which revision of the board do you have? PCChips is one of these manufacturers that in each revision screws something and prints a new manual

Are you sure it's sx 50? Didn't know about that model... I was thinking in dx50

As for the RAM, I can confirm 256MB is possible on an UM8881F chipset. 10 years ago someone gave me a small 486 with a tiny motherboard with that chipset that had been upgraded about as far as I'd ever seen a 486 upgraded (AMD Am5x86DX5 "PR75" (=133MHz) CPU, 32MB RAM, 8MB SiS 6326 PCI video card) and had still been in desktop use (running WIn98SE) up to that point. But at the time I had *a lot* of components lying around, so I decided to go one better.

First I upgraded it to a Voodoo3-3000, then added more memory. And more. And more. It only had two 72p SIMM slots, but every time I gave it a bigger module, it detected it and happily used whatever was in there. Eventually I was running 2x 128MB SIMMs. Unfortunately I hadn't bothered to check whether it could cache this whopping 256MB, but it certainly worked.

I then tried to install Windows XP. This of course failed because XP requires at least the 586 instruction set (and the 5x86 is, despite its name, a simple 486). So I replaced the CPU with an Intel Pentium Overdrive PODP-83. That worked, and about 12 hours of installation later I had a working Windows XP desktop and confirmation that the 256MB was available to the OS.

The 128MB SIMMs were double-sided SIMMs, but I don't remember the exact chip config. Assuming there were 16 chips on them, they would have been 16Mx4 chips.

weldum wrote:

which revision of the board do you have? PCChips is one of these manufacturers that in each revision screws something and prints a new manual

V2.1

BIOS report (WRONG):

colgar imagenes

MSD report (RIGHT):

colgar imagenes

and the computer works properly!! (The P54CM is a Pentium variant for multiprocess mainboard and this is a 486 "monoprocesssor" motherboard ¿¿??)

jcarvalho wrote:

Are you sure it's sx 50? Didn't know about that model... I was thinking in dx50

Completely

subir imagenes foros

The exactly model is A80486SX2-50 (OEM) L4172116 SX845 More info: http://www.cpu-world.com/sspec/SX/SX845.html

Regards

dionb wrote:

As for the RAM, I can confirm 256MB is possible on an UM8881F chipset. 10 years ago someone gave me a small 486 with a tiny moth […]
Show full quote

As for the RAM, I can confirm 256MB is possible on an UM8881F chipset. 10 years ago someone gave me a small 486 with a tiny motherboard with that chipset that had been upgraded about as far as I'd ever seen a 486 upgraded (AMD Am5x86DX5 "PR75" (=133MHz) CPU, 32MB RAM, 8MB SiS 6326 PCI video card) and had still been in desktop use (running WIn98SE) up to that point. But at the time I had *a lot* of components lying around, so I decided to go one better.

First I upgraded it to a Voodoo3-3000, then added more memory. And more. And more. It only had two 72p SIMM slots, but every time I gave it a bigger module, it detected it and happily used whatever was in there. Eventually I was running 2x 128MB SIMMs. Unfortunately I hadn't bothered to check whether it could cache this whopping 256MB, but it certainly worked.

I then tried to install Windows XP. This of course failed because XP requires at least the 586 instruction set (and the 5x86 is, despite its name, a simple 486). So I replaced the CPU with an Intel Pentium Overdrive PODP-83. That worked, and about 12 hours of installation later I had a working Windows XP desktop and confirmation that the 256MB was available to the OS.

The 128MB SIMMs were double-sided SIMMs, but I don't remember the exact chip config. Assuming there were 16 chips on them, they would have been 16Mx4 chips.

This board has an UMC8498F, not an UM8881F chipset

How can I know the maximum amount of memory I can install and this is cached?

Regards

muon wrote:

This board has an UMC8498F, not an UM8881F chipset

How can I know the maximum amount of memory I can install and this is cached?

Regards

The link you gave to the manual had a pic with the 8111F chipset - although now I look at it, that name is on dubious stickers on the chips. Long live PC-Chips and their relabeling.

