[uniwarking]

Hi all. I've completed a recent build and wanted to share the results and also ask for some feedback on the setup and cooling configuration. Thanks to all who have provided help in getting to this point!

System Specs:

• Corsair iCue 4000x Mid Tower Case
• i7 12700k (still stock, plan to overclock)
• Corsair H150i Elite LCD Cooler
• EVGA RTX 3090 XC3 Ultra Hybrid (I plan to overclock)
• MSI Pro Z690-A DDR LGA 1700
• G.Skill Ripjaws 32GB (2x 16GB) DDR4 3600 CL16 (running X.M.P)
• Samsung SSD 980 Pro 1TB NVMe M.2 SSD
• EVGA SuperNOVA 1000 T2
• 2x WD Black 5TB 7200 RPM Drives
• Creative Sound Blaster Z

Overall, I'm super happy with how the build has come out. The 4000x case looks good, in my opinion anyway, but it was a tight fit with the two radiators and all the other content (it doesn't look near as sexy behind the back panel). I'm also not sure if I'm getting the best cooling performance.

Here is how I have the cooling configuration set up currently:

As I've noted in another thread, the 3090 cooling is decent except the memory temp. Even if I pull off the tempered glass front and ramp up all the fans to 100% I still see 3090 Mem temps hitting 82c in Time Spy. At reasonable fan curves I'm seeing 88c and worse yet I'm seeing it hit 95c after extended gaming sessions in Halo Infinite. The GPU stays pretty cool itself in the mid 50c range. Other games don't seem to push it this far. I have a mild overclock of +84 core and +200 memory with 105% power target.

As for the CPU temps, in general it seems to be running ok depending on application. My last PC had a 6700k with AIO and temps didn't jump around like this 12700k does. For general gaming/benching, CPU temps max out around below 60k which seems ok. If I run Prime95 Blended tests for 15+ min I see an average temp of around 60 as well but peaks up to as high as 99c (package and some cores). The coolant temp has hit 42c. This even happens if I put all the fans/pumps at 100%. I've used a few different apps to look at temps: iCue software, GPUz, Real Temp, etc. I guess I would have expected better with this H150i Elite LCD cooler, although I know Prime95 is a pretty extreme application. Are my expectations too high or is there perhaps something wrong with this cooler or my setup? Back in the day, I'd have used Prime95 for 24hrs plus to verify OC stability, I'm afraid to run this thing for more than 15 minutes stock.

Bench Scores:
PCMark 10: 8917
Time Spy: 19699
Time Spy Extreme: 9803
Port Royal 13581
SUPERPOSITION 1080 Med: 31828
SUPERPOSITION 8K: 7573

Thanks!

 

[bscool]

If running Prime 95 with avx enabled yeah it is going to get hot. Never used to test so not sure on temps for P95 with AVX.

Something like Cinebench 23 should have temps in say 80c range with 12700k at defaults I would think.

I have 12900k OCed and in the 80c range depending on clocks and voltage with Arctic 420 AIO for a CB 23 run.

 

[anticommon]

3090 has memory chips on the backside of the card. Not all cards have good backplates that will help to dissipate that heat (or lack proper thermal pads) so the 12gb of memory slapped on the back gets really hot. A fan pointed there might help.

As for everything else, I think the problem is mainly that corsair put glass in front of where your fans go. tsk tsk

Beyond that, CPU's with lots of cores are going to have a lot more variable load than something like a 6700k which either stays under a more consistent load or just has a lower TDP/operating power so you don't see those swings. Watercooling will naturally provide a bit of temperature dampening at the beginning of a load due to being able to utilize the water as a heat reservoir before exchanging any of that energy to the air via the rad (heat exchanger). Once its soaked and the capacity for heat removal of the radiator is saturated the temps will continue to rise. Also, water getting that hot will result in worse cooling performance by virtue of having less temprature delta (difference) between the water passing over the block and the temperature of the block (ie heat removed from the CPU) so a spike load with water at 25c and 42c will look a lot different.

 

[uniwarking]

As for everything else, I think the problem is mainly that corsair put glass in front of where your fans go. tsk tsk

I was interested in this as well, what variables would improve CPU cooling. I just ran Cinebench R23 in a few different configurations. The long term plan was "balanced" fan curves with the glass front to have a quiet and good looking setup. I'm now thinking of getting the "airflow" front panel for this case.

Peak Temp	Score	Front Glass	Fan Curve	Peak Fan RPM
88C	22191	Glass	Balanced	1925
87C	22216	Glass	100%	1925
84C	22198	No Glass	Balanced	1700
81C	22201	No Glass	100%	1925

Something like Cinebench 23 should have temps in say 80c range with 12700k at defaults I would think.
I have 12900k OCed and in the 80c range depending on clocks and voltage with Arctic 420 AIO for a CB 23 run.

Yeah, even with my fans at 100% and front glass removed I'm hitting 81c with no OC. I really had hoped for much more out of the H150i Elite LCD at the price point it's at. It apparently has better fans than the previous H150i... I know part of the price tag is the fancy LCD though.

 

[s1rrah]

I don't own a 3090 Hybrid but I've read that most users of that card have had issues with VRAM temps; I don't think VRAM cooling is a strong point with any of the 30 series hybrid cards according to what I've read and personal discussion with owners of such. BTW ... how is your 3090 an XC3 Hybrid? Did you add the kit after purchasing the regular XC3 cooled model? Just curious...

But to your dilemma and some possible solutions...

