Ribbicipp wrote on 2024-04-04, 06:43:

Ok so i've been browsing around, trying to research this for over an hour now and there are very conflicting facts about what up […]
Show full quote

Ok so i've been browsing around, trying to research this for over an hour now and there are very conflicting facts about what upper limit the Socket 5 can handle in regards to CPU speeds.

I currently have an old AST Advantage 622 with a Pentium 1 100mhz in it. I would like to max this socket out with the fastest Pentium possible.

In another post on this forum it was mentioned that the Socket 5 can handle up to 120mhz, while other websites and wikis says a Socket 5 can handle everything up to 200mhz depending on if it's an Overdrive version or not.

Im guessing it's all about the CPU voltage?

Im quite a noob about these things so please help me clear this out once and for all.

Side question: Can my Socket 5 PC handle a Pentium 1 with 133mhz?

Thanks in advance
/Andy

It will handle a Socket 5 3.3v 133Mhz Pentium 1 just fine, it can also handle all AMD K5 CPUs and a number of IDT Winchip CPUs up to 240Mhz.

Not sure if Cyrix ever made socket 5 CPUs someone else may be able to answer that. There are also a number of Overdrive Pentiums that should also be able to work in socket 5 up to 180Mhz but I have little experience with them, I see that there may also be a couple of MMX overdrives up to 200Mhz that also work in Socket 5.

https://theretroweb.com/motherboards/s/bcm-fm561

You could put a k6 233ANR in it

A pentium 200

Or a 200mmx overdrive

A winchip 266

But your answer is that you can get whatever socket 5/7/ss7 chip you want in it most of them will need an interposer though.

Some may need a bios update. Which may or may not exist.

But the max possible cpu for that board with an interposer is k6-3+ at 400 mhz to 500 mhz depending on how high you can get your fsb.

You may need to so some mods to the board or cpu to set multiplier to setting you want.

Well, if you want to include overdrives, you might as well include powerleap and evergreen adapters. If you have the versions that supply their own current, you can pretty much run any CPU you like. The more limiting factor will be whether or not the BIOS chokes on them.

Ribbicipp wrote on 2024-04-04, 06:43:

Ok so i've been browsing around, trying to research this for over an hour now and there are very conflicting facts about what upper limit the Socket 5 can handle in regards to CPU speeds.

The reason you read conflicting opinions is that there are multiple bottlenecks:

- which FSB can you set?
- which multiplier can you set?
- which voltage can the board supply?
- how much current can the board supply at that current?
- what will the BIOS accept?

The exact answer for a specific board will depend on which of these apply.

To take these one-by-one:
- So5 spec does not specifically say anything about FSB; most boards can be set to 50, 60 or 66MHz, but in the very early So5 days you had boards that only ran at one fixed FSB, or could only be set between two speeds with a single jumper. Check your board.
- So5 specifies a single multiplier pin (BF0) so you can set either 1.5x or 2x multiplier. That would imply speeds from 75MHz (1.5x50) to 133MHz (2x66), assuming the commonest FSB settings are available. However here too some very old boards omit the jumper for this, leaving you fixed on 1.5x and giving you max 1.5x66=100MHz.
- So5 is a single-voltage specification, so there is no difference between Vcore and VIO. You can generally set VR (3.3V) or VRE (3.5V). So you are limited to CPUs that run at single volage and - unless using interposers or CPUs with built-in VRM (like the Overdrives) - that rules out CPUs like the Pentium MMX, K6 etc.
- Regardless of voltage, a board's VRM can handle a certain max current. If you draw more current than the VRM can handle, the VRM might not deliver correct voltage, or worse, it might fail (possibly even spectacularly). The challenge here is that this is almost never specified in manuals etc. In general these boards have linear voltage regulators to make VR/VRE from the PSU's 5V line. If you look up the data sheet for the regulator (big IC on three legs near the socket, usually with heatsink attached) you should find its spec. If you can't find that, look up the TDP for the fastest CPU officially supported. Divide that TDP by its stock voltage to get a lower limit for the max safe current.
- BIOS needs to at least tolerate the unknown CPU. It doesn't matter if BIOS incorrectly detects it, but some boards (Intel OEM...) will refuse to boot with an unrecognized CPU. This can potentially be patched/modded.

So, TLDR, in general a Pentium 133 is the fastest non-overdrive CPU that will run unmodified on an So5 board, but worst-case you may have a board with no 2x multiplier option and a BIOS that refuses to boot with unknown CPUs, limiting you to the 100MHz you have now.

Note I said 'unmodified'. If you're skilled/brave, there are more options:
- do a pin/socket mod to change value of BF0 and BF1. That will let you set any desired multiplier, including ones officially only supported on So7. Using that you can go up to a Pentium 200 (3.0x66), assuming 66MHz FSB is available and BIOS doesn't block it and the VRM doesn't explode under the extra load.
- replace crystal or (if there's a PLL involved) re-wire a bit to get different FSB speeds available, if you're stuck on 50 or 60MHz FSB
- there's no easy fix for split-voltage CPUs other than an interposer.
- if you're worried the linear VRM can't handle the current of a faster CPU, you can replace it. Find something with same voltage specs that can handle more current. If there's a big difference, add a bigger heatsink too while you are at it - linear regulators are inefficient and that extra power needs to be dumped somewhere.
- in case of BIOS limits, look for a different BIOS. In case of Intel OEM boards in So5 era, MR-BIOS is your best bet. Still, these boards usually use soldered Intel flash ROM, so recovery options are limited. Be sure to have the wherewithall for a bootblock recovery if things go wrong.

