11:42:

shevalier wrote on 2025-08-12, 10:19:

My bad. It's not "R" that's needed, it's "R plus". VIA is really making me nervous with these chips naming. […]
Show full quote

My bad.
It's not "R" that's needed, it's "R plus". VIA is really making me nervous with these chips naming.

It seems to me that now this SATA-1 in the chipset is not needed at all.
VT8237 supports ATA133, on which any SSD via an ATA-SSD adapter will work no slower then SATA-150, and without any problems.

Yeah, ide-sata works fine and that's what i am using. But that's extra adapter i need to buy, it is a shame i can not just use sata ports which are there.

Speeds... at least on nforce2 my experience was that sata1 is ~20-30% faster than ATA133 for sequential operations. Absolutely not critical, but still faster.

For older SendForce controllers, there is a utility that switches SSDs to SATA-1/2/3 modes.
It definitely works with Intel 330 series SATA-3. nForce Crash51 Socket754 (Ga-k8n51gmf) began to detect it as SATA-2 instead of SATA-1.
The ancient Kingston 60Gb will arrive in a few days, I will compare it at VIA VT8237 (JetWay K8T8AS).

Archer57 wrote on 2025-08-12, 11:42:

AlexZ wrote on 2025-08-12, 10:46:

See https://www.pcgamingwiki.com/wiki/List_of_gam … t-in_benchmarks

Some of these games have also dedicated downloadable benchmarks that are free and much smaller. Also you can test on games you wouldn't normally play on s939 such as Crisis.

Thanks, i'll look into this.

Crysis... should probably work on this hardware with something like HD3850, but older cards will show single digit FPS, not sure how useful that is...

https://www.techpowerup.com/download/benchmarking/
https://www.guru3d.com/files/category/benchma … s-demos/page-3/

should also be useful

Archer57 wrote on 2025-05-08, 10:28:

Honestly? I have the same impression about all AMD cards (got RX5700XT when it was new, that thing turned perfectly working system into a mess which could not even run a browser without crashing). And from my time with that card and discussions about it on some local forums i came to conclusion that whole "bad drivers" thing may be an excuse for faulty hardware to avoid RMAs. But i want to give it a honest try.

You're kidding right? I have a 5700XT Anniversary Edition - bought it new. Worked flawlessly then, and still does in my sister's PC, 6 years later. There must have been something wrong with your sistem.

I never had an issue with AMD video cards or drivers from the HD7000 series up to today (running a 6900XT), and I've setup countless builds in the last 14 years. On the other hand, I've had plenty of issues with nvidia drivers, like the WHQL certified Geforce 196.75 drivers stop some graphics card fans from spinning. Killed my GTX 260 while playing bloody wow. Current nvidia drivers are lovely. Google "nvidia black screen issue".

Archer57 wrote on 2025-05-08, 09:35:

So... i've been playing around with this card. I do think the card is not faulty - it runs certain games, even relatively new on […]
Show full quote

So... i've been playing around with this card. I do think the card is not faulty - it runs certain games, even relatively new ones, completely fine for large periods of time. The card is from sapphire.

However, in typical AMD fashion, it is unstable as hell. Windows may crash (typically 8E) when changing resolution or starting certain tests in certain versions of 3D mark. Like it can not finish 2001SE or 2003, but does finish 2005.

Obviously regular drivers do not work. I've tried searching around and while i've seen performance tests with different drivers what interests me more is stability. I do not want BSODs, crashes, etc.

I also had a hard time finding a place to download any of this drivers at all, which is not a sketchy "driver pack" type of thing or torrents. Ended up downloading 8.5 driver from HIS website (sapphire themselves are... one of those "nice" manufacturers which remove old drivers) and it works, but is unstable.

This is not related to the system itself - tried in different systems and they all work just fine with other cards (like 6600).

Would be grateful for any advice...

