Page 1 of 1

Alternative cooling methods for temperature bound GPUs

Posted: Wed Dec 04, 2013 3:27 pm
by Canageek
I was reading Scott's article on the R2 290X and was thinking; Has anyone tried things a touch fancier then the standard fan or water loop leading to a fan? For example, sending the water to a tank inside an actual refrigeration unit. Alternatively, an open-loop system where you draw water from a tap, then expel the hot waste water (To avoid wasting it, you could pour it into your hot water tank, in theory anyway. It would make your computer rather immobile though, and rather drive up your water bill; however, it does work quite nicely; this is basically what I do in my lab to dry solvents, a still hooked up to a tap and a condenser.)

While we are on crazy ideas, when I run single crystal X-ray diffraction patterns (In short; take a crystal, shine x-rays on it, find out what molecule it is made of) at work, we have a cold stream on the crystal, which is basically a tank of liquid nitrogen connected to a regulator that blows he evaporating nitrogen at a set rate onto the crystal, looking it down to about 100 Kelvin (~-173 c or -280 F for you Americans). Heck, back when I was doing gamma-ray spectroscopy we had part of the spectrometer immersed in liquid nitrogen (~-196 C or -320 F).

Those would would be rather expensive, (liquid nitrogen costs 0.25-$1/L and you'd need close to 100L a week, if not more) but has anyone ever hooked a modern, thermally bound GPU up to a system like that and seen what the limits of the hardware actually are? Is there a lower limit as to what you can safely cool a GPU to?

Re: Alternative cooling methods for temperature bound GPUs

Posted: Wed Dec 04, 2013 3:33 pm
by Captain Ned

Re: Alternative cooling methods for temperature bound GPUs

Posted: Wed Dec 04, 2013 3:57 pm
by Canageek
Cool. You know, you could probably do a decent job with some isoproponol (rubbing alcohol) and dry ice (-78 C); not as cold, but far easier to work with and acquire.

I think Tech Report North should investigate this; I'd be willing to help teach them how to work with either dry ice or liquid nitrogen.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Wed Dec 04, 2013 4:26 pm
by Airmantharp
A guy named Vega runs a chilled loop and is active on the [H] forums- though I believe he only runs the system chilled for benchmarks, considering the problems involved with condensation throughout the system.

But for inexpensive, repeatable testing, that looks like the way to go, since the greatest marginal expense would be in water blocks to support newer GPUs.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Wed Dec 04, 2013 4:42 pm
by Chrispy_
I ran a tap-fed watercooler when I was messing with a Peltier on my old Thunderbird.

Honestly though, I must be getting old. I don't see the point of going with crazy exotic setups for anything other than a bit of experimental funl. When you can dissipate 300W using an off-the-shelf, 240mm AIO water loop, it seems mad to sink too much time and effort into an overclock that's probably only a couple of percent higher.

After a while, cooling isn't the issue so much as the conductivity of the silicon, at which point you want to drastically alter that using LN02.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Wed Dec 04, 2013 8:14 pm
by Kougar
Most exotic I've tried was a full water loop I'd chilled to -6C. Condensation all over the tubing but it was enough to get a benchmark run on a 4.3Ghz Q6600. Never seen ice melt so fast before :D

If you've not heard of them XtremeSystems is the type of forum that specializes in exotic cooling for all parts of a PC. The stuff they run over there makes DIY watercooling look as standard as stock Intel cooling.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Wed Dec 04, 2013 8:54 pm
by Canageek
Kougar wrote:
Most exotic I've tried was a full water loop I'd chilled to -6C. Condensation all over the tubing but it was enough to get a benchmark run on a 4.3Ghz Q6600. Never seen ice melt so fast before :D

If you've not heard of them XtremeSystems is the type of forum that specializes in exotic cooling for all parts of a PC. The stuff they run over there makes DIY watercooling look as standard as stock Intel cooling.


Thanks, I hadn't heard of that; I'm still saving money to build my gaming PC, and it is going to be stock cooling. I was just looking at things we have around my lab (I'm an inert atmosphere chemist) and wondering what would happen if we used them to cool a PC, since the new GPUs are so obviously temperature-bound.

To avoid condensation you'd want to put a drying agent on your air input, say force the air through some magnesium sulphate or such. Or if you are using nitrogen cooling anyway, just blow some dry nitrogen into the case, instead of wet atmospheric air.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Wed Dec 04, 2013 10:07 pm
by Diplomacy42
I used a regular 9.99 box fan and a custom intake/exhaust made from cardboard, tinfoil and aluminum tape so that I could use cold winter air to add a little thermal headroom. it worked very well. quiet. I always got nervous when it rained though.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Wed Dec 04, 2013 10:18 pm
by Airmantharp
Diplomacy42 wrote:
I used a regular 9.99 box fan and a custom intake/exhaust made from cardboard, tinfoil and aluminum tape so that I could use cold winter air to add a little thermal headroom. it worked very well. quiet. I always got nervous when it rained though.


