Advice on Noisy, Overheating Shuttle PC

[wtx]

I've been having trouble finding good advice for this, so any suggestions would be greatly appreciated.

The build is a ~3yo Shuttle S33G5 with a Q8300 2.5GHz C2Quad + 4GB DDR2 and a 5.25" SATA Bluray drive. I recently added an Asus fanless GT430 and replaced the magnetic drive with a 120GB Sandisk SSD. It sits in a closed but fan vented media cabinet - the space is 9"H x 17"D x 27"W. I'm guessing, but temps inside the cabinet are probably around 80-85F.

The software setup is Windows 7 MC + MediaBrowser + MPC-HC + LAV Filters + madVR. Recently, watching a 1080p movie with the Shuttle's case on caused the system to overheat and restart twice during the movie. A second test with the case off was fine, and I grabbed HWMonitor data as it was finishing. Although it doesn't list CPU Core temps, they were all below 60°C.

• Hardware monitor ITE IT87
  Voltage 0 1.17 Volts [0x49] (CPU VCORE)
  Voltage 1 1.26 Volts [0x4F] (VIN1)
  Voltage 2 3.28 Volts [0xCD] (+3.3V)
  Voltage 3 4.92 Volts [0xB7] (+5V)
  Voltage 4 11.97 Volts [0xBB] (+12V)
  Voltage 5 -2.94 Volts [0x2E] (-12V)
  Voltage 6 -7.62 Volts [0x77] (-5V)
  Voltage 7 4.92 Volts [0xB7] (+5V VCCH)
  Voltage 8 3.22 Volts [0xC9] (VBAT)
  Temperature 0 57°C (134°F) [0x39] (TMPIN0)
  Temperature 1 66°C (150°F) [0x42] (TMPIN1)
  Fan 0 2192 RPM [0x134] (FANIN0)
  Fan PWM 0 0 pc [0x0] (FANPWM0)
  Fan PWM 1 56 pc [0x48] (FANPWM1)
  Fan PWM 2 99 pc [0x7F] (FANPWM2)
  
  Hardware monitor ACPI
  Temperature 0 40°C (104°F) [0xC3C] (THRM)
  
  Hardware monitor NVIDIA NVAPI
  Voltage 0 0.99 Volts [0x3DE] (VIN0)
  Temperature 0 71°C (159°F) [0x47] (TMPIN0)

This looks really hot to me. My best guess as to what's happening is that the GPU is getting really hot decoding the video, along with the rusty old PSU supplying power to it, and because the CPU is idle (GPU doing all the work), the only fan (I.C.E.) is blowing really low RPM and not really circulating enough air. So with the case on it overheats.

Obviously I need to replace the PSU since it gets really hot and is responsible for most of the system noise. But will that be enough to keep this cool? I have some other ideas, but I'm not sure what will be the best move - I could mod in a top or side exhaust fan, replace PSU with picoPSU (will that be enough power?), swap everything to a new case with better airflow (what will take the Shuttle mobo?). Or anything else within reason (not replacing all the components). If you have tackled a similar problem please enlighten me!

Thank you!

-W

[MikeC]

with the Shuttle's case on

By case, do you mean cover? That what it seems like, and my comment assumes this.

Your plan on swapping the PSU w/ a picoPSU may help. Those stock Shuttle PSUs were inefficient (certainly by today's standards) and noisy. The pico puts all the conversion heat outside the case, w/ the AC/DC adapter out of the media cabinet, it should be a little cooler in there.

But if the GPU is overheating due too inadequate airflow, a quiet side fan might be a better, simpler, cheaper solution.

[HFat]

Many GPUs wouldn't mind 70C. I don't know about yours but it shouldn't be picky if it's designed to be fanless. And I don't know if your setup under-reports GPU temperature. But your PSU might be the only part that overheats.

[wtx]

Mike, thank you for the response. Yes, I meant cover. I'm not certain what tripped the restart, but I've read that 70C is hot-ish yet reasonable for a fanless card. I am hoping it is just this shoddy power supply.

Can a picoPSU supply enough wattage for all these parts? I have seen similar configs on SPCR forums using them, but everything I read suggests I should need 250-300W.

I have also seen posts about undervolting CPU, RAM, and GPU to reduce heat and power consumption. Would you recommend this? I don't need max performance.

Thanks again.

