What does CPU-Z say?

Do you have a picture where both cache chips can be seen - and the TAG-RAM chip?

majestyk wrote on 2023-09-15, 22:35:

Do you have a picture where both cache chips can be seen - and the TAG-RAM chip?

i added the additional pictures.

paste whole cachecheck output, speed difference might be there but too small for automagic detection to kick in

Repo Man11 wrote on 2023-09-15, 21:50:

What does CPU-Z say?

good call on that. CPU-Z only reports L1 cache.

Does the BIOS summary screen report "512KB cache"?
If it does this might be because it just reads the jumpering (R11, R12) or because the 2 x 256K cache chips are functional, but L2 cache doesn´t work due to a defective TAG-RAM chip (UT61M256JC-12).

Have you reflashed BIOS with the latest version?
https://theretroweb.com/motherboards/s/efa-p5 … 97-at#downloads

This reminds me of when I picked up the AT system I have that came with a PCChips M520 motherboard and a K5 PR120. The motherboard has fake cache chips, and the cache slot was empty! For however many years it was used, it had no L2, and performed as you'd expect. It still had a 486 sticker on the case, so it had been upgraded from some unknown 486 motherboard to the M520 - I guess is was probably faster than the 486 had been, but probably not by much with no L2.

Should it turn out that the cache isn't repairable, that board might work with a K6-3 or a K6-2+/3+ @ 400 MHz. The Retro Web says the voltage goes as low as 2.2 volts; even without full BIOS support, so long as it will POST you're good. Write allocate can be enabled with software.

majestyk wrote on 2023-09-16, 09:08:

Does the BIOS summary screen report "512KB cache"? If it does this might be because it just reads the jumpering (R11, R12) or be […]
Show full quote

Does the BIOS summary screen report "512KB cache"?
If it does this might be because it just reads the jumpering (R11, R12) or because the 2 x 256K cache chips are functional, but L2 cache doesn´t work due to a defective TAG-RAM chip (UT61M256JC-12).

Have you reflashed BIOS with the latest version?
https://theretroweb.com/motherboards/s/efa-p5 … 97-at#downloads

Yes it does. L2 Cache Type: Pipelined Burst. Cache Memory 512K
Bios Date: July 1st, 1998.

Repo Man11 wrote on 2023-09-16, 12:11:

This reminds me of when I picked up the AT system I have that came with a PCChips M520 motherboard and a K5 PR120. The motherboard has fake cache chips, and the cache slot was empty! For however many years it was used, it had no L2, and performed as you'd expect. It still had a 486 sticker on the case, so it had been upgraded from some unknown 486 motherboard to the M520 - I guess is was probably faster than the 486 had been, but probably not by much with no L2.

Should it turn out that the cache isn't repairable, that board might work with a K6-3 or a K6-2+/3+ @ 400 MHz. The Retro Web says the voltage goes as low as 2.2 volts; even without full BIOS support, so long as it will POST you're good. Write allocate can be enabled with software.

So i checked and i happen to have a K6-2 (AFR) 300. Not the plus . Also i don't have any soldering skills so if that's needed, then well i will have to try and use a different motherboard. You think i can get a K6-2 300 working on this board? Its worth a try i guess.

robert@exchangetest. wrote on 2023-09-16, 13:40:

Repo Man11 wrote on 2023-09-16, 12:11:

This reminds me of when I picked up the AT system I have that came with a PCChips M520 motherboard and a K5 PR120. The motherboard has fake cache chips, and the cache slot was empty! For however many years it was used, it had no L2, and performed as you'd expect. It still had a 486 sticker on the case, so it had been upgraded from some unknown 486 motherboard to the M520 - I guess is was probably faster than the 486 had been, but probably not by much with no L2.

Should it turn out that the cache isn't repairable, that board might work with a K6-3 or a K6-2+/3+ @ 400 MHz. The Retro Web says the voltage goes as low as 2.2 volts; even without full BIOS support, so long as it will POST you're good. Write allocate can be enabled with software.

So i checked and i happen to have a K6-2 (AFR) 300. Not the plus . Also i don't have any soldering skills so if that's needed, then well i will have to try and use a different motherboard. You think i can get a K6-2 300 working on this board? Its worth a try i guess.

I wouldn't give up on getting the cache working just yet - some of the people who post here are electrical engineers, and others are well qualified hobbyists who do things like add additional cache to motherboards just because they can. Maybe there is a broken trace that can be repaired, etc.

