| View previous topic :: View next topic |
| Author |
Message |
steve_v
Guru
Joined: 20 Jun 2004
Posts: 388
Location: New Zealand
|
 Posted: Sat Sep 16, 2023 2:06 am Post subject: |
 |
|
| szatox wrote: | 20 years ago Apple had a line of desktops cooled this way. You might remember those translucent TV boxes with CD trays at the bottom, whatever they were called. AFAIK they were cooling CPUs with heat from CRT monitors placed right above.
If you put your hand above the exhaust, the draft was strong enough you wouldn't know there was no fan inside, yet the only spinning thing you could hear was the hard drive. |
Sounds like the 1st gen "jellybean" iMac. While an interesting design (and CRT-assisted convection cooling was not unique to Apple mind), I'd be a mite dubious of copying Apple's cooling solutions... Historically they have a pretty horrendous reputation.
Early apple IIs were known to literally melt (to the point of a model-wide case replacement program), the Apple III overheated so badly that chips worked their way out of sockets (the old "drop it from an inch off the desk" fix), watercooling on (G5 IIRC) powermacs used corrosive coolant which invariably leaked directly into the CPU socket, the list goes on.
Even their modern designs prioritise form over function to an extent that makes my head hurt.
_________________
Once is happenstance. Twice is coincidence. Three times is enemy action. Four times is Official GNOME Policy. |
|
| Back to top |
|
 |
szatox
Advocate
Joined: 27 Aug 2013
Posts: 3171
|
| Posted: Sat Sep 16, 2023 10:27 am Post subject: |
|
|
> I'd be a mite dubious of copying Apple's cooling solutions... Historically they have a pretty horrendous reputation.
Nothing ventured, nothing gained. Jellybean iMac looks just like what I remember and the units I saw worked really well, so it can be done right.
Convection cooling runs on a very low pressure though, so it must be unforgiving. Gotta scale it properly and keep things clean.
I wouldn't be surprised if other models (or even units which didn't have a bunch of nerdy IT guys taking care of them) failed.
> watercooling on (G5 IIRC) powermacs used corrosive coolant which invariably leaked directly into the CPU socket, the list goes on.
I can imagine that. Just buy a new one, right? Right? Right? 
> Even their modern designs prioritise form over function to an extent that makes my head hurt.
Bruh. I just though of a soap bar and a certain stereotype regarding apple's user base, and min(d)crafted a flashback to redo of healer. Gimme a bucket.
Anyway, pjp, did you find any upgrade ideas you actually like? |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54387
Location: 56N 3W
|
| Posted: Sat Sep 16, 2023 10:33 am Post subject: |
|
|
| John "Maddog" Hall wrote: | | If Apple made bathrooms, when you needed a tap washer, you would have to buy a whole new bathroom |
Why does that quote stick in my mind :)
I did look at convection cooling a couple of years ago bul eventually went for a big heat sink with fans to match.
All heatsinks tend to accumulate 'grot' over the years so derating the cooling solution is important to accommodate the build up up 'grot' until its cleaned.
I don't tend to do that regularly as it involves rolling about on the floor under my desk.
Step 1 is usually take the side off the case to improve the cooling and delay the rolling about on the floor
I've seen passive cooling solutions for up to 100w. That's real watts not salesmans TDP but I would want about twice that to allow for the 'grot' accumulation in years to come.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
pjp
Administrator
Joined: 16 Apr 2002
Posts: 20067
|
| Posted: Sat Sep 16, 2023 4:46 pm Post subject: |
|
|
| NeddySeagoon wrote: | All heatsinks tend to accumulate 'grot' over the years so derating the cooling solution is important to accommodate the build up up 'grot' until its cleaned.
I don't tend to do that regularly as it involves rolling about on the floor under my desk.
Step 1 is usually take the side off the case to improve the cooling and delay the rolling about on the floor |
That's partly why I'd prefer to restore the unit to it's original place, on top of the desk. In a smaller size.
Might grot be a term that refers to dust + other stuff, or is it something else? For the dust stuff, I generally use, unfortunately, canned air. Availability of non-disposable solutions seems harder to come by.