You need to use the drivers your manufacturer supplies for the card. I have a 2600XT AGP and a 1650XT AGP - none of them work correctly with generic AMD Catalyst drivers. For my 2600XT I used the drivers off the HIS website - version "XP_8.401-070727a-050964C-ATI with WHQL". Test system was a Pentium 661 + Abit AS8-V, XP32 and 2GB of ram. Tried Far Cry (1), Doom 3, Quake 4, Company of Heroes and C&C3.

AMD never bothered properly supporting the AGP versions of 2xxx cards. I'm not 100% on this, but I remember all AGP versions of the 2600 are based on a beta driver witch is then modified and re-packaged by each card's manufacturer. It's possible some manufactures did a better job then others with these drivers.

Socket3 wrote on 2025-08-16, 20:17:

You're kidding right? I have a 5700XT Anniversary Edition - bought it new. Worked flawlessly then, and still does in my sister's PC, 6 years later. There must have been something wrong with your sistem.

Nothing wrong with the system. It worked just fine with RX580 for a few years before and still works fine with RTX2080s which replaced 5700XT back then. I never wanted to buy 2080, i thought it was not a good value, but AMD forced me...

And yeah, i am not going to buy another new AMD card for a long while after that experiment. Enough is enough. Good thing intel cards now exist so i can safely say i am not an nvidia fan 😁

Socket3 wrote on 2025-08-16, 20:17:

I never had an issue with AMD video cards or drivers from the HD7000 series up to today (running a 6900XT), and I've setup countless builds in the last 14 years. On the other hand, I've had plenty of issues with nvidia drivers, like the WHQL certified Geforce 196.75 drivers stop some graphics card fans from spinning. Killed my GTX 260 while playing bloody wow. Current nvidia drivers are lovely. Google "nvidia black screen issue".

Yep, nvidia made some fun mistakes too. Another example was their drivers deleting current user's home folder on linux.

That's the reason i update drivers like once a year at most. If you update to each new version you are guaranteed to experience all the bugs. If you skip most updates you skip most bugs. I was using nvidia card when they screwed up with fans, never got the faulty version...

Socket3 wrote on 2025-08-16, 20:17:

You need to use the drivers your manufacturer supplies for the card. I have a 2600XT AGP and a 1650XT AGP - none of them work correctly with generic AMD Catalyst drivers. For my 2600XT I used the drivers off the HIS website - version "XP_8.401-070727a-050964C-ATI with WHQL". Test system was a Pentium 661 + Abit AS8-V, XP32 and 2GB of ram. Tried Far Cry (1), Doom 3, Quake 4, Company of Heroes and C&C3.

AMD never bothered properly supporting the AGP versions of 2xxx cards. I'm not 100% on this, but I remember all AGP versions of the 2600 are based on a beta driver witch is then modified and re-packaged by each card's manufacturer. It's possible some manufactures did a better job then others with these drivers.

If you look through the thread further - a lot has changed since me posting it. Short summary: nothing to do with drivers, everything to do with hardware level incompatibility. I still blame AMD though, because when you have a system (or multiple systems) which works perfectly with every other card including old ATI radeons but then does not work with AMD HD**** stuff the conclusion is pretty obvious.

Also yes, AGP cards work just fine with generic drivers, the only issue and the only "fix" card manufacturers did is the fact that device ID is different and installer does not "detect" the card. There are a few ways around that.

I've since tried the card on asus a8v and it works fine with any driver, apart from various driver bugs, but that's normal.

That said AMD never bothering to make drivers work for AGP cards also shows pretty well how much they care about their customers.

Archer57 wrote on 2025-05-08, 10:28:

Honestly? I have the same impression about all AMD cards (got RX5700XT when it was new, that thing turned perfectly working system into a mess which could not even run a browser without crashing). And from my time with that card and discussions about it on some local forums i came to conclusion that whole "bad drivers" thing may be an excuse for faulty hardware to avoid RMAs. But i want to give it a honest try.

