I agree with feipoa and jolaes76. Use multiple programs like speedsys and cachechk to check your cache.

your cache looks the real deal and your speedsys score is spot on but one way or another you have to find those jumper settings.

feipoa wrote:

Usually if the wrong cache jumper settings are set, you'll get HIMEM errors. Normally, when the cache is not seated properly, you'll also get HIMEM errors. Either the cache is fake (though it looks genuine), or it is somehow disabled, via a jumper, or in the BIOS. I see what looks like a 256 KB for cache during POST, this usually implies that the jumpers are set correctly. If you've confirmed that there are indeed traces leading from the cache sockets to at least somewhere on the board, then look for a setting to enable/disable L2 cache. This is usually found in the BIOS settings.

From what I can guess, you're passing HIMEM because the cache is disabled or non-existent.

There are indeed traces (on the back of the motherboard) leading to the cache sockets. "External Cache" in the BIOS is set to Enabled.

Jolaes76 wrote:

All sort of weird problems can arise, anyhow... for example, I upgraded a Shuttle HOT-419 to 512 kb single-banked cache, jumpered it accordingly, yet the ONLY program that sees and can test the L2 cache is Speedsys. CTCM, Cachechk and the likes ALL report NO cache present...

CTCM? Please excuse my ignorance.

EDIT: A'ite, disregard that. Got it now, and it reports "Secondary Cache (L2): not detected or greater than 2048 KB". Still a mystery.

Any chance cache may be disabled by incorrect turbo switch settings? As I don't have an AT case, I've simply jumpered 2 of the 3 pins of the turbo switch header, trial and error style, until I got Speedsys CPU results that looked close to what I expected them to be. Could it be that my motherboard allows for 3 different settings? Like Slow (low mhz no cache), medium (high mhz no cache) and full (high mhz + cache).
Been spending a lot of time googling for a user manual for this motherboard, but the best I've come up with so far is the datasheets for the chipset, which naturally is of no real help. It is of course generally helpful in finding out how much ram and cache the system can support and such, but gives no clues to jumper settings and such.
Anyone else who have SiS496/497 based motherboards who can share some thoughts? What jumper numbers are related to cache on your boards?
I don't expect this board is very common (Spring Circle P2CI if I interpret the print on the motherboard correctly).

CT magazine's bench program.

http://web.info.uvt.ro/~lcucu/htmldocs/toms/ctcm.html

Also see my attachment for different flavours.

regarding cache: try the jumpers surrounding the cache and tag chip.

I'm "messing" with the jumpers near the tag chip, and it doesn't seem to affect the amount of cache reported at POST, but I've got several other possible settings. Why can't all motherboards have the settings printed on the board? 😁
Asus used to have that back in the days. Then again, brands like that generally have the documentation still available online, so the print on the motherboard wouldn't be that important.

Right, I think I've tried all possible combinations for the jumpers near the tag chip, and motherboard keeps insisting that I have 256KB cache, but cachechk and speedsys disagrees. Not sure what else to try.

Some motherboards are set to turbo 'ON' when the turbo switch is off, some need to have the turbo jumper. If your turbo header has three pins, like

1
2
3

Then there are three possibilities. 1-2 shorted, 2-3 shorted, all off.

feipoa wrote:

Some motherboards are set to turbo 'ON' when the turbo switch is off, some need to have the turbo jumper. If your turbo header […]
Show full quote

Some motherboards are set to turbo 'ON' when the turbo switch is off, some need to have the turbo jumper. If your turbo header has three pins, like

1
2
3

Then there are three possibilities. 1-2 shorted, 2-3 shorted, all off.

Aye, that's what I figured. Also, testing shows that all off = slow (about 47MHz according to Norton Sysinfo or 50MHz according to Cachechk. 15MHz according to Speedsys. 1-2 shorted = slow, same as all off. 2-3 shorted = fast, 100MHz according to Norton Sysinfo and 112.9MHz according to Cachechk. 34MHz according to Speedsys. Neither setting give me working L2 cache.
If the cache chips were fried (static during shipping or some such), would there definitely be HIMEM errors? The system goes through HIMEM just fine.
I have ordered a set of 10 UMC sram modules, matching the specs of my current cache chips. They are 32-pin ones, but the tag module is 28-pin, so I guess I'll have to find a separate tag chip. I'll try replacing just the actual cache chips first though, and we'll see. It's being shipped from Hong Kong though, so I expect slow delivery.

DIP-32, 32kx8 cache chips? Are you sure they aren't DIP-28? I've never seen any jumper settings for double-banked 256 KB L2 cache with DIP-32 chips. Or perhaps you ordered DIP-32, 64kx8 chips, for 512KB double-banked cache? That is fairly standard.