_________________
Quis separabit? Quo animo? |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54387
Location: 56N 3W
|
| Posted: Sat Sep 16, 2023 6:48 pm Post subject: |
|
|
pjp,
'grot' is a English English slang term for all sorts of dirt, often sticky, that accumulates in any neglected area.
Maybe it's not quite right for the dust and hair that accumulates in CPU fan/heatsink assemblies as that is not normally sticky.
When I'm compelled to clean my system internals, the implement of choice is a stiff natural bristle brush.
That avoids the risk of static and/or fan bearing damage that goes with compressed air and the like.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
pjp
Administrator
Joined: 16 Apr 2002
Posts: 20067
|
| Posted: Sat Sep 16, 2023 8:23 pm Post subject: |
|
|
That's what I thought, but wanted to be sure you weren't referring to some sort of chemical reaction.
I'll have to look for such a brush. I usually try to keep fan blades from moving, but haven't always done so.
_________________
Quis separabit? Quo animo? |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54387
Location: 56N 3W
|
| Posted: Sat Sep 16, 2023 8:40 pm Post subject: |
|
|
pjp,
A 1/2" natural bristle paint brush is good.
My wife is an artist, so I have a good choice of brushes to choose from. As long as she doesn't catch me. :)
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
pjp
Administrator
Joined: 16 Apr 2002
Posts: 20067
|
| Posted: Sat Sep 16, 2023 8:59 pm Post subject: |
|
|
:)
_________________
Quis separabit? Quo animo? |
|
| Back to top |
|
|
C5ace
Guru
Joined: 23 Dec 2013
Posts: 474
Location: Brisbane, Australia
|
| Posted: Sun Sep 17, 2023 5:12 am Post subject: |
|
|
Use convection water cooling.
Mount MB vertical. Bolt heat sink with water inlets at bottom (cold side) and water outlet at top (hot side). Mount motorbike or small car radiator outside of house 2 meters or higher then the computer. Connect cold side of heat sink to bottom of radiator and hot side of heat sink to top of radiator using plastic tubing covered in insulation foam. Fill radiator with car radiator fluid to above radiator hot side. Leave space for cooling fluid expansion. Keep hoses vertical to prevent air and water locks. No need for pump or fan.
We used this cooling system in the 70's to cool banks of 1Kw transistorized HF linear amplifiers with 16Kw output power/bank located in very remote locations.
_________________
Observation after 30 years working with computers:
All software has known and unknown bugs and vulnerabilities. Especially software written in complex, unstable and object oriented languages such as perl, python, C++, C#, Rust and the likes. |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54387
Location: 56N 3W
|
| Posted: Sun Sep 17, 2023 10:29 am Post subject: |
|
|
C5ace,
I can see the instructions for powering up a Cray now ...
Turn on the chiller
Turn on the pumps
Verify that cold water is circulating
Apply power in the following sequence ...
:)
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
Ralphred
Guru
Joined: 31 Dec 2013
Posts: 506
|
| Posted: Sun Sep 17, 2023 9:53 pm Post subject: |
|
|
| NeddySeagoon wrote: | I don't tend to do that regularly as it involves rolling about on the floor under my desk.
Step 1 is usually take the side off the case to improve the cooling and delay the rolling about on the floor |
Preach brother! |
|
| Back to top |
|
|
pjp
Administrator
Joined: 16 Apr 2002
Posts: 20067
|
| Posted: Mon Sep 18, 2023 7:33 pm Post subject: |
|
|
| C5ace wrote: | | Use convection water cooling. |
I've considered water cooling, but I always get this niggling feeling that something is amiss with that approach. :)
| C5ace wrote: | Mount MB vertical. Bolt heat sink with water inlets at bottom (cold side) and water outlet at top (hot side). Mount motorbike or small car radiator outside of house 2 meters or higher then the computer. Connect cold side of heat sink to bottom of radiator and hot side of heat sink to top of radiator using plastic tubing covered in insulation foam. Fill radiator with car radiator fluid to above radiator hot side. Leave space for cooling fluid expansion. Keep hoses vertical to prevent air and water locks. No need for pump or fan.
