5800X3D hits 90C for REALLY simple things...

C

CruisD64

2[H]4U
Joined
Mar 6, 2007
Messages
2,464
Hey folks. Let me preface this by saying that I know this chip runs really hot... I've got a bit of a headscratcher, though. It might be totally normal but seems like it shouldn't be. I dunno.

Backstory. I just put together an all AMD system for fun. 5800X3D. 6900XT. 16GB 4400 DDR4. Evga 280mm AIO with Noctua A14's as exhaust. Haven't tried intake.The difference between this system and others I've built is the footprint. I went with a HYTE Revolt 3 mini itx case. It WAS in an Evolv Shift 2 behind a 120mm AIO (knew it would suck but tried anyway). Both the GPU and CPU ran extremely hot so I figured I'd get a case that supports a 280MM AIO which the revolt does and things would be fine, right? Nope.

I noticed some strange things. The CPU would hit 70C just opening chrome...or the Timespy app and other simple things that don't even dent the temps of my main rig which cools with a Kraken X73. For instance, during Timespy the CPU sits around 65C while running the graphically focused tests. Seems normal. Then it gets to the CPU test where the FPS are initally high but tank once the CPU hits 90C. FPS cut in half. Obviously due to throttling. Cinebench hits 90C immediately and never drops. GPU temps never exceed 70C so problem solved there.

I have a spare 3800x laying around so I threw that in. That chip used to run at 4.2 with 1.2v. Was always a cool chip. 62C max in an Evolv X. It did better than the 5800X3D but still not great. 80C running Cinebench and scores nearly as well as the 5800x3d (12,800 with the 3800x vs 13,500 with the X3D). I saw a reddit thread about high temps and part of me wants to say "ok this is normal behavior" but another part says "wait, just opening chrome it hits 70C?"

Currently I'm thinking the AIO isn't making great contact with the IHS. Isn't perfectly flat? Contact isn't tight enough? Seems like a reasonable theory. I've resat the block several times with the same result. I'd love to undervolt and see what happens but you can't. I also don't want to buy another AIO based off a theory. What are your thoughts/experiences?
 
Last edited:
Is the water pump working correctly? Do you have the pump set to max speed? I recently diagnosed two different computers that produced similar issues to what you describe. In one case, the AIO had a very slow leak and too much fluid evaporated so basically only the waterblock itself was cooling the chip. In the other case, someone had plugged the water pump into a case fan header that was set for voltage-based fan speed control and so the pump was not getting enough voltage.
 
Yep, sounds like an internal cooler issue or misconfiguration. If you can monitor pump speed and coolant temps, it'd give an idea of what's going on. In my AIO the cpu temp doesn't go over 80C unless it's a long load and coolant temp goes over 35C, which takes 5-10 minutes on all core PBO load (~205W steady consumption). If your pump speed and coolant temps are good, I'd expect an internal issue with the cold plate (fins blocked), mounting issue or similar.
 
GotNoRice said:
Is the water pump working correctly? Do you have the pump set to max speed? I recently diagnosed two different computers that produced similar issues to what you describe. In one case, the AIO had a very slow leak and too much fluid evaporated so basically only the waterblock itself was cooling the chip. In the other case, someone had plugged the water pump into a case fan header that was set for voltage-based fan speed control and so the pump was not getting enough voltage.
Click to expand...
I believe so. I have it connected to the "pump" header on the motherboard. I'll double check the bios after work to see if it's receiving maximum power.

bobzdar said:
Yep, sounds like an internal cooler issue or misconfiguration. If you can monitor pump speed and coolant temps, it'd give an idea of what's going on. In my AIO the cpu temp doesn't go over 80C unless it's a long load and coolant temp goes over 35C, which takes 5-10 minutes on all core PBO load (~205W steady consumption). If your pump speed and coolant temps are good, I'd expect an internal issue with the cold plate (fins blocked), mounting issue or similar.
Click to expand...
I can feel the liquid moving through it though I'm not sure at what speed. Coolant temps sit around ~30C and don't really fluctuate even under load. That seems odd. You might be right about the cold plate. The cooler is brand new as far as it's never been used but who knows how long it sat on a shelf and might have been going on inside. I will be troubleshooting some more tonight.
 
check your mount/paste and that pump is at max. you should see small burts that will go into the 60s but not the 90s.
 
CruisD64 said:
Hey folks. Let me preface this by saying that I know this chip runs really hot... I've got a bit of a headscratcher, though. It might be totally normal but seems like it shouldn't be. I dunno.

Backstory. I just put together an all AMD system for fun. 5800X3D. 6900XT. 16GB 4400 DDR4. Evga 280mm AIO with Noctua A14's as exhaust. Haven't tried intake.The difference between this system and others I've built is the footprint. I went with a HYTE Revolt 3 mini itx case. It WAS in an Evolv Shift 2 behind a 120mm AIO (knew it would suck but tried anyway). Both the GPU and CPU ran extremely hot so I figured I'd get a case that supports a 280MM AIO which the revolt does and things would be fine, right? Nope.

I noticed some strange things. The CPU would hit 70C just opening chrome...or the Timespy app and other simple things that don't even dent the temps of my main rig which cools with a Kraken X73. For instance, during Timespy the CPU sits around 65C while running the graphically focused tests. Seems normal. Then it gets to the CPU test where the FPS are initally high but tank once the CPU hits 90C. FPS cut in half. Obviously due to throttling. Cinebench hits 90C immediately and never drops. GPU temps never exceed 70C so problem solved there.

I have a spare 3800x laying around so I threw that in. That chip used to run at 4.2 with 1.2v. Was always a cool chip. 62C max in an Evolv X. It did better than the 5800X3D but still not great. 80C running Cinebench and scores nearly as well as the 5800x3d (12,800 with the 3800x vs 13,500 with the X3D). I saw a reddit thread about high temps and part of me wants to say "ok this is normal behavior" but another part says "wait, just opening chrome it hits 70C?"

Currently I'm thinking the AIO isn't making great contact with the IHS. Isn't perfectly flat? Contact isn't tight enough? Seems like a reasonable theory. I've resat the block several times with the same result. I'd love to undervolt and see what happens but you can't. I also don't want to buy another AIO based off a theory. What are your thoughts/experiences?
Click to expand...
Did you microwave it?
 
What are your voltages in HWiNFO?
You can undervolt the 5800X3D no problem. I am using PBO2 Tuner to do a -20 mV all core undervolt on login. Really helps lower temps. -30 mV is max offset for this software.
1652767422452.png


Post for latest version of CurveOptimizer with Task Scheduler support: https://www.overclock.net/threads/c...urve-optimizer-settings.1777398/post-28983886
Some good info regarding 5800X3D undervolting in the 5800X3D owners thread on overclock.net as well.
 
Take it out of the case and bench test it. That will answer your question about the aio working properly. Make sure it’s configured properly, and just bench test it.
 
bobzdar said:
If coolant temps are only hitting 30C it's definitely not a case/airflow issue. It's a heat transfer issue.
Click to expand...
I'd say if it's running a (light) stress test for more than a few minutes and stays under 30c, then there's definitely something wrong with coldplate contact. If it begins to climb, then it might just be hitting those temps at very short intervals, due to increased cache/core activity. In the latter case, reduced frequency/voltage might help, but better TIM application may also help slightly if it was done poorly.
 
Shadowed said:
What are your voltages in HWiNFO?
You can undervolt the 5800X3D no problem. I am using PBO2 Tuner to do a -20 mV all core undervolt on login. Really helps lower temps. -30 mV is max offset for this software.

Post for latest version of CurveOptimizer with Task Scheduler support: https://www.overclock.net/threads/c...urve-optimizer-settings.1777398/post-28983886
Some good info regarding 5800X3D undervolting in the 5800X3D owners thread on overclock.net as well.
Click to expand...
Prior to trying this tool it would peak at 1.3 which I think might be a hard limit. After trying the PBO2 Tuner tool (Thanks for the link!!!) it peaked at 1.18 with a -30 curve set. It took a little longer to happen but still hit 90C. I think the cooler might be bad...

bobzdar said:
If coolant temps are only hitting 30C it's definitely not a case/airflow issue. It's a heat transfer issue.
Click to expand...
Coolant temps sit around 32C at idle and under load hit about 37C after 2 passes of Cinebench R23 back to back.

Nobu said:
I'd say if it's running a (light) stress test for more than a few minutes and stays under 30c, then there's definitely something wrong with coldplate contact. If it begins to climb, then it might just be hitting those temps at very short intervals, due to increased cache/core activity. In the latter case, reduced frequency/voltage might help, but better TIM application may also help slightly if it was done poorly.
Click to expand...
This is what I'm starting to think. As stated above I tried the PBO2 Tuner tool. It make a difference in that it took a bit longer to hit 90C. So I'm starting to think either a defective cooler or IHS on the CPU. Leaning towards the first.
 
jmilcher said:
Take it out of the case and bench test it. That will answer your question about the aio working properly. Make sure it’s configured properly, and just bench test it.
Click to expand...
I did end up doing this with the same result.
 
CruisD64 said:
I did end up doing this with the same result.
Click to expand...

Then it’s either a faulty aio or a very uneven surface on either the aio or the cpu. You can set the edge of a new quality razor blade on each surface and see if there’s any major gaps. If not, and you are sure the aio settings are correct, then you have a faulty aio.
 
did you make sure the pump is running and at full speed(DC 100%)?
you could take the block off and show us a pic of the paste (on the cpu and block).
 
pendragon1 said:
did you make sure the pump is running and at full speed(DC 100%)?
you could take the block off and show us a pic of the paste (on the cpu and block).
Click to expand...
I did make sure of that but I'll check again. I set max power in bios for the pump header and also the evga clc app. Actually, IIRC the paste did not look evenly distributed the last time I pulled the block off. I'll take another look.

I picked up a Kraken Z63 during my lunch break and have it on standby but will double check everything before swapping it out.
 
Would you have a look at this... It appears to be convex in the picture and the razer blade test confirms it. It seesaws over the middle. I haven't put the new cooler on but this has to be it. I'll post back after everything is back together.

PXL_20220518_020001860.jpg
 
CruisD64 said:
Would you have a look at this... It appears to be convex in the picture and the razer blade test confirms it. It seesaws over the middle. I haven't put the new cooler on but this has to be it. I'll post back after everything is back together.

View attachment 474698
Click to expand...
That'd be fine if there were enough pressure to get good contact on the parts that matter. The problem is if part of the heatspreader over the die isn't getting good contact, then your fxcked. That looks pretty obsenely convex, so it may well be the problem.
 
Nobu said:
That'd be fine if there were enough pressure to get good contact on the parts that matter. The problem is if part of the heatspreader over the die isn't getting good contact, then your fxcked. That looks pretty obsenely convex, so it may well be the problem.
Click to expand...
The new cooler definitely helped. At 1.18v it hit 83.4C max during a cinebench run. Ambient of 27.2C. This is with ONLY the cooler changed so the old cooler definitely played a factor. 83c still seems high though and higher than other peoples results. I still need to play with the fans on the cooler as they're currently on a splitter running off a single motherboard header which I'm not sure can fully power two fans. I'm going to try running them full speed off a sata splitter next and see what happens.
PXL_20220518_031900685.jpg
 
Adjusting fan curve took it down another 3 degrees to 80.4 at 1.2V. I think I might be ok now. Just really hoping MSI releases an updated bios that let's me undervolt permanently.

Thanks for all the help everyone. Now to RMA this darned cooler...

EDIT: 3DMark CPU test max temp was 74C with the new cooler.
 
Last edited:
I am seeing similar results with my 5800X3D - https://hardforum.com/threads/5800x3d-gaming-review.2018891/page-6#post-1045363830 - I see that your temps improved - but I also read that HWiNFO may not read the CPU correctly. For example, from my post - AIDA64 reports 79 C when HWiNFO says it's 10 C higher during stress testing. It's a NZXT Kraken X63 with upgraded fans (Noctua NF-A14 industrialPPC-3000 PWM) and Kryonaut. Coolant is equally "cool" as you experienced.

No thermal throttling at all, FYI - so it may be normal behavior? Use as much temp headroom as possible?

EDIT: I see that you're using HWMonitor now HWiNFO - my bad. Still may be a misread?
 
Mine was definitely throttling under high load. And my original post mentioned my cinebench scores which were off. Something like IBT or P95 will still get it to hit 90c though. If that's normal behavior, it's a first for me. Perhaps one of the reasons it's marketed so heavily as a gaming CPU is because games are the least likely to get it hot but benefit the most from cache. 🤷‍♂️ Ideally, I'd love a bios release where I can undervolt it to 1 or 1.1 volt and just enjoy it at stock speeds and decent temps as I'm sure many others are as well. :)
 
CruisD64 said:
Mine was definitely throttling under high load. And my original post mentioned my cinebench scores which were off. Something like IBT or P95 will still get it to hit 90c though. If that's normal behavior, it's a first for me. Perhaps one of the reasons it's marketed so heavily as a gaming CPU is because games are the least likely to get it hot but benefit the most from cache. 🤷‍♂️ Ideally, I'd love a bios release where I can undervolt it to 1 or 1.1 volt and just enjoy it at stock speeds and decent temps as I'm sure many others are as well. :)
Click to expand...
People have been undervolting it. What board do you have?
 