For cachechk to show a proper CPU frequency on a 486, I find I need to use version 4 of cachechk instead of version 7.

feipoa wrote:

DIP-32, 32kx8 cache chips? Are you sure they aren't DIP-28? I've never seen any jumper settings for double-banked 256 KB L2 cache with DIP-32 chips. Or perhaps you ordered DIP-32, 64kx8 chips, for 512KB double-banked cache? That is fairly standard.

For cachechk to show a proper CPU frequency on a 486, I find I need to use version 4 of cachechk instead of version 7.

I'm sure they are DIP-32. The datasheet isn't hard to find if you just google W24512AK-15.
For convenience: http://www.datasheetcatalog.com/datasheets_pd … 4512AK-15.shtml
My motherboard uses 4 chips for cache, so yes, they are 64kx8 chips, but only 256KB cache total.
Hmmm, this got me thinking. It looks like this motherboard may be a slightly "off" design. According to the specs for hte SiS496 chipset, 256KB cache must be 8 pcs of 32kx8 chips. 512KB on the other hand, should be 4 pcs of 128kx8. Perhaps I simply cannot use the RAM configuration that is installed on this motherboard. In that case, the one other photo I've found of the same type of motherboard, has the same invalid cache config.
It could be a long shot, but perhaps I need to look for some 128kx8 chips instead.
EDIT: Also, again according to the specs for the chipset, my tag chip is the size I should be using for 512KB cache, it's a dip 28 32kx8, and apparently it should be 16kx8 for 256KB.
I'm hoping this might be a breakthrough.

I didn't realise your board only has the option for single-banked cache.

So, 4 pieces of 64kx8 for 256KB or 4 pieces of 128kx8 for 512KB. TAG is 32kx8 in both cases.

Here's a pic of the whole board, just in case it's of any help to figuring this out. This is one of the pics the guy who sold it to me provided.

The cluster of 3 x 3pin jumpers near the 85C496 chip does not seem to have any impact on cache settings. In fact, as far as I could tell, changing those settings didn't affect much at all.
I've seen one other large pic (high enough res to actually see any details) of this type of motherboard, and aside from one jumper near the CPU (voltage presumably, as that pic has a DX4 100) every setting is the same.

I only see about 1/3 of the board in that image.

Ahh, right, it's to do with the forum not scaling pics, just chopping them off in relation to your resolution. Just save the file and you should be able to see the whole thing. Not sure how to make the forum display it smaller and make it "click to expand".

Refer to the cache jumper settings on the Zida Tomato 4DPS. This board looks very similar.

They do look quite similar, unfortunately it looks like the jumper settings are completely different. At least as far as I can tell from the Tomato board specs I can find.
I guess I'll just have to wait for new cache chips and try to replace them. Don't really know what else to do, as I don't think I'll find the specs for my own board online.

Right. There is nothing to worry about. As feipoa already said, you will need the 1024k pieces. Then you will try permutating the jumpers settings again and find the proper one for 512k.

Right, is this when I'm supposed to smack myself in the forehead and go "D'oh!" ?
Put the DX2 chip in, and swapped the jumper I assumed was voltage, back to how it was when I got it. Cachechk froze... not a good sign. Speedsys on the other hand, reports L2 cache of 256KB.
Just to try out, I changed the jumper over again, with the DX2, and Speedsys again shows no cache.
Very strange. So that jumper... the only difference between my board when I got it with the DX2 and the single other pic I've found, with a DX4 (i.e. 3.3V part), seemd to have disabled cache.
Anyone got a piece of software that can tell you CPU voltage? 😁
Either this motherboard has been running the DX2 at too low voltage, the motherboard automatically adjust voltage depending on CPU (unlikely) or the DX4 has been happily running at too high voltage.
Just tried Speedsys with the DX4 with the cache jumper enabled. It reports cache naturally, although cachechk still says I've only got L1 cache. 3DBench scored increased by about 33% if my maths serves me right.
I'm just wondering now. Is the DX4 going to take long term damage, assuming it's running at 5V? The heatsink, nor the edge of the cpu sticking outside the heatsink, even gets warm. It's at a comfy room temperature, so overheating seems out of the question.

Run your multi-meter across Vcc and Vss on the CPU to determine what DC voltage the CPU is receiving. I see a voltage regulator on your motherboard, but if the board came witha DX2 installed, it is probably jumpered for 5V. You need to jumper your voltage regulator to 3.3 V, or 3.45 V.

There are people around here and at http://www.vintage-computer.com who run their 3.3V DX4 chips at least for a year at 5V either accidentally or in lack of a voltage adapter. But it is, of course, not healthy on the long run.