Should my power supply be failing in this situation??

Shark974

Limp Gawd
Joined
Aug 4, 2009
Messages
378
I have a 500 watt Antec Basiq PSU. It's spec sheet here http://www.newegg.com/Product/ShowI...siq BP500U 500W Continuous Power Power Supply

You can see it has two 12v rails at 18 amps each.

My system is pretty..almost low end these days. It's a Q6600@ 2.4 (not overclocked) 4GB DDR2, 500GB HDD, DVD drive, and a powercolor 4890 (which is factory overclocked to 950 mhz).

I started getting strange shutdowns while playing Crysis (the only PC game I've played on this rig btw). What would happen is the system would literally "snap" off in an instant, no error screens or anything like that, just black screen and then, the PC would reboot itself.

This happened a few times while playing over weeks, I just chalked it up to a bug, but on one recent session it happened several times, so bad that I had to stop playing.

I knew it was either overheating or PSU. It's almost certainly not CPU overheating using that real temp prog uncalibrated, my CPU temps where like 50C from thermal max. Even uncalibrated that rules out the CPU imo.

The GPU, well first I wondered if it was VRM overheating or core. Turns out theres no way for me to check my VRM's because it's a non-reference (I guess, it does have a red PCB though which I was told means it's reference, but the fan is def non ref) so all the programs wont even read my vrm temps which is shit. But the vrm's do seem to have a heatsink on them, so I'm pretty sure thats not it. Now on the core, it does get hot, running furmark the max I achieved was 89C. But it would get to 85+C under load every time. I also ran Crysis in a window so I could watch temps, and the GPU got to 85C, and that was in a 1280 window so I know it gets hotter in fullscreen 1680. But my idle temps are great, like 54C.

But trying to figure out if it was a temp issue was inconclusive, I couldnt really get the shutdowns to replicate consistently, even though it did shutdown sometimes in furmark too, other times it didn't.

So then I downloaded that OCCT program, and here's where I decided it was PSU..when I was running that prog it has a voltage graph (why it can read my system voltage but other progs seemingly cant I dunno), and if that graph is accurate, I believe it showed my 12v rail rapidly dropping to like 10.7 V when the stress test started. The computer actually didn't shutdown though ( I aborted test after a minute or two though).

So at that point I decided it must be the PSU, but whats bothering me is it seems that PSU really should be sufficient for a 4890. From what I read online. Whenever I google it, forum topics suggest a 500 watt supply is plenty for a 4890, people with bigger rigs than mine being told they use less than 400 watts, and so on. All the online power supply calculators say the same. The amps, as far as I can figure, should be enough to. So I dont get it. A new power supply is 100 bucks, so I dont want to buy one if thats not the problem. Any thoughts?
 

Dangman

Ninja Editor SuperMod
Joined
Dec 15, 2005
Messages
46,062
First: The OCCT voltage reading is crap. No software voltage reader is every accurate. If you want to get an accurate reading, do this: Buy and use a digital multi-meter to monitor the voltages coming from the PSU. Follow the instructions for "Using a multi-meter to check voltages":
http://www.bfgpower.com/troubleshooting.html#DMM

Second: A quality 500W PSU will be enough. However most quality 500W PSUs tend to have and can provide around 35A to 40A on the +12V rail. Your Basiq PSU only has 29A on the +12V rail:
Antec Basiq 500W PSU

In the above review, that Basiq 500W PSU could not provide 498W of power without dying. At most, it could only provide 377W of power before dying. Therefore, the Basiq is not a good quality 500W PSU. It's a decent 400W at best IMO.

Third: So 29A on the +12V rail means that you basically only have 348W of power available for all of the parts on the PC except for the RAM. That means 348W of power is available. The HD4890 can draw up to 190W of power by itself, non-oced' of course. The Q6600, assuming taht you have G0 revision, can draw up to 95W of power. So added together thats 285W of power alone for the CPU and GPU. That leaves you 63W on the +12V rail to power the hard drive (25W at startup), motherboard (Varies by the motherboard IIRC), DVD burner (Roughly 20 to 30W IIRC), and whatever fans and add-on cards you have.

