I recently upgraded a Athlon 64 to an opteron 180. The system is perfectly stable, runs for months at a time, on the Athlon. On the Opteron, the system is stable on low loads, but on high loads, I get random BSOD. The system errors vary, but it is usually blamed on a various device driver, or a IRQ_IS_LESS_THAN_OR_EQUAL, or some other things.

When these errors started I wasn't present at the system when it happened and, when it started I had no idea what the cause was. The errors had no repetitive pattern other than BSOD. I had to rule out a whole bunch of things.

I went through power supplies, Bios settings, memory reseating, removing PCI cards, etc. I bought a new Radeon just in case. I reset everything many times. In all of these configurations my system was stable when I put the athlon 64 back in. Interestingly, in my investingations, I found that memory slot 0 and 1 are much more stable than slot 2 and three. I run two sticks of Corsair DDR 400, for a total of 1 gig.

After installing analysis software, and adjusting systems settings in BIOS, I have found a few things. First, the system draws too many amps for it's setup. I have several different power supplies, and it won't even boot to Bios on any power supply that has less than 45 amps on the 5 volt line. This is a gigantic supply. I thought this supply might have been weak, so I bought a couple of new ones and my system won't even post to bios. It runs fine on these supplies on the Athlon 64. I have one supply with 20 amps on the 5 volts, and another with 35 amps. They are all greater than 500-550 watts. The calculators that I have found tell me that the system should run on about 26 amps on the 5 volt line. I have had a very difficult time finding amp requirements on the AMD web site. The builder guides strangely don't have guidance, just simply to test after the install. The processer have wattage given, and mine can dissipate 110 watts. Basic theory would have max 22 amps at 5 volts. This is if there were no amps from the 3.3v line, which is unlikely.

So I have determined that my Opteron seems to take too many amps.

Another thing is that my opteron runs very hot. It idles about 33, and that is about the case temp. But under small loads gets into the 50's and 60's. But when I put it under full load (virtualdub, half life, Everest test software) the temps get up into the 70's, with processer one getting about 10 degrees C higher than processer zero.

After 2-30 minutes at these temps, processer one will suddenly shoot up into the 80's, or higher, and if I don't shut down in time, I will get a forced shutdown from the processer heat.

All of this has been a huge amount of work. But it was sort of fun. And I have learned a tremendous amount from the process.

Does anyone think that my processer is not bad? It likely has an internal short or some other problem that soaks up power. I think I need a replacement.

I have always used stock heat sink and fan because I did not want to break my warrantee. I have tried a bigger heat sink and fan on the athlon 64 but it made no difference. I don't overclock. I just want stock power. I tried to underclock but it still overheated.

Here is my specs.

Motherboard: MSI K8N Neo 2 platinum. (Opteron 180 approved by MSI).
Video: was a Geforce 6600 GT, but I moved to a Radeon HD 2600 Pro.
Bios: all up to date.
Memory: Corsair DDR400 PC 3200, set to very conservative timings to try for more stability.
Power supply: various ones. Most powerful one and only one that runs the opteron is StarTek 550.

I have become very good at putting just the right amount of heat sink paste on. I use silver oxide from StarTech.

On everest the opteron posts some really impressive numbers, before it BSOD's.

Anybody have any ideas?

John.

Edited: 02/04/2008 at 03:17 AM by marcusjohn

Given those problems you'd had and that the Athlon runs fine I'd RMA it if it was mine

-------------------------
Asus A8N-SLI Deluxe
AMD Opteron 185
2GB Crucial Ballistix DDR 500 2.5 3 3 6
BFG 320MB 8800 GTS OC
150GB WD Raptor
250GB WDC SE16 SATA II
Thermalrite Northbridge Cooler
Thermaltake Armor V8003BWS

Thanks, Opieman, for the feedback.

I went ahead and RMA'd it. It just got there today. The RMA process was very reasonable, and the engineers on the phone were very friendly. I hope I get a good one back.

John.

Sounds like a messed up chip. I have upgraded via the same path as you all the way down to the same video card and had a few issues but nothing that was unsolvable.

My MOBO (ASUS A8V Deluxe) will step the RAM clock down if all banks are loaded. This actually worked in my favor. The Opteron 180 runs at 2.4 (12 x 200) With the ram underclocking by 20% it allowed me to push the CPU FSB up and bring it to 235 (stable) which was very do able with the head room left on the ram.

I think you may want to prepare for another round of issues when your new chip comes in. The opty is a solid CPU. It may be the MOBO. There are not bad chips that make it out the door but the QA on those is very tight.

