Corsair CX series overheating/crashing - anyone else?

[DKlein]

So my old OCZ 500W died on me a couple days ago after 5 good years (just rebooted randomly and upon restart only a black screen, Power_Good was 61mV when jumped on), and I decided to upgrade quality for a supposedly better brand while downgrading quantity a bit and went for a CX430. I suppose I might've gotten worse quality in the end...

My specs:
[size=-2]GTX650Ti-B 2GB (Gigabyte) (134W max - usually around 90% TDP under load)
i5 3350P (65W max - 30W usually under load)
2x4GB 1600MHz 8.8.8.28 @1.5V (GSkill)
SBX card
Gigabyte B75M-D3H motherboard (mATX)
1x SSD (Corsair)
2x 7200RPMs (various)
3x 120mm fans (SilverStone heatsink + Rosewill case)
CD-DVD ROM drive (?)
MicroATX tower Rosewill case (17.5 x 18.5 x 10.25in)[/size]

The CX430 comes with 384W on the 12V rail, so I figured I was more than good to go, especially with every online calculator telling me I needed around 350-360W total. I figure I'm actively using less than 50% of the capacity under load.

What happened?

Immediately after installing it, BF4 after 20min crashes to reboot ("CTRB" henceforth). Other than the grand finale of the previous PSU, it never did that before (though it would "soft CTD" every few hours, but it's EA, so par for the course). So I start digging and testing.

CPU, GPU, and MB temps top out at 55C, 62C, and 38C respectively and stay steady after a few minutes. CPU and memory pass 1hr stress tests without issue (P95/OCCT & MemTest+). Then onto FurMark... CTRB after 10min. Again, the same. Onto OCCT PSU test... CTRB after 8min.

Welp, pretty narrowed down here. Although the other temps are staying constant after 1-2min AND the voltages are, though 12V is fairly low at 11.520-11.448V during load (yeah, it came fresh from the factory running just shy of 5% low on the 12V), down from 11.780 at idle, the big thing I find is with my multimeter-con-IR-probe: I stick the IR probe behind the PSU exhaust and take note.

Temps start low and start going up under load. And up. And up. Past 60C. Up more, 8-10min past test start. Until right around 62C and I get a CTRB. And that's 62C just at exhaust...

Culprit found. The unit just doesn't have the heat dissipation required to keep it cool under heavy load, and thankfully it at least has an overheat protection that's kicking in (so it's not a total failure I suppose). I know the case is a bit tight, but it has decent airflow with the three big fans, and every other component is in the mid-range of temps.

RMA process started. I'm hoping to just get a full refund and will eat the return shipping as a lesson learned, and now shared:

[size=+3]Cheap series from good-brand PSUs are still cheap PSUs. They're a gamble, buy at your own risk, especially with a tight case setup.[/size]

[size=-2](lol, old sig/rig is oooold - haven't been here in a little while )[/size]

 

[DKlein]

Oh, also any recommendations for a 400-450W PSU that's not uber expensive but also doesn't CTRB constantly? I'm not sure if the RMA will require I get another Corsair, so maybe keep suggestions to that brand, but maybe also Antec/SeaSonic/XFX.

80-Plus/Bronze is fine, I don't need any fancy metal standards


Also, does anyone think that the problem might be any other component, or is it pretty clearly the PSU here? I mean, there are no graphical glitches/etc, and so the only other concern I have is the motherboard. I am never quite sure how to test the motherboard or know when it's the root cause. It seems pretty clear that the PSU is here, but I'm never 100% sure on this stuff.

 

[DrMrLordX]

You should always consider the possibility that one or more components in your system could have suffered damage when your OCZ PSU went dead. It's not guaranteed or even probable, but it is possible.

I haven't heard of CX430s having overheat problems, but hey if that happened then it may well be that you had a bad fan or bad fan mount in the unit you bought.

 

[DKlein]

You should always consider the possibility that one or more components in your system could have suffered damage when your OCZ PSU went dead. It's not guaranteed or even probable, but it is possible.