As for how to tell - given the lack of documentation (both from UMC and particularly from PC Chips) I'd say the only way is by trial and error. If the SIMMs are to large, nothing will be broken, it will just fail to POST, or POST but only allow access to part of the memory. Start with small SIMMs and work your way up. Once you have determined how much RAM it will accept, you can do the same with cache. Start with a low amount (eg. 16MB) and do a cache benchmark. Increase to 32, 64, 128 and 256MB and rerun the bench each time. As soon as you're outside cachable area, you'll see the difference.

Word of warning with PC-Chips: some 486 boards they shipped had fake cache, so just black rectangles with no function. They were certainly doing it in the era this board came from. Look here for an example from the M919: http://www.redhill.net.au/b/b-96.html
If that's the case with you M915i, you obviously won't notice any difference in performance if you go past the max cacheable limit of the memory controller...

dionb wrote:

The link you gave to the manual had a pic with the 8111F chipset - although now I look at it, that name is on dubious stickers o […]
Show full quote

muon wrote:

This board has an UMC8498F, not an UM8881F chipset

How can I know the maximum amount of memory I can install and this is cached?

Regards

The link you gave to the manual had a pic with the 8111F chipset - although now I look at it, that name is on dubious stickers on the chips. Long live PC-Chips and their relabeling.

As for how to tell - given the lack of documentation (both from UMC and particularly from PC Chips) I'd say the only way is by trial and error. If the SIMMs are to large, nothing will be broken, it will just fail to POST, or POST but only allow access to part of the memory. Start with small SIMMs and work your way up. Once you have determined how much RAM it will accept, you can do the same with cache. Start with a low amount (eg. 16MB) and do a cache benchmark. Increase to 32, 64, 128 and 256MB and rerun the bench each time. As soon as you're outside cachable area, you'll see the difference.

Word of warning with PC-Chips: some 486 boards they shipped had fake cache, so just black rectangles with no function. They were certainly doing it in the era this board came from. Look here for an example from the M919: http://www.redhill.net.au/b/b-96.html
If that's the case with you M915i, you obviously won't notice any difference in performance if you go past the max cacheable limit of the memory controller...

Thanks for the information

About FAKE cache I read this: http://th2chips.freeservers.com/m915/fake.html

for discovering faked cache:
- Cache chips were soldered in.
- No jumpers for cache configuration (or fixed).
- WRITE-BACK being printed on them

My board has got sockets for cache chips and jumpers for size cache configurations. I think It has a real cache chips

If that photo you posted is from your board then the cache chips are almost certainly fake, notice the "WRITE BACK" text on them. You can run a tool like CACHECHK to confirm that. Good news is they're socketed, so perhaps you could eventually replace them with real ones.

Regarding the SX2 processor, the BIOS is probably unable to identify it but as long as it works fine there should be no problem. Maybe a BIOS update could help in getting the proper name displayed during the startup.

Also, I wouldn't bother with more than 32MB on a 486, but 64 seems to be the max. in this case.

TheMobRules wrote:

If that photo you posted is from your board then the cache chips are almost certainly fake, notice the "WRITE BACK" text on them. You can run a tool like CACHECHK to confirm that. Good news is they're socketed, so perhaps you could eventually replace them with real ones.

All the pictures come from my board. I need change the CPU too, because this model doesn't support cache. I read that in http://www.cpu-world.com/sspec/SX/SX845.html and I only has one 486 class CPU.

TheMobRules wrote:

Regarding the SX2 processor, the BIOS is probably unable to identify it but as long as it works fine there should be no problem. Maybe a BIOS update could help in getting the proper name displayed during the startup.

The PC with this CPU SX works properly and other soft detects right, only the BIOS is wrong.
Does anybody know if the BIOS version in this page http://th2chips.freeservers.com/m915/ is right for my board?

TheMobRules wrote:

Also, I wouldn't bother with more than 32MB on a 486, but 64 seems to be the max. in this case.

32MB per SIMM module or 32 MB total?

Regards

nice cache you got there

Yeah that cache is completely fake AND missplaced. Buy some real chips and align them to the right, that's how it worked on my board. If you don't believe me, just verify with a multimeter ^^

Deksor wrote:

Yeah that cache is completely fake AND missplaced. Buy some real chips and align them to the right, that's how it worked on my board. If you don't believe me, just verify with a multimeter ^^

Where? and What chips?

muon wrote:

TheMobRules wrote:

If that photo you posted is from your board then the cache chips are almost certainly fake, notice the "WRITE BACK" text on them. You can run a tool like CACHECHK to confirm that. Good news is they're socketed, so perhaps you could eventually replace them with real ones.

