VOGONS


PIII-s on a 694T: mysterious memory errors

Topic actions

First post, by tauro

User metadata
Rank Member
Rank
Member

Hello guys,

I'm putting together a Socket 370 system, aiming at the best from that era.
The Motherboard is a MSI 694T Pro VER 5. The CPU is a PIII-s (1.4 GHz 512KB L2 Cache)

All CPUs I tried on this board work perfectly, Coppermines, Celeron Tualatins, but... if I use the PIII-s the system posts, the processor is recognized correctly. It even boots DOS, but the memory is not working well at all.
Speedsys shows "Failed" after testing the extended memory, Memtest86 shows lots of errors, I could go on.

Interestingly, all the other CPUs work fine with my RAM (and I also tried many other RAMs, also different sizes/slots with the same result).

I tried this very same CPU (PIII-s) on a BIOSTAR M6VLR with a 8601T chipset, and... everything works fine.
(This chipset is also detected as PLE133T, by the way).

I also used this CPU on a 815e modded motherboard, and it works perfectly, but there's that crippling 512mb limitation which is a bummer for me, I was looking to triple boot 98, XP and some GNU/Linux distro. That's why I'm inclined to the PLE133T chipset.

So, what else could I try? I'd really like to use the MSI 694T Pro.

I already flashed the latest BIOS, checksum is ok. I'm really puzzled.

Did you ever experience something similar? with this processor? with this motherboard? Any ideas?

Reply 1 of 24, by AlexZ

User metadata
Rank Member
Rank
Member

My bet would be some kind of incompatibility between board and RAM, BIOS setting or bad caps. I would try to set FSB to 100Mhz and CL3 RAM timings. Maybe the board needs a recap.

Pentium III 900E, ECS P6BXT-A+, 384MB RAM, NVIDIA GeForce FX 5600 128MB, Voodoo 2 12MB, 80GB HDD, Yamaha SM718 ISA, 19" AOC 9GlrA
Athlon 64 3400+, MSI K8T Neo V, 1GB RAM, NVIDIA GeForce 7600GT 512MB, 250GB HDD, Sound Blaster Audigy 2 ZS

Reply 2 of 24, by flupke11

User metadata
Rank Oldbie
Rank
Oldbie

Try the board with an overclocked 1GHz Coppermine or an overclocked 1,3 Celeron to stress the VRM and the caps, if the errors return, you've found the culprit. Even if they look ok, the caps might be shot.

Reply 3 of 24, by tauro

User metadata
Rank Member
Rank
Member

Hey guys, it's been a tortuous experience trying to get the PIII-s working with this board.

I tried everything I could. I tried downgrading the BIOS.

I grabbed all the BIOSes from MSI's website, they are as follows
7.0, 7.1, 7.2, 7.3, 7.4, 7.5

There was some encouraging progress with the older BIOSes (up to v7.3), as Speedsys would show "PASSED" when testing the memory. But sadly, no "serious" system would boot. Memtest86 showed lots of errors non-stop (as I mentioned before, if you switch the CPU and keep the same ram, there are no errors).

Then, I removed all the caps, measured them, replaced a couple that I didn't like, but overall all the caps were ok. This took me lots of time.

Turn it on again and... no changes.

Later I found this post here on Vogons which mentions the 694T, someone asking if it is compatible with the PIII-s, which lead me to this cpu-world page where they list all the compatible processors with this board.

Surprise surprise... it's NOT supported 🤡

So... lucky me! I wish I would have known this a couple of days ago.

Anyway, I'd love to know if anybody has solved this and got the PIII-s 1.4 GHz to work with the 694T Pro Ver. 5. Maybe there's some mod that needs to be done. Some guy on the cpu-world page I linked above says that the PIII-S 1.4 GHz is supported with the latest BIOS, which is partially true. It posts, but there are memory issues.
I'm sure some BIOS mod would solve this, because all the other chips work. And I saw some progress in DOS with the earlier BIOSes.

I hope this helps the next unlucky person trying to couple these two pieces of hardware together.

Reply 4 of 24, by AlexZ

User metadata
Rank Member
Rank
Member

694T should support Tualatins. I find it hard to believe that would mean only the 256kb cache versions without -S.

Pentium III 900E, ECS P6BXT-A+, 384MB RAM, NVIDIA GeForce FX 5600 128MB, Voodoo 2 12MB, 80GB HDD, Yamaha SM718 ISA, 19" AOC 9GlrA
Athlon 64 3400+, MSI K8T Neo V, 1GB RAM, NVIDIA GeForce 7600GT 512MB, 250GB HDD, Sound Blaster Audigy 2 ZS

Reply 5 of 24, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
AlexZ wrote on 2022-04-09, 20:45:

694T should support Tualatins. I find it hard to believe that would mean only the 256kb cache versions without -S.

I have seen this before, certain boards (officially) did not support the Tualatin with 512KB cache.