Oh, I had the same problem when I switched from a perfectly stable HD7870 to a RX580.
All tests and games run fine, but as soon as you turn on video playback - a black screen.
As always, the drivers PSU are to blame.
But it's better to write right away that the drivers. Then everyone will nod understandingly.

shevalier wrote on 2025-08-17, 04:10:

Oh, I had the same problem when I switched from a perfectly stable HD7870 to a RX580.
All tests and games run fine, but as soon as you turn on video playback - a black screen.

There is actually a funny bug related to video decoding on AMD cards, not sure when state it is in now but back when i had RX580 and that ill-fated 5700XT it was a thing.

Trying to decode a video within allowed specs in terms of resolution/codec, but way, way too high bitrate caused something to crash triggering VPU recover, and if done a few times in a row - BSOD. Admittedly such video is not easy to come by, essentially has to be specifically made, but still - this is basically a DOS attack on computers with AMD graphics which can be performed through a simple web page.

Nvidia or intel refuses to decode such stuff entirely.

Realistically though i have no idea what the issue was, other than logical conclusion that if multiple cards work without issues in the same system but one does not - the card is the issue.

shevalier wrote on 2025-08-17, 04:10:

As always, the drivers PSU are to blame.
But it's better to write right away that the drivers. Then everyone will nod understandingly.

Yep, easy to jump to "drivers" conclusion and waste time. I am obviously guilty of that...

PSU... perhaps, though very unlikely in my case. PSUs are quite overpowered, modern and generally chosen to be high quality from reputable manufacturers. Also 2080s needs more than 5700XT.

Even more so on this old stuff the thread is about, because the PSU used is modern and likely significantly better than anything available back then. Just to be sure i tried multiple too...

Archer57 wrote on 2025-08-17, 05:26:

Trying to decode a video within allowed specs in terms of resolution/codec, but way, way too high bitrate caused something to crash triggering VPU recover,

All more or less modern equipment has very aggressive power saving settings.
That is, it consumes nothing, and then instantly produces 100500 W.
Have you noticed in reviews of modern power supplies a significant increase in capacity in secondary circuits at the same power?
Here, literally for 650 W models with 5000 μF (2 * 2200 + 560 polymer), up to a wild mix with a total capacity of 10,000+ μF.

shevalier wrote on 2025-08-14, 07:13:

For older SendForce controllers, there is a utility that switches SSDs to SATA-1/2/3 modes. It definitely works with Intel 330 s […]
Show full quote

Archer57 wrote on 2025-08-12, 11:42:

Yeah, ide-sata works fine and that's what i am using. But that's extra adapter i need to buy, it is a shame i can not just use sata ports which are there.

Speeds... at least on nforce2 my experience was that sata1 is ~20-30% faster than ATA133 for sequential operations. Absolutely not critical, but still faster.

For older SendForce controllers, there is a utility that switches SSDs to SATA-1/2/3 modes.
It definitely works with Intel 330 series SATA-3. nForce Crash51 Socket754 (Ga-k8n51gmf) began to detect it as SATA-2 instead of SATA-1.
The ancient Kingston 60Gb will arrive in a few days, I will compare it at VIA VT8237 (JetWay K8T8AS).

Yes, native SATA looks preferable.
The first two screenshots are - J3455M is a quad-core Celeron-class SoC on DDR3

DUT is
Model: KINGSTON SKC300S37A60G
Fw : 605ABBF0
Size : 57241 MB
Toshiba 19nm MLC 16k 64Gb/CE 64Gb

shevalier wrote on 2025-08-17, 06:01:

Yes, native SATA looks preferable. The first two screenshots are - J3455M is a quad-core Celeron-class SoC on DDR3 […]
Show full quote

shevalier wrote on 2025-08-14, 07:13:

For older SendForce controllers, there is a utility that switches SSDs to SATA-1/2/3 modes. It definitely works with Intel 330 s […]
Show full quote

Archer57 wrote on 2025-08-12, 11:42:

Yeah, ide-sata works fine and that's what i am using. But that's extra adapter i need to buy, it is a shame i can not just use sata ports which are there.