I seldom overclock any CPU but the Opteron was so cool and stable I had to push it up. It runs well at 50c and never goes beyond 56 - 57c while thumping at 100%. I am interested in how this works out for you. I hope that you get it going and can enjoy the system. This is that last great fetch for the socket 939. Please post back and let us know.

Best of luck to you.

JQL

I received the new opteron 180. I got a brand new one in a brand new box. Interestingly, the heat sink was a new design.

I put it in. I am rock solid stable now, but I still seem to need to use the most conservative memory settings to keep it stable. For instance if I set the command rate to 1T, it will lock up on high loads. On 2T, timings 2-2-2-7, I can't make it BSOD.

It is stable on everest stability test, which was crashing in seconds to minutes before. I can run superPI to 32M without problem.

Idle temps are 32 degrees. Full load temps are 63 max, solid and steady. (Case temps of 30 or so).

I haven't shut down my computer in days and no problems at all.

My next step will be to run Virtualdub for hours at a time. This can be a torture test as well, but is the main purpose of this computer, as well as being my home ASP/Web/File server/OutlookServer.

Thanks everybody, and especially thanks to AMD for a good processor.

John.

I wonder if I need a better heat sink? I am giving up about 1000 mhz of memory bandwidth on 2T timings. If I can cool the processor more, would this get better? Would it even matter that much? Even though I am not overclocking, I am torturing the processor at full throttle for long periods of time. I wonder if the processor would last longer if I cool it better? I know I would break my warrantee with a better cooler, but would it last longer anyway?

Like you say, this is the last gasp of 939's. I probably will not be using this computer in three years, but maybe the kids will.

John.

Glad to hear you have the new CPU installed and getting some use out of it.

It sounds like it is running hot under a load. Anything in the 60's is hot for the Opty 180.

It will do it and all day long but that is warm for sure.

Let me check mine right now... Case temp is at 21c (70f) and CPU is 35c (95f). I did a few things when I pulled the Athlon 64 out and installed the Opteron.

I installed a new front fan (old one died years ago) and polished the heatsink where it contacts the CPU. I used AMD issued gray thermal compound. I actually scrapped it off the heatsink and then after I cleaned and polished the base of the stock heatsink I heated up the compound and dripped it on to the heatsink. I know you shouldn't reuse that stuff but it is better then the Radio Shack brand (IMO).

The case I use is old. It was built for a P3 system long ago. I have opened a port and installed a tube to allow about 20 to 40% of the air being pulled into the case to directly enter on the top of the CPU (as with many new case designs). This dropped the temp by 5 or 6 across the board.

What to do: Ensure you have good airflow through the case, cool air in the front (or side) and warm air out the back. More fans should be pulling warm air out then there are fans pushing cool air in - don't over do it or you could back draft the power supply, which in effect takes all the heat from the power supply and stores it in your CPU isgust;.

What else.... with the computer off but still warm gently push down on the heatsink and twist it - ever so gently - side to side and ensure the compound has spread out. In the old 486 systems I would really twist but these 939 pins are a lot more delicate. Be careful but give it a try.

One more thing, default the BIOS, run it and see where the temps are. Stability in these chips come first from memory settings then heat tolerance. Get your memory set first - running the fastest (stable timings and clock) then push it up a little at a time. Once it fails pull back down a bit - call it done. I think you are on the right track there.

I would not overclock the FSB. Running as hot as it is you might very well cook it in a short period of time.

Over clocking will or better stated "can" decrease the MTBF of a CPU. There is proof that long term over clocking actually causes the gates in the CPU's to grow little hairs that will short out. IMO anyone that overclocks is not going to keep a system long enough to see the failure from overclocking unless they are extreme and have a catastrophic failure like a power spike or cooling failure and the chip bakes.

Hope this helps. I agree - AMD makes a good chip. (For the money)

Best of luck
JQL

These are some great ideas. I am going to be cutting a hole for a tube that will blow down right on the heat sink. I think I will put a fan on it. This will give the heat sink 21C air, instead of 32C right now case temp. I can do this without messing with my stock heat sink and messing up my warrantee. I will make a tube out of something. My case already has four fans if you count the two on the power supply. One more on the CPU tube won't hurt anything.

Thank you.

John.

If the heat sink fan combo is new heat pipe equipped design you need not change it.
Its VERY good.
Look at the case fans.
disable cool n quiet if its active.

-------------------------
Asuss A8N SLI premium
Opteron 180 S939 X2 Toledo-Denmark
BFG 7900GT
OCZ ddr400 2Gigs