PDA

View Full Version : new, 2nd GPU questions



thebigJ_A
02-03-2012, 07:56 AM
Couple quick questions.

Just got a second 560 ti GPU, identical make/model as the first, for SLI, but the new one (placed on the bottom) runs ten degrees hotter under load. That's not necessarily an issue, but the odd thing is, the maximum fan speed of it is like 300 RPM slower than the old one. Is that something to do with how SLI works, or do some cards just have slower fans, or what? If it's just that the card has a slower fan and that's the way it is, maybe I'll move that one to the top slot.

Oh, and it's getting into the low 80s when I run Furmark, even though I have a very large number of fans in my case moving air. The other is in the 70s. Not sure if I should be concerned, but I digress.

The other thing that's bothering me is, as the temperature goes up in Furmark, I hear a gradually louder buzzing noise, which I believe is from the new gpu, as I never heard it before and it stops the instant I exit the program. I've heard the term "coil whine", but this is more a buzz, and is rather alarmingly loud after a while. Is that coil whine or something worse?

Cheers.

kalelovil
02-03-2012, 08:51 AM
Although the second card may be the same make and model as the first one, graphics card makers do make small changes to their models over the course of production runs (usually to cut costs) without changing any of the branding.

ufc4836
03-03-2012, 10:18 AM
I am waiting to hear the answer.

gimperial
03-03-2012, 02:02 PM
Like I said in the other thread, those temps sounds pretty low, especially under Furmark, so no need to worry about that. My 5850s got to high 80s with Furmark. And the bottom card is always hotter because there's less cool air because of the top card.

As for the whine, are you sure it's not just the 2nd gpu's fan?

(And from the other thread, going by neweggs prices 2 560tis are ~$520, and you could have gotten a 7970/7950 for that money. If it was brand new might have been able to return it, but either way it's too late now, and with the two cards you have a lot of horsepower. Personally (this is just my opinion) the only time I see 2 GPUs are being good value is if you can get two of them used for cheap, so in my case I paid nearly half price for what was effectively a 5970, the most powerful GPU around back then, but you do end up having to deal with a lot more stuff than with 1 Gpu)

Bungle
03-03-2012, 08:58 PM
Normally the top card runs hotter than the bottom card, as heat rises. My first guess is that your top card is the hotter card and you're reading the temps wrong. I owned three different multi-GPU computers back in the day and the top card was always about 10C hotter than the bottom one. If your bottom card truly runs hotter than the top card, you could try removing the heatsink, cleaning the GPU, applying Arctic Silver to the GPU, and re-installing the heatsink. It doesn't really matter, as 80 isn't going to hurt any video card anyway. It's just a curiosity.

The buzzing sound is coming from a capacitor in your computer. It could be the new video card, but it could also be coming from your motherboard or power supply (adding a second video card means your computer draws more power now and it can be straining another part of your system). Buzzing caps are annoying as hell. I don't actually know if it means the capacitor is going to die or not - all I know is that I would demand an RMA if my video card made a squealing or buzzing noise.

thebigJ_A
04-03-2012, 06:09 AM
It's GPU2 that runs hotter, the one that doesn't do any work if I set a game to run on one card. GPU2 also has a slower maximum fan speed. (3300 vs the 3600 rpm the first card I bought had/has)

I've got a 750w Corsair PSU, which ought to be sufficient.

I got up the bravery to run Furmark again, this time for longer, and I didn't hear the noise. However, GPU2 got up to 87 degrees before my PC rebooted itself (GPU1 was at ~76).

Maybe I should switch them and see what happens.

@gimperial I decided a new thread was in order for this, since this isn't really about building a pc for the first time anymore. Together they cost $420.

Odeon
05-03-2012, 07:19 PM
I didn't see this thread until after I replied to the other one here (http://www.rockpapershotgun.com/forums/showthread.php?2818-Help-1st-time-PC-builder&p=98842&viewfull=1#post98842). I remember reading in a thread somewhere (I'm not even sure what forum site it was) that a lot of the SLI/Crossfire setups these days tend to have the same "higher heat, slower fan on the second card" issue. The OP was trying to raise his second card's fan speed using SpeedFan, but the general agreement in the responses was that the second fan runs slower in all cases because of a limitation in the drivers and that there's no way to directly control the second card's fan because it treats both cards as one. The OP even swapped the two cards and got the same exact result, supporting the consensus. Since the driver only allows access to the first card's fan, when the system sees that the second card is getting hotter, it tries to raise the fan speed to reduce heat, but only the first card's fan speeds up, reducing already low-enough heat on that card and the second card continues to get hotter.

