socket 5 and socket 7

[afd00]

hi all:

i have a small question: i am upgrading my friends PC and she has a pentium 133 CPU on a socket 5. i have an old PC with pentium 200 MMX on socket 7. can the 200 MMX work on the socket 5?? i tried it and it physically fits with ZIF but will it actually work???? thanks.........

[deadkenny]

No, there will not be proper voltage support for an MMX chip on a socket 5 board.

[Midknyte]

I concur. The bios of the board probably wouldn't support the mmx cpu anyways.

[Peter M]

Well if the electrical bits and pieces aren't there, why should it? Socket-5 and early socket-7 indeed didn't have "split rail" CPU voltage supply for MMX Pentiums, Cyrix MII, K6 and such.

[Rugor]

It won't work, sorry.

The socket 5 is a 320 pin socket with a plug where the 321st pin would be. The socket 7 cpu is of course a 321 pin design. You may be able to get it into the socket, but it won't work.

[deadkenny]

Originally posted by Rugor
It won't work, sorry.

The socket 5 is a 320 pin socket with a plug where the 321st pin would be. The socket 7 cpu is of course a 321 pin design. You may be able to get it into the socket, but it won't work.

That's interesting. It was my understanding also that there's an extra pin on socket 7. But I would have though that the extra pin was there to prevent later CPU's from being inserted into socket 5 boards. In this case though he said it physically fits. In any case, it probably won't work at all because of the extra pin. As Peter mentioned, there are two 'versions' of socket 7, P54 and P55. Only P55 supports the split rail voltage requirements of the later processors (MMX etc.). Although in that case I believe the later processors will run on an older P54 socket 7. But it will fry them eventually, because the vcore will be 3.3v whereas the MMX spec is 2.8v.

[Peter M]

They won't even work, because not only had the voltage gone down, but also did the amperage go up. Big time.
And they'll fry in an instant, of course. Just a matter of whether the CPU or the board's voltage regulators go boom first.

[causticVapor]

Go to powerleap.com. They have an adaptor that will allow you to plug a socket 7 CPU into a socket 5 board with correct voltage step-down, etc. The have them available with CPUs of up to 400MHz. :t

[Peter M]

... with which you still need BIOS support. As I said further above, finding that a mainboard BIOS supports a CPU that electrically can't work on that very mainboard is highly unlikely. BIOS engineers have better things to do than implement useless stuff.

[causticVapor]

The BIOS only needs to provice the correct multipliers, etc. The voltage step-down process is transparent.

[Peter M]

No, sorry. If the BIOS doesn't know what to do with a certain CPU make and model, the thing won't get far. It's the BIOS that initializes the CPU, ya know. (The multipliers are done inside the CPU btw, it has always been that way. The jumpers just let you feed a configuration number.)

[deadkenny]

There's an article on Tom's Hardware about running a K6-2 on a socket 5 board using such an adapter, so it can work. I've actually run a P55 Pentium processor on a P54 socket 7 board. The processor heated up pretty fast, but it booted. I'm sure that the processor would have fried eventually. A guy I know claims to have run such a setup for a year before frying the processor.

[causticVapor]

It is my understanding that the CPU does dictate the multiplier; however the BIOS can override and manipulate this. The CPU interprets a signal ? from the motherboard and then sets it accordingly... kind of like the 2x->6x for the K6 series...

If the motherboard's maximum multiplier is below the CPU's default multiplier, then the CPU dictates without influence, right?

True, if the BIOS doesn't even know a CPU is there, with the correct pin maps etc, then it wouldn't POST, but the adaptor apparently sets this right -- it has a miniature VRM on it, and if I remember correctly, DIP switches for the core voltage.

Sorry if I was a bit sharp there, Peter. :t

[Peter M]

Sorry, no, it doesn't do that. At all. It's all CPU internal. The association of jumper settings to resulting multipliers is too. E.g. that 2x->6x "remap" is a CPU internal change, the jumper setting that used to mean 2x means 6x for newer K6-2, that's all. CPUs have had different interpretations of the same jumpers for ages - remember K5 (1.5x, 1.75x, 2.0x), Pentium (1.5x-3.0x), Pentium-MMX (2.0x-3.5x), Cyrix 6x86 (1.0x-4.0x)? It's all CPU internal.

What BIOS does is initialize the CPU's internal configuration registers. And for this, it needs to know what a certain CPU make and model is like. If it doesn't, it can't set the CPU up properly, and the system won't POST right - with caches disabled and all performance features not activated, typically, but some BIOSes choose to halt the system with an unknown CPU in.

