Radeon 4670 or PSU not functioning properly?

[milanis]

Recently I've changed my faulty radeon hd4670 for another one. I've also added another hdd. Then lately games like PoP: The Forgotten Sands and Starcraft 2 have started crashing with a grayish/blackish screen with vertical lines and irritating noise. Only the hard reset helps then. I've run some test and concluded that this happens only during prolonged graphically demanding scenes when fps drop below 25fps. If I lower the graphics settings everything runs ok without crashes. I finished PoP with just few crashes somewhere in the middle of the game, while in other parts of it there were none. Same with SC2. Games crash after 2 or 5 or 10 minutes or not at all depending on the above.
I could easily sacrifice visuals for the safety, but still I'd like to know if it's the PSU or the GPU causing the crashes. Any ideas?

AMD Phenom X4 9550 2.21 Ghz
Asus M2N68 Plus
Kingston 3GB (2+1) DDR2 RAM @800Mhz
Radeon 4670 128-bit 512MB GDDR3
2 HDDs
PSU 450w
Win7 Ultimate 32-bit

Yeah, drivers, directx etc are up to date.

 

[happy medium]

It sound to me that your card is overheating.

Grab gpu-z and monitor your temps.

http://www.techpowerup.com/downloads/1837/TechPowerUp_GPU-Z_v0.4.4.html

 

[milanis]

Tnx, I'll test it and post the results.

 

[milanis]

You're right! Tnx! GPU load is round 95% ingame, but it crashes even when the load is lower than that! In Windows, the GPU temp is round 55 °C idle, and it crashes when it reaches 80 °C ingame and stays so some 30 secs! Are these temps enough reason for it to crash and if so, what should I do to fix it? A new cooler maybe?

 

[happy medium]

80c is not too bad.
Try to underclock the card in the catalyst controll center and turn the fan speed to 100percent.
OR
You can use msi afterburner to downclock the card and change fan speed up to 100 percnt.

http://event.msi.com/vga/afterburner/download.htm

What is the name and model of your power supply?

 

[Patrick Wolf]

The 4670 you had before did the same thing? Was it RMA'ed?

If the TIM on the core is fudged temps may seem ok, but the card will still crash. it may need a new, proper application.

 

[Termie]

80c is not too bad.
Try to underclock the card in the catalyst controll center and turn the fan speed to 100percent.
OR
You can use msi afterburner to downclock the card and change fan speed up to 100 percnt.

http://event.msi.com/vga/afterburner/download.htm

What is the name and model of your power supply?

I'll disagree here: an HD4670 should not be operating at 80C. Mine surely doesn't. This is an ultra-low power GPU, so if it's getting hot, something is wrong. And I highly, highly doubt this is a PSU problem - it's just not a high enough load for anything but the worst PSU to be strained.

I do agree with downclocking and upping the fan (one at a time and then together), however, to try pinpointing the problem.

And by the way, are these the same problems you were having with your first faulty HD4670? If so, then perhaps it wasn't the card, if not, man, you have bad luck, because I'm thinking this is another faulty card.

 

[happy medium]

Are you sure temps are peaking at 80c or does it crash and then it says 80c 30 seconds later.
That makes a big difference.

 

[Termie]

Are you sure temps are peaking at 80c or does it crash and then it says 80c 30 seconds later.
That makes a big difference.

Good point. Download MSI Afterburner to monitor temps continuously.

 

[happy medium]

I'll disagree here: an HD4670 should not be operating at 80C. Mine surely doesn't. This is an ultra-low power GPU, so if it's getting hot, something is wrong. And I highly, highly doubt this is a PSU problem - it's just not a high enough load for anything but the worst PSU to be strained.

I do agree with downclocking and upping the fan (one at a time and then together), however, to try pinpointing the problem.

And by the way, are these the same problems you were having with your first faulty HD4670? If so, then perhaps it wasn't the card, if not, man, you have bad luck, because I'm thinking this is another faulty card.

I thought most cards can do 80c without crashing ,but if you have one and say otherwise, then your the man.
55c idle seems hot to me,thats whyI question his 80c peak temps.

 

[milanis]

Yes, it was RMA'ed. Though then I had it crashing to the desktop with several games, so another kind of crashing. I monitored the temps in gpu-z continuously (1 sec interval) for the current issue. When it crashed (the new gpu) I checked the gpu-z log file after restart and the readings were there till the moment of the crash (I was ingame testing it when it crashed). I had Asus 1gb model, now it's the MSI one. Also, the fan speed in the log file reads 165.0 rpm when idle (desktop) and 750.0 rpm when ingame (under stress).
I'll try MSI's afterburner and underclocking.

 

[Termie]

I thought most cards can do 80c without crashing ,but if you have one and say otherwise, then your the man.
55c idle seems hot to me,thats whyI question his 80c peak temps.

Edit: I might be wrong actually. Just checked out a recent that shows the 4670 at 52 idle and 90 load: http://www.anandtech.com/show/2917/13. Wow, that surprises me. I can't believe a 4670 could get that hot.