Various 3090 Hybrid users have had great results simply by replacing the thermal pads on the back of their card between the rear VRAM areas and the back plate. If trying this, use high quality 12 w/MK pads and add some sort of fan (120mm?) pointed across the back plate and be sure to use the right height thermal pad for your card to ensure proper contact between the card and backplate. If the pads for your card happen to be of a non standard height of something like 2.75mm, using 3mm is just fine if you use nicely compressible/squishy pads. In fact, any given thermal pads rated thermal conductivity assumes a compressed scenario anyway so I would avoid the harder variety of pads like the Fujipolys for instance. For pad height measurements, you could easily order a replacement set of pads from EVGA and simply measure them to be sure before buying/cutting higher quality pads. This would be a very easy modification and you could run the extra backplate fan off of either a manual fan controller or control it via PWM/software so that it only runs when you want it too. I do the exact same sort of fan "trick" with my highly OC'd system RAM and it works like a magic trick.

Otherwise, replacing the full thermal pad set on the front VRAM chips can also help a great deal. Just stick with reputable and highly compressible pads from the likes of Thermalright, Gelid or the rarely discussed but well reviewed, Frosty Tech pads, all three of which one can easily find on Amazon. There is also a relative new-comer to the GPU thermal pad scene running an online service called Kritical Pads; he sells high quality, precisely cut, full kits for all of the EVGA 30 series cards and is adding new kits for other cards all the time (Strix, FE, etc.). He rates his custom pad sets at 20 W/mk but I wouldn't pay that much attention to such claims. But they do work good and several users have ranked them better than Gelid and Alphacool, the latter options being those you would have to cut yourself. If you can't find a set of Kritical pads for your card, email/message the owner; he is very open to custom orders and generally responds quite quickly. I'm using one of his custom pad sets for my EVGA 3080 Ti XC3 and VRAM temps are fantastic, generally staying below 76C during hours long gaming sessions and with a fairly healthy memory overclock. Others, especially miner types, have also had very good results with these pads but as with all such haxoring, YMMV. In my own controlled tests, general daily use temps were only slightly better with the Kritical pads but significantly better in mining tests. I don't do mining, BTW ... but it is certainly the best test for VRAM thermal pad performance and you only have to run a 15 minute test or so.

And I'd also highly recommend replacing the stock EVGA thermal putty that is used over the two rows of VRM's on the EVGA 30 series GPUS (you will know the putty when you see it LOL). For thermal putty replacement on my EVGA 3080 Ti XC3, I used the very well known and established 10 w/mk T-Global TG-PP-10 and it has worked fantastic for my own card. It's about $25 and can be found on DigiKey if you care to try it. Unlike the EVGA putty, it does not migrate at all and remains in a "playdo" like state, even after months of usage and can even be (carefully) reshaped/re-used if ever tearing your card apart again. Some users even use the T-Global putty as direct replacement for thermal pads on the GPU VRAM and have found it to be better than pads due to the fact it compresses to precisely fill surface gaps. On the VRM caps, the T-Global putty is even more ideal as it nicely compresses down into the space between the closely mounted VRMS ... this latter fact with putty in general most likely being why EVGA and other vendors chose to use it over that "wierd" area of the 30 series GPU PCB.

Finally, and if you want to go full mod-tard, you can get one of these (admittedly) kind of neat looking all copper heatsink arrangements off of ebay and combined with the simple fan mod described above along with pad replacement on back of the card? I think such would no doubt help quite a bit in reducing overall VRAM temps and GPU temps in general. I would personally go with the most expensive option (6mm @ $50) as it has the tallest heatsink pins.

~s1rrah

 

[JSHamlet234]

It must get pretty hot inside that case with the CPU+GPU both dumping their heat into it.

 

[uniwarking]

BTW ... how is your 3090 an XC3 Hybrid? Did you add the kit after purchasing the regular XC3 cooled model? Just curious...
~s1rrah

Here is the info on my card: EVGA GeForce RTX 3090 XC3 ULTRA HYBRID GAMING, 24G-P5-3978-KR, 24GB GDDR6X, ARGB LED, Metal Backplate. This was purchased straight from EVGA

Thanks for the tips on the potential mods, I've already chatted with Kritical pads. They don't make a kit for my card but it's very similar to the FTW3 Hybrid, just need to trim up a piece or two from the sound of it.

It must get pretty hot inside that case with the CPU+GPU both dumping their heat into it.

Yeah, there is plenty of warmth coming out of the rear exhaust... as I guess would be the case with any 600W+ space heater! I thought about having air exit the top through the GPU radiator but I'm sure that would hinder cooling performance of that unit. All in all, I think this is the best config to keep the CPU and GPU at their coolest while maintaining positive case air pressure, but I'm open to suggestion (primary reason for the post).

 

[JSHamlet234]

Yeah, there is plenty of warmth coming out of the rear exhaust... as I guess would be the case with any 600W+ space heater! I thought about having air exit the top through the GPU radiator but I'm sure that would hinder cooling performance of that unit. All in all, I think this is the best config to keep the CPU and GPU at their coolest while maintaining positive case air pressure, but I'm open to suggestion (primary reason for the post).

It's a good config for your CPU and GPU core temps, that's true, but it's not good for anything else - VRAM temps, VRM temps, SSD temps, HDD temps, etc. For the best component life overall, you'd want definitely want to reverse the GPU fans, and maybe also the CPU fans and replace that front glass with mesh. The fun of this is testing it yourself. Run HWINFO64 and see how your temps are of every component I mentioned during a mixed CPU/GPU load situation. Then reverse the GPU fans and see what happens. Then reverse the CPU and rear fans and see what happens. Then see what happens with and without the front glass. Then you'll have actual data with which to make these decisions.