(Hint: Before you even think of trying to teach me more about how you think these things work, please look at my profile :))

[deadkenny]

Two points:

First, the CPU 'controling' the multiplier was a change that came in with the later Pentiums. I'm pretty sure up through the early Pentiums the processor would actually run at whatever the mobo FSB / multiplier jumper settings dictated.

Seond, although the processor does have to be 'recognized' by the BIOS to some extent, this is a pretty generic recognition. Often K6-2 processors will actually run on old socket 7 boards that were never intended to run such processors. Then voltage support becomes the key issue. But it's even possible to run these processors without the proper voltage support. The K6-2's have vcore requirements between 2 and 3 volts, with many socket 7 boards delivering only 2.8v. But often the K6-2 will run at the incorrect voltage, and can even survive quite a while if cooling is properly addressed. I read an article testing a Tualatin adapter product, where a Tualatin Celeron was run on an old slot 1 board using both a slocket and the socket adapter. The CPU was recognized as a 1.4GHz PII :rolleyes: but it did run and even produced benchmarks in line with what you would expect from such a chip.

[Peter M]

Ah come on folks, this is my profession you're talking about. We're designing mainboards here, I'm a BIOS engineer working on the actual assembly language source code of BIOS. Wouldn't you think I know a tiny bit about the guts of BIOS and CPU initialization?

Originally posted by deadkenny
Two points:

First, the CPU 'controling' the multiplier was a change that came in with the later Pentiums. I'm pretty sure up through the early Pentiums the processor would actually run at whatever the mobo FSB / multiplier jumper settings dictated.


No, clock multiplication (and properties thereof) has bene CPU internal right from the start, which was the 486DX2 btw.


Seond, although the processor does have to be 'recognized' by the BIOS to some extent, this is a pretty generic recognition. Often K6-2 processors will actually run on old socket 7 boards that were never intended to run such processors.


Only if the BIOS chose to go ahead anyway - but all the important features in the "unknown CPU" will be disabled. If not completely unusable, performance will be WAY lower than it should be.


Then voltage support becomes the key issue. But it's even possible to run these processors without the proper voltage support. The K6-2's have vcore requirements between 2 and 3 volts, with many socket 7 boards delivering only 2.8v. But often the K6-2 will run at the incorrect voltage, and can even survive quite a while if cooling is properly addressed.


K6-2, depending on which revision they are, suffer permanent damage from 2.6V and 2.5V respectively, the newer the lower. Regardless of cooling btw, it's the tiny traces on the silicon that slowly fuse themselves when the voltage gets too high. If you've chosen to go above, do as you please, but don't suggest that to others.


I read an article testing a Tualatin adapter product, where a Tualatin Celeron was run on an old slot 1 board using both a slocket and the socket adapter. The CPU was recognized as a 1.4GHz PII :rolleyes: but it did run and even produced benchmarks in line with what you would expect from such a chip.
Well that's because (despite what Intel marketing wants you to believe) nothing much has happened from the original "Mendocino" Celeron to "Tualatin", and also because Tualatin happens to have the exact same internal configuration register set. Lucky draw, but not at all indicative of the situation we have with a K6 on an old Pentium board. These two are different. Very much different.

regards, Peter

[causticVapor]

But often the K6-2 will run at the incorrect voltage, and can even survive quite a while if cooling is properly addressed.

Quite a while, like a month quite a while. This is at 35-40deg C, normal operating temps.

As for the rest, listen to him - he knows a little more than we do; I hate people who try to make it look like they know more than they do. :rolleyes:


BTW Peter - I wasn't trying to teach, just see if I was understanding it all correctly - I may have confused myself at times and looked stupid - but now it all makes sense. :t

[Rugor]

Just to see that I understand you Peter, I take it that you are saying the PIII Tualatin can be made to work in older boards because it has the same P6 core as all the other members of the PII/PIII/Celeron family, and it will have at least most of its functionality. Further, you're saying the K6-2 cannot be made to work properly in a Socket 5 board because not only does it not support the split-rail voltage, but also because the RISC-like core of the AMD chip relies more heavily on some of the changes Intel made to socket 7 than the Pentium does; since the Pentium chip was originally designed for socket 5 and the K6-2 wasn't.

I hope I'm not making things more confusing.