Speeds... at least on nforce2 my experience was that sata1 is ~20-30% faster than ATA133 for sequential operations. Absolutely not critical, but still faster.

For older SendForce controllers, there is a utility that switches SSDs to SATA-1/2/3 modes.
It definitely works with Intel 330 series SATA-3. nForce Crash51 Socket754 (Ga-k8n51gmf) began to detect it as SATA-2 instead of SATA-1.
The ancient Kingston 60Gb will arrive in a few days, I will compare it at VIA VT8237 (JetWay K8T8AS).

Yes, native SATA looks preferable.
The first two screenshots are - J3455M is a quad-core Celeron-class SoC on DDR3

DUT is
Model: KINGSTON SKC300S37A60G
Fw : 605ABBF0
Size : 57241 MB
Toshiba 19nm MLC 16k 64Gb/CE 64Gb

It is curious to see random 4k being so much lower on more modern platform, it is the most important metric after all, especially for old OS which were not designed with queued IO in mind.

Other than that yes, this is ~similar to what i've seen on nforce2/MCP2-S/R. SATA is a little bit faster than IDE-SATA, not critically so, but noticeably, and old SATA-PCI controllers like SIL3112 - significantly worse than both.

You said there is a way to switch SF-2281 SSDs to SATA1? That would be useful for this old boards with VIA chipsets...

https://drivers.dosreloaded.de/index.php?b=Gr … 2FHD3xxx_HD4xxx
This driver usually works fine for me.

Try using Rivatuner or something and disable SBA, Fastwrites etc for Testing.
Be sure your PSU is fine.

if your ATI 9800 already does lock up, something in your configuration is weird/broken.
Could be also your mainboard has bad capacitors.

Archer57 wrote on 2025-08-16, 21:20:

Socket3 wrote on 2025-08-16, 20:17:

You're kidding right? I have a 5700XT Anniversary Edition - bought it new. Worked flawlessly then, and still does in my sister's PC, 6 years later. There must have been something wrong with your sistem.

Nothing wrong with the system. It worked just fine with RX580 for a few years before and still works fine with RTX2080s which replaced 5700XT back then. I never wanted to buy 2080, i thought it was not a good value, but AMD forced me...

Must have been a bad card then...

Socket3 wrote on 2025-08-17, 16:14:

Archer57 wrote on 2025-08-16, 21:20:

Socket3 wrote on 2025-08-16, 20:17:

You're kidding right? I have a 5700XT Anniversary Edition - bought it new. Worked flawlessly then, and still does in my sister's PC, 6 years later. There must have been something wrong with your sistem.

Nothing wrong with the system. It worked just fine with RX580 for a few years before and still works fine with RTX2080s which replaced 5700XT back then. I never wanted to buy 2080, i thought it was not a good value, but AMD forced me...

Must have been a bad card then...

The cause could be anything traditionally associated with AMD video drivers.
For example, a crash in games with a large number of small units.

The episode with the Tyranids in Space Marine 2 or the attack on the Titans in Serious Sam 4.
The game crashed on an AMD processor under Windows 11.
Only one test revealed a problem with the CACHE - AMD Overdrive Utilities.
Cause.
Microsoft was experimenting with distributing tasks across cores.
A new major Windows update has been released - the problem has disappeared.

PS. VPU recovery with black screen - usually these are most often errors in the system memory, during driver execution area.

shevalier wrote on 2025-08-17, 16:29:

PS. VPU recovery with black screen - usually these are most often errors in the system memory, during driver execution area.

I've recently experienced this on 2 different genuinely dead cards - freeze, black screen, in 10-20 seconds popup about VPU recover and crashed game. So it can be caused by different things...

What's amusing is how many attempts it manages to survive/recover, at least this mechanism is working well...

