I am in the middle of a dilemma, whether to upgrade my Custom WC or not…
Since I upgraded my system, I am experiencing higher temperatures while gaming, and a bit when idle/mild work.
Old system: X570, Ryzen 5900X, Radeon 6900XT
New system: X670E, Ryzen 7950X3D, Radeon 6900XT
Cooling: EK-Velocity CPU Block, EK-Quantum Vector RX 6800XT/6900XT Nickel + Acetal, HeatKiller tube + D5 12v pump, AlphaCool 3x120mmx48mm radiator with 3x120mm EK-Vardar PWM fans at full speed. Tube > D5 > Rad > CPU > GPU > flow indicator / fans are pushing air to rad and then out of the box
Temps
Old:
Furmark Stress test 30min: GPU: 65C, CPU 57C
BF2042 online CQ 128p: GPU: 67C, CPU 59C
Idle/Browsing: GPU 34C, CPU 35C
New:
Furmark Stress test 30min: GPU: 73C, CPU Tctl 69C, Tccd1 62C, Tccd2 51C
BF2042 online CQ 128p: GPU: 73C, CPU Tctl 89C, Tccd1 82C, Tccd2 61C
Idle/Browsing: GPU 34C, CPU Tctl 51C, Tccd1 49C, Tccd2 53C
Keep in mind that I am running in Linux and Windows are inside a VM, therefore Windows cannot access CCD2, as it is isolated from the host.
So, my questions are:
- Are these temps acceptable with a room temp <28C ?
- If I change only the 3 fans with Phanteks T30, will there be any difference?
- If I replace radiator and fans with HEATKILLER RAD 480-L (or EK-Quantum Surface X480M) & 4xT30s would be ok?
- Should I add the above rad to the circuit between CPU & GPU?
- Should I create a second loop, one for CPU, one for GPU? (I have most of the parts, except a radiator, also case has lots of space)
1 Like
Ryzen 7000 is really hot because they made the IHS dumbass thick to maintain cooler compat, which means higher temps
You can delid, or sand down the IHS a bit to improve thermals
You can also apply a negative offset to curve optimizer and a negative offset to frequency override in PBO settings
1 Like
Those negative curve offsets are awesome if the CPU takes it. So much less heat.
1 Like
I had forgotten about that. However, I did some tests today, and didn’t make much difference.
CPU stress test: Tctl 90C, Tccd1 84C, Tccd2 72C
1 Like
What’s wrong with that?
AM5 likes to push the temps and stays at 95°C. As long as you don’t see any throttling in performance, I wouldn’t worry. 12th gen Intel and AM5 treat temperatures differently compared to previous generations.
My 7900x@170W only hits 95°C in things like y-cruncher and even then my results are totally normal performance without throttling. Most other all-core loads cap out in the low-mid 80s with my Noctua D-15 and 1-6 core loads boost to the same amounts because of the higher clocks.
1 Like
The other day, while playing BF2042 shutdown completely. It was the 2nd time, so it wasn’t random.
1 Like
Whoops thought I was in the lounge
1 Like
Maybe just add an additional 120mm radiator after the CPU and before it goes to the GPU to help cool off the temps a little bit? That way the elevated CPU temps dont affect the GPU so much.
1 Like
This is one of the options I am considering, just with a 360 instead
1 Like
Before spending money, cause your temps are not that bad. You appear to have more than enough potential cooling.
Are you running Latest Bios?
What is the operating Temp of your liquid? Always over looked
Case Temp? Are you just getting oversaturation on the board.
Have you reseated your CPU Block, anyways. We all can goof this.
Have thought of using a Contact Frame?
1 Like
Id ignore the CPU temp its going to be hotter regardless of cooling capacity.
So just going on GPU temps your coolant is probably 6-8c higher under load. your fans are already decent so If you want the temp lowered you will need more radiator capacity. Based my experience I don’t think 1 extra 120 will make much difference. Id add an extra 360 if its possible in your case. Also loop order does not really mater very much so just plumb it in whatever way is less of a pain.
edit: Also if you are having stability problems first thing id blame is ddr5 anything over 5800-6000 always seems to have problems.
2 Likes
Yes, the latest non-beta
I don’t have a way to measure it, but touching the fittings, the one that comes out of the radiator is cool, while the ones coming out of CPU/GPU are barely warm.
Room temp is always below 29C. I don’t know which sensor is giving me case temp, but I have one temp, labeled “Sensor 2” that is almost always 78C, and the rest are between 29C and 40C
Yes, multiple time and checked the paste (MX-5)
I did, few days ago, TR Secure Frame, didn’t change anything
I ordered one from EK, 360x44. Thinking of installing it between CPU and GPU, as I have a case opening exactly above them. At least it will lower the coolant temp before it gets into GPU block
Stability issue, probably gone once I disabled Curve Optimizer. My RAM is rated 6400 but I am not using EMP profile, so it runs at 4800.