Yes, it was RMA'ed. Though then I had it crashing to the desktop with several games, so another kind of crashing. I monitored the temps in gpu-z continuously (1 sec interval). When it crashed I checked the gpu-z log file after restart and the readings were there till the moment of the crash (I was ingame testing it when it crashed). I had Asus 1gb model, now it's the MSI one. Also, the fan speed in the log file reads 165.0 rpm when idle (desktop) and 750.0 rpm when ingame (under stress).
I'll try MSI's afterburner and underclocking.

Interesting...Afterburner might help track the temps a little more easily, but it sounds like you already have a lot of information. By the way, that fan speed sounds low to me. Afterburner will help you adjust the fan and monitor its percent activity as well. If you get your fan speed up and your temps way down and you still have a problem, I think it's clear there's a hardware problem. Do you have a different model of card you can test, even an older card? I'm just concerned that something else about your system is destroying your video cards (so don't borrow a friend's card!). Maybe you have a PSU that's supplying bad voltages...not an expert on that type of problem.

 

[milanis]

So here are the results:
-MSI Afterburner - fan speed set to 100%, gpu 750 mhz, memory 873 mhz (deafult clocks):
idle is 52-55 °C, ingame crash after some 20 secs at 84-85 °C at the moment.
-MSI Afterburner - fan speed set to 100%, gpu 750 mhz (seems to be the minimum), memory 440mhz (underclocked): idle as before, ingame crash at 80 °C.
Any ideas?
Btw, someone asked, it's stock PSU 450w.
Edit: I can't test it because of the warranty, though even if I wished I don't have another pci-e card, just some older agp cards.

 

[Termie]

I think you have a hardware problem. Is this your card: http://www.tweaktown.com/reviews/1619/msi_radeon_hd_4670_graphics_card/index16.html. If so, you can see that your temps should be much lower.

Two ideas: Is your fan somehow blocked such that it would register at 100 percent; but not turn that fast? You said your fan was 750rpm before. How about at 100 percent?

Second: I think you may have a voltage regulation problem. If that stock PSU is a poorly made one that came with your case/computer, it may be feeding poor quality power through the motherboard. I'm not an expert on that, so you may want to go to the power supply forum to ask about whether that's possible. What CPU do you have and what are the temps on that? Perhaps it's overheating as well.

One way or another, something on your system is causing your HD4670 to overvolt. I've had that problem in the past, in fact, when two programs caused a conflict in the voltage regulator: HWMonitor and Afterburner were both polling for information at the same time, causing my HD5850 to have a voltage spike and crash. Is your temp consistently 80C, or does it suddenly spike according to Afterburner? You may be able to enable voltage monitoring via one of Afterburner's options menus. Go ahead and do that and take a look at what the voltage is over time, both in game and at the desktop.

I'd love to say "it's just a bad card," but I honestly believe something in your system is destroying your video cards, and before you spend any money on a new one, you should really figure out whether something else might be the problem.

 

[happy medium]

I think you have a hardware problem. Is this your card: http://www.tweaktown.com/reviews/1619/msi_radeon_hd_4670_graphics_card/index16.html. If so, you can see that your temps should be much lower.

Two ideas: Is your fan somehow blocked such that it would register at 100 percent; but not turn that fast? You said your fan was 750rpm before. How about at 100 percent?

Second: I think you may have a voltage regulation problem. If that stock PSU is a poorly made one that came with your case/computer, it may be feeding poor quality power through the motherboard. I'm not an expert on that, so you may want to go to the power supply forum to ask about whether that's possible. What CPU do you have and what are the temps on that? Perhaps it's overheating as well.

One way or another, something on your system is causing your HD4670 to overvolt. I've had that problem in the past, in fact, when two programs caused a conflict in the voltage regulator: HWMonitor and Afterburner were both polling for information at the same time, causing my HD5850 to have a voltage spike and crash. Is your temp consistently 80C, or does it suddenly spike according to Afterburner? You may be able to enable voltage monitoring via one of Afterburner's options menus. Go ahead and do that and take a look at what the voltage is over time, both in game and at the desktop.

I'd love to say "it's just a bad card," but I honestly believe something in your system is destroying your video cards, and before you spend any money on a new one, you should really figure out whether something else might be the problem.

I agree 100 percent.

 

[milanis]

Cpu temp idle is some 45c, when under stress it goes up to 70c, never more. The temp of the gpu when ingame slowly goes from idle 54c to 80c where it hangs for some 20 secs and then the crash occurs. Maybe just a coincidence. The gpu from the link seems like mine, but weirdly I can't find a model with GDDR3 512mb which has memory clock 873mhz deafult. Ati Overdrive can set it from 440mhz to 1050mhz max. I see that this card comes with 1600mhz or 2000mhz deafult memory clock, maybe that has sth to do with the crash...
Anyway, I'll check the voltages and see what comes of it.

