Pentium 478 Interchangeability

[bgavin]

This seems like a silly question, but: can a faster P4 socket 478 be tested in a slower socket 478 board?

For example, I have an SL7E8 (Prescott, 1MB, 2.4, 533, 1.25~1.4v) that I want to test in a D845HV board.

The D845HV specifies P4 in Northwood and Willamette specs (speed, fsb, cache). The problem I have is the Willamette specs at 1.75v, and the Northwoods at 1.4~1.5v.

Q: am I going to smoke the SL7E8 testing it in this board?

[gg1978]

Re: Pentium 478 Interchangeability

I wouldn't put a Prescott in a 845HV, the VRM is probably not good enough for Icc.. Intel has used the VID bits ( a pattern of opens and shorts on the CPU package) to set the VRM voltage for a long time, since before the PII anyway. But a Prescott pulls quite a bit more current than either a Northie or Willie, so it might roast the VRM, as the 845HV IIRC only has a two phase VRM anyway..


If the VRM can't set the correct voltage, per the VRM specsheet, it should just turn itself off.. I woudn't quite count on this though, as i've run a mobile P4-M on a desktop board, and it most assurably set the VRM to about .175V too high, which was still safe for a northwood core, but marginally so..

[bgavin]

Re: Pentium 478 Interchangeability

I have been entirely unsuccessful at finding documentation on the Intel site for either PCG (platform compatibility guide), nor a matrix of board features, VRM phase, etc.

I have to dig it all out the hard way, then add it to my ever-growing spread sheet.

If anybody has URL or hard documentation for the above, please contact me.

I only paid $20 for this damn Prescott, so if it sits in the bin until Hell freezes over, no great loss. I bought it by accident, without first checking to see if it was a Northwood or not. It ain't. As noted above, the Prescott might smoke my existing board... can't have that happen in a client environment.

[willawake]

Re: Pentium 478 Interchangeability

gonna need an 865 or 875 for that. maybe bios upgrade also on some boards.

[bgavin]

Re: Pentium 478 Interchangeability

I'm still digging.

Evidently Intel boards are offered in both 2 and 3 phase VRM. The higher number of phases is required for extra current draw and/or temperature. This is probably why the OC boards such as Gigabyte offer 6-phase VRM.

What I need is a 2.4B processor for the client's existing board. The SL79B I want, appears hard to come by.

[bgavin]

Re: Pentium 478 Interchangeability

gonna need an 865 or 875 for that. maybe bios upgrade also on some boards.

I have a new D875PBZ board on the way from an eBay seller. I can use it to test the Prescott, but I'll be damned if I'm going to waste this fine board on that junk processor. I have a 3.4/800 P4 Northwood for the new board.

[gg1978]

Re: Pentium 478 Interchangeability

I have been entirely unsuccessful at finding documentation on the Intel site for either PCG (platform compatibility guide), nor a matrix of board features, VRM phase, etc.

I have to dig it all out the hard way, then add it to my ever-growing spread sheet.

If anybody has URL or hard documentation for the above, please contact me.

I only paid $20 for this damn Prescott, so if it sits in the bin until Hell freezes over, no great loss. I bought it by accident, without first checking to see if it was a Northwood or not. It ain't. As noted above, the Prescott might smoke my existing board... can't have that happen in a client environment.

Here's the processor compatibity list for the 845HV.. Looks like 400FSB, Northie and Willie only..

http://www.intel.com/support/motherb.../cs-008811.htm

[gg1978]

Re: Pentium 478 Interchangeability

Also, as to how i know about the 845HV being a two phase VRM, i have its big brother, the D845WN.. Same exact layout, but ATX...


As for the 2.4B, it will work, but not at full speed.. That's a 533FSB (B suffix) chip, and the 845HV is 400FSB only.. I might have a 2.4A or a 2.6A at home though, if you're interested..

[willawake]

Re: Pentium 478 Interchangeability

I have a new D875PBZ board on the way from an eBay seller. I can use it to test the Prescott, but I'll be damned if I'm going to waste this fine board on that junk processor. I have a 3.4/800 P4 Northwood for the new board.

yeah!

anyway should be lots of other suitable intel boards from that era ready for recapping

[bgavin]

Re: Pentium 478 Interchangeability

As for the 2.4B, it will work, but not at full speed.. That's a 533FSB (B suffix) chip, and the 845HV is 400FSB only.. I might have a 2.4A or a 2.6A at home though, if you're interested..

Acutally, many suffix A chips are 533 FSB. All those below are P4/533

2.40A SL7E8
2.40A SL7FY
2.40A SL7YP
2.40A SL88F
2.40B SL79B
2.80A SL7D8
2.80A SL7E2
2.80A SL7K9
2.80E SL7PK


I have the latest specification update for 845HV, and 400 FSB is indeed the maximum. The two I own are currently at 1.5 and 1.6 GHz. This board will support up to 2600/400/512k Northwood. If you have a 2.6 that hasn't been abused, please PM.