Until the NVIDIA and AMD drivers start taking this into account, there doesn't seem to be much that can be done. Increasing the speed of the 140mm fan could help, but that's only possible with a PWM fan and it would increase overall noise from your PC.

thebigJ_A
07-03-2012, 03:11 PM
I used to have a 120mm fan in the side, and I switched it to a 140. I'm wondering if the 120 might have been better at cooling the GPUs. Sure, it moves less air, but the air it moved was moving faster. I just can't fathom why my cards, especially the second, run so hot in furmark that the pc reboots, even at stock clocks. I could try the re-seating the cooler thing mentioned above, but while I did that for the cpu, I'm frankly scared to start taking apart a graphics card.

gimperial
07-03-2012, 09:35 PM
I didn't see this thread until after I replied to the other one here (http://www.rockpapershotgun.com/forums/showthread.php?2818-Help-1st-time-PC-builder&p=98842&viewfull=1#post98842). I remember reading in a thread somewhere (I'm not even sure what forum site it was) that a lot of the SLI/Crossfire setups these days tend to have the same "higher heat, slower fan on the second card" issue. The OP was trying to raise his second card's fan speed using SpeedFan, but the general agreement in the responses was that the second fan runs slower in all cases because of a limitation in the drivers and that there's no way to directly control the second card's fan because it treats both cards as one. The OP even swapped the two cards and got the same exact result, supporting the consensus. Since the driver only allows access to the first card's fan, when the system sees that the second card is getting hotter, it tries to raise the fan speed to reduce heat, but only the first card's fan speeds up, reducing already low-enough heat on that card and the second card continues to get hotter.

Until the NVIDIA and AMD drivers start taking this into account, there doesn't seem to be much that can be done. Increasing the speed of the 140mm fan could help, but that's only possible with a PWM fan and it would increase overall noise from your PC.

FWIW, with ATI you can control each card's fan speed separately.


I used to have a 120mm fan in the side, and I switched it to a 140. I'm wondering if the 120 might have been better at cooling the GPUs. Sure, it moves less air, but the air it moved was moving faster. I just can't fathom why my cards, especially the second, run so hot in furmark that the pc reboots, even at stock clocks. I could try the re-seating the cooler thing mentioned above, but while I did that for the cpu, I'm frankly scared to start taking apart a graphics card.

Didn't you say your hotter card went to high 80s? Your PC really shouldn't be rebooting when a GPU is running at high 80s, they should be able to go to 100C. What are your CPU temps?

Are you overclocking? That could be another reason for the reboot. 87C for a GPU wouldn't cause a reboot.

Odeon
08-03-2012, 05:33 PM
FWIW, with ATI you can control each card's fan speed separately.
Hmm, I thought I'd read that the same problem existed for both NVIDIA and ATI setups, but since I've never considered Crossfire, I may just be remembering wrong. It's not unheard of for my memory to fail me on details that I pay less attention to. 8-D


Didn't you say your hotter card went to high 80s? Your PC really shouldn't be rebooting when a GPU is running at high 80s, they should be able to go to 100C. What are your CPU temps?

Are you overclocking? That could be another reason for the reboot. 87C for a GPU wouldn't cause a reboot.
I was going to say the same thing about your card's max temp. It's pretty common for video cards to hit 100 C under continuous, intense loads, so unless your BIOS is set to reboot your PC at temps near 90 C, the rebooting would be caused by something else.

thebigJ_A
09-03-2012, 01:49 AM
I'd totally forgotten I'd set the "auto-overclock" in the BIOS. I'd meant that to be temporary, until I had the time to manually find a stable OC. (Been too busy lately)

Could that be it? I mean, I don't have it on the highest setting, and I have an aftermarket cpu cooler. I figured it could handle it for a while.

