5950X Temperature Question

[DarkeVortex]

Hey all,

 

Just finished my first watercooling build, and I have 3x 360 radiators. Got everything hooked up, leak tested, booting to windows - all good! Then I noticed something odd to me - The CPU (5950X) has an idle temp around 50-60C (fans low +  50% D5 pump => ~60C, fans low + D5@100% => 50C). This is my first AMD build, and my first time using a 16c/32t processor - are these temps too high?
 

I searched around some, and got mixed results. Really want to avoid having to drain + disassemble the computer + re-seat the monoblock if possible, so figured I'd post here to try and avoid some trouble!

 

Motherboard is a Crosshair VIII dark hero, and the BIOS is where I was reading the reported temperatures. Seemed to match pretty well with HWInfo64.

 

If you need any more info from my end, please let me know.

[Guest]

Well the transitory rich processor would definitely fair well on a good water loop if the heart is made of all copper instead of a simple cold plate. 

 

But with a transistor rich cpu, doing single core boosting during idle and low usage times, the v-core increase to accommodate the higher single core frequencies.

 

Also, since water cooling still relies on ambient air temp in the long run, knowing the room temp might help..

 

Also if XMP is enabled, this will automatically increase the CPU SOC voltage which also adds additional heat, is it overclocked or you running bone stock defaults.

 

These are some of the variables that would have some people with higher idle temps than others, but in the end, the max load temp is what matters the most.

[YoungBlade]

AMD does idle higher than Intel. It seems to be partly a reporting difference; the AMD CPU temp numbers are not apples-to-apples comparable with Intel numbers.

 

What is your temp when under full load?

[DarkeVortex]

This is still bone-stock. No OC, ram still running at 2133.

 

If this seems ok, I’ll get the ram up to where it should be and then get some load temps. Any suggestions on what test to use / for how long? This is my first time on Ryzen and whole loop water cooling, so not super familiar yet. First to figure out how to get the RAM to run at 3600…

[YoungBlade]

5 minutes ago, DarkeVortex said:

This is still bone-stock. No OC, ram still running at 2133.

 

If this seems ok, I’ll get the ram up to where it should be and then get some load temps. Any suggestions on what test to use / for how long? This is my first time on Ryzen and whole loop water cooling, so not super familiar yet. First to figure out how to get the RAM to run at 3600…

Assuming your water cooling is working properly, it will take a while for the radiator to be fully saturated where the water inside the loop is the same temperature as the CPU. Air coolers can be saturated in under a minute. Water coolers, typically not, and definitely not for large ones.

 

You can use either Ryzen Master or HWInfo64 to get your temps in Windows.

 

The main test I do for load temps is Cinebench R23. You can download it here and let it loop for the multi-core test. By default, it does 10 minutes. I'm not sure whether that's long enough for your setup, but you can set it to run longer, or just run it again.

 

I don't know that this would work for a liquid cooled setup, because it might not take long enough, but I do also run a Windows Defender custom scan of one of my NVMe drives. This is a different type of workload that also stresses the RAM and drive - so the IO die is getting more of a workout. Also, because it doesn't involve a lot of AVX instructions, the CPU is able to boost higher. I see about 200MHz higher on all cores when doing such a virus scan compared to running Cinebench. It does actually use all the cores at once for the majority of the scan on my 5900X, although maybe the 5950X has too much power for that. The higher clocks can result in higher temps. However, it's also a real-world test that has the load level fluctuate as it goes and that makes it useful for testing stability for OC/undervolting.

[DarkeVortex]

Haven't done any stress testing yet, but here's an update:

 

Still at stock. My fluid is at ~55C, as is my GPU, while playing Trackmania. The CPU load is around 6-7% (according to iCUE) and the CPU package temp is bouncing between 75 and 80C... feels definitely not normal, unless I'm that out of touch with Ryzen temps. Hoping someone can tell me I'm wrong so I can avoid a rebuild :P

[freeagent]

I don't have a 5950 but my 5900 is sitting at 27.5 right now.. my 5600 runs even cooler. It is 22c in my space. Just using air cooling with fans at "standard" curve.

[DarkeVortex]

22 minutes ago, freeagent said:

I don't have a 5950 but my 5900 is sitting at 27.5 right now.. my 5600 runs even cooler. It is 22c in my space. Just using air cooling with fans at "standard" curve.

