I just built another myth frontend/slave backend and for the most part everything is fine. Specs:

Asus A8NE mobo (939)
AMD X2 3800
200GB IDE drive
2 Kworld ATSC 110 tuner cards
1 Hauppage 150
Geforce 6200LE
1 firewire PCI card
DVD+/-RW Dual Layer drive
1 USB Rumble gamepad (love mythgames - I'm still addicted to Street Fighter)
Antec 350 Watt SmartPower ATX12V power

When I watch hi-def content for a while, or when I try to transcode and archive HD files the computer crashes and shuts off. Seems to be related to HI CPU usage. (If I try to play a game while archiving an HD file to dvd the thing dies QUICK). Any suggestions on what the problem is? I'm guessing I need a better PSU? Any recommendations on wattage and brand if so?

I've never used one, but some people tend to say that Seasonic is a good PSU brand. They're definitely a bit pricey, but the efficiency and noise ratings (which are supplied by the manufacturer, so may be misstated) look good. You may also have some iffy RAM. My HTPC is essentially useless now and I'm too busy to rebuild it (but my Myth backend is still running - an ages old 1GHz Pentium 3). It's due to a combination of RAM, mobo, and CPU that has become sensitive to cold, sensitive to heat, and sensitive to stray glances of an odd nature. Sometimes it will lock up at the same part of the same video three times in a row, then play through it the fourth time.

So yeah, it could be the power supply, but it could also be something odd between your CPU, your motherboard, and your RAM. There was also a time when certain PCI cards (Creative SBLive) in certain motherboards (Via chipsets) would cause Ethernet data corruption - so basically, try to isolate it to a single component. If the PSU is easiest to try first, then try the PSU, but if you've got another system around, try swapping out RAM. Also think back to when the CPU was installed - is it possible that the heatsink compound isn't making a good enough contact between the CPU and heatsink? The Arctic Silver 3 in my HTPC seems to contribute to the problems, as the theater room is unheated when not in use (so it gets cold).

One final note: you can guesstimate wattage by looking at the specs for all of your parts. Hard drive data sheets typically list a max current draw, CPUs and mobos often have a minimum current per power rail rating (i.e. the 12V line must supply at least X amps and the 5V+3.3V must supply at least Y watts). Take the biggest numbers and add them up. If it seems too high, consider that it's rare for all of your components to be using their max simultaneously, except when you start doing everything at once (HDD reading and writing HD video, video card doing 3D work for a game, CPU transcoding video and playing the game, etc.).

Edit: wow, this was my 200th post. Apparently that makes me a "Senior Member." (Or maybe I just didn't notice the transition when it happened).

I'd also try to see what temps are like. Perhaps the CPU is overheating?

I didn't think the CPU would overheat because the CPU cores rarely go over 60% (usually stay between 45 and 55 when watching hi-def - with load between 1.0 and 1.4). But I finally got around to figuring out how to configuring lm-sensors and it reports my MOBO spirals up above 70C, but my CPU stays around 37. I found this to be suspect so I checked the BIOS, and I'm pretty sure those readings are reversed (the BIOS reported stable MOBO temp and high CPU temp). I'll trust the BIOS. Any idea why the temp would be so hot? This doesn't seem like a ton of load. And oddly I don't hear the CPU fan get louder either... Maybe the ASUS Q-fan isn't working properly. Or maybe I need to replace the fan/heatsink.

I've had a couple of video cards that caused intermittent problems like this. The problem was the heatsink compound not being properly applied to the heatsink (the compound was not evenly spread). When I took it apart and put the stuff on right, I didn't have any more problems. This could be true for the CPU also, but I've always built my own computers so the CPU heatsink compound is always evenly spread.

You do have to take the heatsink off to check this, however, so you better have the compound ready if you do.

That's what I'll do. when I find a good time for my backend to be down for a bit I'm going to check the CPU fan and heatsink. Thanks for all of your help. I'm pretty sure I know what the problem is - now just figuring out what causes it and how to fix it... If nothing else works I'll by a new CPU/FAN combo and an another quiet fan. The case I have has a vent and a space for an additional fan right above the cpu, so there should be a way for me to keep this cooler...

You've got 3 add in cards, a dual core cpu, PCIe graphics card and... how much ram? Ram takes up a surprising amount of power, btw. All in all,, sounds like 350W would be marginal at best for this type of set up.

Try this calculator: http://extreme.outervision.com/
I put in as much info as you provided, and assumed 2 sticks of ram (It is a dual channel mobo) and cam up with a PSU requirement of 381 watts. Seeing as voltages can get funky when at the top limit, you'ld want to give yourself some head room over that, so try a good 420 watt (or greater) power supply.

You've got 3 add in cards, a dual core cpu, PCIe graphics card and... how much ram? Ram takes up a surprising amount of power, btw. All in all,, sounds like 350W would be marginal at best for this type of set up.

Try this calculator: http://extreme.outervision.com/
I put in as much info as you provided, and assumed 2 sticks of ram (It is a dual channel mobo) and cam up with a PSU requirement of 381 watts. Seeing as voltages can get funky when at the top limit, you'ld want to give yourself some head room over that, so try a good 420 watt (or greater) power supply.

I used that calculator and unless I assume I have 100% CPU utilization and a PCIe graphics card in addition to a regular graphics card I don't get to te 381 watts. I get numbers closer to 300. At 60% CPU utilization (which is where my computer has been failing) I get around 280 watts. If the power supply is too weak would it cause the CPU to overheat so quickly?

I was originally suspicious of the PSU because it just seemed underpowered for all I've got, but this calculator makes me think it is in hte ballpark, and the CPU temp has me thinking cooling is the culprit

I tried that calculator. Forget it. It's not even close. Said I need 199W PSU and that was with my machine loaded with 4 tuners, raod controller 3 HDD's etc., etc. The 400W PSU I originally had would just barely boot and freeze with nothing connected but the video card..... Worked fine with a 600W PSU.. AMD tech support even recommended a 475W or higher.