1 Like
For Reference my liquid runs at 27C idle and up to 45C stress. Stating your liquid maybe cool come out of the RAD, would tell me the RAD is doing its job. Though that’s hard to truly judge without sensors. Now, the 78C potential Case temperature concerns me, is that at idle? Do you have a Motherboard Sensor? Have you checked you cases pressure? IE… positive, Negative Neutral. Possible you may be Very positive, and not enough system exhaust. Throwing in another RAD may or may not fix it, it also could just be a band aid on the real problem.
Just interesting information for you. People speak of NZXTs CAM software being bad, and showing wrong Temps. CAM was 8c higher then ASRock Monitor. I installed Hardware Monitor Plus on daughters’ system. I found out that the CAM software was reporting E-Cores Temp, not the Highest or P-cores. My point is quality information is key. This is also where I found that 2 P-cores were running 12 degrees hotter than the other P-cores. This is when I decided to put a contact Frame on. The frame was a game changer. The IHS on latest Gen’s seem to really suck. I plan on Lapping the CPU one day.
Excuses me for typo’s… Had Shoulder surgery last week, using one hand.
2 Likes
This is not case temp, this is the highest reading I am getting but I don’t know which sensor is it, cause label says just “Sensor 2”.
I am using Linux, so I don’t have access to any Windows Software. I am running Psensor app to get my temps. It is accurate, but not fancy as most Windows apps.
Thank you for taking time to reply, and get well soon!
1 Like
Just thought I might chime in as i have a semi similar build. I have 7900x3d and 6950xt both watercooled on a custom loop. Dual 360mm rads, d5 pump, ek gpu block and phanteks glacier cpu block.
My ambient temps are warm (27C plus). I run dual HW labs GTS (35mm thick) 360 rads and my coolant under load tops out at 40C rad intake and 37C exit temps for my coolant.
This is up significantly from my ryzen 3900xt build with the same gpu.
I remember from reading these chips really try to hit the best clocks possible and can live at 95C forever without degradation. Well 89c for x3d models I belive. So for people who dont have as strong cooling solutions they hit the thermal limit before they hit the current limit. In our case (with a custom loop) the chip is able to pump out and utilize full power which inturn means more heat…much more.
My rule of thumb has always been a 120mm solution for each part then add a 120mm rad for OC. NOW with the mich higher wattage limit on newer chips that means more heat, so now I just go with 360mm percomponent.
Under stress tests I can get up to 85C on CPU, GPU can get up to 65C. My case temps get up to 35C, and my rads are all exhaust. With my case I have matching intake fans, so 6 x 120mm fans as intake. This was before I did curve optimizer… I got down to 80C max CPU. and 60C gpu.
With such high ambient temps though this is exspected for me. I think a second rad will really help…BUT dont exspect as cool of temps as any previous generation chips.
Just my experiance. Also… try looking at actual temps on cpu, cores are often much lower at idle but higher IO die temps make cpu seem much higher at idle because they measure a average or highest temp. My cores drop to 30c at idle light load, but IO die hovers at 40c.
Hope it helps give a frame of refrence.
With my older 8086k and 2080ti, i used to max at 32c coolant temps with the same rad setup…cpu maxed at 60c and gpu at 54c. Theres just so much more power in newer companents which equals more heat. Just IMHO.
Thank you for your reply. I already ordered a 2nd 360x44mm radiator and planning to install it next weekend. We will see how it goes.
CU made my system highly unstable… I used -20 for cores 0-7, and I was having constant resets even when loading Windows desktop.
1 Like
I got -15 for ccd0(with cache), then -25 for ccd1 (without cache), the ccd without cache seems to be ok with a more agressive curve optimizer setting…
Side note I had huge stability issues with mine due to expo… I had to manually set volatges and now no issues at all. Im also running 64gb on two dimms… I did a whole post on that. Seems my ASUS x670e is prone to this problem and all the bios updates. Think I had to do one a week for while. Feels kinda like early adoption tax lol. So wont get off subject.
Also, hope you heal up, and good luck. Look forward to hearing how it goes 
I haven’t enable EXPO yet, I go step-by-step. For now I am trying to optimize CPU performance and power consumption.
I didn’t want to upgrade to such a new platform on the first place, but I had to…
1 Like
A good friend of mine, who is expert in Intel watercooling, told me that certain blocks may need different orientation when installed on AM5 socket.
Anybody has some info/ling for my EK-Quantum Velocity?
I think some blocks have different mounting hardware going from am4 to am5 but other than that the orientation should be the same between the two. Presumably you already have the Mounting hardware if it was different.
In the past there sometimes used to be different jetplates you could swap out for different sockets but I don’t really think that’s a thing anymore. (It was probably mostly for going between HEDT and 115X).
