[SOLVED] System shuts down abruptly when GPU load increases to >99% in very short time

Toggle sidebar Toggle sidebar
S

Sharath Chandramouli

Reputable
Sep 19, 2014
17
0
4,520
The desktop configuration as follows:

CPU - Intel 2.9 GHz LGA 1155 G2020 (stock heatsink/fan), TDP = 55W
GPU - Sapphire DUAL-X AMD Radeon R9 270X with Boost OC 2 GB, TDP = 180W
PSU - Corsair VS550
MB - Gigabyte GA-B75M-D3H
SSD - Kingston SV300S37A 60GB
RAM - Corsair Vengeance DDR3 4 GB (stock heatsink)
CAB - Cooler Master Elite 311 Mid Tower (stock cabinet fan)
OS - Windows 8.1 Pro x64
DRIVERS - Latest AMD Catalyst ver. 14.4

This new system was assembled yesterday; to dedicate the computing resources to the BOINC Einstein@Home project. So it is supposed to be run 24/7/365 (at least planned). Am using the ATI OpenCL-optimized BOINC client. I strictly do NOT intend to play games and strictly NO overclocking. All components always run at normal / reference speeds for stability.

PROBLEM: The system shuts down abruptly within 3 - 5 min when the GPU load goes Max (>99%). At max load, the temp stays between 72 - 75 deg. Cel (GPU-Z readings during the death time). This is happening when using BOINC client or FurMark stress test. This does not happen when the CPU alone is stressed (ran a Prime95 stress test whole overnight successfully without any error).

Also, I ran the Windows Memory Diagnostic Tool and found no problem. Just to mention, while I did some research on this problem, I found out that most point at PSU as the high probable suspect, but,

* Bad PSU: Even though Corsair VS550 is entry-level, I hope it delivers the power which the system is expecting,
* Lower Amps on +12V line: Corsair VS550 spec says, it delivers 42A on +12V line at 240V-5A. I live in India, so it is 240V and I plug the PSU mains into a wall socket directly which is rated at 5A. I hope 42A on +12V is sufficient for R9 270X and the CPU.

Am totally confused what could be wrong. Please advice.

-- Sharath
 
Sort by date Sort by votes
englandr753

englandr753

Distinguished
Sep 4, 2007
2,652
8
21,115
Do you have another GPU and/or PSU to test? That would be the easiest method if you have one or can borrow one from someone you know. Id check the connection for the two 6pin power connectors to ensure they are plugged in properly and the GPU is seated. Remove the GPU then reinstall and reconnect the power connectors to ensure the connections are good.

I once purchased an Nvidia 570 and it was dead straight out of the box. That never happened to me before but is always a possibility that you have a defective part.
 
Upvote 1 Downvote
S

Sharath Chandramouli

Reputable
Sep 19, 2014
17
0
4,520
@Outlander_04
1. While not stressing, the system runs smoothly, I do not experience any abrupt shutdown / restart during normal usage.

2. The GPU is intended for purely number-crunching (since GPGPU could be utilized for general computing by taking it's massive parallelism as an advantage) for my BOINC Einstein@Home project using the OpenCL-optimized client.

3. At this time of year in India, the room temp fluctuates between 21 - 35 deg. Cel. in a day.
 
Upvote 0 Downvote
S

Sharath Chandramouli

Reputable
Sep 19, 2014
17
0
4,520
@englandr753, I do not have a spare GPU/PSU (also borrowing from others is also ruled out as my friends' systems are pretty old), but I would definitely try re-installing the GPU into the PCIe slot and the 2x6-pin connector.
 
Upvote 0 Downvote
O

Outlander_04

Titan
BANNED
Nov 18, 2008
20,308
4
74,960


My guess would be that the power supply is turning off due to too high a load . Since this only happens when you are doing something completely artificial that will not occur in day to day use you could simply ignore it .

Or you could investigate cooling a little more
Is the power supply mounted correctly? Does the problem stop if yu run the pc with the case open and use a desktop fan to cool it more ?
 
Upvote 1 Downvote
S

Sharath Chandramouli

Reputable
Sep 19, 2014
17
0
4,520
Phew ! Finally solved it. The real problem was with PSU. Because some damn fool forgot to connect the PSU's fan pin to it's header. At the cost of voiding the warranty, I removed the PSU cover, to my surprise, I found that the fan's cable was floating freely. I connected the fan's cable to it's header on board the PSU and re-connected all the cables.

Now, I ran a full 15-min 1080p Full-Screen burn-out with FurMark without any hiccup. The system is working smoothly now even under the load. Verified that the PSU's fan is spinning properly during the burn-out.

I sincerely appreciate all your suggestions / recommendations. Thanks !

-- Sharath
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom