• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Putting together a water cooled Ryzen 5950X Radeon RX 6900 XT system

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
If you want something demanding in loop then try Final Fantasy XV benchmark. Can run it in a window and it's free. Usually crashes fast if there is any stability problem (the same for RAM and other components). Somehow it feels more demanding for the whole PC than 3DMark.

I ran the benchmark after closing out a session of the Division 2 and got 9824 from max settings at 1440P borderless while iCue and all the other regular apps that are open are running to replicate the configuration I was using for the Division 2. The temps for the coolant got up to 40C and the RX 6900 XT reached 50C and the CPU reached 62C. However while I was playing the division 2 (3 hour gaming session) the temps for respectively where 47C, 78C a 76C at peak. Nothing throttled and while playing the Division 2 the GPU stays pegged at 2673MHz (setting is for min 2600 max 2700) and ~285W. I'm thinking maybe this is simply a limitation of two 280mm and one 140mm radiator and only the 30mm of thickness that the XR5 radiators have and the glass on the front and top of the Corsair 680X case. For real world gaming use this seems sufficient when the room ambient temperature is rather warm from nine other graphics cards in the room mining away.
 

Attachments

  • FFXVBoderlessUltraWhileTrackingTempertures.PNG
    FFXVBoderlessUltraWhileTrackingTempertures.PNG
    930.5 KB · Views: 153
I modified the fan curves for the two fans on the motherboard in the bios and that made a difference in the fluid temperature. Now the fluid is holding at around 43C of ~4C. The graphics card and the CPU are still getting up to the same temperature ranges. I'm hoping this means during the summer when it is a little warmer then it is now that the 4C drop in the fluid now will give me some headroom to maintain these same speeds. The dips in the temperature near the end of the graphs is from switching out of the game and opening the snipping tool to take a screen shot. The temp sensor is the sensor that the XD5 came with measuring the temperature of the fluid from the XD5.
u0obVVN.png.jpg
 
That's great to know they make custom parts for better airflow. I've been putting off putting the glass back on top of mine - glad there are alternatives available.

System looks great! Any chance of getting a picture of it in action ;)? Looked through the thread a few times and I don't see one :(
 
Sure. This was the lighting when the machine first turned on. IÂ’m not one for leaving it like this but the colors seemed so vibrant so I took a picture of it.
F478B11B-1D07-4380-A473-25664E3D1DDE.jpeg

It seems that even the high end power supplies come with RGB now so I went ahead and set it to white.
89E3A706-D273-4BBE-B1C5-6586AAC062C6.jpeg

This is what it looks like as of now with all the RGB set to static white.
1C73B028-90D3-4F1A-9D61-F004FDC3804B.jpeg

In case youÂ’re wondering about the machine with the red fans in the last photo. That is the Core i9-10900K system.

My take away so far. The AMD system is a pain to boot up at times. Although now I have it down to a science. Ryzen Master won’t start because a “driver isn’t installed”. The machine has quirks and not in a good way. But for all of its pain points it’s crazy fast and it is not having issues with running any of my actual work programs like visual studio, Crestron, AMX, creative cloud. It nips at the heals of the 10900K in games on the cpu side of things with very little effort. I’m mostly happy with it. Although if the Gigabyte RX 6900 XT Xtreme Waterforce wasn’t so expensive I’d buy a second one for the 10900K and put them against each other in a game off. Not that it matters which is better. Just the fact that AMD has done so well with Ryzen is amazing.

The Gigabyte Xtreme Waterforce is also amazing. It gets well past 2600MHz in the division 2 and is capable of 1440P 165FPS to match my monitors capability. ItÂ’s impressive.
 
You know your audience ;) Way cool dude! I love the colors in the first picture!

I've always been timid about attempting to do a full water cooled system, but seeing this thread has enticed me.

Oh and to reiterate what EarthDog said several posts back - iCue is really bad. Don't trust it for anything.

I really like Corsair RGB products though so it's worth living with the poor software if that's your thing.

Thanks for the pictures :)
 
YouÂ’re welcome. As software goes I like iCue far better then ThermaltakeÂ’s software. You can set hardware profiles so the lights donÂ’t reset on reboot. Older Thermaltake fans simply reset to the rainbow if the software closed. So every reboot it was a rainbow until I signed in. Also while it is more wires I like that Corsair fans can plug into the motherboard for control or into a Commander pro or the controller for the new Corsair AIO products. I never liked that the thermaltake fans was never directly controlled by a sensor. I did notice that it refers to the rx 6900 XT junction sensor as the gpu sensor which got me thinking it would have been nice to have it report both sensors. But it explains why the GPU reports where off.

I have done a hard tube and a soft tube water cooled system and soft tube was a lot easier.
 
This morning I ran into a wall on the memory speed. I cannot get past 2158 in Cyberpunk, The Division 2 or 3DMark. Just like Jay in Jays2Cents I'm not seeing much in the way of gains in the memory. That said between my RX 6900 XT Xtreme and my Power Color RX 6800 XT Red Devil the memory on my 6900 XT Xtreme is a grand total of 52MHz faster. That is probably 100% a result of water cooling. The GPU speed on the other hand goes from a min/max of 2420/2520 to 2680/2780. That is a 260MHz gain. I am not certain I am at the ceiling yet. That has a power draw of about 310W at 1200mV. That is a 9.06% GPU clock speed gain plus the extra TMUs and shaders. The fill rate is astounding though compared to my Asus RX 5700 XT Strix that I was playing with just this last winter at the start of the GPU shortage. Not to mention the fact that both the Red Devil and the Strix cards make a lot of noise with the fans keeping the cards from bursting into flames at those speeds.
 

