[IDEA] Cooling based on target temperature
As you know, i'm the proud owner of a DV9000 with nvidia 8600M GS video card. Well, two actually, but only one operational at the moment.
My new chip is still on the way from China, and i need the laptop to work as my Lenovo died (another chip i need to wait for...).
Thing is, this laptop has only been lightly used, and never reworked. When i got it it showed only lines on the screen, but letting it sit belly-up for a few hours, then turning it off and back on again while pressing on the case in a certain way, made it work fine.
I've used this trick a lot of times - one of the things you get used to when owning a DV9000 or pretty much anything with bad nvidia in it...
At first, it had some really absurd temperatures but worked great. The GPU would IDLE at 73-80C, and FurMark brought it to 105C within a few minutes, at which point it started throttling, so it fell to 101C, then it resumed at full speed... and so on.
Then my Lenovo died and i took the DV9000 apart to put the T9300 CPU in it. While i was in there, i gave it a good cleaning. With all the dust bunnies out, the GPU started running much cooler, between 55C at idle and a maximum of 83C loaded. This should be good, right? Well, in theory yes, but something funny started happening - the machine started locking up very often, and sometimes it showed lines again, requiring again to leave it for a long while to heat up.
I also found a pattern. Whenever the temperature went over 70C for any extended period of time, the laptop locked up.
I put up with it as i have other laptops, but i really need to work on this one right now. I got fed up with it last night and remembered it worked better before i cleaned the dust... Seriously. So i powered the thing up to BIOS, blocked the air vents and waited. When the fan got to full throttle, i restarted it and went into Windows. GPU was at around 80C. What do you know - working fine. I also locked the powermizer to full throttle, and ran the Windows Experience Index again. The score was significantly improved. Maximum temperature it reached was 98C. No lockups.
But even blocked, the fan was fairly loud, and this must not be good for it so i looked for another option. I unblocked the vents and the GPU cooled under 70C quickly as the fan was near maxed. As soon as it got over 70C once more - bam, lockup. Knowing of the nvidia underfill problem, it would be ideal to keep the chip UNDER 70C all the time, but this is impossible. 70-73C seems to be "the breaking point", but as i found, the chip will work just as fine if not better, if it is constantly running HOTTER than this.
I know how to control the fan in the DV6xxx and 9xxx series via software, so i locked the fan to minimum speed. GPU runs around 70 to 85C this way. And guess what. It hasn't frozen once since i did that, and it's been running for 14+ hours (left it on while i was asleep).
Until the new chip arrives, i don't want to do any work on this board. Not even a reflow. So if it wants to run hot to work properly, i've got no problem with that...
So here's what i thought. I don't want the GPU to burn up, either, so keeping the fan locked to minimum speed isn't the brightest idea. Instead of normal cooling policy (x deg C = x fan speed), i was thinking to make a fan control software based on target temperature.
So let's say i want my GPU to run at 75C. Which is relatively high for idling, to reach this temp the fan would need to run very slow or not at all. But when the GPU is loaded, a slow-running fan will guarantee temps much higher than this.
The fan should aggressively kick in to maintain this temperature, say within a 5-10C window.
This does actually make sense, as by keeping the chip at a constant operating temperature, thermal cycling is minimized and material fatigue would be reduced.
It would be similar to what AMD did in its latest generation GPUs, but they went much farther than just fan speed - the R9 290x continuously adjust clocks, voltage and fan speed to keep the chip at a target temperature - AMD found it works best at 95C, but this is too high for comfort for most users, so this temperature target can be lowered in CCC.
What do you guys think? And yes, it will be getting replaced, but it will be at least 2 more weeks until the new BGA arrives, and my 1920x1200 screen is also on the way and will likely arrive earlier, so i need to keep this one kicking for a little longer.
My new chip is still on the way from China, and i need the laptop to work as my Lenovo died (another chip i need to wait for...).
Thing is, this laptop has only been lightly used, and never reworked. When i got it it showed only lines on the screen, but letting it sit belly-up for a few hours, then turning it off and back on again while pressing on the case in a certain way, made it work fine.
I've used this trick a lot of times - one of the things you get used to when owning a DV9000 or pretty much anything with bad nvidia in it...At first, it had some really absurd temperatures but worked great. The GPU would IDLE at 73-80C, and FurMark brought it to 105C within a few minutes, at which point it started throttling, so it fell to 101C, then it resumed at full speed... and so on.
Then my Lenovo died and i took the DV9000 apart to put the T9300 CPU in it. While i was in there, i gave it a good cleaning. With all the dust bunnies out, the GPU started running much cooler, between 55C at idle and a maximum of 83C loaded. This should be good, right? Well, in theory yes, but something funny started happening - the machine started locking up very often, and sometimes it showed lines again, requiring again to leave it for a long while to heat up.
I also found a pattern. Whenever the temperature went over 70C for any extended period of time, the laptop locked up.
I put up with it as i have other laptops, but i really need to work on this one right now. I got fed up with it last night and remembered it worked better before i cleaned the dust... Seriously. So i powered the thing up to BIOS, blocked the air vents and waited. When the fan got to full throttle, i restarted it and went into Windows. GPU was at around 80C. What do you know - working fine. I also locked the powermizer to full throttle, and ran the Windows Experience Index again. The score was significantly improved. Maximum temperature it reached was 98C. No lockups.
But even blocked, the fan was fairly loud, and this must not be good for it so i looked for another option. I unblocked the vents and the GPU cooled under 70C quickly as the fan was near maxed. As soon as it got over 70C once more - bam, lockup. Knowing of the nvidia underfill problem, it would be ideal to keep the chip UNDER 70C all the time, but this is impossible. 70-73C seems to be "the breaking point", but as i found, the chip will work just as fine if not better, if it is constantly running HOTTER than this.
I know how to control the fan in the DV6xxx and 9xxx series via software, so i locked the fan to minimum speed. GPU runs around 70 to 85C this way. And guess what. It hasn't frozen once since i did that, and it's been running for 14+ hours (left it on while i was asleep).
Until the new chip arrives, i don't want to do any work on this board. Not even a reflow. So if it wants to run hot to work properly, i've got no problem with that...
So here's what i thought. I don't want the GPU to burn up, either, so keeping the fan locked to minimum speed isn't the brightest idea. Instead of normal cooling policy (x deg C = x fan speed), i was thinking to make a fan control software based on target temperature.
So let's say i want my GPU to run at 75C. Which is relatively high for idling, to reach this temp the fan would need to run very slow or not at all. But when the GPU is loaded, a slow-running fan will guarantee temps much higher than this.
The fan should aggressively kick in to maintain this temperature, say within a 5-10C window.
This does actually make sense, as by keeping the chip at a constant operating temperature, thermal cycling is minimized and material fatigue would be reduced.
It would be similar to what AMD did in its latest generation GPUs, but they went much farther than just fan speed - the R9 290x continuously adjust clocks, voltage and fan speed to keep the chip at a target temperature - AMD found it works best at 95C, but this is too high for comfort for most users, so this temperature target can be lowered in CCC.
What do you guys think? And yes, it will be getting replaced, but it will be at least 2 more weeks until the new BGA arrives, and my 1920x1200 screen is also on the way and will likely arrive earlier, so i need to keep this one kicking for a little longer.
Comment