https://www.ebay.com/itm/162771815822

Maybe an alternative to pin cut/hole drill?

You’ll still have to drill the 321st hole in the adapter though.

You've got to be totally crazy to install a split rail 2.9V K6-233 into any unmodified socket5 board without using an adapter.

Anonymous Coward wrote on 2022-03-14, 02:49:

You've got to be totally crazy to install a split rail 2.9V K6-233 into any unmodified socket5 board without using an adapter.

Based on what I’m seeing, 3.2v K6-233 chips seem to exist though?

Anonymous Coward wrote on 2022-03-14, 02:49:

You've got to be totally crazy to install a split rail 2.9V K6-233 into any unmodified socket5 board without using an adapter.

It’s 3.2v core 3.3v I/O (there may be lower volt versions also, I’m not talking about them)

They basically pull double the power than socket 5 is rated for, a lot of socket 5 won't handle a P-200 because they don't have the amps at the reg. Set it for like 166 and power it with one finger on the reg and one finger on the power button, if it gets ouchy, turn it off.

BitWrangler wrote on 2022-03-14, 03:20:

They basically pull double the power than socket 5 is rated for, a lot of socket 5 won't handle a P-200 because they don't have the amps at the reg. Set it for like 166 and power it with one finger on the reg and one finger on the power button, if it gets ouchy, turn it off.

I agree with the finger method 😀 very wise precaution.

I think it will work fine in most cases though 😀 I’ve heard of several people doing it successfully.

Looking the fet specs up prior to the attempt is a good idea, and not too hard.

Single plane socket 7 will often manage it, don't confuse that with socket 5 though, which you may have to socket trick or mod to get the multipliers.

I believe WinChip2 is single 3.3 (or 3.52?) voltage and should work in Socket 5 without modification. It even has 3dnow, its a pretty fast chip.

If you haven't already, you'll probably want to watch this Necroware video where he (among other things) replaced the voltage regulator to upgrade the CPU on an early Socket 7 FX chipset board.

https://www.youtube.com/watch?v=CMiGVQbMC5U

I find the WinChips super interesting. I thought I was buying a WinChip2 based PNY QuickChip, but it turned out to be a C6 in it, and it also seems to either have problems (possibly just with my motherboard?)

I thought an interposer would be needed for one, but maybe not? Sounds like it was sold in both 3.3v and 3.52v varieties, though eBay seems to only have the 3.52. And unfortunately my motherboard has no voltage settings.

It does have a socketed VRM board of some sort, though it’s clearly not a standard one as it has too few pins. Maybe something could be done there, but I’m not sure it’s worth the effort to try to get a replacement made. I’ll check out that YouTube video later though.

Repo Man11 wrote on 2022-03-14, 04:01:

If you haven't already, you'll probably want to watch this Necroware video where he (among other things) replaced the voltage regulator to upgrade the CPU on an early Socket 7 FX chipset board.

https://www.youtube.com/watch?v=CMiGVQbMC5U

Yassss!! We need those things for sale (but a beefier version). Very fun video

Oh cool, that’s a switching regulator. (Less heat/more efficient)

Interesting find for a socket 5 board..

I understand the winchips were kinda the go to for high speed socket 5, I was curious (for this thread) specifically about a k6 233 in socket 5 though

The k6-233 probably gives even the fastest winchips a run for their money doesn’t it?

Higher speed winchips are hard to find btw. (And expensive)

Ha, I forgot the K6-233 was 3.2v not 2.9v like the other ones. It still doesn't change the fact that installing one might burn traces and/or the VRM.
Even on socket7 boards designed to handle it, a lot of people didn't want the 233 because it ran way too hot.

Anonymous Coward wrote on 2022-03-14, 06:08:

Ha, I forgot the K6-233 was 3.2v not 2.9v like the other ones. It still doesn't change the fact that installing one might burn traces and/or the VRM.
Even on socket7 boards designed to handle it, a lot of people didn't want the 233 because it ran way too hot.