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 6 of 24, by shamino

User metadata
Rank l33t
Rank
l33t

The Shuttle AV18 (AV8? whatever it is) doesn't list support for the P3-S either. There's probably other boards with the same limitation. The reason is probably the onboard power circuitry not being able to handle the power draw of those CPUs.

Reply 7 of 24, by tauro

User metadata
Rank Member
Rank
Member
Tetrium wrote on 2022-04-09, 22:52:

I have seen this before, certain boards (officially) did not support the Tualatin with 512KB cache.

It's quite puzzling, because this board (allegedly, according to the site I posted) supports the PIII-s 1.26 GHz (which also has 512k of L2 cache).

shamino wrote on 2022-04-10, 00:31:

The Shuttle AV18 (AV8? whatever it is) doesn't list support for the P3-S either. There's probably other boards with the same limitation. The reason is probably the onboard power circuitry not being able to handle the power draw of those CPUs.

IMHO it's not related to voltage. Coppermine cpu's draw much more power (1.75v for example), while the PIII-s only draws 1.45v, that's even less than the standard Tualatins (1.5v).

Strange stuff, eh?

Reply 8 of 24, by flupke11

User metadata
Rank Oldbie
Rank
Oldbie

Required voltage is lower, but the power draw in Watt is higher for Tualatin.

Still not huge numbers like the P4, the onboard VRM should be able to handle the 32W at 1,5V if it supports the other Tualatins.

Something is tripping up your system. So one of the components must run out of spec when a 1,4 is installed. Can you underclock it? Bring the FSB back to 100 to see how stable the system gets? (This was mentioned before, but not sure if you tried it.)

Reply 9 of 24, by tauro

User metadata
Rank Member
Rank
Member
flupke11 wrote on 2022-04-10, 06:10:

Required voltage is lower, but the power draw in Watt is higher for Tualatin.

Still not huge numbers like the P4, the onboard VRM should be able to handle the 32W at 1,5V if it supports the other Tualatins.

You may have a point here. You say the Tualatins draw more current? It could be, but as you also said, if it can run other Tuallies, why not this particular one being the others slightly more demanding AFAIK? I don't think that's the issue.

flupke11 wrote on 2022-04-10, 06:10:

Something is tripping up your system. So one of the components must run out of spec when a 1,4 is installed. Can you underclock it? Bring the FSB back to 100 to see how stable the system gets? (This was mentioned before, but not sure if you tried it.)

Oh yes, I did try it. The question is, you can't lower the FSB when you insert a 133 MHz CPU, you could bring it up to 200MHz, but not bring it down. I did set the RAM clock to HCLK-33, CAS 3. Maybe I could do some pin mod to try and see if it would work normally with the PIII-s 1.4GHz at 100 MHz FSB but... I'm not encouraged to use it that way. I already spent way too much time with this particular board. It was a great lesson anyway! But in the end, it defeated me 🤕

(so far 😏...)

Reply 11 of 24, by tauro

User metadata
Rank Member
Rank
Member

If this makes sense...

According to this source and the Tualatin datasheet I would have to connect the BSEL1 to a 3,3v pin (but... isn't it grounded?). I'm not sure and I don't want to toast it. I would try this with a Coppermine 133MHz first anyway to see if it works but only if an expert from the forum can confidently confirm what pins I should short to bring it down to 100 MHz.

Maybe it fails because some sort of bad BIOS programming and this would unlock it... I would then proceed to soft "overclock it" to 133 MHz 😆

I'm intrigued once again!

PC Hoarder Patrol wrote on 2022-04-10, 07:05:

Certainly seems like only the 1.4 is unsupported, even with the latest BIOS - https://web.archive.org/web/20030307110646/ht … =MS-6309+Ver5.0

Amazing digging skills 🎖️

Reply 12 of 24, by flupke11

User metadata
Rank Oldbie
Rank
Oldbie

Thinking out of the box, if you have a i815e(p) based Tualatin ready board, I would go with that unless you need the PIII-s also to run demanding WinXp applications which require 1GB memory. It's not a solution for your main issue, but it's a shame not using the Tualaking 😀.

Reply 14 of 24, by tauro

User metadata
Rank Member
Rank
Member
flupke11 wrote on 2022-04-10, 08:50:

Thinking out of the box, if you have a i815e(p) based Tualatin ready board, I would go with that unless you need the PIII-s also to run demanding WinXp applications which require 1GB memory. It's not a solution for your main issue, but it's a shame not using the Tualaking 😀.

You are right my friend.

Hopefully I will soon revive an old SY-7VBA133U with the same 694T chipset which seems to be cooperative with the PIII-s and unleash the 1.5 GB ram this baby is capable of.
98 (ISA sound card), XP (maybe I'll try 7 too), recent Linux distro, gigabit LAN (PCI), 4x AGP 😏

512MB is just not enough for XP or a fairly recent GNU/Linux distro (with GUI). For 98 it's plenty. But a Tuala is overkill for 98.