Oof. guess I gotta rebuild.

[Kinda Bottlenecked]

21 minutes ago, DarkeVortex said:

Oof. guess I gotta rebuild.

 

my 5900x idles at 50-60*c and load temps at 81*C with a dual tower air cooler on stock auto OC. 

[Guest]

8 minutes ago, Kinda Bottlenecked said:

 

my 5900x idles at 50-60*c and load temps at 81*C with a dual tower air cooler on stock auto OC. 

High temp gradient. No big deal, cooler is doing it's job at load! 

[Kinda Bottlenecked]

1 minute ago, ShrimpBrime said:

High temp gradient. No big deal, cooler is doing it's job at load! 

I'm not surprised, this is how zen 3 behaves 

[freeagent]

15 hours ago, YoungBlade said:

AMD does idle higher than Intel.

My 5900X runs like my 3770K

 

pretty easy to cool.. but still going to get warm no matter what you do

[YoungBlade]

20 minutes ago, freeagent said:

My 5900X runs like my 3770K

 

pretty easy to cool.. but still going to get warm no matter what you do

What number are you using? In HWInfo64, CPU (Tctl/Tdie) is consistently 20C over ambient at idle, but the individual CCD (Tdie)s are just 10C over ambient. CPU Die (average) is 15C over ambient.

 

EDIT: I decided to check "true" idle temps for the CPU - as in nothing running except Windows and HWInfo64, closing all other applications. Before, I was looking at temps with Firefox running, which may have kept some of the cores active - who knows what the browser is doing in the background.

 

I let the computer sit until temps seemed to stop dropping, then pressed the reset button on HWInfo64, and waited 3 minutes. Then, I proceeded to record the numbers on some pressed plant fibers via a dark liquid on a stylus so as not to disturb the computer (like how the Ancient Egyptians recorded their data - it's crude, but effective). Here's what I got:

 

CPU (Tctl/Tdie) - Min: 33.4C Max: 34.8C Avg: 34.0C

CPU Die (Average) - Min: 29.1C Max: 31.3C Avg: 29.6C

CCD1 (Tdie) - Min: 26.0C Max: 42.0C Avg: 27.5C

CCD2 (Tdie) - Min: 28.5C Max: 38.3C Avg: 30.8C

Core Temperatures - Min: 24.0C Max: 30.2C Avg: 26.2C

 

It got rather cold here last night - 57F (14C) - and I opened up the windows. Obviously, it's not that cold in here, but it did get down to 65F (18C). So Tdie is 16C over ambient. CCD (TDie)s are about 11C over ambient. And Die (Average) is 12C over ambient. So the CCD (TDie)s stayed the same, while Tdie and Die (Average) dropped a few degrees by closing Firefox.

 

With Intel, I regularly saw average temps in the low 20C range when it got that cold in here. But I don't think this means that Ryzen runs hot. I think it means that AMD reports differently from Intel.

[DarkeVortex]

1 hour ago, YoungBlade said:

What number are you using? In HWInfo64, CPU (Tctl/Tdie) is consistently 20C over ambient at idle, but the individual CCD (Tdie)s are just 10C over ambient. CPU Die (average) is 15C over ambient.

 

EDIT: I decided to check "true" idle temps for the CPU - as in nothing running except Windows and HWInfo64, closing all other applications. Before, I was looking at temps with Firefox running, which may have kept some of the cores active - who knows what the browser is doing in the background.

 

I let the computer sit until temps seemed to stop dropping, then pressed the reset button on HWInfo64, and waited 3 minutes. Then, I proceeded to record the numbers on some pressed plant fibers via a dark liquid on a stylus so as not to disturb the computer (like how the Ancient Egyptians recorded their data - it's crude, but effective). Here's what I got:

 

CPU (Tctl/Tdie) - Min: 33.4C Max: 34.8C Avg: 34.0C

CPU Die (Average) - Min: 29.1C Max: 31.3C Avg: 29.6C

CCD1 (Tdie) - Min: 26.0C Max: 42.0C Avg: 27.5C

CCD2 (Tdie) - Min: 28.5C Max: 38.3C Avg: 30.8C

Core Temperatures - Min: 24.0C Max: 30.2C Avg: 26.2C

 

It got rather cold here last night - 57F (14C) - and I opened up the windows. Obviously, it's not that cold in here, but it did get down to 65F (18C). So Tdie is 16C over ambient. CCD (TDie)s are about 11C over ambient. And Die (Average) is 12C over ambient. So the CCD (TDie)s stayed the same, while Tdie and Die (Average) dropped a few degrees by closing Firefox.

