You are not registered yet. Please click here to register!
groovetek's 172w (currently 80-90w) watercooled peltier progress thread! - i4memory.com - different look at memory
Tags Register Blogs FAQ Members List Calendar Mark Forums Read

Notices

Reply
 
Thread Tools
Old 17-09-2005, 09:30 PM   #1
groovetek
Senior Member
groovetek's PC Specs
 
Join Date: Aug 13 2004
Location: Melbourne
Posts: 1,307
groovetek's 172w (currently 80-90w) watercooled peltier progress thread!

***groovetek's Current Peltier Cooling Results***


>>NOTE: Where the temperature recorded is "-1" degC, it means the CPU temperature is sub-zero, but I have no method of testing how far below 0 degC the temperature is registering, due to the lower limit of 0 degC on the DFI NF4 Ultra-D temperature sensor/monitor.

> Cooling Results Analysis
As you can see, during low voltage/speed/wattage situations, the peltier running at it's current capacity is more than capable of keeping the CPU cool... registering sub-zero temperatures during IDLE, and load temps of no more than 10 degC. stressCPU was used as the method of full load, and has been shown to generate pretty much as much heat as the CPU can produce.

> IDLE temperature analysis:
Clearly, the A64 chips are able to minimize power consumption when idling. This is shown by the somewhat small spread of IDLE temperatures between 1.26V (LOW VCORE) and 1.66V (HIGH VCORE) (6-7 degrees difference).
It is also evident that the CPU frequency plays a rather small role in determining idle temperatures, where the maximum spread of idle temperatures between 1800 and 2800mhz at any given voltage was 2degC (obtained at 1.66V - 5degC @ 1800mhz vs 7degC @ 2800mhz). Load temperatures are a DIFFERENT case though!
Overall it's clear that at IDLE state, the peltier was not being overloaded.

> LOAD temperature analysis:
Here's where the analysis becomes more trivial and somewhat complicated. Let's begin with testing at 1.26V. When clocked at a lowly 1800mhz (3000+ stock frequency), the maximum temperature observed during constant and extended load was only 6 degC. This would equate to 6-8 degC separation between the idle and load temperatures (cannot be confirmed without a means of measuring sub-zero temperatures though). By increasing the CPU speed in 200MHz steps, the load temperatures increased somewhat linearly, until at 2604MHz where the CPU's temperature had maxed out at 12degC. The CPU failed to run stressCPU @ 2800mhz @ 1.26V without locking up after awhile.
NOTE: the maximum separation between idle and load at 1.26V was approximately 12-13degC. 6degC separated the load temperatures at 1804 and 2604mhz.

At 1.39V, the temperature scaled fairly similarly, where the load temperatures increased linearly with CPU speed. Overall, the load temperatures were about 4degC higher than the 1.26V tests at any given CPU speed, and finally, the IDLE temps had readings of 0degC from 2.4ghz and above.
NOTE: the maximum separation between idle and load at 1.39V was now 17 degC. 7degC separated the load temperatures at 1804 and 2800mhz.

At 1.52V we finally start to see the peltier struggling to dissipate all the head generated by the CPU during full load. Whilst the temperature did scale somewhat linearly with CPU speed (do note the temperature increase from 1804 to 2004mhz seems to be consistently larger than the increments that follow (see tabulated results above)), it was blatantly obvious that at full load, the wheels were starting to fall off, as they say. Whilst IDLE temperature was still only 2degC @ 2.8ghz, the full load had actually hit 25 degC! One can only dread the results that will be obtained at higher vcore!
NOTE: the maximum separation between idle and load at 1.52V was now 23 degC. 10degC separated the load temperatures at 1804 and 2800mhz.

Finally, it was time to totally stress/overload the peltier in its crippled state. At 1.66V, even the IDLE temperatures had somewhat jumped up a little, where even at 1804mhz, the IDLE temperature was 5degC. Shockingly, the full load temperature registered was 25degC. It is interesting to note that despite the higher IDLE temperature, the load temperature was the same as that of the 2.8ghz @ 1.52V. THEREFORE it is rather obvious that the relationship between voltage, cpu frequency, and temperature is not proportional.
Increasing the CPU speed to 2ghz jacked up the maximum load temperature to 29degC - the biggest increment recorded at any given voltage from 1 cpu speed to the next. The remaining speed bumps resulted in linear temperature increases of about 2 degC, where at 2800mhz, the IDLE temperature recorded was 7degC, and the full load temperature was a disappointing 39degC.
This is only about 3 degrees lower than what would have been obtained if the peltier wasn't even being used in the first place!
NOTE: the maximum separation between idle and load at 1.66V was now a huge 32 degC. 14degC separated the load temperatures at 1804 and 2800mhz.

> Cooling Results Conclusions
Evidently, peltiers are extremely effective for cooling CPUs under low wattage situations. Unfortunately however, for majority of computer enthusiasts this is not when powerful cooling is required.
So, to sum up the observations of the trends observed:
* Vcore does affect IDLE temperatures slightly, but has FAR MORE SUBSTANTIAL impact on LOAD temperatures on A64 CPUs.
* CPU frequency has little impact on IDLE temperatures; the difference is more obvious in LOAD situations. However, LOAD temperatures do increase, more or less linearly proportional with CPU frequency.
* Difference between IDLE and LOAD temperatures is increased with increased wattage.


I can't wait to get the 24V PSU so that I can run this 172w peltier at its full capacity. The tube from the block to the pump was warm, about 30-35 degrees to the touch (a rough estimate , i know...), indicating that there is some cooling capacity left in the current WC setup. I shall see how it copes when the peltier is running at full power.

Thanks for the feedback guys, I appreciate it.

On a side note: The insulation is working beautifully, I almost feel safe to pull out that piece of paper I placed under the block to protect my graphic card from any drips... but I will leave it on for longer just in case.

~~~~ (original post) ~~~~
hey guys, just spent ALL day doingthis... man, the things u gotta do are just endless...

anyway, FINALLY got it set up - the idle temps are nice at 6 degrees C, thisis 1.63v, 2.9ghz (now stable at 1.5v)... SADLY HOWEVER, at full load this still reaches 28-29 degreees...

the reason behind it is that the pelt is a 172w pelt rated at 24.6v, and currently runing off a 12v molex means the maximum power dissipation is somewhere around 90w... which if you know peltiers - it gets suddenly worse and crap...

soi'll be orderinga meanwell s320 24V psu some time, doesn't cost much i guess... hopefully keepload temps 15 or less... we'll see...

but nonetheless, i know it's nothing special, but just powering on and seeing 3 degrees was quite a fright, haah... never before...
~~~~ (original post) ~~~~

EDIT: please read below for progress and pictures...

EDIT2: Tabulated and plotted the results obtained with peltier running at 12.08V (measured with multimeter), which equates to a Qmax dissipation of approximately 90watts.

Last edited by groovetek; 19-09-2005 at 01:44 PM.
groovetek is offline   Reply With Quote
Old 17-09-2005, 11:04 PM   #2
T_M
Senior Member
T_M's PC Specs
 
Join Date: Dec 07 2004
Location: Jakarta
Posts: 5,335
sweet as man, let us now how it progresses.
wanna see 3GHz+ and some decent volts
T_M is offline   Reply With Quote
Old 18-09-2005, 12:31 AM   #3
groovetek
Senior Member
groovetek's PC Specs
 
Join Date: Aug 13 2004
Location: Melbourne
Posts: 1,307
hehehe... it's clear that the peltier was being overloaded when i ran 1.66v, 3ghz... idle was around 10 deg, but full load shot to 25, then to 30 within 10 seconds, and up to 35 eventually, which it crapped out in stressCPU...

so, once i get the 24V to double it's heat dissipation it should be good... and it'd increase the deltaT as well... although of course the water would get alot hotter...