I’d give it a whirl with mine if I had a socket 5 mobo 😀

What would you recommend as a good alternative. Mmx200 overdrives are the obvious go to aside from winchips, but fast winchips are rare, pentiums are better, and overdrives are pricy.

One thing is the VRM/FET, another is if the traces on the PCB can handle the extra current, which may be doubtful if the regulator needs to be upgraded in order to handle it ..

H3nrik V! wrote on 2022-03-14, 06:35:

One thing is the VRM/FET, another is if the traces on the PCB can handle the extra current, which may be doubtful if the regulator needs to be upgraded in order to handle it ..

are there actually reports of people doing this (with a 233), or are we over thinking?

Because I’ve actually heard of several people successfully doing it. (I’m hoping one of them chimes in)

“Sometimes the problem isn’t the problem, sometimes the problem is your attitude about the problem”

😀 you may be right. I wanna see reports, data, talk to people. Prove it to me, no speculation of doom and gloom! :p

Pull out a mobo and show a heating trace on a thermal camera, let’s look up the ratings of these fets, take temp readings. Data! First hand! Let’s get scientific! 🧪 🧫 🧬

Sphere478 wrote on 2022-03-14, 06:17:

What would you recommend as a good alternative. Mmx200 overdrives are the obvious go to aside from winchips, but fast winchips are rare, pentiums are better, and overdrives are pricy.

You're kind of limited to overdrives and fancy VRM modules.
If you can't get either of those, you're probably best just leaving it alone or maybe a P166. I don't think the K6-233 is going to be a meaningful upgrade over a P166 anyway.

Sphere478 wrote on 2022-03-14, 07:01:

are there actually reports of people doing this (with a 233), or are we over thinking? […]
Show full quote

H3nrik V! wrote on 2022-03-14, 06:35:

One thing is the VRM/FET, another is if the traces on the PCB can handle the extra current, which may be doubtful if the regulator needs to be upgraded in order to handle it ..

are there actually reports of people doing this (with a 233), or are we over thinking?

Because I’ve actually heard of several people successfully doing it. (I’m hoping one of them chimes in)

“Sometimes the problem isn’t the problem, sometimes the problem is your attitude about the problem”

😀 you may be right. I wanna see reports, data, talk to people. Prove it to me, no speculation of doom and gloom! :p

Pull out a mobo and show a heating trace on a thermal camera, let’s look up the ratings of these fets, take temp readings. Data! First hand! Let’s get scientific! 🧪 🧫 🧬

I hear you! 😀 But after 15 years of PCB design with a bachelor degree in electronic engineering, I'll argue that I have a bit of qualified background to assume from 🤣 But agreed, let's see what we have and work from that. If the regulators are dimensioned for a lot more power than a 166 P54C, then it's very plausible that PCB traces are as well. OTOH, if the regulators are only dimensioned for juuuust exactly what it needs, chances are that PCB traces may have been cut, in order to route other signals etc.

I tried an IBM PR233 in an old Socket 7 at 66mhz FSB (166mhz clock) and even after adding additional cooling to the linear VRM it still would overheat and crash.

I’ve often thought a little universal kit should exist for converting/ upgrading linear motherboards to switched replacements

Something else rarely seen was modding a single rail board by simply making the vrm output a lower single plain voltage, a PMMX runs just fine at a 3.0volt univoltage

Ah well pipe dreams

The Winchip problem in general is that with a C6 which is only about 0.66 the performance of a Intel CPU, is that while it's low wattage for the clock, you only get about the same actual computing power per watt. So the fastest C6es are not really an upgrade over a P-166. The Winchip 2 and 2A do a bit better, but for a given power limitation it's still a pretty minor upgrade vs fastest intel CPU with that power limit. If you're stuck with multipliers up to only 2x and can get a 75mhz bus then a PR200 cyrix/ibm 6x86 with an official 2x75 will work, when your official max on the P54 is P133, that's an upgrade, but then you can run a late P-133 at 150 very likely, which makes it a wash.