MS-6541 : Celeron Willamette vs Pentium Northwood

[Zmail]

Dear colleagues,

I have two identical HP EVO310V PCs, with MS-6541 motherboards and Celeron 1G8/400 Willamette CPUs.
Because both of them are working well for more than 8 yeras, I decided not to throw them out (and damage our environment) but to upgrade their CPUs.
I have choosen Pentium Northwood 2G8/533 CPUs.
My problem is, that one of the PCs accepted the stronger CPU without any problems,
but the another one not (both PCs have the latest available HP BIOS ver 3.1. The cooperative MS-6541 is of the ver. 2.1, the unresponsive is of the ver. 1.0.
After further experiments I realised that the MS-6541, ver. 1.0 refused also P4 2G66/533, C4 2G66/533, P4 2G4/400 and P4 2G0/400. From all CPUs, I have got, only C4 1G8/400 and C4 1G7/400 are accepted.
If a stronger CPUs is inserted (of course only with the cooler), the PC does not boot and does not give ani sign of activity (beeps, etc.).
The keyboard LEDs are 1x flashing but Caps Lock, or Num Lock are not switching them on.
The CD-ROM drive LED flashes also 1x and it is possible to open the drive, but the FD LED does not flash and the head moving sound is missing. It seems that the reason lies in some difference between the ver 1.0 and 2.1 of the MS-6541 motherboards.
Thorough visual comparation shows only different placement of two tantal electrolyte SMD chips under the CPU socket (version 1.0 have them in a parallel fashion, ver 2.1 in a diagonal - but the values are the same). My question is, if somebody of you have experience with such strange behavior and if there are some know solutions (besides buying a new comp
Because the layout of both motherboards seems to be exect the same (of course I am not able to inspect the internal layers I can imagine that MS/Compaq/HP had problems only with the stabilitiy and cleaniness of the 1,75V / 60A Vcc source and that they fixed it with some HF woodoo (another difference is a small - 5mm long, seemingly purposeless cut in the GND plane (done by hand, but obvious in the photoresist) ?

Best Regards,
thanks for any proven hint,
z.mail@zmail.sk

[mavroxur]

The motherboard may have an issue with processors that aren't in the Willamette family. I'm willing to bet a Willamette Pentium 4 1.8G/256/400 would work. I wouldn't suspect a "problem" with the board, as it's probably more of a hardware limitation on that version of motherboard. The later version probably added support for the Pentium 4 Northwood family.

[Zmail]

Unfortunately, I do not have any P4 1G8/256/400 and there is probably no reason to search and buy one - because the cache increase would be marginal. By the way, the PC was primaly used as home printer server and sporadic internet browser (it is connected to a LCD TV). The other one is not mine but a borrowed one and is serving similar purposes. My aim was to expand its capabilities by adding a HiRes (1920x1080p) player. Surprisingly there was no problem to add DDR1 memory upgrade and a HD3650 512kB AGP4x VGA card with HDMI output (video+audio). I had some minor issues with unwanted up/down sacaling by PC and TV - I flasely expected that a 1:1 pixel transfer is HDMI native. But now is everything OK and the picture is crystal clear. With the HD3650 is even a C4 1G8/128/400 able to replay 1920x1080p Mkv videos and a older AutoCAD version is running also smoothly (for occasional 3D furniture modelling). So I do not have any practical reason to upgrade the CPU On the other side I bought already some of them and I am very curious why they are not working just in my PC. Because I think, that the motherboard environmat do not "see" the internals of a CPU and if the FSB and voltages are the same (or allowed), there should be no external difference between a C4 1G8/128/400, P4 2G0/512/400, P4 2G6/512/400 or even a P4 2G8/512/533 ??? From the MB point of view, this should be still the same CPU, regardless of its architecture (Willamette, Northwood, Prescott, Cedar Mill), cache size (128/256/512/1024) or internal frequency (multiplicator), is'nt it ? Thus I suspect that the CPU is radiating its very high internal GHz frequencies through its numerous power lines outside its package and that the MSI/Compaq/HP engineers did'nt master their task. I admit that a GHz modulation on a 60A DC power source is a nightmare, somewhere between a micowave oven and a radar On the other side, they eventually masterd it, because the friend's 2.1 version of the motherboard is working well with all mentioned CPUs. The final question is, if the ver 1.0 is amendable or not (I have a precise soldering iron and I am familiar with it
Have a nice day and thanks for the response !

[mavroxur]

It has nothing to do with "radiated gigahertz". Motherboard are very aware of the CPU's internals, and not all motherboards support all CPU's, even if the Vcore and FSB are the same. The motherboard has to support the architecture of the processor. For example, a motherboard may support CPUs of a given family with a 65nm core, but not a 45nm core, even though there were some transitional CPUs that had the same speed and FSB, but were manufactured on the old and the new process. Or, just because a CPU is electrically interchangeable (Pentium 4 LGA775 chip w/ 800FSB and a Core 2 Duo LGA775 800FSB chip) doesn't mean the system will use it. The motherboard has to support the microcode of the given CPU, and the board's chipset must support the processor. There's a lot more that goes into it. The only power limitation would be systems that aren't designed for a higher TDP chip (the regulators can't provide enough current to the CPU.

[gg1978]

COuld be the older model only supports 400FSB models. Are the power supplies the same or equivalent between the two systems??

[mavroxur]

[power supplies]

The power supply was changed from the original 250W to a 350W ATX2 one (the mentioned 350W supply came from a working 3,4 Ghz, double core Presler PC - is also proven
The curiosity of the case is that the Intel 845G chipset, used in the MS-6541 motherboard was designed primary for 130nm (Northwood) CPUs. And I am not trying to put a 90 nm Presler chip there, but exactly a normal Northwood chip. It is rather strange that the PC was delivered and is running with a 180 nm (Willamette) Processor - because there are also some differences between the Northwood and Willamette sockets (pin A26: IMPSEL input versus Vss, pin C3 PROCHOT I/O versus Output only, probably pins L24 and P1 also, because it is not clear if 50 Ohm or 60 Ohm should be selected).
The question probably is, why and what did the MSI engineers change in the 2.1 version of the motherboard ? Because both borads are visually for 99% exactly identical - they didnot even remove the places for not used components ! Only the two SMD electrolyte capacitors under the CPU socket are arranged in a unusual way (diagonally) in the new version and a small cut in a power plane is visible. Because no an element of the motherboard is placed diagonally, I think tihs was not a cheap design change - but from simple electrical point of view (it is a DC plane) there is no sense to do so, because the number and values of the condensators did not change. I admit that there could be also design changes in the internal layers - unfortunately there is no easy way to check it.

[Zmail]

sorry, the last post is mine - i Forgot to fill the username
2 mavroxur (2)
I am trying to put a CPU (Northwood) in motherboard with a chipser (845G) wich was designed exactly for this CPU - regardless if it is FSB400 or 533.
HP claims that the installed ver. 3.18 BIOS should support all Northwood cores - inclusive HT (up to 3GHz).
But the mentioned motherboard (ver. 1.0) is accepting only Celeron 1G8/128/400 SL68D (and slower) and even the P4 2G0/512/400 SL5YR is not running The other board (2.1) accepts everything up to P4 2G8/512/533 (I was not trying to test faster or newer ones)
How is this possible ? Has Intel delivered to some motherboard builders faulty 845G chips, able to run only old Willamette Celerons ?
Thanks, bye.