Hi @PrMinisterGR 33°C is something like 8°C water delta T which I would say is perfect for your loop setup and how much this loop pulls from wall, I expend something around that figure too and usually I target that figure too on most loops More likely you are not flow rate limited but you air starved which explains water temperature drop after removal side panels Second pump helps with flow rate for sure, I done tests with my current loop, in my loop every pump is running at maximum speed and noise is almost nonexistent, with pumps running at lowest possible speed flow rate will drop to 190-200LPH ,running multiple pumps will allow you to run pumps at lower speeds with same or higher flow rate, I would expect you will at the end have flow rate at 0.7-0.8GPM,don't think so you will see double of your current flow rate unless there is other restrictions like 90 degree fittings etc Regarding the Lian Li O11 Dynamic XL front panel it should helps with bringing fresh air to your case which is only positive, good fans helps with that, personally I use Thermalright TY-147B or TY-143 fans on my loop, they're my favourite fans Its up to you where you put second D5 pump, in my loop I have in pedestal where are other dual D5 pumps, depending on loop layout you will should be able to find space where to put pump, when you will be filling the loop, just disconnect 2nd pump till whole loop is filled and then turn in on that pump, nothing more Hope this helps Thanks, Jura
Hi there 1400MHz I can do only in benchmarks, in gaming +1295MHz is max which my GPUs can do stable in games or rendering +1400MHz on VRAM in my view is awesome there and looks really like you have good bin Hope this helps Thanks, Jura
1400 is insane! I just set it to the "official" spec for this memory, which I believe is 21,5GHz (10550), but don't quote me on that one. I'm still waiting for the stuff to arrive. Thanks for the advice and confirming that there's nothing really insane going on. If the @Cave Waverider 's pictures are an indication for my loop, I'll have 3x140mm Noctuas blowing air in, which should be more than enough. It's not yet open panels, but basically one open panel with air coming in actively, so in my head this seems to be even better than the open panels. We'll see. About the pump, I'm thinking of installing it just before the GTS360 and the CPU block. If I'm careful and run it after the loop is filled I won't damage it, and I think I might be able to mount it on the rear fan to do this. In my head, it's the place it will help most, since it's the middle of the loop. I am using six 90 degree fittings, I hope I can eliminate at least three of them, and possibly even five, I think it will be more that enough by then. I'll also take a look at the GPU backplate once more. Thanks again everyone for the patience and help
You cant run it stable for many hours. Not on 1400. Lottery. For example. You can see it sometimes on videos you downloaded. There you can see it perfectly. Games are more tolerant about overclocking. But videos not. Mark my words. Do not believe, that your system is stable over hours. I am AMD style. And i know about things. Congrats by the way that you got those kind of graphic cards. But i am happy with my 6800XT. No offense against NVidia users here. No offense at all. Why videos to test. Because i say ISSO. Do it. Use VLC player. Use the video on repeat mode, catch a line 1 minute repeat it until you see a bluescreen on VLC. Or use UUPdump for example, that wants power. Exactly on the little things you use you gonna SEE, that your overclocking fails. Bluescreen fault page or protected page. That is memory. Because you did overclock.
I mean, it's not a secret that the GDDR6X memory on them is rated for 21Gbit, and it depends on thermals. It's 95C for 21Gbit, and 105C for 19Gbit, so NVIDIA spec'd the cards for 19Gbit. Check it out.
Short update: I redid the whole thing regrading tubing & added the second pump. I managed to get rid of two 90 degree adapters. I mounted the second pump on top of the rear fan. I can now have both of them at 60% and get the same flow as before, but without the noise. With both at 100% I do get 1GPM, but for now 0.5 is fine, and I have zero issues. I also got the courage to tighten up the backplate, and now I have 74C load temps on the VRAM. I am still waiting for the front panel so I can just install 3x NF-A14s and (finally) call it a day.
Offtopic question, guys : do you think it's worth the upgrade from I7 9700k @ 5,1 GHz to i9 11900k ( let's say, OC to 5,1 Ghz ) ? I have the RTX 3090, someone suggested that my CPU is bottleneck the GPU. I play mostly MSFS 2020, DCS World and e few other flight / car / sims Thank you
If you play games at 4K then there is no worry about CPU bottleneck, and I wouldn't upgrade an 8 cores CPU to another 8 cores CPU anyways (though the 11900K has 16 threads) Best option is wait for 12900K which is coming soon.
On Neowin they're implying next year. https://www.neowin.net/news/intels-...-until-next-year-according-to-leaked-roadmap/
Short update from my commute, I'll try to give more details later. In the end, I had to kind of side mount the NF-A14s in the front with the ColdZero plate, as the SR2 on the side is way too thick. They have helped a lot, and I also turned the side fans to throw air in, as the case really is starved. Right now I'm in positive air pressure. The bottom SR2 has fans pulling out, the side SR2 has fans pushing in, and the top GTS360 has fans pushing out. The extra D5 is on the back, mounted on the intake fan, and has enabled me to run the whole thing with both of them at 60%, which makes them virtually inaudible. Temp wise now, things are much better. I'm basically getting open case temps. Water never goes 10-11c above ambient, and the max I've seen the fans spin is around 70% with the profile I have in the bios. That's not inaudible, but it's on the worst case scenario, as the summer here is now quite warm and humid. Far Cry 5 at 4k120 has been the hardest one to cope with, and it was pushing 60C+ GPU core before the new fans. Right now it's at 49-50C for the core, low 60s for the hot spot and low 70s for the memory hotspot. That's with a 390W limit on the card, and a +800 mem overclock. The card now constantly stays around 2GHz during game play. The CPU has a custom PBO profile, as I found it to give better performance than the motherboard limits, despite them being stupendously higher. I have posted what I did and my settings in the Ryzen OC thread for reference. The CPU is now 11C lower than before, and the two cores that the game stresses the most, hit around 75-79C, it was a constant 89C before, where before is the same bios settings without the intake fans. Even in that profile the CPU is still at a constant 120W load, and considering how tiny the sports are, I'm not sure that a lot more can be done physics wise. Keep in mind that the GPU is also dropping another 400W of heat in the system at the same time. All and all, I think I'm done with this for now, thanks for the help and the good words everyone
Been so long since I last imported I completely forgot you had to manually add which sensors you want to monitor!
