welikeyou wrote on 2024-11-25, 00:16:

I recently found a PQFP-100 NG80386SX-33, the fastest PQFP-100 packaged CPU I could find on eBay. This package is the that is used for the cpu of the IBM L40SX. But, this still begs the question, can I put a faster Intel i386sx in a L40SX laptop computer?

PS, I also know that if I want to change the clock speed, I will also need to change the clock crystal and maybe the Oscillator.

PPS, I may not will have also need to change the clock crystal and maybe the oscillator.

Magic 8 Ball says: "Outlook not so good"

Meaning, if there were a list of 386SX machines to do this to, I would put this near the bottom.

Reasoning: Where sources gush that IBM had to pioneer a whole new RF screening approach for this machine, I think "So, it is inherently noisier than contemporary machines, such that it required extreme measures to counteract, why is it putting so much power into making noise???" So there may be deficiencies in the design of the motherboard already, components are also noisier when used at the limit of their specifications, maybe that is a factor too. Thus, it would worry me to try to get this motherboard to go faster.

Also these machines were recalled for burning holes in themselves, so yeah, like I said, probable deficiencies in design.

Then we have the IBM being IBM stuff like timing checks in the BIOS that will need to be patched out. They swear up and down that this wasn't to prevent upgrading, but to provide a check against risk of data corruption due to clocks going out of spec. More like they were deathly afraid of missing a service charge opportunity, but anyway, makes for an obstacle. The instructions for patching around the check on one of the PS/2 machines are given in Upgrading and Repairing PCs, so you'll have to adapt that.

So then there's the known unknowns like did IBM do anything else particularly devious elsewhere in the system, by now they weren't in a computer for everyman mode, but a put thumbscrews to the profit fairy mode. Then the unknown unknowns of whatever wild shit Murphy throws in your way, guess that's the same for many a modification though.

Anyway, unless you can find a mailing list that knows these inside out and has 40 page arguments about the benefits of using a 74ALS vs a 74F chip in a particular location, then chances are slim that it will be easy to accomplish.

Some early 386sx's are only set up for 25Mhz max. If this is one, a 25Mhz CPU will work, but a 33Mhz may cause problems.

If you replace the oscillators from 40Mhz to 50Mhz, then the current CPU should work at 25Mhz and the system be fine.

Going to 33Mhz is a chance.

That's probably the sensible way to test it, the extra heat at 25Mhz should have been accounted for by the "what if an exec is in a hot country with no AC" vs running for testing at home in a 21C room temp, so you can test at 25Mhz, and either leave it like that if you never expect to use it anywhere warmer, or put the 33Mhz chip on for peace of mine in CPU reliability, but the rest of the system may be more limited. However, IBM did throw this thing together in a panic because Compaq got the jump on them for laptop sized machines and wanted something about the same shape in the market ASAP.

The WDC chipset is a 286/386sx chipset, so a bit pedestrian, but is rated to 25Mhz. In a desktop machine you could add a fan and maybe run it at 33, but tighter form factors it's not going to be a wise move.

That is not to say that the machine cannot be speeded up, I just don't think it will take 33Mhz... This guy stuck a 486SLC chip in his for a speed boost... see down the page a bit https://www.thevintagecomputer.de/ibm-ps-2-laptop-l40sx/

I myself would actually prefer that kind of setup where you have to run a utility to enable full speed of the CPU, then you can use the speed for things that need it and keep the slower speed for compatibility.