 

With Intel, I regularly saw average temps in the low 20C range when it got that cold in here. But I don't think this means that Ryzen runs hot. I think it means that AMD reports differently from Intel.

That's an interesting test. I might have to do the same - I definitely am concerned given that I'm seeing 80C at a 3% load, but then again, I don't know if 100% load won't move the temperature much past 80C... (This is the package temperature read off of iCue, i'll be more thorough with HWInfo64 in a few moments here).

I typically have a few things running when I was talking about "idle" temps - GoXLR App, Discord, Synapse, iCUE are the main ones. On my 9700K, idle temps under a 360AIO were maybe 5-10 over ambient, certainly not 40...

 

Any recommendations here? Should I run your true idle test, and also run a load test to see what happens? If so, what numbers would be acceptable or indicate bad contact with the waterblock? I'm going to update my BIOS first, too - saw somewhere that that can sometimes fix random problems, so probably just a good idea overall.

[YoungBlade]

1 minute ago, DarkeVortex said:

That's an interesting test. I might have to do the same - I definitely am concerned given that I'm seeing 80C at a 3% load, but then again, I don't know if 100% load won't move the temperature much past 80C... (This is the package temperature read off of iCue, i'll be more thorough with HWInfo64 in a few moments here).

I typically have a few things running when I was talking about "idle" temps - GoXLR App, Discord, Synapse are the main ones. On my 9700K, idle temps under a 360AIO were maybe 5-10 over ambient, certainly not 40...

 

Any recommendations here? Should I run your true idle test, and also run a load test to see what happens? If so, what numbers would be acceptable or indicate bad contact with the waterblock? I'm going to update my BIOS first, too - saw somewhere that that can sometimes fix random problems, so probably just a good idea overall.

Yeah... that's a lot for low load. I left HWInfo64 running and while browsing these forums and watching a couple of YouTube videos for the last hour or so, my average Tdie only rose to 41.8C - a 7C increase. The max temp in that time was 60.4C, probably when watching a video. So the CPU touched 40C above ambient, but only briefly. In general, it's more like 20C over ambient. That's still more than I used to see for web browsing with Intel, but nothing like what you're seeing.

 

I think there may be an issue with the cooling.

[DarkeVortex]

1 minute ago, YoungBlade said:

Yeah... that's a lot for low load. I left HWInfo64 running and while browsing these forums and watching a couple of YouTube videos for the last hour or so, my average Tdie only rose to 41.8C - a 7C increase. The max temp in that time was 60.4C, probably when watching a video. So the CPU touched 40C above ambient, but only briefly. In general, it's more like 20C over ambient. That's still more than I used to see for web browsing with Intel, but nothing like what you're seeing.

 

I think there may be an issue with the cooling.

Just looked at the BIOS, and the CPU is pulling 1.47V - in the BIOS, so literally nothing is running. Is that abnormal?

[YoungBlade]

4 minutes ago, DarkeVortex said:

Just looked at the BIOS, and the CPU is pulling 1.47V - in the BIOS, so literally nothing is running. Is that abnormal?

That's normal for stock AMD. Actually, the CPU pulls higher voltage the lower the load is. To hit those higher frequencies, it needs more voltage, but when all cores are active, the frequency is lower, and so the voltage drops closer to 1.1V. My max voltage reported is 1.487V and the average is 1.443V - because I've just been browsing the web, so usually only one or two cores are active, and those cores can push to 4.95GHz, which requires more voltage.

[Kinda Bottlenecked]

5 minutes ago, DarkeVortex said:

That's an interesting test. I might have to do the same - I definitely am concerned given that I'm seeing 80C at a 3% load, but then again, I don't know if 100% load won't move the temperature much past 80C... (This is the package temperature read off of iCue, i'll be more thorough with HWInfo64 in a few moments here).

I typically have a few things running when I was talking about "idle" temps - GoXLR App, Discord, Synapse, iCUE are the main ones. On my 9700K, idle temps under a 360AIO were maybe 5-10 over ambient, certainly not 40...

 

Any recommendations here? Should I run your true idle test, and also run a load test to see what happens? If so, what numbers would be acceptable or indicate bad contact with the waterblock? I'm going to update my BIOS first, too - saw somewhere that that can sometimes fix random problems, so probably just a good idea overall.

 

Something seems iffy here. I too have chrome tabs, discord and multiple 3d apps open during "idle" but it doesn't break 65*C on air cooling. What are your load temps like? 

 

When you rebuild the loop, you might want to try giving the thermal paste a good spread over the ihs. 

[DarkeVortex]

Just now, Kinda Bottlenecked said:

 

Something seems iffy here. I too have chrome tabs, discord and multiple 3d apps open during "idle" but it doesn't break 65*C on air cooling. What are your load temps like? 

 

When you rebuild the loop, you might want to try giving the thermal paste a good spread over the ihs. 

Yeah, Im thinking the thermal paste application was sub par, or I could have tightened the block down a bit harder. First time putting one on, so I didn't want to overdo it; probably erred on the other extreme. Either way, looks like today is a drain & rebuild day... Oh well

[DarkeVortex]

3 minutes ago, YoungBlade said:

That's normal for stock AMD. Actually, the CPU pulls higher voltage the lower the load is. To hit those higher frequencies, it needs more voltage, but when all cores are active, the frequency is lower, and so the voltage drops closer to 1.1V. My max voltage reported is 1.487V and the average is 1.443V - because I've just been browsing the web, so usually only one or two cores are active, and those cores can push to 4.95GHz, which requires more voltage.

Gotcha, that makes sense. Going to let the BIOS update finish and see if anything is different, but its unlikely. Are you on air or liquid? If you're on liquid do you know the delta between your CPU temps and your fluid temps? I know it can obviously fluctuate when the CPU is loaded, but seeing the liquid + gpu (loaded) at 56C and the CPU at 85C (no load) just seems like too big a difference for good contact, unless AMD reports really differently

[Kinda Bottlenecked]

Just now, DarkeVortex said:

Yeah, Im thinking the thermal paste application was sub par, or I could have tightened the block down a bit harder. First time putting one on, so I didn't want to overdo it; probably erred on the other extreme. Either way, looks like today is a drain & rebuild day... Oh well

 

Hey, you'll at least have twice the experience after 

[YoungBlade]

Just now, DarkeVortex said:

Gotcha, that makes sense. Going to let the BIOS update finish and see if anything is different, but its unlikely. Are you on air or liquid? If you're on liquid do you know the delta between your CPU temps and your fluid temps? I know it can obviously fluctuate when the CPU is loaded, but seeing the liquid + gpu (loaded) at 56C and the CPU at 85C (no load) just seems like too big a difference for good contact, unless AMD reports really differently

I'm using air cooling: Scythe Fuma 2 and my case is the Phantex P300A with all 4 fans spots populated. I've never used liquid myself - only watched some videos about it.

[DarkeVortex]

Just now, Kinda Bottlenecked said:

 

Hey, you'll at least have twice the experience after 

True Honestly might just wait til my Amazon package shows up today... Mr Bezos is supposed to be bringing me another ball valve that I like the look of more, so may as well take the opportunity to put that in.

Now I'm a little worried about the GPU block as well, maybe I should run a test on that before tearing everything down to see if it's at least behaving ok

[DarkeVortex]

Just now, YoungBlade said:

I'm using air cooling: Scythe Fuma 2 and my case is the Phantex P300A with all 4 fans spots populated. I've never used liquid myself - only watched some videos about it.

Gotcha. Thanks! Regardless, I think the signs are pointing to bad thermal transfer between the CPU + waterblock. The temps are just way too different. The GPU and water temp seem to match much better, which makes more sense in my mind.

[Kinda Bottlenecked]

Just now, DarkeVortex said:

True Honestly might just wait til my Amazon package shows up today... Mr Bezos is supposed to be bringing me another ball valve that I like the look of more, so may as well take the opportunity to put that in.

Now I'm a little worried about the GPU block as well, maybe I should run a test on that before tearing everything down to see if it's at least behaving ok

 

Yeap I would've done the same, if temps are alright on the gpu block right now I wouldn't worry about it.