GPU Temperature Question - How High is Too High

[RockHardly75]

I've set up my first SLI config with two 660's (regular, not Ti).

I noticed some artifacting (some flickering here and there - kind of sporatic) with FarCry 3, so I loaded up a temp monitoring tool and watched it as I played for a while.

I noticed that it jumped up to about 60-70 C pretty quickly, then slowly increased to 80-81 on GPU1 and about 73-74 on GPU2. The fans on both read 52% and 48% respectively. The temps pretty much stayed there for the next 10 minutes or so as I played, so I assumed that was as high as they were going to go.

I've looked on the forums here and seen that it looks like 80-90 is the expected temp range for a graphics card while under load. I did note that some people advised to start getting concerned when the card goes 90+, especially if it goes to 100 and beyond.

Is this true? Is 80-90 C the normal range for an under-load GPU?

 

[scout_03]

what kind of air flow you got in that case ?

 

[Cons29]

60 is fine, 70-ish is ok. but near 80, personally is not ok with me. if this was an 8800gt then it is normal.

are your cards close together? i mean, almost no gap in between?

for reference, my gpu (gtx 670) hovers between 49-50 most of the time, playing bf3 on ultra settings. only a single card though, i reach 54-55 when not using aircondition

 

[Cons29]

yes improve the airflow, add some fans, do cable management, etc. get spotcool or something and help the gpu fans

 

[RockHardly75]

The case is an Antec 900, I have two fans pulling in the front (over the hard drives), one fan pulling in the side, one fan pulling in the back (over radiator for CPU), and the large fan at the top blowing out.

All fans blowing in are 80mm, the fan at the top I think is a 200mm.

The GPUs have about an inch between them - can't do much about that, only have two slots on the board for them.

One of the intake fans in the front is going over 3 hard drives - I suppose I could try to move one of those around. I also have quite a few cables in there, but actually less than I had before (when I used one card) - because my newer power supply lets me get by with fewer cables.

Between the time I posted the last message and this one, I downloaded FurMark and ran a 15 minute test - it maxed out GPU 1 at 83 deg, and GPU 2 I think hit 78 deg.

I did notice that during the whole test, the fans on the GPUs increased as the temp increased, and when the temps were hovering at their max'es, the fans stayed around 50-55%. Should I look for a way to increase these speeds?

 

[unksol]

80 as a peak temp on the top card in SLI is no reason for concern at all. You can certainly increase airflow, turn up the fan etc, but it's not hurting anything

 

[stickmansam]

You could change the fan curve to be more aggressive with a program like MSI afterburner

I would optimize the profiles to your most stressful game rather than Furmark

 

[RockHardly75]

Installed Precision X (from EVGA install disk with card), set fan control to auto, and it kicked the fans up a bit. Ran the FurMark test again, and got max temps of 74/66 (with the majority of the test ranging around 72/64).

I noticed that with this second test, the fans kicked up to about 72% or so vs. 55% the first time when the temps were 80-82 deg.

So - I got about a 10 degree boost.

Any good ideas/guides/etc. regarding cable management? I seem to always have a bird's nest of cables in there.

 

[RockHardly75]

I would optimize the profiles to your most stressful game rather than Furmark

Are you saying this because Furmark may not be as stressful as gaming, or because Furmark is more stressful than what I would normally see?

 

[stickmansam]

It is generally more stressful on your card though Far Cry 3 is not far behind and some say just as stressful it not more

Best to optimize to games since that's your most common use and make sense to gear it towards that

In term of cable management, just keep the cables tucked away behind the motherboard tray if you have room or off to the side to less restrict airflow

 

[hizodge]

I don't see any problems running cards above 80 degrees. Most graphics cards with reference cooling run at 80+ under load and can take all the way up to a whopping 100 degrees, although I personally wouldn't let it get that high.

Having your #1 card run hotter than the other one is very common with Crossfire & SLI configurations, as are occasional artifacts popping up in Far Cry 3. If you're not seeing artifacts showing up in a wide range of games or they're not constant in Far Cry 3, I'd say you have no problems at all.

