You can't upgrade 386DXs, all you can do is oc it (or solder a faster chip if you have the skill). But you'll be stuck with SX CPUs, DX CPUs need a wider bus.

As for the cache, it may not be useful if you manage to get the fastest timings on your ram. On my 386SX25, I managed to maximize what the CPU can do, so cache wouldn't have been useful. However if your bios don't let you do that, or you don't have fast enough ram, cache will be useful

The cache does wonders for 386. You get a amazing boost going from cacheless to 64KB, going fruther up to 256KB gives only a small bump.

Why don't you replace battery with similar new NiMH barrel type battery? They are quite cheap on ebay/aliexpress.

Thanks for the information guys. It looks like 64KB cache to me, so I'm glad to hear it helps quite a bit. This will hopefully deter me from actively seeking a faster 386 system. 😊

I've thought about replacing it with a rechargeable NiMH, but they aren't that cheap, considering how many dozens of CR2032s I could buy for the same price. And I will likely never have to worry about it destroying the board in the future. I have seen 25+ year old lithium coin cells that actually still work, like the ones in my Roland MT-200 synths (still look brand new and they still kept the time and date after 28 years). I trust them, and I have dozens on hand. 😀

SX will still have roughly half the performance of a DX, even if the cache was large. Databus is 16bits on SX while it is 32 bit on the DX.

Ozzuneoj wrote:

I've thought about replacing it with a rechargeable NiMH, but they aren't that cheap

I've got them like 4 barrels for 3$ total from eBay and they work great. 😀

What about this 4-pin "J1" connector near the keyboard connector? It might be a header for an external battery...

Regarding battery...
Then I have used the type here below, on several 386/486 systems. Some say it is a bad idea, and thar the only danger is if they heat up, it will explode. Yet after 1 x 386 and 3 x 486 boards, gotten this treatment. Then I am still to see any of those batteries heating up and none have exploded. Works pretty well for me. Else there are these really heavy duty ones. They are purple, yet I can't remember their name or brand. Yup. Purple and black and expensive as f*ck.

https://rover.ebay.com/rover/0/0/0?mpre=https … %2F222209265422

Tiido wrote:

SX will still have roughly half the performance of a DX, even if the cache was large. Databus is 16bits on SX while it is 32 bit on the DX.

SXs are slower than DXs, but it's not THAT bad. Assuming both systems have cache and run at the same speed, an SX should deliver more like 75% of the performance of the DX.

As for the battery, I swear by these things:

Three cheap NiMH rechargeable AAA batteries and you're good to go. If the board has an external battery header, add connectors, if not solder onto the old battery pads. The wire lets you mount it away from and below the motherboard, so if it starts to leak in future, at least it won't damage stuff.

Anonymous Coward wrote:

SXs are slower than DXs, but it's not THAT bad. Assuming both systems have cache and run at the same speed, an SX should deliver more like 75% of the performance of the DX.

I have a 386SX that doesn't deliver much more than half of what a DX would - at least not with 2 memory modules.
Four modules give it a nice boost, up to ~80% of a DX. At least in Landmark. In normal applications (Like Doom. OK I know it's unplayable but I DID played through it on a 386SX and a dog slow Tseng ET3000 VGA in its days!) the difference is smaller yet tangible. So something fishy is going on in the background...

Another SX type board I had (not anymore, it died on me 🙁) was equipped with a 486SLC CPU and cache like this one and it didn't exhibit this peculiar behavior - but the cache definitely helped a lot compared to an uncached 486SLC which was basically just as slow as a similarly clocked 386SX. I guess a 16-bit 386SX bus is severely constraining the 486 core. The cached SLC performed much more like a 386DX than a 386SX.

Your board looks quite similar to one that gave me some headache in the past (here). If you have the means, please attach a dump of your BIOS chip, and also please test your board with only two SIMMs installed. Being a SX it should work, but mine doesn't.