A k6-2 300 might to work on this board since it supports the K6 233 (but the multiplier is 4.5, I don't know how that would work out), but it wouldn't help with the L2 cache issue. The K6-3, and the plus CPUs, all have on die full speed L2 cache which would make the board's cache irrelevant. I had a look, and there are K6-2+ 450 MHz for sale on Ebay for a reasonable price. A standard K6-2 400 MHz would work, and maybe the raw MHz would make the missing L2 have less of an impact, but overall performance with the K6-2+ would be much better. And the BIOS could be patched to add support for the plus series CPU.

Repo Man11 wrote on 2023-09-16, 15:28:

robert@exchangetest. wrote on 2023-09-16, 13:40:

Repo Man11 wrote on 2023-09-16, 12:11:

This reminds me of when I picked up the AT system I have that came with a PCChips M520 motherboard and a K5 PR120. The motherboard has fake cache chips, and the cache slot was empty! For however many years it was used, it had no L2, and performed as you'd expect. It still had a 486 sticker on the case, so it had been upgraded from some unknown 486 motherboard to the M520 - I guess is was probably faster than the 486 had been, but probably not by much with no L2.

Should it turn out that the cache isn't repairable, that board might work with a K6-3 or a K6-2+/3+ @ 400 MHz. The Retro Web says the voltage goes as low as 2.2 volts; even without full BIOS support, so long as it will POST you're good. Write allocate can be enabled with software.

So i checked and i happen to have a K6-2 (AFR) 300. Not the plus . Also i don't have any soldering skills so if that's needed, then well i will have to try and use a different motherboard. You think i can get a K6-2 300 working on this board? Its worth a try i guess.

I wouldn't give up on getting the cache working just yet - some of the people who post here are electrical engineers, and others are well qualified hobbyists who do things like add additional cache to motherboards just because they can. Maybe there is a broken trace that can be repaired, etc.

A k6-2 300 might to work on this board since it supports the K6 233 (but the multiplier is 4.5, I don't know how that would work out), but it wouldn't help with the L2 cache issue. The K6-3, and the plus CPUs, all have on die full speed L2 cache which would make the board's cache irrelevant. I had a look, and there are K6-2+ 450 MHz for sale on Ebay for a reasonable price. A standard K6-2 400 MHz would work, and maybe the raw MHz would make the missing L2 have less of an impact, but overall performance with the K6-2+ would be much better. And the BIOS could be patched to add support for the plus series CPU.

Well the K62-300 (66mhz fsb + 4.5 multiplier) is working just fine. I think the 32KB + 32KB of level 1 cache likely more than makes up for the missing 512K of cache. However I noticed that now cachecheck is showing both caches enabled (32KB + 512KB) which it wasn't before. CPUI-Z still shows only 1 cache enabled.

I am running W95 OSR 2x with the unofficial service pack. I cant imagine I would have to reinstall windows to get the additional cache to show up?

Thanks,
Robert

This L2 cache issue is very strange. I just looked at the manual and it supports the K6 300, which gave you the necessary jumper setting for the multiplier so that's good news. I also just saw that it also has a setting for 75 MHz bus speed, so that means that you can upgrade to 450 MHz if you ever want to in the future.

Have you checked the setting for JP4? I've never seen a motherboard that had this jumper, but I've never had a motherboard with this chipset. If it is set incorrectly, it may be what's behind this odd behavior.

With this jumper in the wrong position the system wouldn´t boot.

But I would not be surprised if "rasz_pl" had a point (wouldn´t be the first time).
You are using the SD-RAM slots. SD-RAM is faster than DRAM and the speed difference between SD-RAM and L2 cache can get quite small (depending on the chioset, BIOS settings etc.).
cachechk needs a significant speed difference to decide that L2 is present. In your case this might be the case for certain CPUs, for others this difference is too small so chachechk decides that "only L1 cache is present".
You can verify this if you compare the chachchk output at the end of the test when it reports the %-numbers.

More information is always good, so I would also try Speedsys and see what it has to say. If you don't already have it, it's included in Phil's DOS Benchmark Pack: https://www.philscomputerlab.com/dos-benchmark-pack.html

Here's what my M520 looked like with no L2 cache.

Repo Man11 wrote on 2023-09-16, 23:13:

This L2 cache issue is very strange. I just looked at the manual and it supports the K6 300, which gave you the necessary jumper setting for the multiplier so that's good news. I also just saw that it also has a setting for 75 MHz bus speed, so that means that you can upgrade to 450 MHz if you ever want to in the future.

Have you checked the setting for JP4? I've never seen a motherboard that had this jumper, but I've never had a motherboard with this chipset. If it is set incorrectly, it may be what's behind this odd behavior.

SO that one I tried. and if you switch the jumper from default to the cyrix position the board wont make it through a successful post.