We used this cooling system in the 70's to cool banks of 1Kw transistorized HF linear amplifiers with 16Kw output power/bank located in very remote locations. |
Hah! I've seen amplifiers that relied on the majority of the case being finned aluminum. I've wondered what could be accomplished with something like that for a CPU.
_________________
Quis separabit? Quo animo? |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54387
Location: 56N 3W
|
| Posted: Mon Sep 18, 2023 7:45 pm Post subject: |
|
|
pjp,
Amplifiers have the output transistors, the hot bits, bolted to the heatsink.
You can't do that with a CPU as it has to operate at a much higher frequency and has a lot more pins.
Having wires even several mm would compromise performance.
A transistor have 3 pins. A CPU, over 1000.
We all really want silicon on sapphire, with liquid cooling on the back of the sapphire substrate.
That's OK until the cooling fails, or maybe a silicon carbide CPU.
Silicon carbide will work with much higher junction temperatures than silicon. Its also very difficult to work with.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
pjp
Administrator
Joined: 16 Apr 2002
Posts: 20067
|
| Posted: Mon Sep 18, 2023 9:31 pm Post subject: |
|
|
The details may not matter, but wires weren't directly involved. There was an aluminum shim with thermal paste that bridged the components -- the hot part(s) with the aluminum case / heat sink. Some embedded or similar devices appear to do something like that, as far as the external case appears. Maybe it has nothing to do with the CPU.
_________________
Quis separabit? Quo animo? |
|
| Back to top |
|
|
pc_user4
n00b
Joined: 14 Sep 2023
Posts: 12
|
| Posted: Tue Sep 19, 2023 6:32 am Post subject: |
|
|
I would like to note that I have some thin clients which only have a TDP of 10W at most however my E3 1240 V3 has a TDP of 80W. But unless I am compiling, the E3 1240 V3 rarely consumes above 10W or 15W; at idle it can be as low as 7W. The lower TDP doesn't indicate your CPU will be running cooler / at less power when idle or in low CPU use.
My opinion is you can get away with less cooling unless you are really and seriously compiling all day long constantly and need to neutralize heat as efficiently as possible.
I'd recommend taking a peek at Gamers Nexus youtube channel for their review of coolers. Despite the name, it's a very scientific channel dedicated to measuring the efficiency and effectiveness of various PC components. You might find some very powerful coolers for under $40.
If you are not compiling constantly, maybe even the stock cooler which comes with certain AMD CPU's is fine. I have a decommissioned Dell workstation I run Gentoo on; it has no front fan and only the stock Dell CPU cooler. I cleaned it very carefully and put on new good quality thermal paste, and it rarely breaks ~72C while compiling (I have a CPU temperature updated every 15 seconds in my tmux status bar). I am confident that could be another 15C cooler with a very nice cooler, but such a cooler can't be mounted in the Dell chassis, and I'm not particularly worried since once it stops compiling, under my usual daily workload I don't think I've ever seen it break 40C.
Regarding water cooling: when your CPU is at an elevated load for an extended time, it takes longer for the CPU to heat up to high temperatures than it does with air cooling, but when the high load is done, it takes a long time for the CPU to settle at low temperature again. With air cooling, the temperature drops drastically immediately upon the heavy CPU workload stopping. This may or may not be important or even worth considering to an end user. I only don't use liquid cooling because I don't find it provides an adequate benefit for my usage over a $30/40 air cooler which is a giant chunk of metal with two 120mm fans attached. |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54387
Location: 56N 3W
|
| Posted: Tue Sep 19, 2023 7:37 am Post subject: |
|
|
pc_user4,
You can put 100w into a heatsink designed for 10w continuous power... until the thermal inertial is used up.
If its the CPU doing that, modern CPUs will thermally throttle back to 10w, so you end up paying for a lot of performance that you can't use, except for 'sprinting'.
Most laptops do that sort of thing, where the cooling is totally inadequate for continuous use flat out.
Some lithium batteries do that sort of thing deliberately too. At high charge/discharge rates the cells heat up,
They have a thick jacket of something that melts before the battery management shuts the battery down.