Conclusion: A quality 500W PSU is more than enough for your setup. However, you've been pushing your system with a low quality 500W PSU barely capable of provide 400W of power with a rail setup that only allows roughly 350W of that power on the +12V rail where a majority of PC parts draw power from.

Looks like a new PSUis in your future. I highly recommend replacing the PSU before damage can occur (if it hasn't already occurred already). As a replacement, I recommend this PSU:
$68 - Seasonic SS-500ET 500W PSU
 

Shark974

Limp Gawd
Joined
Aug 4, 2009
Messages
378
Huh, the plot thickens and I'm more confused than ever.

Tonight I hooked the multimeter up, and torture tested using two programs, OCCT and Furmark.

Both cause shutdowns at various times.

First, I thought ATI had blacklisted/crippled these programs in the drivers anyway? So why are they even working hard enough to cause shutdowns?

Now, in "idle" the mm is reading 12.36-12.38v
Furmark cause that to drop to around 12.20
OCCT cause it to brutally drop as low as 12.07, but around 12.10.

In absolute terms, thats as much as .3v drop, which I suppose is cause for concern, yet somehow I dont really believe it's the PSU. For one, it shut down in furmark (which causes much less voltage drop) as much as OCCT.

The other thing is temps, I double checked the CPU, it seems ok, those progs dont stress the CPU too much anyway.

In furmark, it gets the 4890 measured by GPU-Z into the 80's C, which I think is acceptable.

However OCCT is absolutely brutal, even in a 640X window, it causes GPU temps to explode, reaching between 100-110C before I'd get scared and shut it down, provided it hadnt blackscreened at some point first.

However, I'm not sure the OCCT tests mean anything, so I cant declare it a temp issue furmark is probably more stressful than any game, yet it still only has me in the mid 80's. Further, at times OCCT can run the temps into the 100's and it doesn't shut down.

But through it all,, there seems little rhyme or reason to the shutdowns. I cant say it shuts down when it hits "x" voltage nor "x" temp either really. Once, it shut down the instant I started furmark,before it even had a chance to heat up. It's puzzling. other times it will run furmark or OCCT seemingly indefinitely.

So yeah, I'm lost really.

Some things I'm wondering:

-It's something weird like some motherboard component overheating
-My GPU heatsink isn't seated correctly (evidence against this, high GPU temps dont necessarily and provably cause the shutdowns as far as I can see).
-I have a card that is defective in some manner.

If I was a mind I'd start throwing money at the problem, buy a new beefy PSU, buy a new aftermarket GPU cooler if that didn't work, start eliminating things. But I'm really not of a mind to put more money into this rig just to play Crysis right now.
 

Ocellaris

Fully [H]
Joined
Jan 1, 2008
Messages
19,077
Huh, the plot thickens and I'm more confused than ever.

Tonight I hooked the multimeter up, and torture tested using two programs, OCCT and Furmark.

Both cause shutdowns at various times.

First, I thought ATI had blacklisted/crippled these programs in the drivers anyway? So why are they even working hard enough to cause shutdowns?

Now, in "idle" the mm is reading 12.36-12.38v
Furmark cause that to drop to around 12.20
OCCT cause it to brutally drop as low as 12.07, but around 12.10.

In absolute terms, thats as much as .3v drop, which I suppose is cause for concern, yet somehow I dont really believe it's the PSU. For one, it shut down in furmark (which causes much less voltage drop) as much as OCCT.

The other thing is temps, I double checked the CPU, it seems ok, those progs dont stress the CPU too much anyway.

In furmark, it gets the 4890 measured by GPU-Z into the 80's C, which I think is acceptable.

However OCCT is absolutely brutal, even in a 640X window, it causes GPU temps to explode, reaching between 100-110C before I'd get scared and shut it down, provided it hadnt blackscreened at some point first.

