We can rebuild him. We have the technology. Help me PC re-build.
August 9, 2018 12:19 PM Subscribe
Let's say you have a not-brand-new-but-just-a-few-years-old gaming computer (the provenance is only important in that we didn't build it) that was a generous gift to someone in your household. For a while or so it worked just fine so despite there being no documentation all was good and YAY gift!
it started shutting down for no obvious reason, we determined it was overheating and the decision was made to see if we could re-build to readjust the air flow since that seemed to be the diagnosis. Sure enough when it was opened in eagerness, the fan wasn't working correctly, so things were moved, the fan was placed better, and everything was seemingly plugged back in and ready to go. Now the fan works but that's about all that happens when we plug it back in.
If you notice the passive voice in the second paragraph it's because I wasn't actively involved in most of it so I don't know exactly what was done. I've been drafted to try to fix it for my my puzzle solving ability and general "not afraid to try stuff" attitude. And so far I've determined that shit was not plugged back in correctly, some stuff may not be powered correctly at all (this isn't a "we need to hold the button to reset" situation).
The issue: My inside-the-box PC support is from 20 years ago and now pretty much limited to swapping out memory or a graphic card because I can look at where it was and where it needs to be. I can install one thing because I can read a thing or watch a video. I can identify a motherboard, graphics card, power supply, etc. But when it comes to having all the custom parts but not knowing where or what should be plugged in (and if what's already done is correct), I'm a little out of my depth.
So where to begin in 2018? I'm fairly tempted to just take it all apart and start again, but even with that, I'll need some general guidance. There's tons of stuff online but what's the best? Or even least bad? (I will even watch videos if it's showing me what to do and telling me what's been done - but a lot of them seem to be "here's this cord which I'm not naming which I'm going to plug into this port that I'm not naming and you can't really see it because my camera can't show what I'm doing and voila it works")
Any and all resources would be appreciated.
P.S. The right answer is probably to find a friend who can help with this or pay someone to do it but now that I've started, I remember how much I actually like doing this, so we're gonna try this tonight for now.
it started shutting down for no obvious reason, we determined it was overheating and the decision was made to see if we could re-build to readjust the air flow since that seemed to be the diagnosis. Sure enough when it was opened in eagerness, the fan wasn't working correctly, so things were moved, the fan was placed better, and everything was seemingly plugged back in and ready to go. Now the fan works but that's about all that happens when we plug it back in.
If you notice the passive voice in the second paragraph it's because I wasn't actively involved in most of it so I don't know exactly what was done. I've been drafted to try to fix it for my my puzzle solving ability and general "not afraid to try stuff" attitude. And so far I've determined that shit was not plugged back in correctly, some stuff may not be powered correctly at all (this isn't a "we need to hold the button to reset" situation).
The issue: My inside-the-box PC support is from 20 years ago and now pretty much limited to swapping out memory or a graphic card because I can look at where it was and where it needs to be. I can install one thing because I can read a thing or watch a video. I can identify a motherboard, graphics card, power supply, etc. But when it comes to having all the custom parts but not knowing where or what should be plugged in (and if what's already done is correct), I'm a little out of my depth.
So where to begin in 2018? I'm fairly tempted to just take it all apart and start again, but even with that, I'll need some general guidance. There's tons of stuff online but what's the best? Or even least bad? (I will even watch videos if it's showing me what to do and telling me what's been done - but a lot of them seem to be "here's this cord which I'm not naming which I'm going to plug into this port that I'm not naming and you can't really see it because my camera can't show what I'm doing and voila it works")
Any and all resources would be appreciated.
P.S. The right answer is probably to find a friend who can help with this or pay someone to do it but now that I've started, I remember how much I actually like doing this, so we're gonna try this tonight for now.
CPU reseat as Huffy Puffy mentioned is one thing. Power supply stuff is the other. Make sure the power supply cables are all plugged in, especially the big connector being plugged in firmly at the motherboard, the small yellow and black wired plug that goes into the motherboard as well and provides the CPU power (usually a 4, 6, or 8 pin connector). You say the fans power but nothing else, no video?
posted by deezil at 12:41 PM on August 9, 2018 [1 favorite]
posted by deezil at 12:41 PM on August 9, 2018 [1 favorite]
Is it dusty, or full of cat hair? Sometimes things overheat and a good clean-out can be important.
posted by wenestvedt at 12:44 PM on August 9, 2018
posted by wenestvedt at 12:44 PM on August 9, 2018
Are you going to use it for gaming or just general browsing and email? If not gaming, and the CPU has built-in graphics, just yank the video card and a lot of your over-heating problems will go away.
posted by JackFlash at 12:46 PM on August 9, 2018
posted by JackFlash at 12:46 PM on August 9, 2018
And so far I've determined that shit was not plugged back in correctly, some stuff may not be powered correctly at all
What kind of, uh, shit? Things like reset/power buttons, blinking lights, USB ports? Probably some of the motherboard headers were disconnected during cleaning. Look up which motherboard you have (if you can't tell from looking at it, CPU-Z will tell you), go to the manufacturer site, and download their terrible manual. The header connections should be detailed in there. (This is arguably the only thing most motherboard manuals are good for.)