For the battery, I strongly suggest using an external pack: 3*AAA cells, either NiMH connected to the same spot where original barrel was located, or alkalines at the external battery header (J1, near the keyboard connector). Here's the board that most closely resembles this one. There are some differences, but the jumper settings are identical from what I found.

Regarding cache, there won't be much difference in performance between 64k, 128k or 256k cache.

64kb cache is a big improvement, but adding more won't be nearly as significant.

Tiido wrote:

SX will still have roughly half the performance of a DX, even if the cache was large. Databus is 16bits on SX while it is 32 bit on the DX.

Like I said, my 386SX lets me max out the CPU bandwidth to the RAM. But so far I've not seen any 386DX allowing to do that, so that means my 386SX 25 has faster RAM speed than my 386DX 40 which has a waitstate that divides by two the bandwidth of my ram.

To clarify, that means my 386SX can get two bytes per clock while my 386DX can get four bytes per two clocks. And since 25>(40/2), my 386SX 25 has faster RAM speed ^^

However my 386DX 40 also have cache to counterweight that slow ram.

This is not true, you can upgrade your CPU.

But u need a CYRIX 486srx2 or SLC or Ti486SLC Upgrade 😉

I had forgotten about another discussion I'd started about this board over a year ago.
Re: Questions about batteries on old motherboards...

Oops!

Anyway, I think I've finally figured out the battery charging. I have found one thing that concerns me though. I disconnected the internal battery charging circuit the other day (lifted one end of a diode that leads directly to the + terminal of the internal battery) which got rid of the 5 volts I was measuring across those battery terminals. Once I figured out how to wire the external battery header (it looks like it should be pin1= positive, pin4=negative) I decided I'd better check that for voltage too when the system is running. For some reason, I'm measuring .04v coming across the external battery terminals with no batteries installed. To me, this looks like it is intended for trickle charging? I still get no continuity between the 5v pins on the AT power connector and the positive external battery pin (where as there was continuity from +5v through a resistor and a diode to the internal battery + terminal), so I don't know what that means. I've read that alkalines can technically be recharged if done slowly, and a .04v charge doesn't seem like it would be enough to do damage, but I want to be sure before I insert the batteries and power on the system.

What do you guys think?

That 0.04V is some sort of leakage current that passes the remaining circuitry (i.e through a chip via ESD diode or something). It isn't a thing to worry about.

Tiido wrote:

That 0.04V is some sort of leakage current that passes the remaining circuitry (i.e through a chip via ESD diode or something). It isn't a thing to worry about.

Wonderful, thank you! It has kept all of its settings with the battery installed for a couple hours now.

For anyone else who finds this board, this page's explanation of the external battery header appears to be correct, except that the wording is too vague. The internal battery would require a jumper on pins 2-3 of J1 or it will not keep CMOS settings. The external battery would require the positive battery terminal to be connected to Pin 1 and the negative on Pin 2 (ground)... apparently this is what it means when it says "CLOSED" for the whole connector 😵 . I am not that great at understanding circuits, but as far as I can tell from probing around, this should be accurate. It is currently running with 3xAAA alkaline batteries and there is no smoke, fire or bad smells escaping from any components, so it appears to be fine. I had already disconnected the diode from the charging circuit (thinking it'd be possible to use a CR2032, but it didn't provide enough voltage) so I am unsure if the charging circuit would have any effect on the external battery header. I don't think it would though.

I think I've finally got this thing ready to use now... at least until some other trace\component decides to go bad later on. Time to move on to the fun part!

The computer seems to be working great, but I ran CACHECHK and noticed something odd. It said that "Megabyte #4 isn't being cached". Is that accurate? Is it normal?

I checked the cache jumpers and they're all correct. I even added another 64KB cache and set the jumpers for 128KB and it still says it. Here's a screenshot:

The system has 4MB of RAM. Running a memory test in checkit shows that all is well.

Maybe the TAG RAM size?