Attachments

  • Division27864GPU.PNG
    Division27864GPU.PNG
    267.9 KB · Views: 122
  • RX 5700XT GPUZ Compared to 1080TI.png
    RX 5700XT GPUZ Compared to 1080TI.png
    444.2 KB · Views: 121
Right now my temps are not bad in real world applications and it is already very silent but I would like to get the GPU junction and the CPU below 60C with the same workload.
X0hkeJ3.png.jpg

This might sound insane but I'm contemplating removing the LL120 intake fans and putting in a push pull 280mm in a intake position. That would give me 980mm of radiator surface and hopefully not reduce the airflow entering the case. I'm thinking this will help reduce the temperature of the GPU and CPU when both are under load. I'm worried that I would need a second pump. I have seen some people put a Corsair XD5 RGB in the rear chamber of the Corsair 680X case. I could try that if needed or I could get a Corsair XD3 RGB and mount it on the rear exhaust 140mm fan as shown in Corsair's stock photo.

-CX-9040003-WW-Gallery-XD3-RGB-21.png
 
I reserved the parts from my local Microcenter. I have some work to do tomorrow. :D I hope this improves my thermals.
 
I don't think you're short radiator, but it will likely help... just not by much.

The XD3 should help with the flow, which you where concerned with previously. I know the XD3 doesn't have as much power in the pump as the XD5 but there simply isn't room in the chamber for it.

Do you think it would do better to convert the system into a dual loop and if so which component would be ok with the XD3, the 140mm and one 280mm radiator?
 
I ended up buying just the XD3 and two fittings. The temperature dropped by a few degrees. However some of the tubes are kinking and I donÂ’t have a clean way to install them as is. I ordered a few more adapters and have a plan to adjust the tubes to eliminate the kinks tomorrow. IÂ’m hoping with them eliminated that it should bring the temperatures down a few more degrees. Then when I get the replacement panels to replace the front and tomorrow that has the air flow style it should bring the temps down a little more. The overall gain should hopefully be a reduced temperature of about 10C for the GPU and CPU. IÂ’m happy with that.

I also have another two temperature probes on order to add into the loop. This way I can see the temperature of the fluid at each pump and after the cpu and gpu.

I do not think I will bother with the third 280mm radiator.

4B9C4223-09E1-4456-8012-13EF2CC6DEB5.jpeg
 
I also have another two temperature probes on order to add into the loop. This way I can see the temperature of the fluid at each pump and after the cpu and gpu.
check the stickys....you won't see a but couple of C difference at any point in a properly radded and flowing loop. The water isn't in the blocks or rads long enough to see increases or decreases more than that. You've got data overload with all the dashboards and such.:)

Don't you have a probe or two already? This conversation feels like deja vu. Lol
 
ItÂ’s possible that the XD5 wasnÂ’t enough to handle the three radiators and two blocks and the XD3 added to the loop increased the flow to where it needed to be. Another nice gain is I no longer need to run the pump at 100% to keep the temperatures in check. There is now a point where increasing the speed of the pumps doesnÂ’t lower the temperature further. IÂ’m happy to adjust the curves of the pumps to keep it as low as possible for max effect when the fluid temperature has increased. That will translate to less noise from the pump.

I should have the loop drained, finalized and leak tested completed late this evening.
 
Here is my best run for Firestrike Extreme before I modified the loop.

3DMark29316ScreenShot1.PNG
3DMark29316ScreenShot2.PNG

Here is the result from tonight after updating the loop. The speeds of the CPU and GPU have not been adjusted.
3DMark29441ScreenShot1.png
3DMark29441ScreenShot2.png

So this was purely the hardware able to function better because of the improved thermal management. The CPU peaked at 72.63C and the GPU peaked at 54C in this test where during my previous best run the CPU peaked at 76.63C and the GPU peaked at 59C. So that is a improvement of 4C for the CPU and 5C for the GPU.
 
Happy to see a temp improvement!! Im more curious to see what the temps were after the loop was saturated/normalized/equilibrium and see what happens temps-wise. Likely a similar story, but at least the test was the same. :thup:

Here is the result from tonight after updating the loop....So this was purely the hardware able to function better because of the improved thermal management.
120 points of almost 30k overall and less than 100 out of 31k (graphics) is within run variance (less than 1/2 of a percent).
 
The best overall score for the machine so far is 21905 in Timespy. https://www.3dmark.com/spy/23463724

I modified the loop a bit. I added in a drain valve but I failed to think it through. Its too high to drain easily. I'll drain it this weekend and move it somewhere lower.

uHoWEcQ.png.jpg

8Pw5iSN.png.jpg

For gaming performance here is Horizon Zero Dawn 1440P max detail.
iR9A75D.png.jpg
 
Back