Fans with two or three-pin connectors should plug directly into the motherboard. The connectors are usually clearly labeled "FAN".
The basics haven't really changed in the last 20 years, believe it or not. Things are more standardized these days so it's arguably easier to put things together.
If you reconnect/clean all the fans and it's still overheating, then yeah you might need to re-seat your CPU cooler, or buy a new one. These can be kind of tricky.
posted by neckro23 at 12:47 PM on August 9, 2018
What kind of, uh, shit? Things like reset/power buttons, blinking lights, USB ports? Probably some of the motherboard headers were disconnected during cleaning. Look up which motherboard you have (if you can't tell from looking at it, CPU-Z will tell you), go to the manufacturer site, and download their terrible manual. The header connections should be detailed in there. (This is arguably the only thing most motherboard manuals are good for.)
Fans with two or three-pin connectors should plug directly into the motherboard. The connectors are usually clearly labeled "FAN".
The basics haven't really changed in the last 20 years, believe it or not. Things are more standardized these days so it's arguably easier to put things together.
If you reconnect/clean all the fans and it's still overheating, then yeah you might need to re-seat your CPU cooler, or buy a new one. These can be kind of tricky.
posted by neckro23 at 12:47 PM on August 9, 2018
Can you find a copy of the mainboard user manual online (and one for the case too)? I'm thinking one of the sets of wires connecting the mainboard to the case buttons might have come undone or been improperly plugged.
Also, get some thermal paste and clean and remount the cpu cooler.
posted by under_petticoat_rule at 12:51 PM on August 9, 2018 [2 favorites]
Also, get some thermal paste and clean and remount the cpu cooler.
posted by under_petticoat_rule at 12:51 PM on August 9, 2018 [2 favorites]
Q: How long does it take a DEC repairman to fix a flat tire?
A: It depend on how many spare tires he brought with him, and how many of those are flat.
What I have found to be an effective troubleshooting methodology for "won't even turn on" PCs is reducible to: 1) start with the minimum set of components that can possibly work then 2) figure out which one is bad.
Minimum components (to get the thing to POST, not to do anything useful) are: mobo, CPU, RAM, power supply. Try just that. If it works, great -- now you can add things back in one at a time until it stops working (and then the thing you just added is probably the culprit).
If not:
Use your eyes and nose. Do you have leaky caps? Does the power regulator look cooked and smell like burnt hair? On a scale from good to bad, those are bad.
Next (as other posters have noted): check the interconnections between components: power supply to mobo, CPU to mobo, RAM to mobo. Remove each. Check connectors for damage or contamination. Replace and make sure they're in there and in there good.
Still no joy? The easiest way to figure out which component is bad is to swap them out. Having a deep junkpile makes this painless. Failing that, try the cheapest/easiest options first. You probably have multiple sticks of RAM. Pull all but one. If it still fails, swap it out for a different one.
Power supply failure is the next likely possibility. My advice would be to get a cheap power supply tester like this one. Mine has paid for itself many times over, and earned lots of goodwill by being loaned out. (False negatives are possible, but this will show you if you have a dead rail, which is the most common failure mode in my experience.)
Mobo and CPU are more expensive to swap. Before you go that far, I'd get a cheap POST test card and see if it gives you any clues.
Most problems are solved by this point. If yours isn't, think about whether you want to replace the mobo and gamble that the CPU is good, or if it makes more sense to just upgrade both.
posted by sourcequench at 12:51 PM on August 9, 2018 [4 favorites]
A: It depend on how many spare tires he brought with him, and how many of those are flat.
What I have found to be an effective troubleshooting methodology for "won't even turn on" PCs is reducible to: 1) start with the minimum set of components that can possibly work then 2) figure out which one is bad.
Minimum components (to get the thing to POST, not to do anything useful) are: mobo, CPU, RAM, power supply. Try just that. If it works, great -- now you can add things back in one at a time until it stops working (and then the thing you just added is probably the culprit).
If not:
Use your eyes and nose. Do you have leaky caps? Does the power regulator look cooked and smell like burnt hair? On a scale from good to bad, those are bad.
Next (as other posters have noted): check the interconnections between components: power supply to mobo, CPU to mobo, RAM to mobo. Remove each. Check connectors for damage or contamination. Replace and make sure they're in there and in there good.
Still no joy? The easiest way to figure out which component is bad is to swap them out. Having a deep junkpile makes this painless. Failing that, try the cheapest/easiest options first. You probably have multiple sticks of RAM. Pull all but one. If it still fails, swap it out for a different one.
Power supply failure is the next likely possibility. My advice would be to get a cheap power supply tester like this one. Mine has paid for itself many times over, and earned lots of goodwill by being loaned out. (False negatives are possible, but this will show you if you have a dead rail, which is the most common failure mode in my experience.)
Mobo and CPU are more expensive to swap. Before you go that far, I'd get a cheap POST test card and see if it gives you any clues.