I was only interested in testing the Prescott in one of these boards, but can see it would be a mistake. I will do the Prescott testing in the D875PBZ when it arrives. I'm bidding on another one, maybe a couple as service boards for dead systems. D875PBZ is a good replacement for anything in the 478 socket family, Prescott included.

[gg1978]

Re: Pentium 478 Interchangeability

Acutally, many suffix A chips are 533 FSB. All those below are P4/533

2.40A SL7E8
2.40A SL7FY
2.40A SL7YP
2.40A SL88F
2.40B SL79B
2.80A SL7D8
2.80A SL7E2
2.80A SL7K9
2.80E SL7PK


I have the latest specification update for 845HV, and 400 FSB is indeed the maximum. The two I own are currently at 1.5 and 1.6 GHz. This board will support up to 2600/400/512k Northwood. If you have a 2.6 that hasn't been abused, please PM.

I was only interested in testing the Prescott in one of these boards, but can see it would be a mistake. I will do the Prescott testing in the D875PBZ when it arrives. I'm bidding on another one, maybe a couple as service boards for dead systems. D875PBZ is a good replacement for anything in the 478 socket family, Prescott included.

All the "A" chips you just listed are Prescott.. Only one was a 2.4B "northie" ... Intel changed the designation with the introduction of Prescott.. "A" became 533FSB, and they added others..

[bgavin]

Re: Pentium 478 Interchangeability

SL79B is reputed to be a Gallatin core on 0.13 micron, but I cannot firm it yet. It is the M0 stepping, which I understand is an EE chip that didn't pass the tests. This is vague information, also. It is difficult to find hard documentation on the Intel site.

My point was not about the above begin non-Prescott. Only that a number of A-suffix are 533 FSB. Intel had a number of naming inconsistencies over the product.

If you have a URL or PDF for stepping documents, please post it.
Thanks for helping me keep my data accurate.

[zandrax]

Re: Pentium 478 Interchangeability

SL79B is reputed to be a Gallatin core on 0.13 micron, but I cannot firm it yet. It is the M0 stepping, which I understand is an EE chip that didn't pass the tests. This is vague information, also. It is difficult to find hard documentation on the Intel site.

My point was not about the above begin non-Prescott. Only that a number of A-suffix are 533 FSB. Intel had a number of naming inconsistencies over the product.

If you have a URL or PDF for stepping documents, please post it.
Thanks for helping me keep my data accurate.

I found some reference in the P4 Erratum Datasheet (page 29): it's a Gallatin core at 99% because it shares the same CPUID and stepping. It's similar to the Gallatin Xeon 2.4B SL72D (same fsb, multiplier, stepping and cache size; vcore is higher and TDP too): don't forget that Gallatin P4s are Xeon cores in a socket 775 package.
Intel name rules is disappointing becuase change with every processor family: for Pentium 4 an A suffix means a cpu based on a new core but sharing fsb and multiplier with an older one (e.g. Northwood 2.8 and Prescott 2.8A, both 21x at 533 MHz) while a B suffix means a same speed, same core but different fsb (e.g. the Gallatin 2.4, 12x at 800 MHz, and 2.4B, 18x at 533 MHz). E and F suffix identifies some Prescotts: the first should means a 32 bit cache (addressed up to 4 GB) while the second identifies Intel's x86-64 implementation. The C suffix, instead, identifies latest Northwoods, which were HT enabled, from same speed and same fsb previous not HT cpus.
Unfortunately for us, Inteld didn't use a D suffix and moved to a 3 number index, such as BWM cars, before ending the alphabet

Zandrax

[bgavin]

Re: Pentium 478 Interchangeability

Wow, lots to understand. Thanks for continuing the update of my education.

Do you have a URL or PDF that documents the various process vs. code names? For example, you have info on the Gallatin core, but I can only find unofficial stuff on non-Intel web sites.

[zandrax]

Re: Pentium 478 Interchangeability

Do you have a URL or PDF that documents the various process vs. code names? For example, you have info on the Gallatin core, but I can only find unofficial stuff on non-Intel web sites.

Nope, core names should be for internal use only and never appear on official docs; however they are quite known because Intel spokesmen mentione them at press releases.
A good starting point is reading their CPUID: in the errata datasheet I posted CPUIDs change when there is a core or a stepping change. For istance, Willamette B2 cpus have a 0F07h CPUID while C1 ones have a 0F0Ah ID and D0 ones a 0F12h; for Northwoods, B0 stepping has a 0F24 one, C1 a 0F27h and so on.

Zandrax