mitro's adventures in overclocking

[mitro]

Well today I got a new processor, an Opteron 165. I’ve been watching the trends lately to see what models of the dual cores were doing well and recently its been the 165s. And I knew what steppings were doing the best. For those who have no idea what I’m talking about, steppings are a code on the CPU that refers to the week manufactured and the fabrication plant it came from (as well as other info). Many people were getting good results with 165s with the 0610 DPMW stepping. So I took a chance and ordered a 165 from Monarch Computers and guess what shows up:



Time to swap my best 3800 X2 out of my main computer (“Project SR-71”) and see what this sucker will do! :P

[mitro]

So the first step was to lower the overclock of the computer to a level that I know where it would start up no matter how good or bad the new chip would be. Which was 2.5GHz (278x9). THEN shut off the computer and removed my 3800 X2.



Notice the 3800 looks a little different? Its integrated heat spreader (IHS) has been removed. More about that later, since the 165 will soon look the same.

[mitro]

So I hit the power button and she started right up (Always a tense 10 seconds). I always test stability with two instances of Prime95, or the prettied-up version of it: SP2004.

I’ve found that if I can get 5-10 minutes of SP2004 that it will ATLEAST boot into windows 100MHz higher at the same voltage. This speeds up the process of finding the limits of the processor. Also, at this point I’m running the memory on a low divider so I can push the limits of only on thing at a time: the CPU.

Well here we are at 2.5GHz on stock voltage:



I also am not liking the cpu temperature as its 5C higher than my 3800 when its getting 1.5v. I'll take care of the temps later. :P

[mitro]

The next 100MHz, still on stock voltage:



Ahh an FX-60 for $800 less

[mitro]

How about a 50% overclock? OK… no problem:



At this point I know this is a better chip than even my best 3800 X2.

[mitro]

Now things are just getting silly! 2.8GHz at stock voltage:

[mitro]

Well at 2.9 and stock voltage I found the limit. SP2004 crashed after 3 minutes. OK… time for a little more voltage then. I set the voltage to 1.4 in the BIOS but it was still reading 1.36v. So other than the “auto” voltage its off by -.05v. So then I set it for 1.45v (1.4v reported):



Back in business. Those temps though…

[mitro]

Well after trying 3GHz and failing with voltages up to 1.5 (reported) I knew it was time for this processor to go naked! And by “naked” I mean it was time to remove the integrated heatspreader (IHS). The IHS does work to spread out the heat from the core, but that’s what we have heatsinks or whatever for. The IHS just adds one more layer for heat to pass through to escape and no matter how perfectly its done, it always retains more heat.

This is something I do not recommend and should ONLY be done if you can afford to lose the processor, because there’s a high probability of that if you don’t know what you are doing. If you want to know more about how to do it then search Google, because I’m not going to be held responsible. :P



In the process of doing this I bent a pin on the bottom side and didn’t know it. I managed to fit the CPU back in the socket (which made it worse) and almost broke the pin off after taking it back apart and realizing what I’d done. It would have been a real shame to have killed this one.

[mitro]

Well… the results are extremely good. An approximate 10C drop in processor temps and much more stable. 3GHz with 1.5v (1.45v reported):



As of right now it’s been Priming at that speed/volts for 3 hours error free! Woo!

I think that’s it for today. Tomorrow its time to get ready for the Championship. :P

EDIT: I stopped Priming after 6 hours with no errors and fired Boinc back up. If I have any errors all I'll need to do is bump the voltage up .025v.

[NVM]

good stuff. i was going to give it a try myself but i'm stuck with stock cooling.