I'm currently using a 5700 XT with a Kraken G12 bracket to hybrid cool the GPU. VRAM and VRM temps are under control, the GPU Edge Temp never hits above 65-70c BUT the Junction or Hotspot temperature seems to Skyrocket even with an undervolt.
I'm steady at a 1105 mV undervolt with +20% Power Limit and the Memory OC'd to 900 Mhz.
Core Clock is kept at 2100 Mhz max and I have a custom fan configuration with the VRAM / VRM fan running and the push / pull on the radiator set to run at 100%.
The only thing I could think of was that I wasn't making good contact on the GPU but that doesn't quite make sense if I'm still achieving strong edge temps. With the junction behaving the way it is my hopes of getting the maximum performance out of this card are pretty low.
Before I ran these numbers today I went ahead and re-seated the AIO using MX-4 thermal paste. If anyone has any idea of what I could do to improve that Junction temperature so I can finally game / benchmark my card without any worry please let me know!
You didn't mention what Junct.temp Why 110-Degree Temps Are Normal for AMD’s Radeon 5700, 5700 XT
110* is in no way normal for junction temp with a sufficient cooler. 110* happens in some situations with the reference cooler, however it’s not ideal. Sapphire RX 5700 XT Pulse Review: Thermals, Noise, & Overclocking - YouTube
See 4:00 for a chart comparing the reference 5700 xt to the sapphire pulse edition.
G12 is not suited for rx 5700, it doesnt cool all the necessary parts, if you want to use hybrid look on ebay for an accelero-hybrid-ii, which works the best (i have it too on my second RX) or hybrid III. Hybrid II is using the same backplate cooler as Extreme IV (also tested on RX 5700 XT) and the temps are really good, with your undervolt junction will stay between 50-65 degress and core temp between 50-67 (Extreme IV), i also managed to OC the core to 2100 (Extreme IV), with Hybrid the temps might be a bit lower.
The hybrid arctic cooler doesn’t look like it really does anything differently than the g12. Its been tested and proven that graphics card VRMs can be cooled just fine using a fan without heatsinks. Vram doesn’t really need to be cooled for the most part, it usually runs at acceptable temps with or without cooling.
The hybrid arctic cooler doesn’t look like it really does anything differently than the g12.
Lol its a huge difference, its like letting ice cream on sun vs in the freezer, that passive cooler on the backside is very important, if you want to keep the G12 you should by an additional cooler blowing air directly on the backside of your graphic card, at 110 junction your graphic card will start throttling due to high temperature.
The card doesn’t come with anything to cool the backside from the factory though. I’m not saying that you’re wrong, I don’t have any way to test it, but I don’t really see how a heat sink is going to cool the components through a fiberglass pcb.
I wanted to wait a few weeks before I put in an official reply updating what I've experienced so far.
I ended up fully removing the G12 and reapplying thermal paste, ensuring that it was getting maximum coverage with firm and even tension on over the GPU. I have heatsinks on the VRMs and VRAM.
While depending on the game I will still occasionally see spikes into the 105-110C range for the Junction (hotspot), the average temp does not ever get over 85C and the overall GPU temp stays around a nice 67-75C under full gaming load after several hours.
Mind you I ALSO have a 240mm rad for my CPU that blows ACROSS the motherboard, out the back where the 120mm rad is for the GPU. So not only do I have hot air going through the GPU rad BUT it's still maintaining incredible temps. Overall I'd say that the G12 looks incredible in my new case and I'd be more than happy to share if anyone is interested.
Those temps still seem rather high for water cooling. What size rad are you cooling the GPU with? Do you have a super quiet fan curve?
Edit: I didn’t see you said you have a 120mm rad, I suspect that’s probably causing the relatively high temps, especially since it’s got warmed up air going though it from the cpu rad. However I do wonder if perhaps the GPU die is slightly recessed in the chip, causing the water block to put more pressure on the edges of the chip rather than the die. That’s probably not the case though, just a thought that crossed my mind.
That is too hot and as far as i know there was a review on youtube testing the G12 and it failed for RX the cards and it even didnt finish some benchmarks, so G12 absolutely not, but of course it is your decision.
I have a 240mm rad too, but the hot air should go out out of chasis, so hot air from your CPU blowing on your motherboard, woah.
G12 might look "incredible" but your temps are "incredible" too.
Like i said Extreme IV or Hybrid II, core was never above 67, Memory temps between 50-65 degrees, Junction never went above 80 (only when i used overclock to 2100 and a higher voltage), junction stays always between 45-70, depends on usage.
"Its been tested and proven that graphics card VRMs can be cooled just fine using a fan without heatsinks" That is so not true!!! But lol, if you want to have a good locking cooler and a damaged card, fingers crossed, but please do not advice it to people, no offence, but it is a very bad decision.
You can watch any of the g12 reviews and they’ll almost all show that vrm temps are well within operating temp without a heatsink. If the vrm needed a heatsink the g12 would include one. Motherboard vrms barely need heatsinks as long as you’re not drawing crazy power from them, that’s why they usually use solid hunks of aluminum as heatsinks. With active cooling a heatsink is not needed. As far as that video by Timmy Joe where he couldn’t get a g12 to work, he showed that vram and vrm temps were well within acceptable limits. He said himself he couldn’t figure out why he was having problems. My guess is he wasn’t monitoring junction temp and for some reason it was getting hot. A simple thermal pad or stronger mounting pressure could be the solution.
I got it working! Well to be completely honest I’m still waiting for my g12 bracket in the mail, but in the meantime I ziptied my AIO to my GPU die. I’m maxing out at 41 degrees during stress testing. The reason AIOs are hit or miss with this card is that they have to have DEAD flat cold plates since the die is so small. If the cold plate has a convex/cone shape too it (all current Asetek pumps do because that shape’s supposed to make the best contact with a flat IHC) or if the mounting pressure isn’t dead even on all 4 corners the temps will skyrocket. I had to deck my coldplate, but once I did I saw my load temps drop from high 70s/80s down to low 40s.
I stand corrected, delltech was right about how important ram cooling is with GDDR6. Unfortunately GDDR6 has a much, much higher TDP than older versions of vram (at least Samsung’s does). With heatsinks directly on the chips, on the back of the card behind the chips, and with a fan blowing directly on them I’m seeing temps up to 90 degrees (even higher during certain stress tests). Granted that’s with my 5700 (non xt) overclocked to 2100 MHz and junction temps around 80-90 degrees (edge temps top out at 55 degrees). Stock my ram temps top out in the mid 70s, but the whole point of water cooling this card was for overclocking stability and noise. I’ll be ditching the AIO route and going full open loop (something I’ve always wanted to try anyway).
Sorry not overclocked to 2100mhz exactly, the target frequency is set to 2100mhz but it sits around 2050-2060mhz.
Something interesting I’ve noticed is once I overclock to 1900mhz or higher the edge temps max out at 55 degrees regardless of pump speed or fan speed. Junction temps will continue to rise according to how far I overclock it, but again once they top out fan speed and pump speed make no difference in temps.
Btw it looks like that AIO or watercooled RX 5700 xt cards do not have flickering and bsod problems, only if you do not use the 4pin fan connector on the graphic card. I did have a lot of problems with it, but after i have switched to Alphacool Eiswolf which doesnt use the fan port at all no issues....
If you plug any fans up to the card’s fan header I highly recommend that you download the morepowertool from Igor’s Lab to adjust the RPM range so the card knows the range of the aftermarket fan(s). I have my card header unplugged and confirm, the card won’t have any issues with that header open.