New 390x causing shutdown on anything above Light settings

[phrost]

Hey guys I recently got a new GPU, a 390x to replace my old 280x and now my computer crashes when I run Folding@home at anything above Light settings (Average of Full).

So I made a couple of logs (in the latest version of hwinfo 64bit), one of it not crashing on Light setting and one of it crashing on Full settings (Sunday and Monday respectively). I'm not really the greatest with hardware issues, so if anyone has a free minute, could you take a look at my logs and let me know what the problem is? And maybe how I can fix it? Thanks!

Sunday: https://docs.google.com/spreadsheets/d/ ... sp=sharing

Monday: https://docs.google.com/spreadsheets/d/ ... sp=sharing

some additional info. I'm running Windows 10 64-bit enterprise edition. I have an FX 8320 CPU and a Solid Gear Neutron 750watt PSU. I'm now going to reinstall the latest Catalyst drivers and Folding@home from scratch. I think the problem is that my motherboard is overheating though.

[Nathan_P]

My guess would be the PSU is not upto the job of powering everything at full load, can we have full list of hardware.

[bruce]

My guess would be the PSU is not up to the job of powering everything at full load, can we have full list of hardware.

Quite likely.

Second guess: The fans are not up to the job of keeping temperatures in the acceptable range at full load.

[phrost]

Thanks for the reply guys. Here's a screenshot of my specs through hwinfo:

http://i.imgur.com/8rDLg52.png

or in text:

Gigabyte 990FXA-UD3 motherboard Socket AM3r2
AMD FX 8320 8-core CPU
8GBx2 Crucial DDR3-1600 / PC3-12800 DDR SDRAM UDIMM
XFX Radeon R9 390X 8gb
1 240GB SSD
1 750GB HD
DVD+RW
Solid Gear Neutron silver rated 750watt PSU
Microsoft Windows 10 Enterprise Edition

I'm not sure what else to tell you. I'm also worried that it's a heat issue. I'm discussing with the hwinfo coder how to properly make logs. He just updated the program to fix a bug that affected logs, so I'm about to make another log of my computer crashing. Hopefully I can get to the bottom of this eventually. Once again, I appreicate your help!

[jrweiss]

Get CPUid HWMonitor to show your system, CPU, and GPU temps real-time. The Overdrive page on CCC will also give you your GPU load and temp.

What is your case cooling setup -- fan size, in/out orientation, positions?

Peak power draw of the 390X is ~360 Watts, so the PSU should be able to handle the system. Getting rid of the waste heat is another problem. Though there is an opening in the backplane, it does not appear that much (if any) hot air is exhausted through it. Rather, most of it is exhausted right back into the case. You will have to ensure a free flowpath for that hot air out of the case.

First, ensure you have positive pressure in the case - your intake fans supply more air than your exhaust fans (including PSU) can handle. Then use slotted backplane covers in 1 or 2 slots immediately adjacent to the GPU slot covers. That will help direct the hot GPU cooler exhaust immediately out of the case.

Second, assess if you need more or better case fans. Many OEM fans on cheaper cases have relatively poor airflow. Invest in some Noctua fans, which combine great air flow with quiet operation (even my Lian Li fans, which are relatively well rated, were not up to Noctua specs). Start with the front intake fans (see note above about positive pressure).

Then, assess if you need to upgrade or re-orient your CPU cooler. Ideally, its fan will point right at the back or top exhaust fan, so its heat will go out the case without warming up other components. Horizontal coolers such as OEM Intel coolers may simply scatter hot air through the case.

[phrost]

Hey jrweiss, thanks for the response. Here's a screenshot of what you suggested: http://i.imgur.com/QjPiYB1.jpg

As far as heating and fans, I'm not quite sure about this stuff, but the guy that helped me build my computer is actually coming over in about twenty minutes. He's going to take a look at some stuff that's not quite right inside the case, such as the PSU being installed backwards and one fan not being connected at all. Hopefully, we'll get to the bottom of this.
Also, I do have one Noctua NF-A14 PWM 140mm premium fan. I appreciate you helping me out as you seem to know what you're talking. If you wouldn't mind chatting with me.. I would be so grateful. I use steam to chat most of the time but if you're available elsewhere... please let me know.

http://steamcommunity.com/id/phrostb/

[Nathan_P]

Is that cpu overclocked?

[jrweiss]

The screenshot looks like it was done with the GPU at idle. Other temps (relative high GPU min temps, with GPU fan still at 1400 RPM) suggest it started recording max/min temps not too long before the shot was recorded, and went to idle shortly after that.