Must be nice to live in such a place. In Texas, we get all four seasons, on Thursday.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Wed Dec 04, 2013 11:41 pm
by zenlessyank
Prolly get laughed outta here, but I set my tower sideways in front of a window A/C unit with the side cover off and cool the whole friggin thing at once. The air is already dehumidified and clean and gives the user the opportunity to OC their rig with every part in it. A decent Window unit is about 200 bux that will work fine. But as usual, bigger is always BETTER!!!

Re: Alternative cooling methods for temperature bound GPUs

Posted: Thu Dec 05, 2013 12:53 am
by anotherengineer
Airmantharp wrote:
Diplomacy42 wrote:
I used a regular 9.99 box fan and a custom intake/exhaust made from cardboard, tinfoil and aluminum tape so that I could use cold winter air to add a little thermal headroom. it worked very well. quiet. I always got nervous when it rained though.


Must be nice to live in such a place. In Texas, we get all four seasons, on Thursday.


lolz whats that
hot, hotter, hot and humid, and hottish with rain/hurricane?

I'm around Timmins Ontario, about 400 miles north of TO, or Rochester NY, we sometimes get a day or two during the winter when the mercury dips to -40C/-40F, I guess you could put your whole tower outside and run cables into the house. It would completely eliminate condensation and even air coolers would probably not hit positive numbers. Maybe one day when a pc is ready for the boneyard I will give it a shot.

But in all practical sense reality, there are fairly inexpensive off the shelf aftermarket coolers, like arctic cooling, etc.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Thu Dec 05, 2013 1:44 am
by I.S.T.
Airmantharp wrote:
Diplomacy42 wrote:
I used a regular 9.99 box fan and a custom intake/exhaust made from cardboard, tinfoil and aluminum tape so that I could use cold winter air to add a little thermal headroom. it worked very well. quiet. I always got nervous when it rained though.


Must be nice to live in such a place. In Texas, we get all four seasons, on Thursday.


Jesus Christ you are not kidding.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Thu Dec 05, 2013 1:52 am
by Orthello
I plan to liquid chill two R290x lightnings to hopefully near 0c temps under load. Under load i think they will be slightly higher. Currently my 7970 lightnings are just under 0c at 1420 mhz xfire under load, or 1450 mhz single card i can get to.

My rig / chiller thread here, avg liquid temp -25c no condensation due to air tight cooling chamber with ~ -23c chilled air. pics near the bottom http://forums.extremeoverclocking.com/s ... 97&page=13.

Only a month or so to wait for those lightnings i hope ...

Re: Alternative cooling methods for temperature bound GPUs

Posted: Thu Dec 05, 2013 3:13 am
by Arclight
I personally think that a normal water cooling solution is more than enough to cool an overclocked self overclocking R9 290x for 24/7 use.

A more exotic cooling system for 24/7 use scenario would be custom made phase change cooler. That said, i''ve only seen people use such a device for the CPU, idk if someone makes them for GPUs as it's more complicated with the mounting system. Plus you'd have to cool the board components separetely with heatsinks and maybe a fan.

Edit: Since phase change will bring the GPU way under 0 degrees Celsius, i must ask, do GPUs suffer cold bugs?

Re: Alternative cooling methods for temperature bound GPUs

Posted: Thu Dec 05, 2013 4:05 am
by puppetworx
Orthello wrote:
My rig / chiller thread here, avg liquid temp -25c no condensation due to air tight cooling chamber with ~ -23c chilled air. pics near the bottom http://forums.extremeoverclocking.com/s ... 97&page=13.


Holy crap! I've dreamt about doing something like this, that whole thread is like porn to me.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Thu Dec 05, 2013 12:44 pm
by Canageek
I.S.T. wrote:
Airmantharp wrote:
Diplomacy42 wrote:
I used a regular 9.99 box fan and a custom intake/exhaust made from cardboard, tinfoil and aluminum tape so that I could use cold winter air to add a little thermal headroom. it worked very well. quiet. I always got nervous when it rained though.


Must be nice to live in such a place. In Texas, we get all four seasons, on Thursday.


Jesus Christ you are not kidding.


Heh, that reminds me of how we cooled a cooling test rig I worked on when I was in Chalk River Ontario; We drew water up from the bottom of the Ottawa River and ran it through a heat exchanger, which is the same way they cool the nuclear reactors there. The hilarious thing, is in the winter the water at the bottom of the river is too cold to pipe into the reactor, and they have to pre-heat it so that the thermal difference isn't too great for the pipes to handle.

Which would be a fun way to cool a PC if you lived in a houseboat. Waterproof case, radiators running to the outside of the case, place it at the bottom of a river, run cables up to your houseboat. Free, very good cooling.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Thu Dec 05, 2013 2:06 pm
by madmanmarz
Nothing beyond water cooling is very practical for everyday computing nor worth it in terms of price/performance. Not only that but you have to make room for extra stuff, and for what? A couple extra hundred Mhz? Then you have condensation to worry about. They still use water cooling for engines so that's enough for me!