Well, that's a bit what I'm worried about. The RAM seems fine, at least not the cause of these crashes, as does the CPU. I'm going to test out an older video card later today to check if the GPU is causing it, but I'm not sure why that would force a CTRB all the time, rather than just artifacts/glitches/CTDs/etc. I mean, the system runs games flawlessly up until the CTRB. And like I said, the old PSU went out quietly, with a whimper, no magic smoke or sparks or anything, just "off" suddenly.

The only other component that would be a possible issue is the motherboard, and I don't really know how to isolate that for testing, or isolate the PSU from it.

The fact that the PSU just keeps climbing in temp while everything else doesn't is what's got me fairly certain it's at fault, especially since this is a fairly typical PSU-style issue.

 

[DKlein]

Swapped vid cards to an older one (150W, so slightly more power hungry), a GTS 250 1GB (DX10.1 I think).

OCCT PSU test DX11: 12V rail barely drops past idle (11.78V); everything runs stably for 45min.
OCCT PSU test DX9: 12V rail drops past 11.20V; CTRB after 8min, right on cue.

So it's not the video card, leaving the MB or PSU.

Honestly, even if the PSU isn't causing the CTRBs, 11.20V is crap and needs an RMA just for that. 430W should be sufficient for a 150W single GPU and low-wattage CPU setup. This one clearly isn't. Guess I just got a lemon. I thought Corsair had good QC, but I guess budget lines are cheap for a reason...

Also not sure overheating is the exact problem, at least not unit-wide (might be something specific to a 12V rail component, though), since I was seeing higher temps from the exhaust during that DX11 run. Who knows, either way it's getting returned.

 

[PhIlLy ChEeSe]

Where did you get the corsair, can you take it back? Try leaving the case open while stress testing, 11.2v is perfectly acceptable it drops cause its being stressed. You need to worry if 11.2v with out stress.

 

[VirtualLarry]

Where did you get the corsair, can you take it back? Try leaving the case open while stress testing, 11.2v is perfectly acceptable it drops cause its being stressed. You need to worry if 11.2v with out stress.

5% low would be 11.4v. So I think that's out of spec.

 

[Charlie98]

You should always consider the possibility that one or more components in your system could have suffered damage when your OCZ PSU went dead. It's not guaranteed or even probable, but it is possible.

I haven't heard of CX430s having overheat problems, but hey if that happened then it may well be that you had a bad fan or bad fan mount in the unit you bought.

FWIW, I ran one of my i5-2500K rigs with a GTX560Ti 448 on a CX430M for a while... OC'ed to 4.3GHz but left the GPU stock, and didn't have problems... but I would not have been surprised if I had. The CX430-series is a decent entry-level PSU (I have 3 in service right now...) but it is not intended for severe service like a gaming rig.

There is also the possibility, like DrMr mentioned, that you simply got a bad unit, and/or you have a component issue... your rig is roughly 4 years old, mind you, and you just had your old PSU fail.

 

[PhIlLy ChEeSe]

5% low would be 11.4v. So I think that's out of spec.

If he's got a bad board, couldn't a component be pulling it out of spec's? He said it checks OK till he turn's it on?

 

[master_shake_]

i've got a cx500 in a corsair 250d with a 670 and overclocked 2600k and never had an issue.

 

[lehtv]

i've got a cx500 in a corsair 250d with a 670 and overclocked 2600k and never had an issue.

And? :|

 

[DrMrLordX]

5% low would be 11.4v. So I think that's out of spec.

It is . . .

If he's got a bad board, couldn't a component be pulling it out of spec's? He said it checks OK till he turn's it on?

It could. The board could be misreading voltage inputs entirely, or the VRMs could be doing an awful job, or something else. Unfortunately, the only way the OP is going to know that is to use a multimeter or something to test the PSU while it's isolated from the rest of the system AND while it's still at (or close to) the maximum operating temperature (62C? ouch!)

OP should probably either test his unit with a multimeter or swap in a known-good PSU and see if it experiences the same problems (with an eye towards operating temperature). I'm not trying to shill for Corsair or anything, but I am not aware of overheating being a common problem for either revision of the CX430. 62C is a pretty ugly temp for a consumer-level PSU (that's like . . . 12C higher than Oklahoma Wolf's/jonnyguru's hotbox testing), and that can affect power delivery.

