Hi,
I'm curious about the temperatures I should expect with my setup.
It's a TR 2950x cooled by a NH-U14S TR4-SP3 in a Fractal Design Define R6.
The idle temperature is 37°.
When I start tasks using 32 threads, the fan reaches 1500RPM, the temperature reaches 68° in a few seconds (30 ?) and goes down about as fast when the job stops.
When the temperature is at 68°, I see that about half the "cores" are at 4.4GHz and the other half at 1.7GHz.
Does this seem normal ?
Thanks.
Solved! Go to Solution.
If your Noctua CPU Cooler doesn't have a second fan you may want to install one. Here is a chart from Noctua concerning the TDP Watts of TR4 CPU and their recommendations: AMD Ryzen Threadripper TDP guidelines .
Hard to say, hurricane. What are you using to measure temperatures? If not Ryzen Master (RM), please install it and post a screenshot. RM is the AMD reference for CPU temperature and this is what you compare to the 68 C limit. If these are valid temperatures you are measuring, then you processor is throttling to control the temperature. Sorry but a quick look at your cooler seems to say it is not sufficient. I have an Enermax Liqtech TR4 240 on my 1950X and plan probably a 280 for a build with the AMD 2990WX. There is also a 360. The numbers reflect 2 or 3 120mm fans or 2 140mm fans. Does the Noctua cold plate cover all four of your chips? This is a specific feature of the Enermax. I am not selling these, but am happy with mine. I suspect many cooler vendors now have cold/water blocks that cover all four chips. Maybe the core speed distribution will indicate what cores are covered by your cold plate. Please let us hear. Thanks and enjoy, John.
EDIT: According to elstaci, 2950X has only two chips not four as I stated. The geometry problem is the same.
Thank you for your reply. I'm using the cpuid/hwmonitor program. It seems to give the correct value (takes the 27° offset into account).
The cooler plate covers the whole cpu.
Yes, Master Ryzen reads the correct temperature of Ryzen CPUs as stated by MisterJ
If you are showing, and it is accurate, 68c is the maximum operating temperature for the Ryzen 2950x. 37c at idle is very good but 68c under load is not. The CPU is basically starting to overheat and throttle.
Found this Reddit thread concerning the same situation you have. They mentioned about adjusting the CPU voltages and configuring the other BIOS features to keep the Ryzen CPU temperature lower under load: https://www.reddit.com/r/Amd/comments/8n4bni/info_about_ryzen_5_2600x_safe_temperatures_and/
Make sure that the Noctua is firmly attached to the CPU and it doesn't wiggle or have any sideways movements. Do this very carefully. Any type of movement or wiggling indicates it doesn't have a tight firm contact with the CPU. It is a big CPU Cooler and it is possible with the computer case being moved around it may have loosened up a bit.
If your Noctua CPU Cooler doesn't have a second fan you may want to install one. Here is a chart from Noctua concerning the TDP Watts of TR4 CPU and their recommendations: AMD Ryzen Threadripper TDP guidelines .
Thank you for the suggestion. I'll acquire another fan and see how it goes.
Thank you for your reply.
The radiator is firmly fixed.
I suspect the problem could be with the thermal paste.
I know the first time I tried I didn't put enough as there were clean "holes" on my first try. I imagine I may have put too much on my second try.
But if the thermal paste was the issue, I would expect the idle temperature would be higher but to be honest I don't know what to expect on that front.
Having a thin layer of thermal paste has never been an issue over the years, until now. The surface to cover is really big.
Given the difficulty I had, I'd rather be sure it's the culprit before doing it a third time.
This User with a FX 8350 CPU was overheating with a Wraith MAX CPU Cooler. https://community.amd.com/message/2868030#comment-2868030?q=cpu%20overheating . Turn out the Wraith was not making proper contact with the CPU surface.
He found out was when the CPU was overheating he put pressure on the Wraith CPU Cooler and the Temperature immediately went down.
That was the reason I suggest to check your CPU cooler to make sure it is making proper contact.
By the way, I found the best way to spread Thermal Paste is to use a small flat edge or Credit card to spread the thermal paste evenly over the entire CPU surface (paint it on) before installing the CPU Cooler.
Note: By any chance did you read the Thread from Reddit concerning configuring BIOS to make your Ryzen run cooler : https://www.reddit.com/r/Amd/comments/8n4bni/info_about_ryzen_5_2600x_safe_temperatures_and/
I usually use a credit card but Noctua gives another way to do it. (Drops of a given size at specific point(s))
That's what I did for my 2500g and 2700x and it worked perfectly.