I used that calculator and unless I assume I have 100% CPU utilization and a PCIe graphics card in addition to a regular graphics card I don't get to te 381 watts. I get numbers closer to 300. At 60% CPU utilization (which is where my computer has been failing) I get around 280 watts. If the power supply is too weak would it cause the CPU to overheat so quickly?

I was originally suspicious of the PSU because it just seemed underpowered for all I've got, but this calculator makes me think it is in hte ballpark, and the CPU temp has me thinking cooling is the culprit

First, when decoding or transcoding mpeg (as you state you're doing when it crashes) you WILL occasionally run the cpu at 100%.

Second, Most 939 mobos do not have integrated graphics (I don't know of any, but then again I don't know everything), so yes, you have a PCIe Graphics card.

Third, 60% cpu utilization does not (at least in this dimension) imply 60% psu utilization. In fact, if compute intensive task are off loaded to pci cards and gpus, you're psu could easily max out while the cpu is sitting almost idle.

Forth, you never did mention how much ram you have, and what type. Bare in mind, low cost double sided ram will count as 2 sticks, even though it physically occupies one slot.

Another thing to remember is tv cards use much more power then typical pci cards, as they have on board processors for encode/decode operations. The 5 watts allowed in the calculator is pretty weak.

Finally, I did make an error in that I entered a dual processor config AND identified a dual core processor.

The symptom described is almost always indicative of psu shutdown. I own a psu tester myself and have to tell you that literally 6 out of 10 psu's I buy get returned because they are not up to snuff. It's the one part of the system people seem to think they can save money on, which makes it the first place I look when someone has a problem.

Also, is your PSU a Smart Power or a Smart Power 2.0? (i.e., 20pin or 24 pin?) Just asking... but if you answer 20 pin and you have ANYTHING installed other then a graphics card... well, you know what to do.

I'm not a noobie, I've built more then a few of these.... get a bigger power supply.

Regarding overheating... AMD CPU's are VERY heat tolerant. you can run them at 70c indefinately, and in a pinch you could use them to make pasta (just kidding, kinda). They don't "shut down" when they "overheat" as much as they are turned off by the system bios because it's gone beyond a certain preset limit. DO make sure they thermal paste is properly applied, but unless you overclock like an idiot and run Prime95 24 hours a day with the cpu fan burnt out, this is seldom a problem.

You are far more likely to overheat the south bridge chipset. In fact, I would venture that overclocking aside, 100% of genuine overheat problems are south bridge chipset problems, except no one looks there. Most HTPC will have a passive cooler on the south bridge because the little fan coolers make an astonishing amount of noise. Those chip are more sensitive to overheating.

*** If you know what you're doing, you can safely raise the cpu shutdown temp 5c or so. Overclockers do it all the time. In your case i wouldn't recommend it... your not overclocking and the temps seem ok. NEVER raise the chipset threshold.

I don't think you read my response properly. I know I have a PCIe card (I bought it and installed it). what I said was getting at is that calculator has a separate slot for an onboard video and PCIe and the only way I could get to the number you stated was to say I have both (along with 100% CPU utilization), which I don't.

I have monitored my CPU and as I stated in an earlier post post it only never goes over 60% when doing the activities I described where it shutdowns. It never reaches 100% doing the activities I have described. I'm aware certain activities can push it that far but that does not characterize the behavior that has been shutting it down. If it went to 100% I wouldn't be as surprised to see the CPU temps go up.

I didn't feel the need to mention the number of sticks since I now how to use a dropdown box and select the number of sticks myself using the calculator. There are 4.

It's a 24 pin power supply. I tend to agree with you that is the power supply, as the title of this thread suggests this is my first guess. However, the CPU temp spiking so quickly gave me reason to think it might be a cooling problem (the CPU shutdown temp is 72), but I know a weak power supply might be causing that too. At this point, I just have to test using a new PSU and/or checking the heatsink. When there is a good time for me to have it down I will.

I think I'll skip raising the CPU shutdown temp and try to get it to behave rpoperly. at 60% utilization for 15 minutes it seems to me it shouldn't raise up to 70C.

I'm not a newbie either, and although my degrees and previous professional experience are in software, I'm familiar with hardware and have built my fair share of computers. But I appreciate your thorough responses.

In short - swap out the PSU and check the heasink connectivity. which is what I've been planning to do.

For what it's worth, my CPU running with a Zalman heatsink (supposedly quiet, but hot) never gets over 60C. If you decide to replace your CPU cooler, I would recommend the Blue Orb - it's giant (weighs over a pound), but keeps a Pentium D 820 at 35 degrees.

The actual purpose of my post is noting that if your motherboard is supposed to vary the CPU fan speed with CPU temperature, it may not work while Linux is running. "Back in the day" my laptop's fan wouldn't come on at all when Linux was running, unless it was already on when Linux booted. (I actually burned my leg the first time I tried playing a DVD, which is how I discovered the problem...)

Thanks for the recommendations. I have been monitoring my CPU RPM with lm-sensors and it does seem to increase (I guess it must not be a loud fan, I couldn't hear it increase at all!). Anyway, time to go shopping for a PSU and maybe a cooler.

For those who care, after swapping out heat sink, PSU, RAM, etc. etc. I finally found what I think is the problem. My firewire PCIe card apparently was the problem. Sometimes I would have a hard time even booting up. Took that card out, and no problems booting or anything so far. I'll do more rigourous testing soon. I need to get a new PCIe firewire card, however, cause i use that to tune control my cable box which is very important.