However, I'm not sure the OCCT tests mean anything, so I cant declare it a temp issue furmark is probably more stressful than any game, yet it still only has me in the mid 80's. Further, at times OCCT can run the temps into the 100's and it doesn't shut down.

But through it all,, there seems little rhyme or reason to the shutdowns. I cant say it shuts down when it hits "x" voltage nor "x" temp either really. Once, it shut down the instant I started furmark,before it even had a chance to heat up. It's puzzling. other times it will run furmark or OCCT seemingly indefinitely.

So yeah, I'm lost really.

Some things I'm wondering:

-It's something weird like some motherboard component overheating
-My GPU heatsink isn't seated correctly (evidence against this, high GPU temps dont necessarily and provably cause the shutdowns as far as I can see).
-I have a card that is defective in some manner.

If I was a mind I'd start throwing money at the problem, buy a new beefy PSU, buy a new aftermarket GPU cooler if that didn't work, start eliminating things. But I'm really not of a mind to put more money into this rig just to play Crysis right now.

you are fishing way too hard here, voltage does not always drop to dangerous levels before a system can blip out. your power supply is not sufficient for your system to run reliably and safely.

if anything, try underclocking you cpu AND video card as much as you can to reduce power draw, then run those same tests again.
Posted via [H] Mobile Device
 

RamonGTP

Supreme [H]ardness
Joined
Nov 9, 2005
Messages
8,150
A DMM cannot measure ripple, which often times happens on a stressed PSU before the voltage drops below spec.

Sounds like you know what your problem is, and I certainly think it's powersupply related
 

Shark974

Limp Gawd
Joined
Aug 4, 2009
Messages
378
Yeah I guess. Is the manner of shutdown any proof here? I mean, it black screen, reboots itself every time. No graphical corruption or other type of errors precede this.

I mean that sounds more like PSU to me.

I still think 500 watts should be enough, but, oh well I guess.

The underclocking test sounds cool, except it will reduce heat as well, if that was the problem.

But it almost has to be PSU, so I think it's fairly settled. Time to pony up 100 bucks I guess :(
 

Dangman

Ninja Editor SuperMod
Joined
Dec 15, 2005
Messages
46,062
I still think 500 watts should be enough, but, oh well I guess.

A 500W PSU is enough if it's a quality PSU. However, as I showed earlier, your 500W PSU isn't a 500W PSU at all nor quality

And no, you don't need to spend $100. The $68 500W PSU I linked to an earlier post is more than enough.
 

Shark974

Limp Gawd
Joined
Aug 4, 2009
Messages
378
The review you linked only showed it failing at 100v, not at 120v. Aren't our wall sockets 120v? Not only that but it's difficult to find why even a 400w supply should fail this build (Antec Basiq passed all the ~400w tests).

As your 520w linked psu, thanks but if I do get a new PSU i'm going to get one that can support my next build somewhere down the road as well, so probably a 650 watt+ unit.

That way whenever I got to build a new system in the next few years, hopefully I wont have to upgrade the PSU again as well, saving me money. So thats why I'm looking at ~$100.
 

jeremyshaw

[H]F Junkie
Joined
Aug 26, 2009
Messages
12,512
The review you linked only showed it failing at 100v, not at 120v. Aren't our wall sockets 120v? Not only that but it's difficult to find why even a 400w supply should fail this build (Antec Basiq passed all the ~400w tests).

As your 520w linked psu, thanks but if I do get a new PSU i'm going to get one that can support my next build somewhere down the road as well, so probably a 650 watt+ unit.

That way whenever I got to build a new system in the next few years, hopefully I wont have to upgrade the PSU again as well, saving me money. So thats why I'm looking at ~$100.

Our wall sockets, loose voltage over distance from your power substation (IIRC).115 is a more realistic estimate.
 

RamonGTP

Supreme [H]ardness
Joined
Nov 9, 2005
Messages
8,150
Well it's a moot point really. The PSU is failing, so whether or not it should doesn't matter becuase it is.
 
Top