couldnt you just use a fan to cool the piwc (or something like that)
or put some passive cooling on it??
just a suggestion

Yes, I have; and I have since relocated the fans. New (much lower) temps on CPU and PWMIC will be added to the article shortly.



http://www.madshrimps.be/?action=get...&articID=3 22

The article has been updated : http://www.madshrimps.be/?action=get...&articID=3 93

Good article Lazy I like this perspective andn I learned alot in prepartion for my own A64 3800 X2 arriving Tueday.

Hope yours would have better memory controller:)

Monarch has it now down to $285

http://www.monarchcomputer.com/Merch..._Code= 120339

Although this may seem like a criticism or an anti on-die MC (Memory Controller) manifesto, it is not. I believe the on-die MCH is one of the best things to happen to the modern processor and by the same token a step=sideways because it essentially "integrates" something else onto the core. The principle behind this is anlogous to integrated graphics into motherboard chipset's. There's not a knowledgeable person on the planet whom would choose integrated VGA over today's add-in cards. While this could be resolved the motherboard would become a behemouth cost over a 1000USD and if the graphings fails the whole baby is shot. Same thing in the Audiophile world. not only are tube or transistor high end amps seprate from the preamp, stereo is considered inferior toi using mono-blocs since the transfomer isn't shared. this has an infinate reduction as many Audiophiles Bi-amp and even Tri-amp, using three mono-blco vaccum tubes amp per side for a total of six, each amp drving a stage in the speaker thus eliminating the crossover and another series of circuits which come between the end-user and the music.

My hypothesis is the on-die memory controller while having obious advanatges also has it's diadvnatges. That is memory controller related issues now occur within the proceesor. Whehe once a motherboard (and it's NB-MCH) could be discarded (RMA a faulty mobo), switched out (get a new/different mobo), cooled (H20 cool or phase-chnage cool the NB-MCH), modified (self evident) etc. this is literally impossible unless you believe your going to open up you core and begin tinkering. There just aint no screw drivers that small.

With dual core and a shared memory controller we're now doubling (at the very least) the number of problems which can occur. it;s almost as if were buying the cow to get the milk, and a CPU is not the cheapest cow. Every processor sold with an on-die MC brings with it an entirely new set of poptentiasl problems.

1.) We know Intel prcessors (Prescotts/P IVs) overclock much higher and i assert those reasons are;

A.) have a differernt architecture (AMD braniac vs Intel speed demon)

B.) just like overclocking a Pentium will depend on the quality of the North Bridge MCH; since A64's have incorperated the MC on-die this has become an additional obstacle and one of the primary reasons we're seeing more failures amoung AMD CPU's.

C.) where Athlon XP's overclocked much higher (relative to the A64) because they eschewed (or raher were Pre on-die MC) they overclocked much higher. the PR (Perfromance Rating) system was still necessary based on architecture, however processor failures were much less frequent.

Sharing the memory controller amoung what are essentially two independent cores may in fact double (at least) the chance problems can occur and may even reduce the processors life by half. I would assert that at about three - five years were going to see many failed early A64's and the same with early X2.s.

Finally now that we have established there is a correlation between CPU failures where VTT (DDR voltage is 1/2 VTT) is excessively high and Vcore is low (default). This is due to high cuurent on one side of the diode damagin it and allowing that to surge directly into the CPU. By raisng Vdimm you balance the current on both sides of the diode. The bottom line here agin is heat, rasing voltages on both sides of that protection circuit increases heat. This IMO is an ad-hoc fix at best.

I would not go as far as calling a lesser overclock a "failure".


good find on the Vcore/Vdimm issue

The Company consists of President and Vice President; perhaps two VP. Putting the President next to the VP will certainly be better than having the VP sitting in another building;)

While the VP of Sales (Video card) be better off hitting the road to get more sales:D (the weak link).

The short fall of Intel P4 is not to intergrate memory (controller) into the CPU. While nothing is perfect; the mem controller built in is closer than Intel's approach. Intel knows it, just too screwed up to admit.;)

Besides, ocing is never the given.

When I said "failure" I meant literally, a higher incidence of CPU failure, meaning dead, fried, stodged, KaPut, a pile of burning embers, finito, seen it's maker, ascended into heaven, crossed over, deceased, sleeps with the fishes, "No mor love you long time".

Lazy I do understand what you are saying and to a large extent I agree, however, this is still early technology in the overall scheme of things. I don't think the on-die MCH has been pefected and what's making this difficult is that the CPU is every shrinking. It's like trying to play catch up.

While your analogy of then President and VP of the processor core company Inc. may be accurate in theory, it's missing something.

The President and VP are working together prblem is theor essentially sharing a Secratary, or a single phone line, or office toilet, or what have you. Point is their sharing something each used to have for themselves, and their now putting twice the workload on that thing they share. The end result being the arbitrator between whom now has twice the workload begins to get overworked, begins to "sweat" over-heats. One day it just quits.

now for the "VP of Sales" i think what we have is an redundant postion. lets say the company decixes to outsource it's "weak link" which happen to be Visual Department. Now delegating that responsicility to an outside souce which specializes in this area, quality increase ten-fold.

What do you think about my writing a quick article on this?