[Peter M]

Not quite. In the required BIOS initialization, Tualatin Celerons are identical to their older brothers - so if a BIOS is daring enough to apply what it knows to that much newer CPU, it'll happen to be fine. HOWEVER, Tualatins are electrically different from the older chips, in that the core _and_ I/O voltage differs. You might be able to trick the board to supply the right core voltage, but since the I/O voltage is fed by the chipset, there are no tricks other than having a Tualatin compatible chipset. There are adapters that pull a few tricks to make it work, but you're risking damage to the CPU or unstable operation.

In Pentium vs. K6-2, there have been both quite dramatical electrical changes (3.4V, 6A on Pentium, K6-2 at 2.2V and up to 14A), and a totally different internal register set. So even if you solve the electrical stuff, a K6-unaware BIOS has no chance of guessing it right.

What AMD did to socket-7 reaches beyond what Intel did. Intel just upped the allowed power draw for the original move from 5 to 7. Then, in a second step shortly thereafter, the split-rail voltage stuff was actually introduced (2.8V for Pentium-MMX). Somewhere around that time Cyrix went 75 MHz (on VIA and SiS chipsets). AMD then later went beyond that with 2.2V support, and yet again shortly thereafter, when Cyrix had already moved to 83 MHz (on SiS and ALi chipsets) and the first AGP boards appeared (VIA and SiS), AMD came up with "Super-7" with yet again stronger CPU power supply and 100 MHz.

regards, Peter

[deadkenny]

Originally posted by causticVapor


Quite a while, like a month quite a while. This is at 35-40deg C, normal operating temps.

As for the rest, listen to him - he knows a little more than we do; I hate people who try to make it look like they know more than they do. :rolleyes:

Is this directed at me? WTF!?!

First you contradict Peter, now you're kissing his feet. I hate people that do THAT!:mad:

[deadkenny]

Originally posted by Peter Missel
Ah come on folks, this is my profession you're talking about. We're designing mainboards here, I'm a BIOS engineer working on the actual assembly language source code of BIOS. Wouldn't you think I know a tiny bit about the guts of BIOS and CPU initialization?



No, clock multiplication (and properties thereof) has bene CPU internal right from the start, which was the 486DX2 btw.



Only if the BIOS chose to go ahead anyway - but all the important features in the "unknown CPU" will be disabled. If not completely unusable, performance will be WAY lower than it should be.



K6-2, depending on which revision they are, suffer permanent damage from 2.6V and 2.5V respectively, the newer the lower. Regardless of cooling btw, it's the tiny traces on the silicon that slowly fuse themselves when the voltage gets too high. If you've chosen to go above, do as you please, but don't suggest that to others.


Well that's because (despite what Intel marketing wants you to believe) nothing much has happened from the original "Mendocino" Celeron to "Tualatin", and also because Tualatin happens to have the exact same internal configuration register set. Lucky draw, but not at all indicative of the situation we have with a K6 on an old Pentium board. These two are different. Very much different.

regards, Peter

OK, chill dude.:cool:

First, I'm not even contradicting what you're saying, and certainly not questioning your expertise. I'm sure you know far more about the inner workings than I'll ever know (or even want to know). However, you're talking theoretically and I'm talking empirically. Sure, if the BIOS isn't especially designed to properly recognize a particular processor, then it might not work. However, in practice, it often does. If it's not completely or properly supported, then it might not function optimally. But in practice it's often good enough. We are talking about socket 5 / 7 and Pentium 'classics' and K6 AMD's, not exactly leading edge stuff here.

As far as me 'recommending' to others that they run K6-2 processors at out of spec voltages, well I think I've given them the information to make an informed decision. Yes it will work. Yes it will damage the processor. How long that takes does partly depend on cooling. I wasn't suggesting that cooling will eliminate the damage, but the fact is heat is a factor and the cooler the better. If someone from a third world country has just spent a week's pay on a K6 processor, then don't risk it. On the other hand, if someone has just salvaged one from an old junker, and is interested in 'playing' with it then feel free to give it a shot. I don't have good stats on the life expectancy of a K6-2 running out of spec, but if you do then I'd be happy to hear it. Obviously it will be a function of how far out of spec the voltage is. If running at 2.8v, then I wouldn't expect a 2.2v to survive long. A 2.5+ might hang in there.

I do have one question regarding the multiplier. You say that the CPU has always 'controlled' the multiplier. I was talking about the multiplier being 'locked'. Is that what you mean by 'controlled'? Because I'm sure I heard / read people claim early Pentiums were not multiplier 'locked'. If all these processors have always been locked, then what was the point of putting jumpers on the motherboard in the first place? What functions does that serve if the CPU controls the multiplier regardless of what the mobo is set to?

