My head is swimming from reading here off and on for several days about these. 
I've had more than a dozen mostly tower here for my own use or just passing through over several years beginning long after the extended warranty ran out. Right now I have 4 working (3 tower, 1 SFF), one that won't post (tower), and the subject SFF that shuts itself off not long after navigating a cleared CMOS and starting memtest86+ from a Knoppix CD. By removing the OEM cooler and fan and using a tower set in their place, it will run longer, but even that bigger heavier heat sink gets blistering hot pretty fast.
I recapped my first (tower) 270 long before finding this forum, somewhere around the time I first found Carey Holzman's Abit cap article, probably more than five years ago. It's still working nicely. It's one of only 2 I've ever changed the 3 big Rubys on. The other, which still won't POST (A, C, D green, yellow B), I just did today, but that's a story for another day.
It's the 3 1800uf 16v Rubys showing no evidence of swelling or leakage I haven't changed that I think this post is mostly about. This SFF with 2.4GHz SL6WF Northwood P4 is service tag 6XYGT31 (case sticker and BIOS), which from what I gather is near the border between the later batches and the bad Nichicon batches. It came to me with a nasty 2200uf 10v OST in the PS that I changed early on, before knowing OST's reputation, and a bunch of other OSTs in the PS, including the big 470uf 200v pair, so I've since quit using that PS at least until I can get the motherboard running stable, after which I can get rid of more PS OSTs. In the mean time I'm testing with a recapped Earthwatts 430 that behaves well in every system tried that works with other power supplies new or recapped.
The biggest Nichicons on the motherboard (between card slots) were HM 1000uf 6.3v I forgot to check dates on before misplacing, and 4 a0332 HM 820uf 6.3v, since replaced with new HN. Everything higher uf is or was Rubycon. Biggest uf Nichicons remaining are 220. The others changed were the 9 1800uf 6.3v along the CPU socket (of which about half were swollen leakers), their two siblings by the 20 pin plug, and the about to burst 1500uf 6.3v near AGP and RAM slots. The 5 220uf 16v between the big Rubys and the 4-pin power plug are 85C Sanyos. The 680uf 10v near most 820s has the Matsushita symbol and looks OK.
I looked fairly hard and could find no obvious evidence of any burnt components.
I really would like this one SFF to be able to replace the working SFF I often use in case it decides to quit on me. One reason I even have these GX270s is they are among the last of the x86 machines with AGP slots. The GX280s and others newer have PCIe. Much of what I do is assuring Xorg and the DEs that run in it remain compatible with older hardware as they continue to evolve. Fewer working machines with AGP slots means more work, changing video cards and configuration settings to match different cards instead of changing machines, when working to pin blame or bisect regressions.
So, with 8 of the caps Tomcat's kit contains not yet replaced, my big interests at this point are:
1-is the hot CPU and shutdown what's more or less expected from a combination of poor PS impact on motherboard components, and 3 baked (but otherwise nice looking) original 1800uf 16v Rubycon MCZs surviving my partial recapping?
2-Is there any reasonably easy place to probe what voltage the CPU is seeing in this condition (since a cheapskate Dell BIOS won't provide any such information) in the short time between power up and overheat shutdown?
3-which caps not yet replaced, must I replace to get the CPU temp down, or in any event make it stop shutting down if CPU temp is not the reason for the shutdowns?
4-is there likely some other problem part causing overheat that replacing more caps won't solve?
5-which caps not yet replaced, not counting answer to #3, if any, should I replace?
6-should I quit trying?
Thanks for reading, and whatever answers can be offered!
I've had more than a dozen mostly tower here for my own use or just passing through over several years beginning long after the extended warranty ran out. Right now I have 4 working (3 tower, 1 SFF), one that won't post (tower), and the subject SFF that shuts itself off not long after navigating a cleared CMOS and starting memtest86+ from a Knoppix CD. By removing the OEM cooler and fan and using a tower set in their place, it will run longer, but even that bigger heavier heat sink gets blistering hot pretty fast.I recapped my first (tower) 270 long before finding this forum, somewhere around the time I first found Carey Holzman's Abit cap article, probably more than five years ago. It's still working nicely. It's one of only 2 I've ever changed the 3 big Rubys on. The other, which still won't POST (A, C, D green, yellow B), I just did today, but that's a story for another day.
It's the 3 1800uf 16v Rubys showing no evidence of swelling or leakage I haven't changed that I think this post is mostly about. This SFF with 2.4GHz SL6WF Northwood P4 is service tag 6XYGT31 (case sticker and BIOS), which from what I gather is near the border between the later batches and the bad Nichicon batches. It came to me with a nasty 2200uf 10v OST in the PS that I changed early on, before knowing OST's reputation, and a bunch of other OSTs in the PS, including the big 470uf 200v pair, so I've since quit using that PS at least until I can get the motherboard running stable, after which I can get rid of more PS OSTs. In the mean time I'm testing with a recapped Earthwatts 430 that behaves well in every system tried that works with other power supplies new or recapped.
The biggest Nichicons on the motherboard (between card slots) were HM 1000uf 6.3v I forgot to check dates on before misplacing, and 4 a0332 HM 820uf 6.3v, since replaced with new HN. Everything higher uf is or was Rubycon. Biggest uf Nichicons remaining are 220. The others changed were the 9 1800uf 6.3v along the CPU socket (of which about half were swollen leakers), their two siblings by the 20 pin plug, and the about to burst 1500uf 6.3v near AGP and RAM slots. The 5 220uf 16v between the big Rubys and the 4-pin power plug are 85C Sanyos. The 680uf 10v near most 820s has the Matsushita symbol and looks OK.
I looked fairly hard and could find no obvious evidence of any burnt components.
I really would like this one SFF to be able to replace the working SFF I often use in case it decides to quit on me. One reason I even have these GX270s is they are among the last of the x86 machines with AGP slots. The GX280s and others newer have PCIe. Much of what I do is assuring Xorg and the DEs that run in it remain compatible with older hardware as they continue to evolve. Fewer working machines with AGP slots means more work, changing video cards and configuration settings to match different cards instead of changing machines, when working to pin blame or bisect regressions.
So, with 8 of the caps Tomcat's kit contains not yet replaced, my big interests at this point are:
1-is the hot CPU and shutdown what's more or less expected from a combination of poor PS impact on motherboard components, and 3 baked (but otherwise nice looking) original 1800uf 16v Rubycon MCZs surviving my partial recapping?
2-Is there any reasonably easy place to probe what voltage the CPU is seeing in this condition (since a cheapskate Dell BIOS won't provide any such information) in the short time between power up and overheat shutdown?
3-which caps not yet replaced, must I replace to get the CPU temp down, or in any event make it stop shutting down if CPU temp is not the reason for the shutdowns?
4-is there likely some other problem part causing overheat that replacing more caps won't solve?
5-which caps not yet replaced, not counting answer to #3, if any, should I replace?
6-should I quit trying?
Thanks for reading, and whatever answers can be offered!
Comment