4GHz on AMD's 65nm process!

[ozzimark]

This would be a first, no? Since it is, it's time for the story of how this came to be, with an excessive number of smilies.

Before I get into it, here is the final result of my playtime for those who are lazy and don't want to read the whole post.
http://valid.x86-secret.com/show_oc.php?id=266059


The chip arrives, and is passed onto the quality control department. It passes!




A closer look at the processor itself. Heatspreader quality is rather poor, nothing that a few passes over some 600 grit can't fix




First up was some testing under a TDX. A pretty hard limit for stability at 3.36GHz was found with just a bit of grumbling at the temperatures



With the same ambient temps (like 63ºF ), 3.65Ghz was observed with 1.525v. More voltage was worse. I suppose heat was getting to it, so I opened a window. After dropping the load temp to 40ºC, I went for another maximum suicide run on water, and turned up with 3741MHz, which tied SillySider03 for the 5000+ Black Edition WR yesterday. I wasn't satisfied with simply tying with the fastest speed. All out is what I wanted, so all out is what I did.

Well, as all out as you can go with a 65nm AMD chip that is. Knowing very well that battling the cold bug with dry ice was going to be a very fun activity, I picked up some dry ice on the way home from work today.


(yeah, I know. Blocks of ice are lame to work with)


And set up the alu tube I have.




First thing I did was find the minimum temperature the chip could run at, which ended up being "238ºC" as reported by smartguardian, which is really -18ºC, but being a Brisbane, who knows how accurate the temperature sensors are.

After a few practice runs, I managed to get this validated:
3933.83MHz

Not a bad start, and exceptionally close to my goal for about an hour's worth of work. I'm getting hungry and need some dinner at this point, but quitting is for losers, so onward I push.

Pushing the temp down a few degrees to -12ºC and the voltage up to 1.6v + 103% for 1.648v (for those not familiar with the DFI NF590 SLI motherboard, it offers an INSANE number of vcore options), I could almost touch the magical number
3990.39MHz

Those last 10MHz were the most difficult 10MHz ever. After trying every voltage combination from 1.6v+102% to 1.6v+106% in 0.05 to 0.1% increments, and a few runs at slightly different temps with a few of the better voltages, I finally managed to break into the 4GHz region and stay stable enough to take a screenshot of it

4009.86MHz




The story ends here... for now