Sign in with
Sign up | Sign in
Your question

HIS HD4870 issue - PSU or poor cooling?

Last response: in Graphics & Displays
Share
July 16, 2009 8:46:24 AM

Hi thanks for reading in advance,

I built a system a month ago:

GIGABYTE GA-EP45-UD3R w/ E5300
4GB ram, one 500MB SATA hard drive and one SATA DVD drive, stock cooling on everything, no overclocks
HIS 4870 512MB ( http://www.newegg.com/Product/Product.aspx?Item=N82E168... not a terribly impressive heatsink)
Thermaltake Purepower 500W PSU ( specs here: http://www.newegg.com/Product/Product.aspx?Item=N82E168... )

I've been having issues when playing games on full resolution and settings where I get the VPU recovery window popping up, and occasionally my computer shuts down completely. I'm fine with say, WOW, but Stalker: Clear Sky is unplayable, I get such an error every 10 minutes. I tried out FurMark out of curiosity, and found it makes my computer shut off completely within 3 seconds of starting the test.

To fix the problem, I tried severely underclocking the card in CCC from stock speeds (750/900) to 600/800. So far it's been much more stable, and I can run FurMark for a few minutes at least.

I don't want to try running it for hours, though, because I'm still concerned about my temps, here are readings from GPU-Z with the 150Mhz UNDERCLOCK:
Idle:

Load (about two minutes of furmark windowed at 1024):

(not entirely sure why the fan is reported as running at 1 RPM, the fan IS running)

That temperature for MEMIO (whatever that is?) can't be normal, is it? Also, when I tried setting the GPU clock at 650 instead of 600, furmark built the main GPU core temperature up to 92C after about a minute or two, which is where the computer abruptly shut itself off.

So I can't tell if my issue is with the PSU or the crappy cooling situation with my video card. I thought it could be a power issue because I don't get any kind of artifacts when the card heats up, the computer just suddenly restarts itself. Also, my power supply only has one 6-pin PCI-E plug wheras the 4870 needs two, so I have a 4pin-> 6pin power adapter of questionable quality for the other power slot on the video card. It's been pointed out to me as well that the PSU is kind of old and dosen't supply as much 12V power.

I'm just stumped as to what actions I should take to try and get my card running at the normal speed it's supposed to run at, or if i need a replacement or a new psu... any suggestions would be appreciated

Thanks
July 16, 2009 9:38:06 AM

Download Rivatuner, and set your fan speeds accordingly (set a rule for games to increase the fan speed to 75%)
July 16, 2009 1:28:48 PM

The first thing to try is more cooling. What happens if you take the side off your case and aim a common fan into it?
How hot is the air coming out of the PSU? It isn't really a 500W model, and may be unstable at 3D gaming loads. I'm not sure that's the "500W" model on the QVL list at ati.amd.com for the 4870. A complete shutdown suggests your PSU can't hack the load. It would probably be a good idea to replace it with a quality unit. I would suggest an Antec Earthwatts 500 or a Corsair 550VX.
Related resources
July 16, 2009 1:44:39 PM

+1 for better psu. Either way, get one with a combined 12v rail amps in the mid 30A ranger or higher.

+1 for better fan control. CCC 9.6 has a fan control if that doesn work, rivatuner.
July 16, 2009 2:27:44 PM

Onus said:
The first thing to try is more cooling. What happens if you take the side off your case and aim a common fan into it?
How hot is the air coming out of the PSU? It isn't really a 500W model, and may be unstable at 3D gaming loads. I'm not sure that's the "500W" model on the QVL list at ati.amd.com for the 4870. A complete shutdown suggests your PSU can't hack the load. It would probably be a good idea to replace it with a quality unit. I would suggest an Antec Earthwatts 500 or a Corsair 550VX.


take a look at the amps on the two 12V rails, some of the lowest I have seen on a 500w PSU
July 16, 2009 4:46:18 PM



High temperature in any electric dispositive is symphtom of a incorrect application of power or a ungrounded power supply, I guess You need a greater PSU, try with a PSU minimum 650 Watt, if I 'm right maybe can crash (with sparks and all the stuff) your actual PSU :o 
July 16, 2009 4:55:29 PM

Yep I think it's PSU. If your video card was too hot it would rev the fan like crazy and throttle the clocks back but would still continue to run. My GTX 295 has gone as far as 88c but have never had a problem with it.
July 16, 2009 6:47:40 PM

Thanks a lot for your suggestions!