I will say that you may just have a bum unit. I had a CX400 in a secondary machine that gave up the ghost some time ago which I never had the chance to RMA for various stupid reasons. It's out of warranty now (bleh). It was cheap, and it did what so many cheap PSUs did: it just failed. Up until then, it worked great. I find it difficult to believe that your 500W OCZ unit was able to handle an operationally hostile (read: hot and cramped) environment that is simply too much for a CX430, so my first guess is that it's got a bum fan or something. It's still worth it to test the rails independently, especially right after a crash while it's still running kinda hot. Once it shuts down, though, it's gonna cool off pretty quickly, so you'll have to move fast. For that reason, it might be more practical to just test another unit in there.

 

[master_shake_]

And? :|

and that computer is better than his and my corsiar cx500 power supply is just fine.

so when he says...

Cheap series from good-brand PSUs are still cheap PSUs. They're a gamble, buy at your own risk, especially with a tight case setup.

it's more likely he got a bum unit.

 

[DKlein]

your rig is roughly 4 years old, mind you, and you just had your old PSU fail.

Nothing in my rig is 4 years old. The old PSU was 5 years old. Everything else ranges from 2 years to 2 months old. I should've replaced it sooner, but I usually aim for 4-5 years of service from a PSU with no other issues (all voltages were still good under load a couple months back when I last checked). I probably should've factored in the heat the last one operated on (tropics for 4 of those years), and ditched it at the 4 year mark. Lesson learned: replace old PSUs before they go bad, probably once the warranty is up. Why risk it?

If he's got a bad board, couldn't a component be pulling it out of spec's? He said it checks OK till he turn's it on?

It's okay right now as I'm typing this. It only has issues under heavy load when the 12V line gets pulled down heavily. For instance, when I ran OCCT PSU test on DX11 (on a DX10 card) it only got pulled a little and was mostly stable. When I ran it in DX9 mode, it got pulled down heavily and crashed within 8 minutes.

It's fine for non-gaming use. It's been on for 8 hours today, handling my wife's online flash games. It's just that 12V draw from the GPU that's crashing it. Whether that's a fault with the board or the PSU, I'm not 100% sure, but that out of spec 12V rail makes me think it's the PSU rather than something that got blown from the old PSU wearing down a MB part.

so when he says...

it's more likely he got a bum unit.

Hence the "gamble" I'm guessing poorer QC is one reason the CX is cheaper than an HX or AX. That and having cheaper, lower QC parts in them as well. When everything is top of the line and QC'd heavily, your odds of getting a lemon are lower.

 

[DKlein]

It is . . .



It could. The board could be misreading voltage inputs entirely, or the VRMs could be doing an awful job, or something else. Unfortunately, the only way the OP is going to know that is to use a multimeter or something to test the PSU while it's isolated from the rest of the system AND while it's still at (or close to) the maximum operating temperature (62C? ouch!)

.... It's still worth it to test the rails independently, especially right after a crash while it's still running kinda hot. Once it shuts down, though, it's gonna cool off pretty quickly, so you'll have to move fast. For that reason, it might be more practical to just test another unit in there.

Thanks for the advice. At this point it's more academic, though, since I've started a return/refund RMA. Like I said, even if the PSU isn't causing the crashes, a 12V rail out of spec on a brand new unit isn't acceptable to me.

Do you think a 134W GPU and ~30W CPU system should be too much for a 430W Corsair with a 384W 12V rail? I don't, but I'm not an expert I guess.


I'm not going to test the PSU isolated because it literally took me 25 minutes to get the 24-pin into the motherboard last time, requiring me to completely take it out of the box and disconnect everything in the process. The plastic just wouldn't fit the holes or something (no pins were bent). Just a ridiculous amount of pressure was required. Given the unit's a dud anyway, I don't want to risk damaging the motherboard with another plug-in attempt.

What I CAN do is plug the multimeter probes into the backsides of a 12V pin and black pin and check that the readings from the motherboard are accurate. Well, I can try to do that at least. We'll see in 1 sec...