The latent heat of fusion of the jacket material serves to absorb the heat and extend the time that that the high charge/discharge rate can be sustained.
Once its all melted, its temperature starts to rise again.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
pc_user4
n00b
Joined: 14 Sep 2023
Posts: 12
|
| Posted: Tue Sep 19, 2023 9:37 am Post subject: |
|
|
| NeddySeagoon wrote: | pc_user4,
You can put 100w into a heatsink designed for 10w continuous power... until the thermal inertial is used up.
If its the CPU doing that, modern CPUs will thermally throttle back to 10w, so you end up paying for a lot of performance that you can't use, except for 'sprinting'.
Most laptops do that sort of thing, where the cooling is totally inadequate for continuous use flat out.
Some lithium batteries do that sort of thing deliberately too. At high charge/discharge rates the cells heat up,
They have a thick jacket of something that melts before the battery management shuts the battery down.
The latent heat of fusion of the jacket material serves to absorb the heat and extend the time that that the high charge/discharge rate can be sustained.
Once its all melted, its temperature starts to rise again. |
NeddySeagoon,
I sure wouldn't recommend putting a 10w rated heatsink on a 100w CPU, but maybe if the user has no better way to melt down that metal.....
I'm glad you mention the 2nd point about the laptop cooling because now I have a reason to mention my video card is a HP OEM card, with a mobile GPU (Radeon R9 360M) placed onto a full size PCIe 16x card with dedicated GDDR5 and proper cooling.. The card performs drastically better than it would in a laptop with their very poor cooling mechanism leading to thermal throttling. This model is the Radeon R7 350. There is no competition in R9 360M in a laptop setting and this card I got for an extremely low price. I can use it for PCSX2 at 1080p and Blender with great performance on my Gentoo PC.
In another PC, with a 6th generation Intel Celeron with a very low TDP, I found when putting on a very large heatsink meant to support 2 fans, that average temperatures were lower with the fans removed than with the fans installed (I logged temperatures across 2 days of use in either case and plotted the data in LibreOffice). |
|
| Back to top |
|
|
C5ace
Guru
Joined: 23 Dec 2013
Posts: 474
Location: Brisbane, Australia
|
|
| Back to top |
|
|
steve_v
Guru
Joined: 20 Jun 2004
Posts: 388
Location: New Zealand
|
| Posted: Tue Sep 26, 2023 3:00 am Post subject: |
|
|
| NeddySeagoon wrote: |
Amplifiers have the output transistors, the hot bits, bolted to the heatsink.
You can't do that with a CPU as it has to operate at a much higher frequency and has a lot more pins.
Having wires even several mm would compromise performance. |
There are other reasons a CPU and something like an amplifier have different cooling requirements as well, probably most importantly power transistors tend to have higher tjmax ratings (often 150-180C for BJTs, and as high as 200C for power MOSFETs), and as thermal transfer rate is largely proportional to td you can run a smaller heatsink hotter and still get the same effective power dissipation.
The same applies anywhere else thermal energy is transferred too, so most power electronics gets away with aluminium heatsinks, mica washers, and el-cheapo silicone grease, whereas modern CPU coolers have copper heatspreaders (if not all-copper construction), heatpipes and fancy thermal pastes.
Getting rid of 200W when your tjmax-ambient air differential is 70C is a very different proposition to when it's 150.
| NeddySeagoon wrote: | | We all really want silicon on sapphire, with liquid cooling on the back of the sapphire substrate. |
Saphire? Bah, synthetic diamond substrate all the way, 40x the thermal conductivity and all that.
We can all dream I guess. 
_________________
Once is happenstance. Twice is coincidence. Three times is enemy action. Four times is Official GNOME Policy. |
|
| Back to top |
|
|
pjp
Administrator
Joined: 16 Apr 2002
Posts: 20067
|
| Posted: Wed Sep 27, 2023 2:06 am Post subject: |
|
|
Are there any applications using either sapphire or synthetic diamond?
_________________
Quis separabit? Quo animo? |
|
| Back to top |
|
|
|
Gentoo Chat