 

[milanis]

Speedfan reports only one fan, fan1 working at 3400rpm. In gpu-z log file it says 165.0 rpm, but in sensors tab it doesn't read the speed. Also, the speed in the log file is always 165.0rpm except ingame when it's 750rpm, even when I set the fan at 100%. Could it be that the sensor measuring the rotation is not attached?
Afterburner's core voltage meter is dead, even after I unlock voltage monitoring under general tab. It says not all cards allow this. Is there some other way to monitor the voltage?

 

[netxzero64]

You're right! Tnx! GPU load is round 95% ingame, but it crashes even when the load is lower than that! In Windows, the GPU temp is round 55 °C idle, and it crashes when it reaches 80 °C ingame and stays so some 30 secs! Are these temps enough reason for it to crash and if so, what should I do to fix it? A new cooler maybe?

i got a friend who owns a HD4670 PCS version, whenever I played Heroes of newerth after 15 mins it crashes, only to discover that the idle temps reaches 76c, because the fans where stucked between a wire so it was only on passive cooling of the pcs, and when it reaches above 90c, the computer automatically reboots... then what I did was just get the wire out of the way then the fan was up and running again...

your 4670 should have no problem staying at 80c... it is weird that you experienced crashes on games with that mainstream card... it shouldn't be that way...

like others have said, try first increasing your fan speed, if this solves the problem then surely the temps are the ones causing the crashes...

 

[Termie]

Speedfan reports only one fan, fan1 working at 3400rpm. In gpu-z log file it says 165.0 rpm, but in sensors tab it doesn't read the speed. Also, the speed in the log file is always 165.0rpm except ingame when it's 750rpm, even when I set the fan at 100%. Could it be that the sensor measuring the rotation is not attached?
Afterburner's core voltage meter is dead, even after I unlock voltage monitoring under general tab. It says not all cards allow this. Is there some other way to monitor the voltage?

Ok, a few more ideas for you:

(1) GPU fan monitoring is wonky. Just stop looking at those readings and put your hand next to the fan as you adjust the fan speed with Afterburner. Do you feel an increased amount of air flow? At 100%, is it a lot of air? If not, then you have a fan problem. Take out the card, clean the fan and check for obstructions.

(2) The fact that your temp gradually goes up tells me it's not a voltage spike, which I originally thought. There must be a reason that your HD4670, which should be a low-temp GPU, is getting hot. What is your case cooling like? Can you run the case without the cover, or even put a house fan next to it to get some cool air in there. Your CPU temps are high, but not extremely high, so it's hard to say if you have an overall cooling issue. I assume you have stock cooling on your CPU, but what CPU is it?

 

[milanis]

I don't feel any change in the airflow with my hand from either side of the case or the back of it. First I set Afterburner to 25% fan speed and then change to 100%, airflow's the same. The pc is under warranty as a whole and opening the case will invalidate it. Guess I'll have to RMA it, I just don't know what I should tell them isn't functioning.

 

[Termie]

I don't feel any change in the airflow with my hand from either side of the case or the back of it. First I set Afterburner to 25% fan speed and then change to 100%, airflow's the same. The pc is under warranty as a whole and opening the case will invalidate it. Guess I'll have to RMA it, I just don't know what I should tell them isn't functioning.

Ok - I thought you had changed out your previous HD4670 by accessing your case, but I guess not. Too bad you can't open the case without invalidating the warranty. You probably won't be able to feel a change in the fan speed without opening it. Did you hear any difference? A GPU fan at 100% should really make a noticeable amount of noise.

You certainly have a valid claim that the system is not working as designed. Good luck.

 

[tweakboy]

That is too high of a load 90c. Its going to kill your card sooner then it should have to go.

If your bored try put foild on your card and put an egg and watch it cook. good yellow and good white.. then make a sandwhich and eat it.

Couple things you can try. Make sure your fan is spinning at 100 percent during gaming, you need to use riva tuner ... Then open yoru side case and have a 74c or less room. Youll get 60's to 70's c ... good luck gg and of course gb

 

[milanis]

Termie and happy medium are the winners! For those who might have a similar problem: I've got my rig back from the service yesterday and after several tests I can say it works as it should! They changed the PSU, so I guess that's what had caused the problems! Tnx to everyone for your help!

 

[Termie]

Termie and happy medium are the winners! For those who might have a similar problem: I've got my rig back from the service yesterday and after several tests I can say it works as it should! They changed the PSU, so I guess that's what had caused the problems! Tnx to everyone for your help!

Interesting - good to hear that they fixed it for you. After all of our discussion, I honestly thought it was a graphics card problem (despite my initial hunch that multiple VGA failures could only mean one thing - bad PSU). If it's working fine and the only thing they changed was the power supply, that old PSU must have been really messed up. I've never heard of a PSU leading to overheating graphics cards, just crashes (due to insufficient amps). It must have had some pretty bad voltage regulation.

Anyway, congrats on the working system.

 

[slash196]

I was going to say your PSU. The fact that you didn't give us a brand name tells us it was probably a low-quality PSU, which can cause problems and even damage your hardware. So let this be another lesson: Your PSU is not like the gas in a car, it's like the oil.