anyway, just to confirm the explanation, i tried stock 3200+ speed and voltage, and idle reads 255 degC, ahaha... i'm assuming sub-zero, since it displayed 0 degrees, but then went to 255, ahha...

full load was 9 degrees - that's HEAPS lower than 35...

but, god i almost shat my pants when i looked into my case and the copper plate had so many droplets of water on it... lol... i hadn't really done any condensation prevention on the plate yet - had only cut out neoprene for my mobo and used dielectric grease on mobo... haha so i spent the last hour cutting neoprene for the block/plate, and also have layered RTV sealant around surface of CPU core...

will get back to it tomorrow... should be good... i also lapped the base of the copper plate as best as i could (when i get back to uni though... i'll take the plate to the mech eng departments' lapping machines to give it a REAL lap job, ahha...)
groovetek is offline   Reply With Quote
Old 18-09-2005, 01:39 AM   #4
T_M
Senior Member
T_M's PC Specs
 
Join Date: Dec 07 2004
Location: Jakarta
Posts: 5,335
nice.
can you photo the setup, before and after condensation proofing?
T_M is offline   Reply With Quote
Old 18-09-2005, 08:13 AM   #5
groovetek
Senior Member
groovetek's PC Specs
 
Join Date: Aug 13 2004
Location: Melbourne
Posts: 1,307
yeah... far out... i woke up this morning realising that so long as any part of the copper plate is exposed, it WILL condense water... this is quite a problem indeed...

even last night with cutting up bits of neoprene, it's not quite covering the whole thing, but last night in my dreams (lol... jk... ALTHOUGH IN MY DREAM, I did somehow just slap on this awesomely cut side panel with 2 windows and neons attached to it, lol) i figured out that the best way would probably be to take it apart, and wrap the plate in neoprene except the tiny bit that touches the core, and the square on the other side where the peltier is pressed against. unfortunately, part of the problem is that where the wires of the peltier are it's really difficult to insulate - i might have to go nuts with the rtv sealant arond there to try and seal up any gaps...

edit: ok the last thing i have to decide is whether or not to buy the 24v psu... basically here's the thing - there is a 226w @ 15.2v peltier i can get, but it's 50x50mm, and at 12v that'd be ideal for efficiency and Qmax (it'd be similar to my current pelt running at 24V)...

however if i DO get the PSU, it means i'm sticking with the current peltier (which is rated at 24.6v, the 24V PSU can be adjustd +/- 10%...).

the 40x40mm is better in some ways in that the base of my block, has a ridge section cos of fricken socket A ppl (lol... give up already, PLEASE )... meaning the actual width of the contact area is 40mm exactly. The lenght is not a problem, about 55mm. The problem with the limited width is that if i used a 50mm pelt, 10mm of the pelt would not be in contact with the block's base, unless i found a piece of metal perfectly sized to fill that ridge bit, which is unlikely.

ALTERNATIVELY, i could sand down the block so much that the ridge part is totally flat with the base, haha... meaning i'll have a nice flat area to work with...

logic tells me i should go with the 24V PSU though and stick with the 40x40mm (not to mention more cooling power is more concentrated into smaller area.)

btw, does anybody know how to calculate wattage of a64 90nm chips? i know stock is like 45w or something, i'm hoping 3ghz @ ~1.6V doesn't exceed 100W, cos if it does, then even the pelt running at full pelt (hehehe, go the pun!) would not dissipate all the heat.

sigh... if only my old venice was still around - i'd be doing 3.15 at least on 1.5v... imagine with the IHS off...

STILL...
i'm pretty happy with my current chip, in that i know it is tolerates up to 35-36 degC at 3.0ghz, so, at least even if the peltier can't dissipate 100% of the heat, it can easily keep it stable enough for 24/7 use if need be... (for safety reasons though, I'll probably only plug the molex into the 24V during benchmarking, to minimize condensation... or heck - even modify a molex plug for a 5V output)...

