Recently swapped a 3500 Venice (had been running at 2.6ghz) out of my rig and replaced it with a 148 Opteron. When making the switch I entered the bios and returned all the CPU multi, Vcore, VID settings etc to default before removing the 3500. All of my memory options and timings were not disturbed at all as the board has been rock solid. I then installed the 148 and mounted a Thermalright XP120 and rebooted without any issues. The board was supplying the CPU with betweeen 1.35 and 1.36Vcore to the chip at default settings.
This is my third opteron, I have had excellent results with my other 148 (running 2.9ghz+ in an asrock 939 Dual at slightly over default Vcore around 1.4V according to the Asrock bios and a few monitoring utilities), my second a 146 (running 2.6GHZ in an EVGA NF41 SLI AX ((Jetway clone)) at pretty much default Vcore as well).
Heres where this gets a bit interesting. After my initial boot I decided I would see just how far I could take the chip at default Vcore, not too darn far surpriseingly enuff. I thouhgt it may be a limitation of my memory whch I had been running 1:1 with the 3500 Venice but it hadn't caused me any issue previously so I started thinking onchip memory controller. I ended up feeding the opteron pretty much the same voltage I had be feeding the Venice just to get to 2.5ghz
CPU startup Vis 1.425
CPU Vid 1.425
CPU specila Vid control 110%
Needless to say these voltages fed to the opty raised my temps significantly even with the XP120. I set my RAM to 166 (yielded around 194 mem bus speeds) and met with better success then moved to 180 around on the RAM (yielded 230mhz membus), 250FSB on the CPU and all seemed well but still required significant voltage (those listed above. I left the machine running overnight folding after completeing sevearl loops of memtest without error. Sat down this morning thinking it would be nice to try lowering my voltage because of my temps and guess what. I slowly lowered my voltage between boots expecting a non booting issue like I encounterd last night. Guess what, now running 250X11, mem bus 230, 7-3-3-2.5 1T Vcore 1.35 CPU special Vid control @104% HTX4 =1000. I guess maybe there really is something to burning a chip in as my CPU temps under load with the XP120 (folding) is 47C which is a far cry from the 60+ temps I was getting before I started lowering my Vcore. I should also mention that I am running 2X1gig OCZ 3200 (OCZ4001024ELPE) and that the machine is rock solid at 2750.
A couple of questions.
1. Why would a chip that required 1.425V, CPU special vid control of 110% sudenly require significantly less voltage to maintain the same clock less than 24 hours later????
2. Any suggestions as to how I can get this beast to 3.0GHZ without watercooling?
I am going to attempt to get the membus to 250mhz from its current setting so that its runnning 1:1 with the system bus, I'llmost likely try relaxing my mem timnigs a bit to do that but I can see where I'm gonna have to start dropping the HT from X4 to X3 in all likleyhood. I've heard differeing takes on performance hits from this. Does the system truely suffer a alarge performance hit when going from X4 to X3 on the HT bus?
Any help would be appreciated.