I came across this thread which mentions a newer version of HWinfo that can report a new GPU hotspot temp. https://www.overclock.net/threads/3090-gpu-hotspot-temperature-30c-delta.1777947/ Many sensors on the GPU are averaged to produce the GPU temp but the hottest one of these is now exposed. This temp is supposed to be 11 to 12C hotter than GPU temp, worst case up to 22C. My temp difference was 23C so I decided to fix it. Its around £60 to replace all the thermal pads with decent ones so thought I'd have a go keeping the old pads and do another little mod to help bring memory temps down. So I took apart the card, only exposing the GPU side pads. The GPU had a small area without much or no paste, this explains the large hotspot delta. Paste used is Thermagic ZF-Extreme, heated with hot water which didnt help one bit lol. What a pita but I worked out a good technique: Drag the paste very slowly over the GPU with the flat of the spatula, pushing a little hard to get a smooth thinnish coating. Use a stanley knife blade to recover the wasted paste from the spatula and tackle the next part of the GPU with it. I forgot to take a picture of the finished result, soz. After fixing the paste, my hotspot delta dropped to 10C, better than others reported! Perfect. Then onto improving memory temps. I found my old 1080ti cooler in my box of bits, the GPU cooler section is ideal for attaching to the rear cooling plate of the 3090. 1.5mm Gelid Ultimate thermal pads and zip ties served well attaching it to the cooling plate, nice and sturdy. With a slow 230mmfan blowing down the gfx card from the motherboard front, this took 6C off memory temps during long benchmark runs. I'm quite pleased with that for the minimal cost of £12 for the Gelid pads and no change to the card apart from paste. It should keep my memory temp around 90C when gaming, it previously maxed at 94C with ambient 21C. Ambient is a bit warmer, I'm a lot happier running my card on hot days now. I have no doubt the 1080ti cooler will do a lot better with a direct fan but enough has been done. I may find a 140mm fan to rest against it. Thus ends my anecdote
Yeah the GPU temp and Hotspot delta is 12C on my watercooled 3090 too, sound like 10-12C delta is good indication of thermal paste application. I have a couple tubes of Thermalright TF-X and Thermagic ZF-EX, for these viscous pastes I use the 5 dots (1 center and 4 corners) method, heat it up with a hair dryer and mount the cooler, never went wrong LOL. Did your GPU temp drop any degree? You didn't mention about it
Thanks for the paste tip, if only I had a hair drier I didnt want to risk parts of the die not getting any paste like before so put myself through it to get a good result. The method used is actually quite economical, very little waste and not much needed. Regarding GPU temps etc: I messed up the figures using different Afterburner profiles by accident. I need to do better testing cos its 4am here lol. I'll have a go tomorrow. For now it appears to work better than initially reported. Memory temp may be a little better, I'll know more tomorrow. Core temp is for sure lower but I need to find where I wrote down my results to be precise. Previously this card would not overclock even slightly unless power limit was raised a lot, that was tested with ambient 21C. Now it does this with current ambient 26C: With default Power limit =100% (maxes at 121% [450W]) it ran Cyberpunk @ +140MHz on the clock with no issues. Core clocks at default were around 1820MHz, now they well exceed 1900MHz and temps are very good, cant remember what atm, I'll verify tomorrow. It registered 366W and is always power limited. Previously to get well over 1900MHz I needed to boost the power limit a lot. It would play Cyberpunk just short of 2000MHz but I dont know what fan speed that was with. I'm yet to test maximum overclock. More to come later.
It appears I didnt log anything sensible prior to the mods, probably cos the card had no manual overclocking value. But I remember a few things clearly so can compare those. Before the mods, with room ambient 22C it used to idle over 41C with fan on auto. At GPU=41C ish the fan would come on and bring it below 41C, rinse/repeat. Now with room ambient 25C, 3C higher, it idles around 39C, the fans dont come on at all. Thats at least 5C improvement at idle. Before the mods, with room ambient 22C I couldnt max the cards core clock and beat 2000MHz with Cyberpunk, it ran around 1950 to 1980MHz. With ambient 25C, now it sustains 2025MHz at GPU=68C with fans 100%, pushing 440W. The fans are surprisingly quiet, nothing like previous cards with max fan. I swear it is cooler too but have no evidence of my old temp sadly. HWinfo results are included to show max memory temp (at max fan, power target and overclock) and the GPU core hotspot difference. Max memory temp is lower with 100% power limit and auto fans.
