Dear all,
I am finally upgrading my home PC after some 8 years.

I do not want to hijack other similar threads and therefore I am taking the risk to open a new thread. Apologies if it is not appropriate. Moderators merge it with similar ones if they wish to.

That said: the system (Ryzen 2700x, MSI x470 Gaming Pro MB) works under Kubuntu and is running the special application.
All seems set correctly and no problem under this respect. At least so far.

But: I am experiencing overheating problems. My other rig (INTEL, stock fan) is running in Winter 7 CPU tasks (+1 for -nobs) without a glitch. Hence, all CPUs used.
With this I cannot run more than 2 tasks without overheating.
Just for the sake of noise (but it should also be more efficient), I even opted for water cooling of the CPU (Thermaltake Water 3 Performance C) in this configuration.
No OC (yet, and probably never if it goes on like this).
I maybe not run all 16 threads, but 2 with all fans at 100% seem a bit too few to me...

I also tried prime95 (under Win10 just for the cause and better control real time all cooling parameters. A real pain, I had to disconnect the PC from the net to avoid continuous uncontrollable updates in order to do anything), and I am borderline for the FFT tests with all fans and pump maxed. The most strenuous test is out of reach with more than 2 threads, just like with Seti.

I could not test Seti under Win10 since at the time of tests the project was not distributing WPs, as noted in the panic thread. Bad luck.

I am in the average pack or something is not quite right in what I am experiencing?

Is it possible that a water-cooled (however bad the cooler I am using is) AMD is overheating more than an air cooled Intel?

This is not only about Seti, but also to understand if the system has any flaws. Running Seti helps a lot to spot problems in our systems and run them to the edge, so both aims go hand by hand.

Thank you in advance, and good 2019 (my 20th...) crunching!

Sleepy

What is hot? What are you using to monitor it? Was the temp the same in Windows?

I has a temporary Corsair 120mm Liquid (similar to yours) on my 1700X and it is happy. Certainly more than 2 thread capable.
I see that has preapplied paste on that model, did you use that or your own?

---

You could just have a duff AIO out of the box. Have you tried using the stock box cooler? You should be able to run all 16 threads at stock frequency without overheating. If it overheats on just 2 threads something is wrong.

Bad application of thermal compound or bad mounting of the AIO onto the cpu IHS?

What do you consider overheated temps? You are aware that most monitoring applications don't report the true temperature of the cpu. AMD in their infinite silly wisdom reports an artificial temperature that is 10° C. higher than actual to force a more aggressive fan curve for cooling for the 2nd generation Ryzen cpus. It was a 20° C. offset on the 1st generation cpus. It is a 27°C. offset for the Threadripper cpus.

The only Windows based monitoring applications that report the true "Tdie" temperature of the cpu minus the offset is HWinfo64 and Ryzen Master. So if you are relying on any other monitoring program that only report the Tctl artificial temperature, just do the mental calculation to subtract 10°C. from that.

For a 2700X on a Corsair H105 240mm radiator or a H110i 280mm radiator AIO I see about 75° C. Tdie temperature when running Seti or Prime95 on all threads.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Not trying to be insulting ,but did you remember to remove the plastic wrap from the bottom of the cooler before installation. Again just checking, it can happen to anyone.

---

Not trying to be insulting ,but did you remember to remove the plastic wrap from the bottom of the cooler before installation. Again just checking, it can happen to anyone.

Ooohh! . . . that's a good one that has caught out quite a few people. Especially the XSPC Raystorm Pro cpu block users like myself. Didn't think to add that to the possibilities. Don't know if that is same thing with Thermaltake. Could try a google search on the TT150 AIO and see if that is a common mishap.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

I don't know what you consider to be "hot."

