If you can hear a manic laughter wherever you are on Earth, that'll probably be me right now. Idle temps are low single digit C. Waiting for GPU driver to update then I can start benching
The chiller is set to 2C and hasn't gone into idle at all, so this is probably on its limit. Since turning on the PC it actually went up to showing 3C. My house is getting warmer, so I've now got windows open. Keep in mind the chiller by itself is pushing ~500W when running...
Isn't the way the condensation is lit up by RGB beautiful?
Initial results:
Y-cruncher - previous watercooling got me to 4.3 needing 1.225v for stability. With chiller, I already forgot what voltage I needed to get 4.4 stable-enough, but 4.5 wasn't having any of it and CPU temps were already into the 80's at that point. I didn't undo some limits in bios at this point, so it might be worth a revisit.
Cinebench R15 - previous watercooling got me to 4.9 at 1.35v. With some bios tweaks, I was able to run 5.0 at 1.40v. 5.1 wasn't happening. Temps again were well into the 80's.
I was able to improve times on the above two but 100 MHz isn't much. I also had a quick go at Cinebench R11.5 but wasn't able to beat my previous sub at 4.9.
Because liquid metal solidifies at low temps, I had swapped to Kryonaut under and over the IHS. I was generous in application, but I always have concern about if it might have pumped out especially as I had to slide the IHS back into position when fitting it into socket, so could some have pumped out? Min-max core delta under load was ball park 13C.
So I could re-apply Kryonaut in case I had pump out. Another alternative is to go back to liquid metal, but limit the chiller to ~11C or so to prevent LM freezing. It'll be an interesting tradeoff... cooler liquid temps but worse TIM, or warmer liquid better TIM.
As for the chiller, set to 2C target temp, it never reached it and turned off. It was close, but not quite making it, especially once I turned on the PC, then floating around 3 to 4C. It might be more energy efficient of me to regulate to 6C, so I lose a few degrees but I'm not draining 500W in the background and dumping it into my room continuously... if I went for the warmer water + LM route, this would obviously not be a problem.
That's awesome Mack! Just keep an eye on the condensation, it can cause real problems. You're not likely going to get it on the bottom of the mobo behind the CPU since you're not that cold but I use Vaseline there. Some pipe insulation goes a long way as well for the hosing.
That's awesome Mack! Just keep an eye on the condensation, it can cause real problems. You're not likely going to get it on the bottom of the mobo behind the CPU since you're not that cold but I use Vaseline there. Some pipe insulation goes a long way as well for the hosing.
Mack just my .02. Make sure you insulate really well when using cold water. In a way it's almost harder to insulate for the condensation when not going sub zero. In addition you may want to try putting that cooler outside in the colder months to really get some good water temps. On Johans and Scotty's suggestion I've been using windshield wiper fluid on my rad and it flows a lot better then Glycol.
I ordered some extra bits today, 1/2" - 3/8" barb adapter so I can use my 3/8" aquarium pump with 1/2" watercooling tubing. Strangely UV purple was cheaper by a big margin than others so that's what I'm getting 3m of. Also getting 4x connectors so I can add both CPU and GPU block if needed.
And elsewhere, I ordered a coolbox, to act as a better res than an aquarium
I didn't order an AM4 waterblock yet, so I'm still limited to testing on 115x and Intel HEDT platforms for now.
I'd like to get a test bench at some point to hold things down better when running inverted. I'm living in fear that I'll step on my X299 mobo and parts which is still lying inverted on the floor from last weekend... I'd have to undo the tubing to move it. Having longer tubing will help a lot there. On that note, forgot to order tube insulation.
Has anyone killed equipment from condensation? In theory it is pure water so shouldn't be conductive. At worse, dry it out and should be fine again?
I cannot say I have killed anything from condensation per se. Sometimes it's hard to tell exactly what the cause was, mine usually is a bit too much voltage. That all said I haven't killed many parts considering how many I've benched.
One thing to note though, a lot of times when moisture gets in, it isn't always easy to recognize that it's the moisture is the cause. Just to be clear I am talking strictly benching and pushing the limits. You'll just start noticing the setup will be acting strange, crashing when it wasn't previously on certain benches. Known working profiles will crash for no reason etc...
Most of the time just shutting down and letting things dry will take care of it. On 1 occasion I did lose a board that was wet, though I was pushing the voltage on it pretty hard so I'm not too sure exactly what let the magic smoke out but....
Yes... killed a board or two back in the day from condensation... which isnt pure water. Sitting on the tube sliding down over all the components before it lands on the board.
Noted about condensation. Would still be surprised if it picked up that much stuff to be enough to cause a problem, but I don't want to test it out!
Cool box arrived as well as the tube and fittings. The tubing colour looks rather horrible, so no wonder it was cheaper than the others. Will have to set up a better mount for the lot on the weekend.
All I know, Mack, is that many people have killed things from condensation. It isn't pure water as it is coming from the air already with impurities as well as touching everything before it hits the board. It doesn't take much to let the electricity travel.
This will give you an idea of what I do. That white foam has been upgraded to some higher quality closed cell foam. After it's all tightened down I wrap the top of the block with a highly absorbent rag. If air can get to it, it will condensate. Trust me it's a bigger issue than you may realize.
There is now a test version of Prime95 that supports AVX-512. The chiller system is the only one I have that supports AVX-512...
On first running default small FFT stress test, hottest core temps went over 100C with CPU at stock other than cache 50% overclocked without touching voltage. I had previously wondered if my TIM application wasn't good, so I redid that, and no change. Under this load condition, there was significant differences in core temps. Coolest cores 0, 2, 5 were ~20C lower than the hottest. If it isn't due to the cooling, maybe it is something to do with some cores being next to other hotter cores. Especially if the 6 core part I have is cut down from the I think 8 or 10 core LCC die. Maybe some cores just have idle neighbours to allow heat to spread better.
Since I wasn't running set up for near zero temps yet, I only set the chiller to 15C in the hopes of preventing condensation, which worked as intended.
As this thread is more about the chiller than AVX-512, I'll save that for another thread some time.
Mackeral, just making sure I'm on the same page. What chip are you running? This is the delidded one? I'm assuming you're speaking of the 8086k, just want to be sure. In addition, did you change the TIM under the IHS or on top, in post #16?
Did you swap out the Tim under the Ihs? That does seem to be a pretty substantial difference in temps. Though, your theory on the die and placement of the cores may have some substance to it.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