CruisD64 said:
MSI MPG B550I
Click to expand...
I haven't seen that board mentioned, specifically. Seems like each board has different features blocked. AMD bios often have 2 or 3 different places, where you can affect the same things. I would dig around a bit in your bios, to be certain. Your bios may have a feature available, which they forgot to block.

Otherwise, PBO Tuner or a similar app, might let you change voltage in windows.
 
chameleoneel said:
I haven't seen that board mentioned, specifically. Seems like each board has different features blocked. AMD bios often have 2 or 3 different places, where you can affect the same things. I would dig around a bit in your bios, to be certain. Your bios may have a feature available, which they forgot to block.

Otherwise, PBO Tuner or a similar app, might let you change voltage in windows.
Click to expand...
I'll look again but I believe it to be blocked. I previously had a 3800x in there where I could adjust the voltage. PBO Tuner works but the lowest my voltage will go it 1.18. I'd love to try 1v like that one dude...
 
CruisD64 said:
I'll look again but I believe it to be blocked. I previously had a 3800x in there where I could adjust the voltage. PBO Tuner works but the lowest my voltage will go it 1.18. I'd love to try 1v like that one dude...
Click to expand...
1.18 should still be a nice decrease from the 1.35 default. I would like to hear your temp data stock vs. 1.18
 
PBO 2 Tuner can be used to set core optimizer values to -30 to keep temps under control. Been using it myself at -30 for all cores with no stability issues and boost values are the same as without it. Dropped from 91C to 81C on cinebench R20 temp and boosted score a little as well. Haven't messed with boost limits at all in tuner as I'm not interested in running the chip any hotter than necessary at this point in time.
 
PBO Tuner helped a little but boost voltages still got higher than what was expected. I was lucky enough to be noisy on MSI's forum and they released a beta bios with (-) voltage offset. That has worked out really well! A slight loss in performance but temps are in the ~70's with the settings that I used. Below is the link to the thread covering it.

https://forum-en.msi.com/index.php?threads/kombo-strike-beta-bios-for-r7-5800x3d-cpu.376532/

The boards that currently have this setting are as follows:

7C92 MPG B550I GAMING EDGE WIFI
7C92 MPG B550I GAMING EDGE MAX WIFI
7C91 MPG B550 GAMING EDGE WIFI
7D13 MEG B550 UNIFY-X
7D13 MEG B550 UNIFY
7C34 MEG X570 GODLIKE
7C94 MAG B550M MORTAR WIFI
7C94 MAG B550M MORTAR
7D50 MEG X570S ACE MAX
7D51 MEG X570S UNIFY-X MAX
 
Try Ryzen Master software for temps. Remember there is alot packed into the 5800x 3D die. So the heat output is alot. On my 5600 and Scythe Mugen in an ITX case Under full load with normal all core workloads I can hit 4.6ish (4650) is max boost) and temps stay in the mid 60's to low 70's but my heatsink is cool to the touch...
 
CruisD64 Hey man, slightly off topic but I wanted to ask exactly what kinds of temps were you getting on your 5800X3D when it was in the evolv shift 2?
Asking because I have the Evolv Shift 2 Air and a 120mm beQuiet PureLoop cooler with a phanteks T30 fan on the radiator. I really want to upgrade to the 5800X3D because I know I'm going to undervolt the CPU and play with the PPT, TDC, EDC limits to get temps lower (I don't mind losing up to 10% performance under load). I'm also going to upgrade my case fans once noctua launched their new 140mm NF-A14.
 
Old thread, but I manually overclocked and dropped my 5600x temps a good bit. Done with two different cpus and two different motherboards. People claiming manual over clocking with AMD is dead, but it seemed to make my rigs run cooler and they still perform well according to cinebench.
 
PBO is back for some mobos in BIOS. I did a -20 for best cores and -30 for the rest. Saw my darktide in game temps drop almost 10C. Prime 95 effective core speed went up from 4.0 to 4.2ghz for me at 90C throttle temp, so definitely working fine.
 
Slade said:
PBO is back for some mobos in BIOS. I did a -20 for best cores and -30 for the rest. Saw my darktide in game temps drop almost 10C. Prime 95 effective core speed went up from 4.0 to 4.2ghz for me at 90C throttle temp, so definitely working fine.
Click to expand...
If ASUS ever gets around to updating my mobo variant, I'll probably only go -5 or something. I had lots of low-load BSOD problems with curve optimizer and my 5600X.
 
I had a launch 5600, tweaking that early on was tricky. I was able to drop the top cores further to -25. -30 crashed it in low load use, so at least I know the top cores need bw -29 and -25.
 
You must log in or register to reply here.
Back
Top