Pics? Seems radiator or pump related. Could be an air bubble that goes to the pump and then is forced out. That sounds very electrical under just the right conditions. Easy to check just using your computer on its side changing the orientation of the pump relative to the rad
3 Likes
Oh my gosh! I am a bit star struck!!! @wendell, Thank you for your response, I did rebuild my loop and made a couple other modifications/optimizations. The issue has been seemingly resolved as per my most recent threads on this post. I just wanted to say that you and the channel are such a great resource and that watching your videos has taught me so much, and has helped inspire me to build my own āgwork stationā, as well as document the process. Thanks to L1 I have found the joy in building my own custom systems and hope to build more, as budget and use cases allow, lol. Just really want to say that I enjoy your videos and the content that you share has been very educational. The forums and community around them have been nothing but supportive, inciteful, and welcoming.
2 Likes
Yes, the cyberspace janitor walk among us here.
1 Like
Hello, everyone I just wanted to post an update. I just really want to say I couldnāt be happier with this PC build. after all said and done this is what I ended up with. still not complete 'next up is a graphics card currently still rocking the igpu of the ryzen 7900x.
-
caseā¦ Sliger Cerberus x(grey with a custom painted top)
-
motherboardā¦ asus b650e gaming wifi
-
processor ryzen 7900x
-
32 memory gskill expo 6000 30 38 38 96 timings with expo enabled
-
5 terabytes storage 2 terabyte Samsung 980 pro boot drive, 2 terabyte western digital project drive and a 2 terabyte crucial SATA for a media dump
-
cooling custom loop with 2 rads, a hardware labs 240mm and hardware labs 120mm both powered by Noctua NF-A12x25 PWM fans also a single Noctua NF-A12x25 PWM for case intake and two slim Noctua NF-A12x15 for exhaust for res I have a ek flat 120mm flat res d5 pump combo.
building was super fun, and teardown and building it again with the modification I made I think are well worth it. I believe I have learned a ton not only with building the machine but with tuning the machine as well, after months of tweaking reading forum posts and then getting tired of it and putting the system into a state of optimized default I feel that I have it tuned to a point that is really fast, stable and quiet.
** current System tunings**
-
loadline calibrations everything set to high with very fast vrm switching.
-
Precision Boost: per core: with core 0 and 1: -32 offset core 2: -29, core 3, 4 :-28, core 5: -26, core 6: -29,core 7: -25, core 8-12:-29
-
PPT 195 Watts
-
TDC 144 Amps
-
EDC 255 Amps
after mucking around and so much research and sometimes hair raising experimentation I came up with these tunings. with these tunings I can run a stable prime95 small FFTs with Avx 512 enabled at a cool average temp of 85c with an average all core clock of 5ghz or around. which allows me to run a fan curve that can keep the machine practically silent. Through some experimentation the optimal pump speed is 55 - to 60 percent, anything less and it looks like the loop is running to slow for any the fans to remove heat from the loop and I and I lose performance( from 40 to 55 pump speed I have seen about 200mhz difference in performance effective core clocks when letting the CPU hit TJmax during prime 95 heavy load) Rad fan curves are set according to water temp. I have found with the loop currently cooling only the CPU I found that full heat saturation within the loop with everything going full bore sits at about 35c water temp. so I have the rad fans sitting at 40% for anything that is under 26c, and maxing out at 100 % if loop goes to 50c. currently even running prime95 small FFTs with with avx512 at 12 min per iteration, at the current tunings my loop saturates at about 33c and fan speed tops out 65%, still giving me a very respectable average active clocks, give or take 5ghz with effective clocks running very close to active clocks, I can safely say that I am not clock stretching, but I think that is due to my loadline calibration settings which are almost extreme, I might in the future play with those a bit more and see if I can lessen the settings and maintain my current performance and stability, case intake and exhaust fans set to custom mix sensor. That is set to the average temp of the mainboard and VRM temps. those set to 40% fan speed and if average between VRMs and mainboard ever reach 70c those will kick in to about 75% but even under heavy load I havenāt seen my VRMās get up to 60c and my mainboard at 26c. so system running cool and quietā¦ I have benchmarked cinebench 24 and interestingly my cpu gets hotter than even prime 95, small FFTs I have seen the processor top out at 90.1c during a cinebench multicore render I get my score at current tunings sits at 1675 for cinebench 24 multi and Ill have to run single core render again, and some additional benchmarks and stress testing. I did try and do a manual OC with the Ryzen and temps got scary hot. I think with a 1.248 core VID I hit a maximum of 106c during prime 95 small FFTs. but it never hit that temp for long average temps were only 101 percent of max during that run, but never again. That was not running cool and quiet. I also tried to use the Asus Auto OC switching but could never got to work with extreme workloads. I could run a 55.75 on ccd0 and a 53.75 on CCD1 with a 1.34 Core Vid and run Cinebench where temps topped at 90c at the highest with active clocks and effective clocks running Cinebench all core at 5.4 ghz, but I could never get it to switch properly with prime95 small fft with avx 512 where it was supposed to switch from a manual OC to default PBO. I just couldnāt get it to run correctly. So I just concentrated on dialing in pbo for nowā¦ and for the time being have given up on my manual overclocking exploits on AM%, and concentrated on PBO modifications and tuning, am happy with the work I didā¦ leave comments and discussions welcome.