PC Hoarder Patrol wrote on 2022-04-10, 09:24:

Someone here has made it work with a pin modded cpu & some BIOS editing - Re: Socket 370 to Tualatin Converter!?

Wow! Great news! Great sleuthing!
So it can be done! 🏋️
I hope the guy is still around and answers my PM. I wish I knew how to mod the BIOS or I could find it somewhere.

Reply 15 of 24, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie

I once owned an Asus TUV4X (694T chipset) that officially supported Tualatin-256 and 1.13GHz PIII-S. And indeed, out of the box, it had loads of trouble with the 1.4S, typically hanging at POST. Now I'm going off memory here, but there was a setting in the BIOS I had to turn off - I think it was CPU/DRAM back to back transaction. With that option disabled, it worked absolutely fine with the 1.4S.

Don't know what it is about the full-fat Tualatins that make them so much harder to drive (they use like 2w more than a 1.4GHz Celeron, and at a slightly lower voltage (-50mV). So sure, resulting current draw is higher, but it's not a night and day difference).

"A little sign-in here, a touch of WiFi there..."

Reply 16 of 24, by shamino

User metadata
Rank l33t
Rank
l33t

In comparing the datasheets, it looks like the 512KB chips have much tighter tolerances for voltage. The tolerance for clock stability is also tighter, don't know what other differences there are.

Surprisingly they don't show any appreciable difference in current draw. Bigger cache makes a huge difference with the different versions of the PPro and P2 Xeon but not with the Tualatin somehow, at least on paper.
My 1.4GHz 512KB definitely drew the most power from a kill-a-watt meter of any s370 that I've seen anyway.

Reply 17 of 24, by shamino

User metadata
Rank l33t
Rank
l33t

You mentioned pulling and checking some caps. Did you put the old ones back in? I don't think MSI used good caps in this period. Don't know if that's the issue but it sure acts like the power circuitry is getting overwhelmed and losing stability when you use the most demanding CPUs. You may find over time that it gets progressively less stable with lesser CPUs.

Reply 18 of 24, by tauro

User metadata
Rank Member
Rank
Member
shamino wrote on 2022-04-10, 18:28:

In comparing the datasheets, it looks like the 512KB chips have much tighter tolerances for voltage. The tolerance for clock stability is also tighter, don't know what other differences there are.

Surprisingly they don't show any appreciable difference in current draw. Bigger cache makes a huge difference with the different versions of the PPro and P2 Xeon but not with the Tualatin somehow, at least on paper.
My 1.4GHz 512KB definitely drew the most power from a kill-a-watt meter of any s370 that I've seen anyway.

Interesting, thanks for the technical as well as the empirical data.

shamino wrote on 2022-04-10, 18:42:

You mentioned pulling and checking some caps. Did you put the old ones back in? I don't think MSI used good caps in this period. Don't know if that's the issue but it sure acts like the power circuitry is getting overwhelmed and losing stability when you use the most demanding CPUs. You may find over time that it gets progressively less stable with lesser CPUs.

Yes, I checked each and every one of them. I replaced four as they were out of spec, they are marked 1500uF 10v but they measured around +1900uF. Even if that's not pretty in principle, we already know that the limiting factor here is the BIOS.

There was also a single cap that was supposed to be 2700uF but measured around 3900uF, something like that. I suppose that's a mislabeled 3300uF. I'm not sure.

I will recap it in it's due time.

Meanwhile, I guess I better start getting the tools to crack that BIOS! Maybe this (changed the link) is going to help?

Help...
🥺

Reply 19 of 24, by tauro

User metadata
Rank Member
Rank
Member

After struggling with tutorials in different languages, trying to find files with broken links, learning how this process is in general (I had no idea at all), with lots of fiddling and some luck, I managed to get the 06B4 microcode (latest Tualatin) into the 7.5 (latest version) BIOS for this motherboard. First dumping the microcodes from another BIOS, merging them into a single file, and then adding it to the last BIOS using AMIBCP76.exe.

So... Good news! Right? Well... no. 😧

CTMC confirms that I'm using the right microcode, 6B4, which is great but in terms of progress nothing changed. Memtest keeps crahsing.

Apparently adding the correct microcode didn't help at all, so the problem must be deep rooted in the BIOS, somewhere. It's definitely related to the RAM configuration, low level stuff. I couldn't say where it is but I can say I can't fix it 😆

It was a nice process modding the BIOS, I learned about some things that are fun, like changing the splash screen and other things. Nevertheless I only learned a fraction about it all. There is a lot of different software regarding BIOS modding available and it's hard to get (especially if you don't know what you are looking for).

Anyway, I'd appreciate if anybody has more info or ideas, I'm out of them. This is it for me with this motherboard.