T_M, what do you think of this idea:
* Buy one of these : http://pccasegear.com.au/prod1933.htm
* For one of the plugs, swap the 5v and 12v
* Remove one of the molexes, and replace it with one from the 24V PSU
* Have the peltier's power cable leading out the back of the case, so I can choose which one to plug into depending on what I'm using it for...

(I know someday this is going to lead to my computers death if I don't insulate my components and copper plate properly... i can just see it now - shutting down the computer after a tiring as day of benching and forgetting to switch it over... next day sister uses my comp all day with the pelt going at 172w, and mostly chip idling - GG )

Last edited by groovetek; 18-09-2005 at 08:35 AM.
groovetek is offline   Reply With Quote
Old 18-09-2005, 08:19 AM   #6
eva2000
Administrator
eva2000's PC Specs
 
Join Date: Jul 22 2004
Location: Brisbane, Australia
Posts: 23,008
twitter.com/i4memorycom Facebook Page livestream.com/i4memorycom
Blog Entries: 42
hmmm where's the pics
eva2000 is offline   Reply With Quote
Old 18-09-2005, 08:39 AM   #7
groovetek
Senior Member
groovetek's PC Specs
 
Join Date: Aug 13 2004
Location: Melbourne
Posts: 1,307
By eva2000
hmmm where's the pics
I DID take some pics... but sort of lazy to put em up.. but I might as well - can't do anything while the RTV all over my chip is drying. (of course i left the core exposed).
groovetek is offline   Reply With Quote
Old 18-09-2005, 09:01 AM   #8
groovetek
Senior Member
groovetek's PC Specs
 
Join Date: Aug 13 2004
Location: Melbourne
Posts: 1,307
ok... here we go:
Hehehe, everything disassembled! Ready to begin:

Argh.... gotta drill holes into plate (mind you, thermaltake's block has made it damn easy to put a pelt onto it, the mounting plate has lots of holes for any purpose):

Cut out some neoprene to insulate rear of mobo:

Of course, gotta use some dieletric grease over the components before slapping the neoprene and bracket on. And, I cut out neoprene for the other side of the mobo too, to fit nicely around socket:

Umm... didn't take many pics after the above ones, but took these last night after i saw droplets of water all over the plate (only happened when CPU speed/volts was lowered) - TIME FOR CONDENSATION PREVENTION!:

Sort of getting there - but I've got a better idea now (as talked about in few posts above). I'll be taking off the plate and insulating that individually. Also, as you can see, I've put a nice layer of RTV sealant across my CPU - letting it dry now... it's been only 8 hours... and check out the mess on my table from all the bits I've been cutting (most of it is from trimming though)... GEE hope i don't run out of rubber (only bought 1 small piece each of 3mm and 6mm... from clark rubber - nice and cheap ):


Well that's it so far... occupied me for at least 4-5 hours yesterday... but basically, once I set it up nicely - the move to a 24V PSU wont' require any work at all...

Oh, yeah, imma go take a pic of my case now - I've recut tubing and stuff so that I can mount my rad on the top of the case instead of onto the back (dont' want warm air from inside passing through my rad, nor do I want hot exhaust air from rad entering my case, so the only REAL solution is to move it completely (lol, I bought some long threaded bolts, just haven't been bothered using it yet - haha last night I was using the shenqiwei RC car remote control (bottom left) to hold up the rad to give the fan some room to suck air):

Last edited by groovetek; 18-09-2005 at 09:04 AM.
groovetek is offline   Reply With Quote
Reply

Bookmarks

Tags
172w, 8090w, groovetek, peltier, progress, thread, watercooled


Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools


Similar Threads for: groovetek's 172w (currently 80-90w) watercooled peltier progress thread!
Thread Thread Starter Forum Replies Last Post
A better "Handle" :-D mag3 Asus Intel motherboards / CPU 6 05-02-2010 04:17 AM
**i4memory SuperPI Mod v1.4 Rankings** T_M General Benchmarks 1006 07-08-2006 01:54 PM


All times are GMT +11. The time now is 08:51 PM.

no new posts