Peltier cooled Opty 170, only 2.4Ghz? Help needed

Recently I've been building myself a computer and stumbled across this website and saw a lovely list of opteron dual cores doing serious clock speeds. Obviously I wanted a piece of the pie for myself, so I went and built my own 939 PC.
Im currently running:
an Opteron 170 OSA170DAA6CD CCBWE 0551 UPMW
Mobo: Gigabyte K8N Pro-SLI, Bios rev F9
Ram: 2 gig Kingston Value
Power Supply: 550W Antec True Power
Graphics: Winfast PX7800GT
Monitor: Dell 2405FPW

Then one of my co-workers pointed me in the direction of peltier cooling and two weeks later i got it all together and working. Its ugly as sin and the peltier "only" pumps 127 watts but i have a pair of 236w peltiers on order. The current peltier keeps things at 2 or 3 degrees at idle, but depending on Vcore it can stabilise at 20~40 degrees under load.

When I try and overclock the cpu it runs to 2.4 smooth at stock 1.35 volts, but when I start upping the voltage in the bios I cant seem to get it stable for more than a few minutes at anything higher. I have turned the HT ratio down and set the ram to half speed so that they dont hold me back.

My questions are these, I see a lot of people with fancy DFI and Asus motherboards and was wondering whether or not the cheap gigabyte board is holding me back?
Similar story with my cheap ram, is it making the system unstable at higher FSB's?
When I do get the Clock to 2.5 Ghz and run SP2004 on both cores I find that core 0 drops out first every time, whats going on here?
Is the Stepping of my opteron from a poor batch?
This is my first post on a forum ever, so any help would be VERY much appreciated!

i don't think that peltier cooling is appropriate for today's cpu's. it just gets overwhelmed by the heat. you see this by your 2-3c idle temps, and 40c full load temps. you could probably get 40c full load with just simple water cooling.

if you have the pelt running off of your main powersupply, it may be hurting the overall strength of the psu, resulting in a low OC. however i am just speculating, others who know more about ocing a64's would be able to provide more insight.

about the pelts, i would check out the cooling section, maybe post a thread there.

Um, is this peltier cooling with a waterblock, or with a heatsink & fan?

Just get a Big Typhoon, Ninja or Thermaright Sink. Leave VCore at stock and you should hit 2.5 easy.

Funny you should mention peltiers being overwhelmed humanmeatbox! When I first fired it up the idle temp at stock volts was 10 degrees below zero! but when i overvolted it to 1.7 with realtime temp graphing and full load, it displayed a perfect Fourier conduction cooling curve, which would have topped out at about 90 degrees, obviously i didnt leave it at 1.7 for long !!!
FYI the peltier is running off a cheap 600watt powersupply, so plenty of scope for adding more peltiers. It is cooled by the asetek waterchill kit at the moment and it does a marvelous job, the water is barely above room temp. The peltier really isnt the problem at the moment, the fact that the cpu is unstable is. Ive had a poke around what other opty 170 users have and see them with DFI LP Ultra boards and Asus boards and wonder whether I might be better off with an Asus K8N SLI 32 Deluxe.

AlabamaCajun, um, if its not stable at higher volts at 2.5 then i doubt leaving it at stock volts will work. I'll give it a shot in the meantime however.
Thanks for the early replies Guys, ill keep on it!

What are you trying to run your memory at?

Two problems:

1) contact. What contact do you have with the cpu and how much mounting pressure do you use?

2) Peltier just isnt strong enough. My Typhoon does ~38c load at stock vcore on my FX, CCBWEs run cooler.

getting 2.5-2.6 on stock voltage is normal. You should see 2.7- 2.8 ish with more volts.

It takes a monster of a pelt for dual core AMD chips now.. they make them but man do they suck the juice.
Around 400w at optimized voltage should to the job..

hmm, the peltiers i have on order do 15V at 24A, about 370watts, so that should do the trick.
I am forced to agree with fhpchris on the contact issue. I have a copper block that is 5mm thick sitting on the cpu. It has been milled flat and clamped down with the peltier on top and the waterblock on the hot side of the pelt.
I have read that some of the caps that cover the core itself are not properly stuck down by AMD so that one core will fail first during overclocking. I am tempted to try and remove the damn thing but I am 99% certain that it will destroy the chip in the process. Has anyone else managed to get this cap off?
My other idea is to mill a better cpu/peltier copper block. One with an indentation in it that covers the sides of the processor as well! That way I can draw heat from the top of the CPU and all four sides, the only problem i forsee with this is that once it is stuck on with thermal paste it will be impossible to remove the chip from the block. Oh well, she'll be right!

Just consentrate on the center of the heat spreader (thats were the heat is consentrated) so milling it to fit would be a waste of energy on your and the TEC's part.. I wouldn't advise popping the top off before getting IHS on results, your already working with some potentialy damaging hardware.

Try a thicker Cold Plate. A 1/4 to 3/8" thick Copper plate should help alot, as this gives some "Reserve" cold for when the cpu temp spikes. I used to run pelts all the time, all the way back to the P3 450 days, and found that the cold plate is key.

apart form all the stuff said above about the cooling: my opty 170 is from the exact same batch as yours. mine will go to 2.5 GHz @ 1.225V (0.125 below stock, 35°C load with a big typhoon), fully stable. after that i need lots of voltage to get it any higher. currently at 311x9 @ 1.525V. core 0 is also the weaker on mine. so i wouldn't say its a bad stepping, not terrific, but decent. now assuming that our procs are pretty identical, your mobo is probably is holding you back, seeing as mine will go to a bit more than 2.6 at stock V. if its possible to increase the voltage on the chipset i would do so, might help stability. i guess you could test if its your board by dropping to the 9x multi and bumping the "FSB" to 266 and testing for stability. if you fail its the board as the proc is still at 2.4 GHz