Most problems are solved by this point. If yours isn't, think about whether you want to replace the mobo and gamble that the CPU is good, or if it makes more sense to just upgrade both.
posted by sourcequench at 12:51 PM on August 9, 2018 [4 favorites]
Power supply failure is the next likely possibility. My advice would be to get a cheap power supply tester like this one. Mine has paid for itself many times over, and earned lots of goodwill by being loaned out. (False negatives are possible, but this will show you if you have a dead rail, which is the most common failure mode in my experience.)
I have the PSU tester linked there and I want to explain a little more about it, because the documentation that comes with it is essentially nonexistent (although there is this tutorial video, which was very helpful to me). This thing tests whether the PSU is providing the correct voltages, but it does not test this under a realistic system load. In my previous question this tester showed my PSU as working when it in fact did not. So be warned.
posted by number9dream at 1:03 PM on August 9, 2018 [1 favorite]
I have the PSU tester linked there and I want to explain a little more about it, because the documentation that comes with it is essentially nonexistent (although there is this tutorial video, which was very helpful to me). This thing tests whether the PSU is providing the correct voltages, but it does not test this under a realistic system load. In my previous question this tester showed my PSU as working when it in fact did not. So be warned.
posted by number9dream at 1:03 PM on August 9, 2018 [1 favorite]
One thought, before even testing the power supply do an inventory of components to check if the ratings are sufficient. As for heat, run it all opened up in a cool room with a strong fan blowing on the components.
posted by sammyo at 2:29 PM on August 9, 2018
posted by sammyo at 2:29 PM on August 9, 2018
The POST test card sourcequench mentions is unlikely to work in any PC built in the last decade or so, as PCI slots are extremely rare and ISA slots are basically non-existent.
posted by Aleyn at 5:29 PM on August 9, 2018
posted by Aleyn at 5:29 PM on August 9, 2018
I agree with sourcequench's advice except for two minor things:
Minimum components (to get the thing to POST, not to do anything useful) are: mobo, CPU, RAM, power supply. Try just that.
Minimum components I always start with don't include the RAM; when tracking down unknown hardware problems I always start with only mobo, CPU and power supply and verify that I hear the "bad RAM" beep (often a long continuous whine, but Google will tell you what your mobo's beep codes mean) when I switch the thing on.
When I don't, the single most likely cause of that in my experience is a bad power supply.
I bought one of those $10 power supply testers too, but I have yet to have it correctly diagnose a bad PSU.
Most of what goes wrong inside power supplies is bad capacitors, just like it is for the voltage regulator sections on mobos. In a PSU, dried-out and therefore ineffective filter capacitors will not stop the PSU providing correct voltages to a tester that only checks it under no-load conditions. What they will do is allow massive amounts of switching noise to appear on the power rails as soon as the mobo draws any real current, and mobos don't like that, not one little bit.
Even the minimal amount of power required to run those parts of the POST that don't rely on the presence of RAM is closer to a realistic load on the PSU than is provided by a $10 tester, which will really only tell you whether the PSU is so grossly faulty that connecting it to a mobo risks damaging it with over-voltage.
If I do manage to get a bad-RAM beep out of the board, next step is to run it with just a single stick of RAM and verify that I can now get the beep code that says it's missing its video card.
And so on. Yes, it's very much like the DEC flat tyre joke. It's also the fastest way to get this done.
posted by flabdablet at 7:50 PM on August 9, 2018 [3 favorites]
Minimum components (to get the thing to POST, not to do anything useful) are: mobo, CPU, RAM, power supply. Try just that.
Minimum components I always start with don't include the RAM; when tracking down unknown hardware problems I always start with only mobo, CPU and power supply and verify that I hear the "bad RAM" beep (often a long continuous whine, but Google will tell you what your mobo's beep codes mean) when I switch the thing on.
When I don't, the single most likely cause of that in my experience is a bad power supply.
I bought one of those $10 power supply testers too, but I have yet to have it correctly diagnose a bad PSU.
Most of what goes wrong inside power supplies is bad capacitors, just like it is for the voltage regulator sections on mobos. In a PSU, dried-out and therefore ineffective filter capacitors will not stop the PSU providing correct voltages to a tester that only checks it under no-load conditions. What they will do is allow massive amounts of switching noise to appear on the power rails as soon as the mobo draws any real current, and mobos don't like that, not one little bit.
Even the minimal amount of power required to run those parts of the POST that don't rely on the presence of RAM is closer to a realistic load on the PSU than is provided by a $10 tester, which will really only tell you whether the PSU is so grossly faulty that connecting it to a mobo risks damaging it with over-voltage.
If I do manage to get a bad-RAM beep out of the board, next step is to run it with just a single stick of RAM and verify that I can now get the beep code that says it's missing its video card.
And so on. Yes, it's very much like the DEC flat tyre joke. It's also the fastest way to get this done.
posted by flabdablet at 7:50 PM on August 9, 2018 [3 favorites]
« Older Help me teach myself partial differential... | Ceramics studio programs for the less privileged? Newer »
This thread is closed to new comments.
If nothing at all happens, and you’ve been messing around with the fan that sits on the CPU/heat sink, I’d re-seat the CPU chip.
posted by Huffy Puffy at 12:24 PM on August 9, 2018 [2 favorites]