I'm using the Wraith Prism that came with my 2700X and I left the auto-OC feature enabled on the mobo. When I'm doing h.264 encoding and task manager shows 'x264_x64.exe' using 98% of total CPU for an hour, I pull up CPU-Z and it shows that I'm running 3950mhz (from the base of 3700). If I then pull up the Gigabyte toolbox to load up SIV for the reported temps, it SAYS 82°C for CPU, but I know it isn't actually 82°C. I know there's an artificial offset on the temp reporting specifically to manipulate the fan curve, but in my situation, it seems to be doing its job, because the CPU feels that it is well within the thermal envelope to OC itself, rather than downclock and shutting cores off.

SIV via the Gigabyte toolbox is the only thing that seems to be able to read the sensors on this board properly, as well. Speccy seems to think that the CPU is somewhere around 130°C under full load, 105 at idle. Same for the stand-alone version of SIV. The one that is on the Gigabyte website and is to be installed via the Gigabyte App Center seems to report properly.

When I've been idle and even a single-threaded task starts to run, the fan goes to 100% right away. The plus-side of a single single-threaded task is that one core will OC to 4250 usually.

it seems though that the fan cooling profile is that anything above more than about 25% load on any one given core/thread will kick the cooling profile to 100%. I'm not complaining.. the Wraith Prism is doing its job as it was designed.



If you're running extremely hot with barely any load, then yeah, it sounds like there's a problem, obviously. Is the radiator as warm as you would expect if the CPU seems to think it is near melt-down temp? How about the water block itself? Is that really warm (be careful when touching it to find out.. I made the mistake of touching the heatsink on an old AthlonXP that hit 105°C and died.. there was a sizzle and it hurt) ?

Basically.. is the heat even being transferred away from the CPU at all? If the block isn't well over 40°C, then you're not getting enough heat transfer to it, which means the thermal paste either isn't present at all, or there's way too much, or as someone else suggested, there might still be that protective plastic film on there. If the block is really hot but the rad is nowhere near that warm (specifically near where the hot liquid should go into it to be cooled), then you're not getting fluid circulation. That could either be a bad pump, or for some reason, the header on the mobo for the pump isn't working properly.

have you at least TRIED the Prism that came with 2700X? What is the result of that? Does it cool properly and allow you to use all cores, even if it is a little louder than you would like?

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

Speccy seems to think that the CPU is somewhere around 130°C under full load, 105 at idle.

I'd say those temperatures are actually in Fahrenheit. If it were Celsius, the CPU would be dead- given that it's rated for a 85°c maximum.

---

Grant
Darwin NT

I would have thought so, too, but it clearly says C.


Meanwhile, this is what the Gigabyte version of SIV says:


I have no idea if "81" is even what it really is or not. I suspect reality is probably near 65. But at any rate.. this is the cooler that came with it, and even at these temps, the CPU feels safe enough to OC itself above base of 3700. It wouldn't do that if the temps were too high.

edit: and the cpu is 12nm, not 14 like Speccy says. Not many tools can read this hardware very accurately, i've noticed.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

I would have thought so, too, but it clearly says C.

I hope the rest of their information is better than that data.
Looks like a bug report is in order.

Scratch that- a quick search shows it's been a known issue for some time, and they've not bothered to fix it through several releases.

---

Grant
Darwin NT

Yeah I'm not bothering to worry about it. Like I said, not many utilities read this hardware even close to properly. Gigabyte's SIV reads the temps fine, but says my ram is "2200mhz" and that the sticks are 1066. Speccy and CPU-Z also say that about the RAM.. but the UEFI menu says not only are they rated for 3200, but that's what they're running at:


So... i'm not worried about it. I know everything is working fine.

Anyway, sorry for the semi-hijack. This isn't about my rig, but I have the same CPU and some knowledge about the temps that OP was asking about, so I thought it was relevant.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

Dear all,
thank you for all your answers.

While I am further investigating (but now is time to work, so just a brief update), I clarify that I have not built the system myself. Just because I was afraid I would be not good enough to clip the CPU to the MB and then do the thermal thing correctly. Nemesis.