I'm not sure where a setting for rebooting at a certain temperature would be to check if it's set to 90 degrees.

Mistabashi
09-03-2012, 02:12 AM
I was going to say the same thing about your card's max temp. It's pretty common for video cards to hit 100 C under continuous, intense loads, so unless your BIOS is set to reboot your PC at temps near 90 C, the rebooting would be caused by something else.

It depends entirely on the GPU - some are rated up to 105c, so generally they'll set the auto-shutdown at 100c. However, some are rated for 95c or even 90c, so the threashold would be set lower. It's not just the GPU itself that determines this either, all the components on the board will have a max stated operating temp, and it's common for card manufacturers to use cheaper components that aren't rated above 85.

The temp threshold for shut-down is determined in the card's firmware, not the BIOS, so you can't change it unless you flash the firmware (possible, but you'd be pretty daft to do it). 90c isn't considered a 'safe' operating temperature for any GPU, if your card is running at these temps for any length of time it's going to have a significant impact on it's lifespan.

@OP: Are your cards in adjacent PCI slots? This is often the case with ATX boards, and it means that the intake from one card is partially blocked by the second card, so it'll get less airflow. Can you take a picture of the inside of your case so we can see where all the fans are etc?

gimperial
09-03-2012, 12:11 PM
I'd totally forgotten I'd set the "auto-overclock" in the BIOS. I'd meant that to be temporary, until I had the time to manually find a stable OC. (Been too busy lately)

Could that be it? I mean, I don't have it on the highest setting, and I have an aftermarket cpu cooler. I figured it could handle it for a while.

I'm not sure where a setting for rebooting at a certain temperature would be to check if it's set to 90 degrees.

The setting for reboot-at-a-certain-temperature is for CPUs not GPUs, but without seeing your BIOS it's hard to say where it is, just that it could be somewhere there (but then again might not, depending on the BIOS/manufacturer).

You can always try run prime95 (which you should be doing if you're overclocking anyway) while monitoring temperatures, and if your CPU goes above 75 and/or shuts down, then that's the culprit.

The max temp for the 560ti is 99C by the way, so 85 is a safe temperature for furmark.

Odeon
09-03-2012, 07:39 PM
I'm Odeon and I approve of the last two messages.

It wasn't until after reading them that I realized what I'd seen was BIOS settings for warnings and auto-shutdowns for high CPU temps, not GPU temps. GPU boards have their own sensors that may not be directly accessible by the BIOS, but either way the card's software would determine when to shut down a PC based on GPU temperature. So soory aboot that bit.

It's possible that your auto-overclock enables settings that are too aggressive for your hardware, but if I'm remembering correctly from your other thread, it doesn't seem like your parts should be very sensitive to moderate overclock settings. Still, it's definitely worth a shot to try turning the auto-overclock off and going back to stock settings to see if there's any difference in temps.

I just re-read what you'd said about your cards having different max RPMs, so if turning off auto-overclock doesn't make the difference, try swapping your cards. And if you haven't already tried it, disable SLI and try running Furmark on one card at a time to see if you can isolate the cause to one of the cards.

thebigJ_A
10-03-2012, 05:06 AM
Oh man, I am so stupid!

The wire from my 12-in-1 (or however many) card reader was pressing up against one of the two fans on the bottom GPU, preventing it from turning at all. That explains the buzzing, and how the buzzing stopped. It must have been spinning against it until it got stuck. And it explains the different RPMs, I think.

So, I tied that wire out of the way and ran furmark again. The temps on both cards leveled off in the high 70s, and the RPMs were the same. I also got a lot higher fps for some reason. Like, 60 where I was getting mid 30s before.

I ran Prime95, though I didn 't have time today to run it for very long, but the highest temp I saw for any of the cpu cores was 76 degrees. That sound right?

I feel like I dodged a bullet. I don't want to think about what could have happened if I hadn't noticed. Now I know to be very careful about where things are in relation to each other. Thanks for the help, everybody!

Odeon
11-03-2012, 05:29 AM
Awesome, I'm glad you got it figured out and that it was an easy fix. Always look for the simplest solution first and go from there. ;-)