I would replace all caps , add little heatsinks on vrm, add some passive cooler to North and South bridge Chip

Gotta replace those caps, first of all. Even if it's NOS, capacitors have a shelf life, and the K7S5A is a bargain basement model that had never included quality capacitors to begin with.

Heatsinks on the VRM would help, along with a top-down CPU heatsink that blows air in the direction of the VRMs. The stock chipset heatsink will be attached using thermal adhesive as these boards (and other ECS/PC Chips boards of the period) do not have holes or clips to attach their chipset heatsinks. Possibly you can mount a Zalman NB32K or similar on there, but you would need some new and strong thermal adhesive to keep it from falling off.

Definitely going to want new caps on that board. I replaced the large ones once on mine 8-9 years ago and now some others need replaced. I'm amazed that such a cheap board has lasted nearly 17 years though without any other failures. It was used in three or four different computers up until it was retired about 4 years ago.

Edited.

What's the risk of letting caps die? Will they take out the board and/or other components?

I would replace the caps, but probably not worry about heatsinks for the VRMs.

The K7S5A was never an overclocking board, putting lots of heatsinks around the place is just overdoing things.

squiggly wrote:

What's the risk of letting caps die? Will they take out the board and/or other components?

All of the above.

Generally caps die gracefully, but in extreme cases they can short out MOSFETs with the latter burning/melting through the PCB. The currents involved with that sort of stuff can kill pretty much anything on a board, so bad news.

Once again, this is rare (as are eye-endangering fountains out of dead caps powered up), but nonzero chance. Re-capping reduces that to zero chance (if you do it right).

canthearu wrote:

I would replace the caps, but probably not worry about heatsinks for the VRMs.

The K7S5A was never an overclocking board, putting lots of heatsinks around the place is just overdoing things.

It was never built as an overclocking board, but with a 3rd party BIOS it does allow for some overclocking. I updated mine to allow for slightly newer CPUs and more BIOS options. It just made a good board even better.

That said, I wouldn't worry about VRM heatsinks either unless you're going to go into this specifically trying to push a classic budget board to its limit.

Edited.

Been forever since I used it but as I recall, there was a popular hacked BIOS called the "cheepoman" BIOS, or something like that. Don't remember what features it added, and maybe there are others.

I once bought a used EpoX board where one of the Vcore caps had shorted. It mostly came back to life after replacing some of the attached MOSFETs, but it still had problems after that.

If you remove the chipset heatsink, try to do it with a twisting motion. Try not to pull or pry very much because that has more risk of damaging the BGA solder joints.
I have an ECS board where some goofball added a heatsink to a heatsink.

Edited.

ninkeo wrote:

Cheers, I'll take a look someday. At the moment, I'll leave it on the stock BIOS but if I want to get a bit more control, I'll track one down

Be very careful with that stock BIOS if it's an early one: when the K7S5A was first released it had dodgy memory stability and a nasty FDD data corruption bug. Both could be fixed by a BIOS update, but if you did that from floppy (never a good idea...) the data corruption killed your BIOS.

I ran one of these as main system for a few years. The later ECS BIOSs were perfectly adequate, but the earlier ones were not. At the very least I'd upgrade to the most recent official ECS BIOS (and NOT from floppy - so also not from Gotek!). IIRC cheeopman mainly added some OC-options, or rather just un-commented the relevant lines in the original ECS BIOS. If you don't want to OC, you gain nothing significant.

Edited.

dionb wrote:

IIRC cheeopman mainly added some OC-options, or rather just un-commented the relevant lines in the original ECS BIOS. If you don't want to OC, you gain nothing significant.

Y'know, I'll keep an eye for a K7S5A if I ever come across one again, and test some SiS735 BIOSes for it, then report the results.

I managed to successfully run a Iwill KK266 BIOS on a K7VZA (although it won't POST from soft restarts - hard restarts do work) and everything seems to work except soft resetting.

I also have a P4P800 SE for the same reason, and a few stock vanilla MBs (with bare minimum to get working - this means NO Deluxe features like POST voice etc.) to try BIOSes out.

As soon as I get my hands on a K7S5A I'll recap it then report back my experiences with other BIOSes (I have several spare DIP32 chips and PLCC32 chips so keeping a original wouldn't be a problem.) as to which one works and which doesn't, and which one would suit for OC without modding. (as in simply flashing a stock BIOS image from another SIS735 board and then resetting instead of getting a modded BIOS)

Edited.

The K7S5A Pro is a different beast. It shares the same chipset, but other than that it's a different board, requiring a different BIOS.

One thing to note with the SiS900 on the Pro: it's a decent enough 10/100 NIC, I always felt it had marginally lower CPU usage than the ubiquitous RTL8139s. However if you intend to run Windows XP on it, be aware that default drivers are present but they have a bug: the drivers read the wrong ROM address when determining the MAC address, so it will always be set to 00:E0:06:09:55:66. This will obviously lead to conflicts if you have more than one on your LAN. Solution is to install SiS' drivers, which read the correct ROM address for the real, unique MAC of the adapter. Older OSs don't have built-in SiS900 drivers, so don't suffer from this bug and newer ones don't have the bug - this is XP only.

Edited.