[Rugor]

I'm going to take a stab at answering part of that.

Since they implemented CPU multipliers they have been coded into the CPU. However, these codes have been subject to being overriden by the motherboard.

With the Pentium II 300 (100FSB) Intel introduced multiplier locks as opposed to settings. These could not be overridden by the motherboard's jumpers. Ever since then Intel processors have had locked multipliers while AMD has generally had multipliers that can be unlocked.

Hope this helps.

[deadkenny]

OK, thanks Rugor. Then I stand corrected, since in my earlier post I used the terms incorrectly. I said 'controlled' when I meant locked. So basically I was commenting on the ability of the mobo setting to determine the speed of the processor, which I think is clear from the rest of my post. Frankly, I don't know enough about the internal workings to appreciate the subtle differences. Bottomline, all I care about is what the processor runs at, and whether or not I can change it by changing the mobo settings.

[Peter M]

Originally posted by deadkenny I do have one question regarding the multiplier. You say that the CPU has always 'controlled' the multiplier. I was talking about the multiplier being 'locked'. Is that what you mean by 'controlled'? Because I'm sure I heard / read people claim early Pentiums were not multiplier 'locked'. If all these processors have always been locked, then what was the point of putting jumpers on the motherboard in the first place? What functions does that serve if the CPU controls the multiplier regardless of what the mobo is set to?

Controlled does not equal locked, of course. The "locked" ones do not take a setting from the outside anymore, they are factory fused to one single setting.

What I was about is the common misconception that the mainboard and BIOS are anything to do with the multiplying in itself. They don't. All that's done is feed a setting number into the processor. What (if anything) the CPU does with that number is entirely up to the CPU core.

As for the potential overvoltage damage, the K6 datasheets are pretty clear about this. The "absolute maximum" voltage, above which physical damage creeps in no matter what else you do, is 2.6V for the first iteration of 2.2V/2.4V K6-2, and 2.5V for the later 2.2V/2.3V ones. Given that when you "set" the voltage to a certain regulation center point, you typically get +-0.1V fluctuation, your sanity headroom ends at 2.5V and 2.4V settings respectively. Not too much, is it?

[deadkenny]

Thanks for clearing up that distinction. As I've said, I should have used a different word, such as determined by. Contrary to accusations by certain other people, I was not trying to come off as 'knowing more than I do'. I was merely trying to make the point that for the early Pentiums, the board setting is actually the multiplier ultimately used. For later (locked) Pentiums this is not the case, the processor will use the factory fixed multiplier regardless.

Regarding the voltage, agreed. But that's also AMD covering their ****. As processor design has progressed, with repeated die shrinks, the voltages have gotten lower and the tolerance of high voltages has gotten less. Still, people run over spec voltage for purposes of OC'ing. The fact that a K6-2 can survive 2.8v when the max rating is 2.6v is to me less amazing than a MMX chip surviving 3.3v on a board that doesn't support split-rail, and yet it's been done (thats half a volt!) No way I would do this with my primary system, but to me an early K6-2 or MMX chip isn't worth much.

[causticVapor]

Originally posted by deadkenny


Is this directed at me? WTF!?!

First you contradict Peter, now you're kissing his feet. I hate people that do THAT!:mad:


Err, no, I didn't intend to contradict -- and I'm sure you didn't either. And you're taking a generalized statement personally. I don't like your tone of voice, sir.

[deadkenny]

Originally posted by causticVapor



Err, no, I didn't intend to contradict -- and I'm sure you didn't either. And you're taking a generalized statement personally. I don't like your tone of voice, sir.

Ummm, you said that you HATE people that try to look like they know more than they do, and it appeared to be directed at me. So, you care to clarify? Seeing as you apparently made a totally unprovoked and unjustified attack on me, I think my 'tone' is well justified.

[causticVapor]

:confused:

Wasn't directed at any particular folk


sheesh :rolleyes:

[deadkenny]

Well, in the context it's pretty difficult to read your 'hate' post in any other way. Also, you didn't exactly post back in a timely manner to correct the impression you have left.

If any of the mods are really bored and want to read though this thread, I'd be happy to hear their opinion on it. I'm referring to the post after Peter 'corrected' me where causticVapor says 'listen to him...' then 'I hate people...'