I have not quoted temps I read because I looking up in Internet (and here) I understood that there was a bit of uncertainty depending the utility used to monitor it.
For example, with an "upgraded" sensors , I read under Kubuntu temps up to 60 °C when the CPU cuts off for overheating.
Under WIN10 and using the MSI utility, I reach some degrees above 70°C and then bang!
In any case, from what you say, it seems there must be something in my system not behaving normally (or it is me, as part of the system ;-) ).
Whatever the temperature reading, you do not speak about thermal cut offs, while for me it is very easy to reach one.

Now, the frequency. I said I was not OCing, But I remember seeing something like 4.1 GHz, which is not really stock. Since I was not asking the seller to OC it, which I would do just in case, I was not giving any attention to this aspect. If it is the case, then I will check if this is just a momentarily boost, or I am in OC without yet wanting it.

And I will carefully put my fingers in the right places to check local temperatures.
Tubing in any case gets barely lukewarm and the radiator is never hot. At most, very mildly lukewarm. I have to check the CPU block itself.
Tonight after work.

Thank you! I will keep you posted.

Sleepy

I still think the pump isn't working. Replace with stock air cooler to test. With Linux you have the stock k10temp kernel mode driver supplied with kernels > 4.15.0 and should report the temps as something like:
k10temp-pci-00c3
Adapter: PCI adapter
Tdie: +75.0°C (high = +70.0°C)
Tctl: +85.0°C

That is with the lm-sensors package installed and a sensors command run in Terminal. You also can load a specific driver for the SIO monitoring chip on the motherboard. I am not familiar with what Gigabyte is using for their SIO chip. You will have to find that with a sensors-detect command as root. Then you can see whether it is one of the standard ones with driver already installed for your distribution or whether you have to go to github and find the driver there and install it.

My ASUS Ryzen boards all use the ITE8665E SIO chip. My ASUS Intel board uses the NCT6779 SIO chip as well as my Asrock Threadripper board. The optionally installed SIO chip kernel mode drivers report much more that the basic cpu temp. They report the voltages and fan speeds just like the monitoring chip does under Windows.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

It turned out they gave me my new PC OCed at the limits, without telling me.
There even is a profile in the BIOS saved with these settings, just in case I had the doubt I had inadvertently changed something (which I did not, because my aim was to have the PC working in standard conditions first, and, just in case, test some OC only afterwards).
They probably tested deeply the cooling and forgot to revert to stock conditions.
Accurate assembling and bad delivering at the same time.

My bad I have not checked it immediately, but I was so sure a PC cannot be given to a customer OCed, but just in vanilla conditions, that it did not come to my mind.
I arrived to notice this after asking you. Many of you were reporting their clock speeds, in a way or the other, and it was all about 3700 MHz and thereabouts.
I was running at 4150 MHz!

So, 2+2 and after checking and a couple of basic tests, everything seems normal now.
I am at stock speeds (for now) and I can pass the most strenuous torture test of Prime95 with the cooling humming very lightly.
As for SETI, I can now run using all CPUs, with temperatures, as from "sensors" of about 56 °C. (I can choose the working point in any case by trimming the fan curves, I think I still have space to reduce it).
Then I will decide if I will limit this, but for the moment all is good.

How come they can give you a new OCed PC without telling you escapes my mind... :-(

In any case, thank you for your help. I got out from this cul de sac reading your suggestions.

In the course of the process, I also had to discover the sweetness of Win10. But this belongs the other specific thread.

Thank you again and I will post more if anything relevant occurs.
I think that at the end the figures published here can be helpful as reference for others.

Sleepy

I'm glad you found the answer to the heating issue.

At least now you know you can change to 4150MHz easily :)
Your smallish 120mm cooler might be able to handle it with 2 fans in a push-pull configuration. Just get a High Static Pressure fan for the 'push'
It would be a cheap upgrade.

---

To have all the cores running at 4150Mhz requires PBO level 3 or 4 likely and the Vcore necessary for that was probably north of 1.42V. More volts = A LOT MORE Watts/Heat and more than your AIO can reasonably dissipate as configured. More fans could help.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)