Also those system with 5700XT is a workstation. So ECC, it can not have memory errors. They'll either be correctable/corrected or uncorrectable/detected and it'll BSOD immediately.

Not sure what the issue was, i just got very, very pissed off and replaced the card. Unlike this old stuff i can fool around with for months i need that system to work and be reliable, not randomly crash when i open a browser or something.

Archer57 wrote on 2025-08-17, 17:41:

Not sure what the issue was, i just got very, very pissed off and replaced the card. Unlike this old stuff i can fool around with for months i need that system to work and be reliable, not randomly crash when i open a browser or something.

Things are really bad now.
Chips are tested in very extreme conditions.
For example, 9001 GHz is unstable, 9000 is stable.
They classify it as 9000.
This applies to everyone now: Intel, AMD, NVIDIA.
So a seriously broken video card/processor from any manufacturer is a given.
2% of defects are simply built into the business model.
I want to say that blaming everything on "AMD drivers" is just mauvais.
It either works or it doesn't.
Sometimes the reason is not obvious.
Sometimes the hardware buggy itself.

And I will whisper sedition.
Asus can do nothing - to be or to seem, the choice is obvious.
AMD video cards - Sapphire, AIB manufacturer №1.
I apologize for offending someone's religious feelings.

shevalier wrote on 2025-08-17, 16:29:

The cause could be anything traditionally associated with AMD video drivers. For example, a crash in games with a large number o […]
Show full quote

Socket3 wrote on 2025-08-17, 16:14:

Archer57 wrote on 2025-08-16, 21:20:

Nothing wrong with the system. It worked just fine with RX580 for a few years before and still works fine with RTX2080s which replaced 5700XT back then. I never wanted to buy 2080, i thought it was not a good value, but AMD forced me...

Must have been a bad card then...

The cause could be anything traditionally associated with AMD video drivers.
For example, a crash in games with a large number of small units.

The episode with the Tyranids in Space Marine 2 or the attack on the Titans in Serious Sam 4.
The game crashed on an AMD processor under Windows 11.
Only one test revealed a problem with the CACHE - AMD Overdrive Utilities.
Cause.
Microsoft was experimenting with distributing tasks across cores.
A new major Windows update has been released - the problem has disappeared.

PS. VPU recovery with black screen - usually these are most often errors in the system memory, during driver execution area.

Serious Sam 4 is a buggy mess. As for SM2, I completed the campaign at launch on my main pc - Ryzen 5700x3d, Gigabyte Aorus B550M Elite, 16GB of corsair vengeance LPX @ 4600MHz CL18 and my Powercolor 6900XT Red Devil. No crashes or stuttering of any kind. In fact I've never had a hitch with this build. Ever - and it's housed in a compact enclosure. 1200w Corsair 1200hx, Chieftec CI-02B-OP, Thermalright "Le Grande Macho" RT. It stared out as a 5700g with no dGPU in 2020, then it got a RTX 3060ti, then it got a PSU upgrade, and finally the configuration is has now, with fast ram, x3d CPU and the 6900xt. I'm also running the same install of windows 10 I put on it in 2020. Best PC I've owned since my i7-3930K.

I don't upgrade graphics drivers too often. I'd say I do it once every 6 months or so. Stil rocking windows 10, and I'll be doing so until steam refuses to run on it, as it did with windows 7. Then I'm thinking of switching to Linux Mint. I've been running Mint Cinnamon on my PC at work as well as my bedroom PC, and I've been messing around with it on several other machines, including my laptop. The only thing I'll miss is Battle.net.

shevalier wrote on 2025-08-17, 17:55:

AMD video cards - Sapphire, AIB manufacturer №1.
I apologize for offending someone's religious feelings.

Mine was a sapphire card 😀 As is HD2600XT from OP.

Nowadays they make decent cards. I do hate them for their old coolers on X and early HD series though, those were not good for multiple reasons.