Opteron 165 overclocking help

[phanna]

Hello. I just recently purchased a new opteron 165 cpu, stepping ccbbe 0610 dpmw, which i heard overclock pretty well. My system setup is as follows:

Epox 9npa+ Ultra
2x1GB OCZ EL Platinum XTC DDR400 2-3-5-2
WD raptor 74GB, Seagate 120GB
Radeon X800XL
Soundblaster Audigy2 ZS
16X NEC DVD burner
Antec NeoPower 480W
XP-90 for cooling with 80mm fan, 2 120MM fans and 1 80mm fan in case.

I previously had a venice 3000+ in here that would do 8.5x306 (2600MHz) @ 1.5V.

Anyway, so I plug in this new Opteron, and start stepping up the speed. I got it all the way up to 2.6GHz on 1.35 volts dual prime stable for 12hrs. So far so good. Now, is the strange part. At 300x9 (2.7), this thing will not pass dual prime for more than 3 mins. I tried voltages all the way up to 1.55, was afraid to go higher, but it didnt seem to make a difference anyway. I found it very odd that it would go so high with basically stock vcore, then just hit a wall all the sudden, with voltage increases not helping. Has anyone had a similar issue with these things? Or is there just a part in my system that may be holding me back? I dont think it is the memory because i use the 2:3 divider so it stays at 200, plus the memory was never an issue with my old chip. Maybe i need a better power supply? or is this just the chips max? Any suggestions are welcome. Thanks!

 

[lyssword]

maybe ur mobo's fsb maxed out

 

[myocardia]

Well, since you already know that your motherboard and RAM will handle those speeds, that leaves only two things: your processor, and your power supply. Chances are very good that your psu is what's holding you back. Unfortunately, the only way to find out is to buy a better one.

 

[nyker96]

I think His Neo 480 should be able to handle the load. My guess is the CPU hit a wall.

 

[Noubourne]

The NEO 480 works OK for me, I get 2.8 out of mine at near stock volts, but I tend to hit 60C on the second core if I push more than 1.4v through it (1.375v+104% in BIOS, reads 1.38 or so in CPU-Z which is stable dualprime 8hrs), and that causes errors - pretty sure it's the TEMP that is holding me back. 2 cores seems to get pretty hot.

Been thinking about going water. With 1.45v and 3.1Ghz I can boot into Windows and do SuperPi 1M, but I can't do dual-prime for even 10 seconds, and it crashes. I bet with water I could be a lot closer to 3Ghz. Thinking about it...

 

[SparkyJJO]

What stepping is your opteron? What are your temps? If your temps are getting too high it can cause prime to fail. Might want to use coretemp as well, that is how I found out that my IHS was badly mounted.

 

[phanna]

Thanks for the replies. I know its not the HT bus, as I have taken it up to 333 before and it was fine. I also know its not the memory. I could try a different PSU at some point, I guess its worth a shot, but the one in here is pretty decent, and I really dont have all that much stuff in this box. I tried CoreTemp, and it reports each core at 27C idle and 43C while dual priming. Something I just noticed though, and I'm not sure if this is normal or not. In the BIOS, I have the voltage at 1.35. In Core Temp, it reads 1.35 as well. In CPU-Z, it reads about 1.38 under load, and 1.41 while idle. In USDM (epox tool) it reads 1.39 under load, and about 1.43 while idle. Does anyone know whats up with that? Which tool is right? Why are they all over the place?

 

[dennilfloss]

This wall phenomenon is quite common.

 

[SparkyJJO]

coretemp reads the CPU VID - what you have it set at in the bios. CPU-Z reads actual, and a little variation between two different applications is normal. Sounds to me like your board overvolts the CPU a little, I know mine does and it is fairly common actually. The voltage dropping a little under load is normal because you are putting more stress on the PSU, mine fluctuates about that much.

Lucky you on having a good IHS mount

 

[Munky]

I have the same "wall" issue with my Opteron 165. I know none of my other components are holding me back, so it must be a common occurence with Opterons.

 

[aka1nas]

Thats about right for that chip on most boards. IME, On higher end overclocks, trying to isolate component speeds doesn't always work reliably as the OC on the chip affects the memory controller, etc.

 

[phanna]

Yea, I think it's just the CPU. I'm not upset about 2.6 at all though, was hoping for more, but 2.6 will do me fine until I upgrade to quad core sometime in early 2008 or so. It's still a nice processor for 150 bucks Thanks again guys.

 

[Noubourne]

8.5 multi?

I thought you weren't supposed to use half multipliers on A64 chips? Just noticed that you're using that.

I guess I would be surprised if 2.6 was all you could get out of a CCBBE. What temps do you get at load again?

If you leave the divider the same and kick the multi down you can test your RAM at that bus speed for stability. If it still fails, RAM would be the culprit.