But here it's really not that easy.
And since too much thermal paste acts as an insulator, that would be the same as not doing proper contact.
When I look for precision boost overdrive in the BIOS, I get a warning so I don't get in there.
I really don't want to do anything out of specifications. Be it for over or under clocking.
I'm almost certain the issue is with the thermal paste. The cooler is supposed to be one of the best air available.
I'll try the additional fan suggestion first but I think I'll have no choice in the end.
The second fan did help my CPU cooler stay a little bit cooler under 100% load and it does maintain a few degrees cooler while idling (Push - Pull Configuration)
But with my Hyper 212 EVO you can connect any fan as long as it is the proper size - 120mm. Turns out when I purchased my Cooler Master HAF Tower it came with an extra bracket and 120 mm Fan to be used for additional cooling of a GPU card. I removed the fan and installed the brackets that came for installing the second fan on the CPU Cooler. The Fan I installed was not a PWM Fan but a case fan with a 3 pin connector instead of 4 pin connector. But still connected to my motherboard's 4 pin AUX CPU port. The fan runs 100% all the time which is not a problem.
The GPU bracket with fan was too large to install with my GPU card so it was gathering dust since I bought the Computer Tower case.
Your CPU Cooler may require a specific Fan from the manufacturer to be able to install it on your CPU Cooler unlike my CPU Cooler that any 120 mm fan would fit.
Yes, if you are not comfortable in using Precision Boost than it is best to leave it at "Default" rather than configure it improperly and making your computer unbootable.
I've run a bench just before opening the case and the heat-sink felt hot trough my gloves, that was a good omen.
I've just tried with a second fan (NF-A15 PWM) and in an open box the maximum temperature reached is 58°C
I've removed every unused tray bay from the R6 and, in the closed box, the maximum temperature is 60°C for node 0 and 53° for node 1. Every core is around 4GHz
(I wonder if it's normal to have a 7° difference between both packages.)
If all proposed solutions were helpful, installing a second fan was the correct answer for my problem.
Thanks again to everyone.
Didn't read far enough down the Noctua page but found this that may explain the Ryzen Temperatures:
Copied from Nuctua TR4 website from previous comment:
Attention: Very good case ventilation is recommended for all Ryzen Threadripper based systems regardless of the cooler model being used.
Thermal throttling of the VRM: After some time of prolonged full load, the CPU clock starts to drop off to very low values intermittently, is the cooler insufficient?
This behavior is usually due to overheating of the voltage regulation module (VRM), which provides the CPU with power. To stay within a defined temperature threshold (defined by the component and motherboard vendor) and to cool down the VRM, the clock speed of the CPU might be reduced to very low values. When the temperature of the VRM is back within a safe range, the CPU clocks will be increased back to the expected values until the temperature limit is reached again.
Internal tests at Noctua have shown that not all motherboards on the market are able to power the 250W models at prolonged full CPU load without the VRM running into thermal throttling. In case you experience the behavior described above please, check the VRM temperature (e.g. via HWinfo) and provide additional cooling to this area. In doubt, please contact your motherboard manufacturer for further information.
High CPU temperatures: Even with a high-end cooler, the performance is not what I would expect, what can I do?
First of all, please note that results in reviews might show different results as many tests are conducted in open test bench configurations outside a chassis, which allows a better supply of fresh air and thus allows the coolers reach lower temperatures than in a real-world setup inside a chassis.
Secondly, processors emitting heat-loads of 180-250W or even more (with Precision Boost) create more challenges for case ventilation than lower power mainstream CPUs as it is very difficult to remove so much heat from the chassis. For best performance, it is crucial that sufficient amounts of fresh air can enter the chassis and that warm air leaves the chassis as fast as possible.
Internal testing at Noctua has shown that many PC cases are rather restrictive when it comes to air intake and exhaust due to dampening materials, doors, covers or a rather closed design approach in general. Increasing the number of system fans can help to a certain degree but will not provide much improvement if intakes and outlets are too resistive. The best cooling results are to be expected from cases which use a rather open approach (e.g. mesh walls and intakes). Unfortunately, Noctua cannot provide detailed recommendations for individual system configurations as there are too many options and variables available to give general recommendations. Please refer to general case cooling guidelines such as this one for information how to setup the airflow inside your case.