Just for reference, I'm running a rather lenient fan profile that let's my 2x 7970s get up to 85 before drastically ramping up the fans all the way up. The #1 card is typically around 76-82 degrees and #2 about 5 degrees lower under full load.

 

[Jonathanese]

one fan pulling in the back (over radiator for CPU)

I would turn that around. I have an Antec 900, and after putting some numbers together, that would have the best push/pull ratio.




For a while, I ran my 660 alongside my 9800GTX+ (PhysX/CUDA), and my temps were at about 82C using the heaven 4 benchmark windowed. After downloading Precision X and monitoring my fan speeds, I realized that the fan was running at only about 42% at 82C. I think that's insane. I don't mind noise when my GPU is that warm. The 9800GTX+ used to spin up to 80% when the gpu would hit 70C, and it had a higher Tmax rating (108C vs 95C).


So I made my own curve in Precision X with more aggressive fan settings, and re-ran Heaven, this time with a 800MHz QDR overclock on the RAM. It struggled to push the upper 60's. So that made all the difference.

Also, the Antec 900 is a great case for custom fan speeds. If your motherboard supports SpeedFan, and you aren't already using it, I would download that right away. With a bit of tweaking, you can control almost any of your fans automatically based on any combination of temperature sensors. You'll want to set it so that your fans run up to 100% when your GPU is getting above 80C. For me, it gets there at about 75C.

Basically, as long as you take those steps, you're maxing out the cooling when the GPU gets warm. So that's the best you can really do without resorting to aftermarket cooling ad TIM.

 

[scout_03]

been looking at your antec 900 case all the fan are tri-cool so they have a switch for the speed from low to high try them at med ,there is also a diuct in that case for cooling the twin graphic card.Iwould put the front and side as intake and the back and top as outake or get some other fan with higher cfm to reduce the gpu heat .

 

[RockHardly75]

Played with the system again tonight - didn't benchmark at all, just fired up FarCry 3 with all the bells and whistles and with the increased fan profile I had set-up last night.

I ended up not getting temps above 68 degrees - and it stayed mostly between 66 and 68. I haven't messed with the cabling - I will do that maybe this weekend or so. Even though my temps seem to be under control with the increased fan speeds, I still need to get the bird's nest in my computer under control.

I've noticed another issue - the text on my monitor begins to ghost a bit after a while. I don't notice it at start-up, but after it's been on a while, if I look at the text in Steam (especially), I see little shadows of the text trailing behind it. I noticed it first in Black Ops 2 (similar font color with grey background to Steam). Just to verify it was my monitor and not my card, I screen shot it - noticed the problem in the screen shot, and then moved the screen shot to my other monitor and the problem went away.

My main monitor is connected via VGA cable and my secondare monitor is connected by HDMI. From what I've read on-line, it sounds like I need to bite the bullet and go ahead and get a DVI cable for my monitor.

It's strange, though, I never noticed this ghosting thing before - I'm wondering if I have just been lucky or not as observant - or could it be the effect is more noticeable in the signal from these Nvidia cards over the VGA cable as opposed to how my old ATI card handled it? Actually - now that I write that - I didn't notice this issue with just the single Nvidia GTX 660.

Anyway - I'm thinking the DVI cable is the answer - at least, it's the cheapest option at this point.

 

[RockHardly75]

By the way - thank you all for your replies and advice. I'm continually impressed with the quality information I can find on this site every time I come looking.

 

[scout_03]

two things you could try is to interchange from dvi and to hdmi on the monitor so the one using the hdmi become the dvi and the dvi gone use the hdmi ,then do the same with the graphic card .

 

[RockHardly75]

I actually picked up a DVI-D cable tonight, and after playing the same stuff I've been playing for about the same amount of time, I haven't seen the ghosting at all. I think that took care of it.

 

[scout_03]

if the ghosting are not back then the cable is the one who made it .

 

[Trogilon3]

Recently I noticed my graphics card was running very hot. I took it out and looked inside it and saw that the heat sink inside was covered with dust.

I took the fan off, cleaned it out, put it back in, and ran it. Cleaning it out took 15 degrees off the temp.

That will be the first thing I check the next time I see it running hot.