[MikeC]

Those power estimates are way bloated. I run a PC w/ a Q8400 with multiple drives and a much hotter video card and don't think I've seen higher than ~160W AC on my kill-a-watt ($20 power meter. well worthwhile!)

A GT430 has a TDP of 56W, much lower at idle or video play. My guess is in HD media play, your system draws well under 80W. A 150W picopsu w/ 12-16A 12V adapter should be fine -- http://www.mini-box.com/picoPSU-150-XT

[wtx]

Okay, based on these suggestions I think I will go with a picoPSU and add a case fan, just to be safe. I'll pick up a kill-a-watt also.

Would you recommend top exhaust, gfx side, or psu side for the fan? Would it also be advisable to go up to the PW-200-M picoPSU for some breathing room?

I should note the power consuming components besides the CPU, RAM, SSD, and GFX:
Bluray drive
PCI Wifi card
USB Bluetooth adapter
USB external HDD (2x sometimes)
USB IR receiver

I apologize if I sound skeptical of power consumption. It just worries me, having been used to seeing only bigger and bigger PSUs going into desktops.

[MikeC]

GPU side fan, imo.

As for the power... it is your $$.

I've been measuring & monitoring computer/component power for 10 years, and the recommended PSU capacity has never been less than 2x the actual peak power of any system. IE, when someone official says you need a 300W PSU to run such&such, 150W is the typical max peak you will see.

Headroom is good to have, but I always ask myself what I do with the machine. On my HTPC, for example, I'm using a 125 TDP Phenom 2 quadcore. 2 HDDs, 8GB RAM, IGP w/ one of the AMD chipset boards. Total overkill. All I do with it is manage media files, download them, stream them, sometimes web surf. Real measured AC power? ~40W idle, maybe 70W peak. Right now I am using a fanless Seasonic X400 but I actually had it running with a 180W junk PSU for a while, with the fan pulled off it, no chassis. Changed it due to electronic whine from the PSU, but the thing could easily be run w/ a pico + 90~120W adapter.

[wtx]

It looks like they are both priced at about $50, unless I am misunderstanding something. Maybe the 200W needs an ATX extension?

http://www.mini-box.com/PW-200M-DC-DC-power-supply
http://www.mini-box.com/picoPSU-160-XT

In the past I have used this box for encoding, which is why I am concerned about power availability. It's unlikely that the GPU would be heavily engaged during encoding (I'll probably be connected from RDP), but I was assuming even without it it could go over 160W.

Also, with the old PSU out of the way there would be a great spot in the back for 40mm fan (location of current PSU exhaust). I know it's not the ideal location, but it would be much more convenient not to have to mod the cover. Unless you think that will be inadequate...

Thanks.

[MikeC]

In the past I have used this box for encoding, which is why I am concerned about power availability. It's unlikely that the GPU would be heavily engaged during encoding (I'll probably be connected from RDP), but I was assuming even without it it could go over 160W.

No way. Probably not even 100W.

Also, with the old PSU out of the way there would be a great spot in the back for 40mm fan (location of current PSU exhaust). I know it's not the ideal location, but it would be much more convenient not to have to mod the cover. Unless you think that will be inadequate...

Just remember, that 40mm fan in the PSU is the main source of nasty fan noise in your system now, and it cannot move much air. Try w/o any fan for the GPU first, after the PSU change. If the GPU still seems too hot (>90C), the fan mod can still be done.

[wtx]

Since the current PSU is rated at 250W, it sounds like I'd be fine with the 150W picoPSU. The 160+W is backordered for a couple weeks anyway. And I'll wait and see on the fan.

I am still scared to run on such low juice! I will report back with results. Thanks for your expertise, Mike.

[wtx]

Installed the picoPSU and I think it's working fine. However, I am still having some restarts, and it's not clear if it's heat, PSU, or GPU -related.

Every crash can be traced back to bugcheck code: 0x116, VIDEO_TDR_ERROR. My assumption was that this is probably an nVidia driver issue, but I have been through several different driver versions over the past few months, and I continue to get this error. I could be wrong, but I have been ruling out heat/PSU issues because I thought that in those scenarios the mobo would just shut the system down, leaving no bugcheck code for the OS.

Any insight or suggestions on how to troubleshoot this would be a big help.

[HFat]

The first thing to try would be a reinstall or another OS since this could be a software issue.

Your GPU may be defective. Make sure it's properly seated.
You could try to cool it better to see if it makes a difference.