Your app doesn't give fan speeds or PWM % for them. Are you sure they are running at 100% when Folding?

You should let the computer idle for a while (15 minutes?) so all temps stabilize at their minimum idle values. Take a screenshot of the temps. Load the CPU only with a F@H WU. Let it run until temps stabilize again; record it. Stop the CPU WU and let the machine go back to idle temps. Start up a GPU WU only. Wait for temps to again stabilize, watching the temps closely. If the machine crashes, try to remember the key temps. Restart the machine, start the GPU WU again, and take a screenshot just before it gets to the crash temps. If it doesn't crash with just the GPU WU running, restart the CPU WU and repeat the process.

Backwards PSU? How?
Fan not connected? Which one?
If you put the LNA on the NF14, remove it.

No steam here. Skype jrweiss98155 for chat.

[phrost]

The screenshot looks like it was done with the GPU at idle. Other temps (relative high GPU min temps, with GPU fan still at 1400 RPM) suggest it started recording max/min temps not too long before the shot was recorded, and went to idle shortly after that.

Your app doesn't give fan speeds or PWM % for them. Are you sure they are running at 100% when Folding?

You should let the computer idle for a while (15 minutes?) so all temps stabilize at their minimum idle values. Take a screenshot of the temps. Load the CPU only with a F@H WU. Let it run until temps stabilize again; record it. Stop the CPU WU and let the machine go back to idle temps. Start up a GPU WU only. Wait for temps to again stabilize, watching the temps closely. If the machine crashes, try to remember the key temps. Restart the machine, start the GPU WU again, and take a screenshot just before it gets to the crash temps. If it doesn't crash with just the GPU WU running, restart the CPU WU and repeat the process.

Backwards PSU? How?
Fan not connected? Which one?
If you put the LNA on the NF14, remove it.

No steam here. Skype jrweiss98155 for chat.

Ok thanks, I've been so busy lately haven't had time to really dig into the problem. I'm not sure about the PSU but we decided to just leave it how it is, also we connected another fan so now I have 4 fans. I don't think it's backwards but the GPU fan is pointing into the case as opposed to out onto the floor/carpet. I'll add you on skype later today after I run some more tests. I'm going to try GPU and CPU only. Also, need to remind myself what "LNA on the NF14" means. Later today!

[jrweiss]

LNA = Low Noise Adapter shipped with Noctua fans -- essentially a resistor to limit top speed.

Out of town now through Sat 19 Sep.

[phrost]

LNA = Low Noise Adapter shipped with Noctua fans -- essentially a resistor to limit top speed.

Out of town now through Sat 19 Sep.

Ok, I think I may have installed that. It looks like all the other cable extenders in the Noctua box right? Also, added you on skype, so hopefully we can have a chat when you get back. I also have some updates. Running GPU only on Full didn't make my computer crash, neither did running CPU only on Full. Neither did running both on full on my old 280X. But on my 390X, running both on Full or even Medium makes my computer crash. I also took logs for many of the configurations mentioned above, and I can post them somewhere, maybe pastebin? I'd rather just make some graphcs to show you guys instead.

[jrweiss]

The LNA is color coded (see the Noctua docs). Make sure you have only a simple extension installed, if needed.

Except for the initial header that shows your configs, logs are unlikely to show much, if the GPU WU will Fold fine alone. It sounds like more of a computer problem, whether configuration or heat. Post that part of your log here within the

Code: Select all

tags

Ensure you have 1 CPU core free for the GPU; use the F@H client to set the CPU slot to 7 cores.

A description of your case and cooling setup, possibly with pictures, might help.

[bruce]

The TDP of the 280X is 250W and the 290X is 290W. This you're probably adding 40W more heat than you were ... and drawing 40W more power from the power supply.

You said you could run the GPU on full with the CPU slot paused. Was that the 290X? Most likely you're right at the maximum power your P/S amd/or your fans can handle and adding the CPU, too, is just too much.

[7im]

And or one of the rails is overloaded if that PSU has more than one.

[Nathan_P]

Definitely sounds like the PSU to me, the fact that the old gpu can fold with the cpu and not both together with the new gpu installed is a major pointer. Personally at this point I would go out and buy a decent PSU from your favourite e-tailer and by decent I mean a big name brand - seasonic, corsair, superflower etc.

Before you buy though check out some reviews from a tech site, if you want some recommendations drop me a PM.