gdjacobs wrote on 2020-08-28, 00:03:

No matter what, I probably wouldn't use them in anything critical, anything which undergoes any kind of stress, or anything which is difficult to service. They definitely do fail (although not so often as the ChongX fake/clone caps). They're like the low end of the cut rate, although definitely better than absolute garbage caps.

In the 5 or so years I've used them, I've put them on motherboards, power supplies, LCD monitors, video cards and some consumer electronics. Not had a single instance of failure so far. If they were going to start dying, they would have started letting go years ago, especially in hot box power supplies. It's pretty fair to say they're middle of the road caps that do their job fine in most general purpose applications. Of course there are some applications you definitely wouldn't want to use them in, like RF gear or CRTs.

gdjacobs wrote on 2020-08-28, 00:03:

I think your time is usually more valuable than the few cents you'll save, but even if you want a more economical option there are much better choices available.

The price difference is actually pretty substantial, it's not just a few cents, or I'd not bother buying them. As an example, I can get 100 ChengX 47uF 16V caps for .019 cents a piece, or $1.91 shipped to my door. The cheapest I can find from Mouser for 100 of the same rated (but not same brand) is $19.47. Digikey is actually a bit more at $20.21. So that's about 10.5x the cost to use bargain basement name brand capacitors. It's well worth the cost for me because if I were to have to stock dozens of different cap values for "emergency can't wait" repairs, it'd easily be hundreds of dollars.

"Bad cap" websites inflate the cost further, since they're incorporating their time and overhead of buying and sorting capacitors for you and their website costs, etc. They have their place making cap kits for the random guy trying to fix some old machine with minimal experience, but aren't a source for people that know what they need.

The $8.00 CPU is working fine.

I discovered that this PC Chips M930LMR has unofficial support for 533 MHz bus CPUs. In the CMOS settings the CPU PnP Setup menu allows you to set the memory and bus speed, but the choices are limited to 100/100, 100/133, 100/166, 133/100, and 133/133. I really wish it had a 133/166 setting, because running the memory at 333 with the 2.8 400 CPU narrows the performance gap between it and a 2.8 533 chip so much that the 2.8 533 isn't really worth bothering with. I also discovered that though it will POST with a CPU with Hyperthreading (3.06/533/512), it will cause errors and won't load Windows. And it wouldn't even POST with the 2.8/533/1024 chip I tried.

Repo Man11 wrote on 2021-05-22, 22:50:

I discovered that this PC Chips M930LMR has unofficial support for 533 MHz bus CPUs. In the CMOS settings the CPU PnP Setup menu allows you to set the memory and bus speed, but the choices are limited to 100/100, 100/133, 100/166, 133/100, and 133/133. I really wish it had a 133/166 setting, because running the memory at 333 with the 2.8 400 CPU narrows the performance gap between it and a 2.8 533 chip so much that the 2.8 533 isn't really worth bothering with. I also discovered that though it will POST with a CPU with Hyperthreading (3.06/533/512), it will cause errors and won't load Windows. And it wouldn't even POST with the 2.8/533/1024 chip I tried.

This moterboard is probably the same as ECS P4S5A. Even BIOS is probably interchangeable.
Modded BIOS for P4S5A is mentioned here
There is a big chance that all non-HT Northwood 533MHz CPUs (512kB cache) will work, but Prescotts are no go (those have 1MB cache).
On the other hand Prescotts are much more power hungry, hotter and difference in performance is negligible. I think there is no point to use them in Socket 478.

zyga64 wrote on 2021-05-23, 12:34:

This moterboard is probably the same as ECS P4S5A. Even BIOS is probably interchangeable. Modded BIOS for P4S5A is mentioned he […]
Show full quote

Repo Man11 wrote on 2021-05-22, 22:50:

I discovered that this PC Chips M930LMR has unofficial support for 533 MHz bus CPUs. In the CMOS settings the CPU PnP Setup menu allows you to set the memory and bus speed, but the choices are limited to 100/100, 100/133, 100/166, 133/100, and 133/133. I really wish it had a 133/166 setting, because running the memory at 333 with the 2.8 400 CPU narrows the performance gap between it and a 2.8 533 chip so much that the 2.8 533 isn't really worth bothering with. I also discovered that though it will POST with a CPU with Hyperthreading (3.06/533/512), it will cause errors and won't load Windows. And it wouldn't even POST with the 2.8/533/1024 chip I tried.

This moterboard is probably the same as ECS P4S5A. Even BIOS is probably interchangeable.
Modded BIOS for P4S5A is mentioned here
There is a big chance that all non-HT Northwood 533MHz CPUs (512kB cache) will work, but Prescotts are no go (those have 1MB cache).
On the other hand Prescotts are much more power hungry, hotter and difference in performance is negligible. I think there is no point to use them in Socket 478.

Thanks for the tip! I'd love to try out that Cheepo BIOS, but finding a live link appears to be a challenge. But it looks as though the stock P4S5A BIOS has the 133/166 setting, so that would be just what I was looking for.

It looks like there is a fly in the ointment - the P4S5A doesn't have the onboard network port that the M930LMR does. The P4S5A DX has the same one, but that's a different chipset.

I did a little more checking, and when you look at the BIOS downloads for the P4S5A, the very earliest one specifies that it's for a board with onboard LAN, and it's 256KB, the same size as the 930LMR one is (all of the others are 450 to 475 KB). So I tried it, and was happy to see that I had the 133/166 memory setting I wanted. But it was unstable when running benchmarks (repeated CTD), so I ended up going back to the Amptron BIOS. I'll have to see if I can modify the Amptron BIOS to get that memory setting.

Check your BIOS chip (2MBit vs 4MBit). Maybe you can flash bigger BIOS on it ?
If not, you can always buy chip with bigger capacity and try hotflash (with Uniflash - tried that many times with success).

Though ECS lists two BIOS for the P4S5A that have on board LAN, the one I'll link to here is the only one that actually does. I tried it again, and I discovered why it wasn't working for me; the M930LMR BIOS has no setting for the multiplier, but the P4S5A does. Since the CPU is locked, you can't change the actual multiplier, but if you don't have it set correctly it makes the system unstable! Set it to match the CPU, and all is well. Being able to run the memory at 333 is a nice little performance boost, and all else seems fine, so I don't know why you wouldn't want to use this BIOS with an M930LMR motherboard.

Edit: More testing revealed a problem - there is some hardware difference in the motherboards so the board cannot properly detect the CPU with the P4S5A BIOS. Every time you power cycle it, it defaults to 100 MHz FSB. Heavy sigh.