All the pictures come from my board. I need change the CPU too, because this model doesn't support cache. I read that in http://www.cpu-world.com/sspec/SX/SX845.html and I only has one 486 class CPU.

Where did you read that? All this page says is it doesn't support write-back caching. It's fine with write-through cache. The performance difference between WT and WB seemed big at the time, but after so much time it's utterly irrelevant.

TheMobRules wrote:

Also, I wouldn't bother with more than 32MB on a 486, but 64 seems to be the max. in this case.

32MB per SIMM module or 32 MB total?

What do you intend to do with this machine?

I think you could go beyond 64MB (as I said, I was able to get 256MB on another 486) but unless you want to run Windows 98 on it (which would be painfully slow due to the CPU, regardless of how much RAM you cram in) there is really no sensible reason to go past 16MB. That's enough for DOS and WFW3.11 and running pretty much anything ever released for those platforms.

muon wrote:

Deksor wrote:

Yeah that cache is completely fake AND missplaced. Buy some real chips and align them to the right, that's how it worked on my board. If you don't believe me, just verify with a multimeter ^^

Where? and What chips?

On the right side of ram you can see 2 rows of 4 chips... Fake fake fake... Remove them and install new ones

dionb wrote:

First I upgraded it to a Voodoo3-3000, then added more memory. And more. And more. It only had two 72p SIMM slots, but every time I gave it a bigger module, it detected it and happily used whatever was in there. Eventually I was running 2x 128MB SIMMs. Unfortunately I hadn't bothered to check whether it could cache this whopping 256MB, but it certainly worked.
...
The 128MB SIMMs were double-sided SIMMs, but I don't remember the exact chip config. Assuming there were 16 chips on them, they would have been 16Mx4 chips.

I too have confirmed that double-sided 128 MB SIMMs work in at least one board which contains the UMC UM8881F northbridge; it was in a Biostar MB-8433UUD. This board can be modified to accept 1024K (I've done the mod on 2 boards now) and with 1024K set into WT:L2 mode, it will cache all 256 MB. The board also works with 4x64 MB.

Could you let me know which UM8881F/8886BF -based motherboard you were able to get a Voodoo3 working in? To-date, I have only been able to get a Voodoo3 working in a 486 motherboard containing a SiS 496/497 chipset.

jcarvalho wrote:

muon wrote:

Deksor wrote:

Yeah that cache is completely fake AND missplaced. Buy some real chips and align them to the right, that's how it worked on my board. If you don't believe me, just verify with a multimeter ^^

Where? and What chips?

On the right side of ram you can see 2 rows of 4 chips... Fake fake fake... Remove them and install new ones

Where = Where Can I buy them?
What = What model the chips Should I buy?

feipoa wrote:

dionb wrote:

First I upgraded it to a Voodoo3-3000, then added more memory. And more. And more. It only had two 72p SIMM slots, but every time I gave it a bigger module, it [...]

Could you let me know which UM8881F/8886BF -based motherboard you were able to get a Voodoo3 working in? To-date, I have only been able to get a Voodoo3 working in a 486 motherboard containing a SiS 496/497 chipset.

Will dig down in my old documentation, but not sure I actually wrote down anything about this one... around the same time I was doing a memory performance benchmark on every Socket 5 / 7 chipset I could get my hands on, which I did document pretty well. This 486 was just a one-off joke. No idea what I did with the board, but I certainly haven't had it since moving in 2008...

muon wrote:

V2.1 […]
Show full quote

weldum wrote:

which revision of the board do you have? PCChips is one of these manufacturers that in each revision screws something and prints a new manual

V2.1

BIOS report (WRONG):

colgar imagenes

MSD report (RIGHT):

colgar imagenes

and the computer works properly!! (The P54CM is a Pentium variant for multiprocess mainboard and this is a 486 "monoprocesssor" motherboard ¿¿??)

I would say that PC Chips BIOS is fishy, but you probably knew that going in.