Well, that's interesting. The motherboard reads 11.520, while the multimeter reads 11.98 under load. Motherboard is 11.952 at idle, while multimeter reads 12.11. Ex-Tech is the brand for the latter, so I assume it's fairly accurate. Certainly the difference is much lower, whatever the initial accuracy. Might it be the motherboard after all?

 

[DrMrLordX]

The plot thickens!

Your RMA is a good move since there's no reason why a decent 430W PSU should be running hotter than a 500W PSU in the same installation (in my opinion anyway). And no, the listed power draw of your components shouldn't push that 430W unit very hard at all.

As for the multimeter readings, those are very interesting. Good on you for having a multimeter handy and knowing how to use it. How hot are your VRMs running right now? You're looking at some serious loss in voltage there . . . or maybe the motherboard is just misreporting voltage due to a damaged/inaccurate sensor.

 

[DKlein]

As for the multimeter readings, those are very interesting. Good on you for having a multimeter handy and knowing how to use it. How hot are your VRMs running right now? You're looking at some serious loss in voltage there . . . or maybe the motherboard is just misreporting voltage due to a damaged/inaccurate sensor.

Thanks. I can't tell the heat directly with the IR probe because the motherboard VRMs are all covered by the CPU heatsink clips (it's a large SilverStone setup). I can't really check VRM temps with the heatsink off without fusing the CPU to the MB, so ???

All I can say is that the motherboard registers a "CPU" temp (different from the actual CPU) as 43C (might be the area around the CPU, VRM possibly???) and the "motherboard" as 38C. Both are very constant once they reach peak temp, so they aren't climbing after that. As is the CPU core temp of ~55C. I'm guessing the "motherboard" temp is the PCH? It's next to the GPU, so that might be heating it up a bit, but it's also directly in front of the front 120mm intake fan, so I'd say the 38C is probably what the PCH is at.

All I can say otherwise is there were no obvious visual signs of problems with the VRM when I last had the motherboard out to look at it (I always check that). I guess the thing wouldn't even turn on properly if a cap/etc was clearly busted, though, so we're looking for something a bit more subtle.

This is where I always get confused. How do you determine if it's the MB or not?

I'll try running the burn-in test again with the side off the case, see if that helps.

 

[lehtv]

and that computer is better than his and my corsiar cx500 power supply is just fine.

so when he says...

it's more likely he got a bum unit.

Can't make generalizations from a sample size of 2. Assuming we have no reason to assume anything about how many units are lemons and how many work fine, you could equally well make the conclusion that "so when you say your unit works fine, it's more likely you just got lucky".

 

[DKlein]

I did 25min of testing with the case open. Well, that's about 15-17 minutes longer than I managed before, so progress?

Temps were basically the same for everything (kudos to me for designing an air cooling system that works as well with the case open as closed). All I noticed was that the PSU was running a bit cooler, but measuring that is a bit inexact by comparison.

I tried to see the heat around the VRM, or close to it. About 45-47C on the board.

Voltages from the multimeter again were right around 11.98V while the MB reported 11.520V.

 

[DKlein]

Hmmm...

I took out the CPU to clean up around and look more closely at the VRM and reapply thermal paste.

Turns out a few pins were bent. Might've happened during the initial PSU component testing - I had to swap out CPUs a few times to check if that was the issue. This might be the culprit. Or maybe it's been that way for years and works okay.

I guess bent pins could do this? Maybe?

 

[DrMrLordX]

If you can identify what the pins do from a pin diagram, that might help. Merely bending them shouldn't mess with voltage levels like that though (I wouldn't think).

Anyway, see if this url helps:

http://www.xtremesystems.org/forums/showthread.php?275021-lga1155-pins

It is possible that the problem is inside the VRMs themselves which would not reveal any outward sign of trouble. But at this point, I am speculating. You may wish to talk to Gigabyte and see if they have any recommendations.

 

[DKlein]

Thanks, I'll look into that diagram tomorrow. Gigabyte's tech support wasn't great the first time (when the PSU died I missed the Power_OK pin so I thought the PSU was good and went to the MB initially as the culprit, and started an RMA about the original no-POST issue). We'll see how it goes, I'm restarting that ticket now.