[Troubleshooting]PSU clicks and video freezes for fraction of a second...

[Enigmahack]

Sorry in advance for the wall of text for a new user. I could definitely use some of your folk's expertise here.

I've never had this issue before until I upgraded the video card coolers from stock to Icy Gelid rev 2. Here are my system specs:

ASUS Maximus VI LGA Formula motherboard
i7 4770K @ 4.4ghz (1.30v)
Two R9 290x's in Crossfire, with Gelid coolers instead of stock
16 gb of 2133 G-Skill TridentX RAM
3TB HDD, 256GB Samsung 840 Pro SSD
XFX 850 Black edition PSU
H220 CLC
Fractal Design R4 case with window
Windows 8.1

What happens: When in higher-intensity 3D applications (3DMark '13, Valley Benchmark, etc) the screen will pause for 1/5th to 1/7th of a second and the PSU makes an audible click. It sounds like someone is tapping a pen against a piece of metal, but it's gentle. It's loud enough that my girlfriend could notice the issue as well. This is an intermittent issue, happens maybe 3 times a minute, sometimes more or less depending on the scene being rendered. When playing less-intense games like Portal 2, there's no issue from what I can tell. I thought it might be the HDD but the sound is coming from the back of the PC, not the front where the drive is mounted. I'll confirm later though.

Some notes:
The video cards are stock clocks for both memory and GPU. (1000mhz/1250mem)
This only started happening after going from the stock Radeon coolers to the Gelid coolers.
This doesn't happen at all for a single card configuration. (Regardless of which card I use)
Temperatures under full load on both cards are pretty much identical: VRM1 = never higher than 80C, VRM2 = Never higher than 45C, GPU = never higher than 70C. I could pull the power information if needed later when I'm in front of the PC.
I can reproduce the issue intentionally when the GPU's are running full capacity.
I've tried various drivers from 13.11 to 14.3 beta drivers. The issue remains in either case.
I've plugged the computer from my UPS/Conditioner into the wall directly, and tried different sockets just in case is was my house not giving enough juice.
I remounted the heat syncs for both cards ensuring all temperatures were well within acceptable limits. (This includes reapplying thermal paste, ensuring tight fits, etc.


What I suspect: I think what might be happening is the video cards, fans, case fans, and everything together under full load is just too much for the PSU to handle... But it only started happening after upgrading to the new coolers.
The reason I upgraded however was because the stock coolers were letting the GPU get up to 90-95 degrees and then started to throttle the GPU back to keep it cool. I assume this would use less power, not pulling on the PSU as much.
The new coolers keep the cards MUCH cooler by 20 degrees or so, which means they can sit at 100% while being used instead of throttling back. I'm going to do some more testing tonight by purposely either underclocking the cards, or artificially dropping the thermal ceiling to about 65C and get the cards to work in a similar fashion.

The new heatsyncs have two fans on them each, where the stock only had one. The other note is that there are 2 140mm fans, 4 120mm fans, and the H220 Closed loop cooler. (I included the two fans on this in my count)

The only part where I'm stuck is... Has anyone heard of this before, or can confirm my suspicions? Is it also possible that this brand new (Less than 3 months old) PSU could be dying? I would have suspected that it should be more than capable of handling two 290x's... but when I run this system through this site: http://extreme.outervision.com/psucalculatorlite.jsp
It shows the power consumption to peak at 849W which means 0 overhead.

What I would REALLY like to try is throw in a 1000W PSU and see if it keeps happening. I just don't have the disposable income to do that unfortunately (unless I'm going to buy it knowing that it's the problem)... And I fear that the crap-brand 1000w PSU's near me would be less stable than the XFX I have now.

Any insight, information, troubleshooting steps that I might try would be appreciated... I'm willing to do just about anything. Hell, I'd be willing to consider tossing the stock coolers back on if it would make a difference but there's a reason I wanted them off in the first place. If someone wants pictures/videos, I'd be happy to include those too if it'll help.

Help please!

TL;DR PSU clicks and when it does, the video out completely stops for a fraction of a second but it's a noticeable stutter. Something's wrong because it didn't do this before. Help me please!

[Arne_Saknussemm]

Hi Enigmahack 😄

If the PSU clicks etc. and everything runs smooth in single card then it does make one suspect the PSU is not handling power demand.

Have you tried running dual cards downclocked and CPU at stock?

Also these coolers have twin fans whereas the original had only 1..how are the fans powered...are they powered separately via molex with connection to card for PWM signal or are they drawing power from the card?

[NemesisChild]

Welcome to the forum!

Only way to know if it's the PSU:
1. Use a voltmeter to measure the output on the +12v rail under load.
2. As you suggested, swap out the PSU for a quality 1000W+ like Corsair, Seasonic, EVGA, etc.

Those PSU calculators are usually not very reliable for real world power requirements.
It does sound like a PSU issue, confirmed by the problems when running crossfire and gone with a single GPU.

[Enigmahack]

Arne Saknussemm wrote:
Hi Enigmahack 😄

If the PSU clicks etc. and everything runs smooth in single card then it does make one suspect the PSU is not handling power demand.

Have you tried running dual cards downclocked and CPU at stock?

Also these coolers have twin fans whereas the original had only 1..how are the fans powered...are they powered separately via molex with connection to card for PWM signal or are they drawing power from the card?

Thanks for that. I was actually going to try that tonight - I was going to downclock my CPU back to 3.9 and if necessary, drop the core clock 50mhz on each card until things calm down again.

The fans are powered via the cards themselves. They DO happen to have Molex connectors that came with them but the instructions said they weren't necessary to run.

I was thinking the video cards are possibly pulling too much power which they can handle on their own but not in tandem... Perhaps it would be prudent to attach the Molex and see what happens instead? (Though I suspect the results would be the same)

Someone I work with also mentioned that it's probably power, and that my cooling actually could be suffering slightly because of the lack of power. Who knows.

NemesisChild wrote:
Welcome to the forum!

Only way to know if it's the PSU:
1. Use a voltmeter to measure the output on the +12v rail under load.
2. As you suggested, swap out the PSU for a quality 1000W+ like Corsair, Seasonic, EVGA, etc.

Those PSU calculators are usually not very reliable for real world power requirements.
It does sound like a PSU issue, confirmed by the problems when running crossfire and gone with a single GPU.

I was thinking that the PSU calculators typically aren't terribly accurate... That said I was told it was a more reputable one. That said, it's probably not 100% on the money. Even if it's not however, I still see with varying configurations that 100% power peaks are all above 870w+ with my configuration. Even if it's not accurate, it's too close for comfort I would say.

I have a voltmeter/multimeter, but I'm not experienced enough to know how to measure the +12v rail. I do have software monitoring tools that might help show the draw of power but that's likely not accurate enough because we're only seeing what the PSU is feeding the components right?

As a total aside... How in the blue heck did you get your 4770k to 4.8 stable? (bows)

Thanks for the info both of you. I was hoping I wasn't on the right track but I've been getting several confirmations from several sources that it's not looking good.

[NemesisChild]

That's correct, software monitoring programs aren't accurate/reliable when measuring +12v rails.
I have a good chip but only use 4.8 for benching, high voltage/temps are the limiting factors.

Good luck, let us know how it all works out for you.

[Zka17]

Try with a single 290X... 850W for 2x290X is just simple not enough...