Thanks for all the answers guys. Very helpful indeed.
So basically what im getting is that there is no real "limit" to the clock speeds of the CPU. It's all depending on different factors. No wonder I was confused with all the conflicting facts.

Since im not used to modifying or building retro PCs, im not going to attemt to modify my system in order to squeeze the most out of it, but instead just go with the fastest CPU possible without modifying FSB, BIOS, voltage etc. and if I understand correct, the Socket5 will handle a Pentium 133 without modifying it (if the motherboard can handle it by changing the multiplier).

If it's correct that this is your motherboard, then I'm not able to find any clear documentation online. However there's a jumper block just above the CPU socket (in between it and the SiS 5512 chip) that seems to have legends on the silkscreen. Could you read what they say and post them here? The little chip they are next to looks like a PLL chip that generates the clocks for the system, so I strongly suspect this is how you set CPU speed. Otherwise there are a few others around the CPU socket too.

Could you read what the legends are next to the jumpers I indicate here?

I'll check it tomorrow and get back to you with pictures. Many thanks for your time and effort on getting to the bottom of this, dionb!

Number one concern on a Socket 5 is figuring out which CPU will blow up your regulator and stay one below that with adequate regulator cooling.

There is typically not a lot of headroom, and having blown up a BCM socket 7 board trying a K6 on it which I thought could be got under the regulator max specs, I would say that BCMs are not overbuilt.

This page is helpful for a quick reference of how much power various CPU need, https://www.pchardwarelinks.com/elec_pentium.htm
For socket 5 though, regard the wattage as scores out of 15 for how likely it is to blow the regulator.

You're gonna look at the winchip tables and think "OMG that's PERFECT!!!" then realise they are rare from not having a huge market penetration then being thrown out for being useless, so they fetch hundreds of dollars more than their performance is worth. Realistically, don't replace an intel CPU with one unless it's at least 60% higher clock than the intel... which isn't really realistic for actual performance gain vs price.

Even then, we haven't really got a number, for max wattage for socket 5 spec, some boards were made before 2x CPU and might not even support those, some boards were transitional to socket 7 or using up old sockets and are STOUT for their apparent type. So it's all a fuzzy wuzzy muzzy area in the first place, then acted upon by that wibbly wobbly timey wimey stuff of how much top end current/wattage handling has been lost by component ageing and amount of use or lack thereof to the present day.

BCM was a pretty low-end OEM supplier indeed. The current P100 draws 3.25A. P133 is only slightly above that, at 3.4A. That's less than P120, as the P133 was made with a more efficient 350nm process. But it's still 4.6% more draw. To be completely sure that the VRM could handle that, look up the code on the regulator IC (screwed to the big flower-shaped heatsink under the CPU socket) and see what its datasheet says.

Incidentally, the Winchip 2A 266 draws slightly less than the P100 at 3.3V, and despite Winchip CPUs being a lot slower clock-for-clock, at that speed it would definitely outperform a P133. But... aside from cost of acquiring them, I've never really had any luck getting any to work on So5 motherboards, nor the original WInchip. I have a C6-200 here that works fine in a later motherboards that would accept the (much) faster P200, but won't boot on any of my old exotics. That said, I haven't tried it on my 5511 based motherboard. I might just give it a spin to see - although tbh the limitation is more likely to be BIOS than chipset dependent.

Trashbytes wrote on 2024-04-04, 06:49:

Not sure if Cyrix ever made socket 5 CPUs someone else may be able to answer that.

I believe the Cyrix 6x86 were socket 5. They were certainly the right speeds. Not sure about 6x86L, they might be their first socket 7.

Right, tried with my SiS 5511 board and I have good news and bad news:

Good news is that my board turns out to be a BCM SQ599, in other words it's basically the babyAT version of the exact same motherboard as OP, so should be very similar.

Bad news: board won't boot. At first I thought it was bad VRM, but I was measuring incorrectly and the CPU is getting 3.5V (VRE by standard...) so that's not the problem. After some wiggling I got C4 output on POST card which suggest a video BIOS problem. Only problem is that VGA is the one thing where the boards differ - this uses UMA (shared system memory) instead of dedicated RAM chips - and it also doesn't have a separate VGA BIOS, assumedly the system BIOS also does VGA duty. I tried flashing with a different BIOS (I have A04, A06 is online and gives exactly same behaviour). So that's not it either. In any event, I shouldn't be hijacking this topic with my issues - suffice to say that I can't confirm anything with my board right now.

But...

This board has a manual online and it supports up to P200, with two jumpers for FSB (JP16 and JP17 on this board) and two jumpers for multiplier (J14 and J15, for some inexplicable reason named like connectors, not jumpers, and in the middle of the row of connectors next to the CPU). That means despite physically being So5, this board has implemented the multiplier jumpers for So7. If yours does too, you're in luck.

dionb wrote on 2024-04-04, 22:10:

Right, tried with my SiS 5511 board and I have good news and bad news: […]
Show full quote

Right, tried with my SiS 5511 board and I have good news and bad news:

Good news is that my board turns out to be a BCM SQ599, in other words it's basically the babyAT version of the exact same motherboard as OP, so should be very similar.

Bad news: board won't boot. At first I thought it was bad VRM, but I was measuring incorrectly and the CPU is getting 3.5V (VRE by standard...) so that's not the problem. After some wiggling I got C4 output on POST card which suggest a video BIOS problem. Only problem is that VGA is the one thing where the boards differ - this uses UMA (shared system memory) instead of dedicated RAM chips - and it also doesn't have a separate VGA BIOS, assumedly the system BIOS also does VGA duty. I tried flashing with a different BIOS (I have A04, A06 is online and gives exactly same behaviour). So that's not it either. In any event, I shouldn't be hijacking this topic with my issues - suffice to say that I can't confirm anything with my board right now.

But...

This board has a manual online and it supports up to P200, with two jumpers for FSB (JP16 and JP17 on this board) and two jumpers for multiplier (J14 and J15, for some inexplicable reason named like connectors, not jumpers, and in the middle of the row of connectors next to the CPU). That means despite physically being So5, this board has implemented the multiplier jumpers for So7. If yours does too, you're in luck.

Thanks for further investigating this.
Here are some pictures of the jumpers i could find around the CPU.

Bf0 and bf1 are your multiplier jumpers

The 100mhz setting I believe is 66 fsb

The bf may be missing pull ups you may have to add a resistor for that.

With two BF pins you can do 233mhz with supported cpu like winchip or amd k6 233anr

The anr is kinda power hungry though.

With interposer you can do up to k6-3+ 400mhz but bios may not support it.

I was a bit worried when I saw the "75MHz 90MHz 100MHz" as that implied you only had one (1.5x) multiplier and could only modify FSB (50/60/66MHz), but looking on you're in luck 😀

BF0 and BF1 (on the fourth pic, next to the socket) are literally the names for the multiplier pins for So7. That means you have the full range from 1.5x to 3.0x.

If (big if!) your VRM can handle it and if its BIOS doesn't go belly-up, it would be possible to run anything up to a P200 on this.

Now, as for the VRM... the silkscreen on the second pic give the relevant info. The votlage regulator is the LP2951 IC, whose output is boosted by the big D45H8 transistor screwed into the heatsink. Basically the LP2951 provides the voltage reference but very low current (0.1A max) and the D45H8 then amplifies that for CPU usage. Conveniently, ST still has the datasheet for that transistor online.

It says:

Collector current 10 A
Collector peak current 20 A

In a linear regulator, I in = I out (with the difference in power between V in and V out being dissipated as heat, hence the huge heatsink), so 10A is also the max CPU TDP it can support, although it is supposed to survive 20A

Comparing that to Pentium specs is simple arithmetic - the P100 you currently have is rated for 10.1W. At 3.3V that means 3.1A, so no problem. That Pentium 200 - like all Pentiums >=133MHz - has a more efficient design, so despite 100% higher clock, it only uses 50% more electricity - 15.5W. At 3.3V that means 4.7A. So it sounds like that should be a safe bet in theory. In practice... this is almost 30-year old tech and if the transistors are still OK, the electrolytic caps around them are probably drying out. I would not recommend pushing it. But with the draw of the P200 being less than 50% of spec, I'd probably risk it myself. Note however that I have a lot of old crap and so am perhaps less risk-averse than someone with just one ancient system...

Seems like the old k6-233 was only 8-9 amps at 3.3 volts.

Add extra cooling to the vrm and it might work.

Cyrix pr166 might work as well if he can’t find pentium classics.

I couldn’t tell but it looks like the board may have an undocumented FSB setting, if that setting is 75mhz he might have some good pentium classic overclocking overhead

The cyrix 166 was one of the more hungry ones. I can’t remember for sure, but I think maybe the highest? I forget where the cyrix chips got less hungry again. Was it 200/233?

If you can find one, this could be fun.

Thanks for the help. Glad i have some options with CPU upgrades on this board.
I recently got myself a P1 133mhz (https://www.cpu-world.com/CPUs/Pentium/ ... -133).html) and im going to try and install it into my system. Gonna upgrade later on when i find myself a 166 or 200mhz P1 for sale at a reasonable price. Im thinking i'll stick with the pentium processors for my PC and not go crazy since I have no previous experience working on these older systems. I can build a new, modern system just fine, but there seem to be some key differences when working on new vs old systems.

As for installation of the 133mhz, how to i need to change the jumpers from how they are set as of now for my 100mhz?
Anything else i need to take into consideration or tweak before installing the 133 CPU?