It's hard to rule out a PSU issue since you've never had an opportunity to test yours. If the graphics card can't get the power it wants, I suppose it could crap out like this.

Ideally, you would try this graphics card on a different PC to see if the issue goes away but I guess you wouldn't be asking here if you could do that.

[wtx]

Thanks, unless I can be fairly certain that something in the OS is broken, I don't think it's worth it to reinstall just on a lark -- more of a Hail Mary than a diagnostic step.

Does the 0x116 error seem consistent with either GPU overheating or lack of PSU power?

I have stressed the GPU and couldn't force a crash. I'll try forcing it again today with AIDA64 running -- hopefully that will provide some clues.

[lhopitalified]

Does the 0x116 error seem consistent with either GPU overheating or lack of PSU power?

Sounds like a GPU problem. See http://mikemstech.blogspot.com/2011/12/ ... error.html.

Do you notice if it hangs right before crashing? (And is it only crashing on video playback?) You might check to see if there's a decoder or file issue -- improperly encoded video files can cause hiccups in the decoding process.

You might also try tweaking the decoder settings to use the CPU for everything and see if the playback performance is ok.

[wtx]

The GPU is a fanless GT430 I bought specifically to run LAV filters and madVR. However, since I replaced the Shuttle PSU with a picoPSU, I haven't had the system restart or shutoff during accelerated video playback. It's only happened recently during playback of recorded TV in WMC (but that could simply be a coincidence). But to answer your question, when I have observed this it doesn't freeze, just immediately restarts.

It seems like other GT430 owners would have this problem if it were just the driver. I will try more aggressive stress test and see if I can narrow it down to a heat issue. However, I've been told by several different folks that these fanless cards can get super hot (70+C) and still be okay.

Thanks for the link. I suppose I could try disabling TDR and see what happens.

[wtx]

I found the problem -- the Asus GT430 finally overheated and died. It may have been an unusually hot card, but I don't want to take any chances when they send me another one, so I plan to put a fan in front of it.

The problem is that there is almost no clearance between the inside of the case and the heat sink of the card, so I assume I would have to mount the fan on the outside of the case? Can anyone give me any pointers? Maybe it's trivial, but I've only seen instructions on mounting to the inside, with a grill on the outside.

Thanks!

[Pappnaas]

Other than cutting apart the side of the cover to make room for a fan and air?

Not really I'm afraid.

[tim851]

To be honest, you should not have a passive video card in a Shuttle cube. Sandwhiched against the side panel, there is nowhere for the heat to go and in those cubes, it receives almost no airflow, esp. if the CPU fan spins slowly.

Sell that replacement GT 430 on ebay and buy a card with a fan. It won't be that much more expensive and a lot less trouble than cutting a hole in the Shuttle case (thereby reducing resale value) and mounting an external fan.

[wtx]

@tim851 - I realize now that this card is not intended for a Shuttle-type case, but there wasn't any indication it would be a problem from their product specs and documentation. Swapping the card for one with a fan is probably the more sensible solution, but I already ordered the fan parts.

I would have taken your advice if I had seen this yesterday, but I wasn't thinking. I don't plan on reselling, so it's not a huge issue. As long as I can cut that hole.

[wtx]

Thanks for the patience guys, the addition of a big 'ol fan has eliminated all of my overheating issues. Based on reports of near-silence I went with a Scythe Kama Flow 2 120mm x 25mm fan, and it's virtually inaudible. I also added a rubber gasket just in case. The tricky part was cutting the hole. This is the hole saw I used:



It's very tricky cutting with a saw that big on a thin wobbly piece of metal. I drilled a pilot hole, then proceeded to drill knowing I could lose a finger if I wasn't careful. It took some finesse, and you can see my lack thereof demonstrated on the side of the case.



Battle scar aside, everything is working well.



However, I did encounter a very strange issue regarding video playback that had not happened before. Some time after installing the replacement graphics card, audio started stuttering dramatically for some fullscreen streaming video, and dropped frames as well. I tried serveral troubleshooting steps, including a clean install of updated graphics drivers to no avail. Strangely, removing the graphics card, booting without it, shutting down and then adding it back made everything go back to normal.

This has happened twice now, and those steps seem to be the only way to fix it. Has anyone seen anything like this? If it happens again I may take a video - it's hard to describe the symptoms - and post a new thread.