Bad Mount on IHS-less X2 - fixed!

Since upgrading my setup a month ago, I've been having stability issues. Games would crash, random lock-ups, and forget benchmarking; everything was flaky.

I loaded up CoreTemp ( http://www.ocforums.com/showthread.php?t=466707&highlight=coretemp ), and saw some pretty messed up temperatures. Even at load, Core 0 (1 in CoreTemp) was running super hot. At load, forget it - up to and over 70 C. And we're talking after just a few seconds at load.

High load temps (notice the Prime95 run - Core 0 at 70C, Core 1 at 47 C):







This was a great indication that I had not seated the Storm properly on my IHS-less X2 4200. So I checked and, sure enough, the Storm was bumping into the socket. I had test seated the Storm and CPU on the same model motherboard, and didn't have any problems. But after a close examination, the older MSI board I had tested on had a different socket than the newer one. We're talking about a 1mm difference in the length of the socket; on the older model, the block cleared the socket without a problem. On the newer model, the storm was bumping the socket.

So, I popped the storm off and removed the CPU. After cleaning both and re-applying a very thin layer of AS5 to the core, I popped the top of the socket off the MB (and almost broke it in the process). I then grabbed an exacto, and after marking how much I needed to remove, shaved away the excess material (very carefully, mind you). I'm not sure, but I think the socket tops are made of Nylon - very easy to screw this job up.

After putting the socket back together and re-seating the CPU and block, I booted up. Clearly, a huge difference:





I was even able to immediately squeeze another 50 Mhz out of the processor, and still keep temps lower. I'll be trying to up the OC later today. Lesson learned: Always double (triple!) check your block for proper mounting.

I should add that CoreTemp made it very easy for me to determine that I had a poor mount. I doubt if the block was even touching one half of the CPU core. Since the BIOS temp is, I believe, an average of the two core temps, I though I was just getting poor temps in general. In reality, I had one core cooling well and the other core almost not cooling at all.

Looking pretty solid now:

Those temps are at 1.4vcore? I just want to know for reference purposes. Mounting naked cores can by tricky. It's almost like you have to work with needle-nose under a microscope. What you posted is good to know. You may have possibly prevented someone from burning up an expensive dual core.

I used two methods when mounting the block on my DTR (no heatspreader either).

1) Mount the block once and then remove it and check the pattern in the thermal paste to make sure there's a full "finger"print.

2) Slip a piece of paper between the block and the cam box of the zif socket to make sure there's clearance.

My DFI has a lower profile socket, so that there's still about a millimeters' room between the block bottom and the cam box, so all went well there.

It's nice that you didn't burn up that purtifull processor, it could have gone much worse than that...

ƒÓÒl said:

2) Slip a piece of paper between the block and the cam box of the zif socket to make sure there's clearance.

Click to expand...

This method seems to work well; it's difficult to see how much clearance there is with everything mounted and in the case.

There was at least some contact with the CPU, but the heat draw was far from ideal. Prime95 would error out after 30 seconds or so on Core 0.

kimochii72 said:

Those temps are at 1.4vcore? I just want to know for reference purposes. Mounting naked cores can by tricky. It's almost like you have to work with needle-nose under a microscope. What you posted is good to know. You may have possibly prevented someone from burning up an expensive dual core.

Click to expand...

Temps are 1.400 VCore, possibly 1.400 + 3.3% due to the wacky MSI overclocking settings.

CPU-Z shows the VCore in the 1.460 - 1.470 range...