I always had the idea to setup my PC(s) in a closet or small space, with all the fans on full blast, radiators mounted externally, and possibly even with a portable air conditioning unit in there (when I was younger I had one in my room blowing at the front of my PC, it was awesome (except loud). Run all the latest cables through the walls/attic that I will ever need and have just the monitor and input devices on the desk.

But even that is overkill. On decent watercooling, and even with mild overclocking, temps usually drop from 40-70C with stock HSF on the CPU to 30-40C, and the results are usually even better for GPU's (either that or their cooling solutions just suck), from 40-100C down to 30-40C. Everything runs almost silently, your room doesn't get as hot and these setups are usually reusable for many rigs to come; so a good loop is well worth the investment. My current CPU block has seen at least 5 CPU's and my GPU block has seen about 4 GPU's.

So yeah I really don't see why anything more is needed.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Thu Dec 05, 2013 2:49 pm
by Forge
zenlessyank wrote:
Prolly get laughed outta here, but I set my tower sideways in front of a window A/C unit with the side cover off and cool the whole friggin thing at once. The air is already dehumidified and clean and gives the user the opportunity to OC their rig with every part in it. A decent Window unit is about 200 bux that will work fine. But as usual, bigger is always BETTER!!!


Well, I gotcha beat, pard.

Many years ago, at a now-legendary LAN party location, this guy, I'll call him Keith (cause that is his name), he decided that his little FOP38 or whatever we were running back then simply wasn't good enough. He got some PVC and some dryer ducting, and his ~2K BTU AC unit, and ran the cooled exhaust directly into the rear exhaust of his case. Reverse all the case fans and poof, he was the first guy I personally knew who ran extreme cooling 24/7. His temps weren't that great, they weren't as cold as the watercooler I used to float ice cubes in, but he could run that setup endlessly. He only ever had to worry about condensation and such if he shut down and opened the case, or if the AC unit got turned off.

I don't recall what CPU he was running, it wasn't that great, and the cooling didn't yield much, but it was highly entertaining.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Sat Mar 31, 2018 9:32 am
by whitzmusic
Canageek wrote:
Is there a lower limit as to what you can safely cool a GPU to?


Of course. There's a specific 'operating temperature' range for almost every electronic equipment imaginable, and GPUs are most certainly not an exception.

Check this article out for further reference: https://www.pcmech.com/article/how-cold ... -hardware/

"WARNING LEVEL TEMPERATURES:
Ambient temperature below 35 F/1.7 C: Generally speaking it’s too cold to operate at this point. You’re dangerously close to freezing and that’s when the physical properties of computer hardware change by flexing (usually). Not a good idea to operate a computer below this mark."

However, keeping liquid nitrogen and other similar extreme stuff aside, I can suggest a few 'practical' things to keep your GPU cool even under high load:

1. Customize the fan-curve so that the fan spins faster quicker than usual. Usually the easiest and most effective solution.
2. Double-check airflow around the GPU. Improve your cable management if there's still room for improvement.
3. Try a spot-cooling solution like these, spot-cooling fans inside the case, etc.
4. Try (slightly) under-volting the card, unless you HAVE TO maximize performance at the same time.
5. Place your system close to a dehumidifier-equipped full-blown air conditioning unit.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Sat Mar 31, 2018 11:20 am
by Chrispy_
It depends on the GPU but often the limit of the cooling is due to insufficient cool air getting to the intake fans of the card. With open coolers especially, the hot air just gets recycled on one side (the side that is next to the PCIe slot) so only half of the heatsink is using air that's cooler than the fins - the hot-half is useless, basically.

Pre-built computer OEM are not shy on using ducting to great effect. In fact, thermally-contstrained servers simply couldn't work without ducting.

Just use flexible ducting from an intake fan or two to your GPU's intake. Hot melt glue, duct-tape, or hell - you could even 3D-print yourself something to hold it in place:

Image

Re: Alternative cooling methods for temperature bound GPUs

Posted: Sun Apr 01, 2018 7:49 pm
by Forge
Four years and change, that's got to be pretty close to the record for a thread necro.

And for the record, Keith was running a K6-2 or a K6-III at that time. He might have just gotten onto a Duron or Athlon, but I don't think he had, when the AC incident went down.

Re: Alternative cooling methods for temperature bound GPUs

Posted: Sun Apr 01, 2018 8:07 pm
by just brew it!
Forge wrote:
Four years and change, that's got to be pretty close to the record for a thread necro.

I'm pretty sure there have been longer ones, even ignoring spam. (Most of the really "deep necro" incidents tend to be spammers, so you only see 'em if you happen to look at the forums before the mods swing the ban hammer.)

Re: Alternative cooling methods for temperature bound GPUs

Posted: Sun Apr 01, 2018 8:10 pm
by Forge
I don't doubt it, but none came to mind. Call it my personal record, then.