Jason Dunn
02-20-2007, 12:50 AM
Ah, the irony of it all. Yesterday I posted about the crazy eBay sale of bent and trashed CPUs (http://www.jasondunn.com/unique-cpu-245). Over the weekend I assembled a new Shuttle, an SN25P (http://www.memoryexpress.com/index.php?PageTag=&page=file&memx_menu=EmbedProductDetail.php&DisplayProductID=6052&SID=), which was to be my photo/video editing workstation. I put in the AMD 64 X2 Dual-Core 4400+ CPU, a socket 939 chip that I paid around $550 for in 2006, which now sells for $215. I wanted to maximize performance, so I started to overclock it - the Shuttle allows for a fine degree of overclocking, including voltage bumps in one-half volt increments. I rarely do overclocking with voltage bumps, because I'm worried about damaging the CPU. So I did my normal bus-only overclocking. I got about 100 mhz out of the CPU, but I wanted more, so I bumped the CPU voltage up half a volt and juiced the bus speed a bit more. I got it up to around 2480 mhz or so (stock is 2200 mhz), but in testing it wasn't very stable - it would lock up the system after a few hours. I let it run overnight on Friday night and Saturday morning I was staring at an ugly bluescreen (and this is running Vista Ultimate).
I played with it some more on Saturday, and it seemed to be more or less stable, but on Sunday it crashed again so I dropped the voltage back down to normal. I dropped the bus speed down to nearly normal, keeping around 50 mhz of extra speed. Everything seemed normal. Then late Sunday afternoon I went down to my office and tried to use the machine but the wireless mouse and keyboard wouldn't work. I re-associated the mouse and keyboard, but it didn't help. I pressed the power button on the machine and nothing happened. That's when I realized that it was actually frozen (I should have looked at the system clock). I pressed the reset button and it rebooted, but wouldn't post - the fans were stuck in high RPM mode, just like the Velocity Micro was doing when it had the CPU problems. My shoulders slumped when I realized that I likely destroyed the CPU with my overclocking attempts. I tried to remove the heat sink, but it wouldn't budge despite my gentle wiggling, so I had to apply a bit more force than normal. It came free with a big sucking sound (how appropriate) and tore the CPU free of the socket without me releasing the socket lever (which was impossible to get at without first removing the CPU cooler). Almost a dozen pins were bent and maimed by the bad removal. I don't know if I put on too much thermal paste, but I've never seen anything quite like that before.<!>
I still had the 3500+ socket 939 CPU, so I put that into the Shuttle and it booted up perfectly. So it's definitely the CPU, and between the pins being bent/broken and the fact that it wouldn't boot, it's a dead CPU. That was a rather expensive learning experience. :-( On the plus side, since I no longer have a socket 939 to power the Shuttle, I don't need it any more - I returned it to Memory Express today. That means I can get a new Shuttle, with a new, modern CPU. I've asked my Shuttle contact for an SD37P2 (http://hq1.shuttle.com/products_page03.jsp?PLLI=508&PI=639) - fingers crossed...
I played with it some more on Saturday, and it seemed to be more or less stable, but on Sunday it crashed again so I dropped the voltage back down to normal. I dropped the bus speed down to nearly normal, keeping around 50 mhz of extra speed. Everything seemed normal. Then late Sunday afternoon I went down to my office and tried to use the machine but the wireless mouse and keyboard wouldn't work. I re-associated the mouse and keyboard, but it didn't help. I pressed the power button on the machine and nothing happened. That's when I realized that it was actually frozen (I should have looked at the system clock). I pressed the reset button and it rebooted, but wouldn't post - the fans were stuck in high RPM mode, just like the Velocity Micro was doing when it had the CPU problems. My shoulders slumped when I realized that I likely destroyed the CPU with my overclocking attempts. I tried to remove the heat sink, but it wouldn't budge despite my gentle wiggling, so I had to apply a bit more force than normal. It came free with a big sucking sound (how appropriate) and tore the CPU free of the socket without me releasing the socket lever (which was impossible to get at without first removing the CPU cooler). Almost a dozen pins were bent and maimed by the bad removal. I don't know if I put on too much thermal paste, but I've never seen anything quite like that before.<!>
I still had the 3500+ socket 939 CPU, so I put that into the Shuttle and it booted up perfectly. So it's definitely the CPU, and between the pins being bent/broken and the fact that it wouldn't boot, it's a dead CPU. That was a rather expensive learning experience. :-( On the plus side, since I no longer have a socket 939 to power the Shuttle, I don't need it any more - I returned it to Memory Express today. That means I can get a new Shuttle, with a new, modern CPU. I've asked my Shuttle contact for an SD37P2 (http://hq1.shuttle.com/products_page03.jsp?PLLI=508&PI=639) - fingers crossed...