I downloaded rivatuner to jack up the fan speed, and I also tightened the screws on the stock heatsink which seemed to be really loose. The GPU now runs 10-20 degrees cooler under load.

However, in the process of tightening the screws, I broke a screw off of one of the heatsinks seperate from the whole fan assembly (probably my stupidity, but damn, those were the flimsiest screws). Now in GPU-Z it shows the "VDDC Slave #1 and #2" at much higher temperatures than before. I tried the same Furmark test that I could successfully perform underclocked before, and I watched as those VDDC temps climbed up to 152 C and then, poof, display goes black.

So maybe I solved the problem by breaking it - Would it make sense if the voltage regulators overheating could be causing my problem, or is it still most likely the PSU?

Either way, I'm going to have to replace that heatsink - Should I get something like this to stick onto the voltage regulators:

http://www.newegg.com/Product/Product.aspx?Item=N82E168...

Or will a whole aftermarket cooling package be a better idea?
July 16, 2009 7:08:36 PM

Exactly which part did you break?
Perhaps you could take it off, remove the broken screw then pop down to you local small store and see if they have a dead card you can 'cannibalise'.

Not sure if the heatsinks you linked to are OK, take a look at these as well, they
look more compact for the cooling surface:

http://www.newegg.com/Product/Product.aspx?Item=N82E168...
July 16, 2009 7:28:13 PM

It was the black bar heatsink that covers these things (not my pic):

http://forums.extremeoverclocking.com/attachment.php?at...

It was attached by two spring-screw things at the top and bottom, I broke one of those off, the heatsink is now just loosely attached on one end... with tape :/ 
July 16, 2009 9:45:22 PM

Hmmm, if that's what overheated, I'm not sure any throttling due to GPU overheating would have happened. Hopefully there's been no permanent damage.
July 16, 2009 10:18:04 PM

I'm thinking about getting these to put on each individual... thing. (not sure the terminology to use for those, FETs?)

http://www.newegg.com/Product/Product.aspx?Item=N82E168...

They seem kind of small though, not sure if they'd help. Otherwise, I could wait till the 30th and order these (the only US site I've found selling these)

http://www.endpcnoise.com/cgi-bin/e/std/sku=zalman_zm-r...

Is there a way I could tell if any permanent damage has been done? I haven't had any additional problems beyond what i'm describing. If it's damaged I could try getting it replaced, but something tells me HIS wouldn't want to take it back with a broken heatsink.

Thanks again for the help.
July 17, 2009 4:00:54 AM

I have recently buy a crossfire setup with 2 radeon 4850 OC at 675 MHz by Asus (4850 TOP). While in single card setup, the temp at loads never get higher than 75C, in crossfire the temps went crazy and I was getting over 100C at loads.

It was clear that the thermal solution wasn't good enough for a crossfire setup. It causes crash due to overheating and securities. Even if the 4850 cards are designed to go that high, it was Asus fault for not designing and programming the whole damn issue correctly that led me to find an alternative.

If your card doesn't have a whole heat dissipator over the board, you are lucky because a simple 120 mm fan at one inch of the card blowing fresh air will drop your temps by easily 20%. It saved me 2 RMA and a lot of trouble at no cost since I already owned an old fan.

Try to put a fan and check back the results.
July 17, 2009 4:05:55 AM

Edit... I should learn to read a little more...

It seems your fan as died out... but what is troubling me is seeing your memory getting so high in temperature. These used passive cooling... it should never get to that level...
July 17, 2009 9:45:09 AM

@ Hoborg: try taking the whole thing down to your local, small computer store, it's likely they will have some dead hardware kicking around with the right, or at least similar fastners on it.
!