NW 'M0'-step Voltages...

[Lancelot]

I'm trying to squeeze every last bit out of my good old 'M0'-stepping 2.8C and she seems to do 3.85Ghz 275FSB at 1.75Vcore. I know this would be an insane Voltage but the M0 is basically a binned Extreme Edition chip and should be able to handle slightly more abuse because of this?! I'd like to know who has been running an M0-step chip with insane Voltages (1.7V +) for quite some time now on water without issues?

[hUMANbEATbOX]

even northwood EE's are prone to S.N.D.S.

i would not go near 1.7v with my M0, but hey, its up to you ;)

[batboy]

even northwood EE's are prone to S.N.D.S.

i would not go near 1.7v with my M0, but hey, its up to you ;)

Yeah, what he said.

[CoreGamer]

Yeah SNDS can be proven to happen with high voltages, and although you may think your chip could take more voltage, you may be on the edge of an ugly blue screen ;)

[microfire]

Why don't you try for low voltage, works for me (1.35v), if it dies then SDNS is a myth.

[Stratcat]

My 'MO' experience:

It was w/i the first month or two of the 'MO's coming out (don't remember the exact dates - ca. mid 2003?). I paid a 30 U$D premium to have it hand picked from Azzo Computer, in Tennessee, USA. They had ordered all they could find, when it became apparent the 'MO's were a hot ticket item.

I initially ran my 2.8 'MO' @ 3.6 using ~1.7Vcore, and after several months, had to increase Vcore to ~1.725 - 1.75V, to maintain a 3.6 OC.

It was a Malay chip, and would got to 3.5, or so, very easily, but I had to crank the Vcore to get to 3.6. I wanted 3.6, 'cuz many peeps were getting it...so I wanted "in" too! My cooling was a Zalman 7000cu, and temps remained below ~45*C, for the most part.

Within another month, or two, the OC started dropping again, just a few FSB, a bit at a time, but more-or-less continuously. Finally, after about a year total, the OC was at 3.4, where it then remained stable for another year, or so. The chip was "retired" during this summer at it's steady 3.4 OC.

So, yes, GNDS affects 'MO's like it does a "regular" Northwood. I also had similar experiences w/several other Northwoods in my crunching farm. The experiences are all virtually identical. I didn't worry about using high Vcore (1.7 - 1.75V), due to the temps generally being low (I was using a lot of all cu Zalman's, at the time.

In retrospect:

I would never push any "P4C class" Northwood 'MO', or otherwise, >1.65Vcore, if you'd like the chip to maintain it's OC for any length (>3 - 6 mos) of time. If ur mobo has a Vcore "overshoot" or "droop" tendency (where you crank more Vcore to compensate for droop under load), I'd mod it.

From my experiences,

Strat

+++

Ohhh...BTW, my understanding is the early "original" P4C 'MO's were based on the Xeon Gallatin core. You may want to research this. Not a 100% sure, it's been a while, and things probably have changed.

[hUMANbEATbOX]

Ohhh...BTW, my understanding is the early "original" P4C 'MO's were based on the Xeon Gallatin core. You may want to research this. Not a 100% sure, it's been a while, and things probably have changed.

that's true, but the m0 was a cut down EE, which was originally a cut down xeon. if that makes sense. :\

[Stratcat]

that's true, but the m0 was a cut down EE, which was originally a cut down xeon. if that makes sense. :Thx.

Actually, yes, it does make sense.

So, apparently we're all right! ;)

I can live with that. :cool:

Strat

[microfire]

If that is correct, then do all non-EE Northwood M0 circuits have electrical current flowing through the disabled L3 cache?
If not, then the entire cpu core should in theory behave differently to an actual EE, thus more voltage would be welcomed to feed over 300% more transistors of the Gallatin M0.

Northwood: 55 million transistors
Gallatin EE: 178 million transistors

Where is the actual proof, that it is a cut down EE.

[Stratcat]

Hi microfire -

I cannot comment on the accuracy of hUMANbEATbOX's statement, except to say if he made the statement, I have no reason to disagree, unless I know otherwise, which I don't.

So I will not debate the accuracy of his statement.

But that's not why I'm penning this post.

I just wanted to touch on one issue:

Vcore (Vcc) != Icc.

The voltage (Vcore/Vcc) for identical transistors, in similar operating conditions, will remain constant, whether there is 1, or 300M. It is the current requirements (Icc) that will change.

And since power is the product of voltage & current, the power draw/dissipation for the entire aggregate package of transistors (the CPU) will also increase w/an increase in the # of transistors.

But voltage will remain constant, irrespective of the # of transistors, as stated two lines above.

This is somewhat analogous to a Prescott needing more current on the 12V PSU rail (and a more powerfull PSU) than a Northwood, in terms of 12V current capability (and therefore, power rating in Watts). Yet the 12V rail remains at 12V, irregardless of the CPU, even tho the Prescott has more transistors than the Northwood, requires more current, and has higher power draw/dissipation.

The same applies to the Vcc of any given transistor type. The Voltage requirements will remain the same for any # of similar transistors; It's the current requirement that will vary with the transistor count.

So, Vcore is not an indicator of the # of transistors of a specific model processor, in a given processor model line. But current & power dissipation, may be an indicator, tho this is further complicated by differences in steppings & process refinement and control.

Just thought I'd mention this.

Strat

[hUMANbEATbOX]

while i don't have solid proof ;), i have had my m0 since they were the hot chips, and i was quite up on them at the time.

here's a thread at xtreme to back up my claims, although the links to the pics are all dead, its quite an old thread...
http://www.xtremesystems.org/forums/showthread.php?t=24091

[microfire]

Never is one core the same the exact identical

[hUMANbEATbOX]

i'm not saying they are identical, i'm saying they are both based on gallatin, AFAIK.

[Circaflex]

my last m0 only took 1.625 volts to get 270 fsb stable

[Lancelot]

my last m0 only took 1.625 volts to get 270 fsb stable

Your LAST M0?! Your current chip in your sig is the same M0 as mine!

[Circaflex]

well i havent updated my sig haha but that chip died after a lighting storm hit around my house, killed my mobo and burnt everything up :( and my "replacement" m0 hasnt been pushed as this mobo sucks and the voltage fluctuates horribly.

[microfire]

well i havent updated my sig haha but that chip died after a lighting storm hit around my house, killed my mobo and burnt everything up :( and my "replacement" m0 hasnt been pushed as this mobo sucks and the voltage fluctuates horribly.

How about ripping the heatspeader off your dead M0 and posting a photo (if you still have it).
Then compare the core to a non-M0's Northwood with the heatspreader removed (should be plenty of dead ones around to find a photo or someone with one). In theroy the die of any non-NW M0 core should be smaller than a any M0.

Maybe we can figure out why so many people think there EE M0's chips appear to handle higher voltages and survive longer (could it be that Intel's binning process is advanced so as the EE's are unlikely to die, or maybe there are less EE M0's than NW M0's so hence more dead NW, or possibly the disabled cache puts more strain on the NW M0's