PC Building

[mike71]

I've tried to day to stress test my build with cinebench on the CPU and I got high temperatures, not on cores, but on the package and CCD#0.

I have a Ryzen 7 7700X with a Noctua NHD15-S as cooler with an extra fan added (Fractal Aspect 12). Mobo is a Gigabyte Gaming X AX B650 and the graphics cars its' a Sapphire Radeo 6700 XT

I have as a case a pop XL Silent by fractal design with added three Fractal Aspect 14 PWM fans, one  on the rear and two on the top pushing air outside, and the three stock Fractal aspect 12 sucking inside the air from the front.

The problem is that when i run Cinebench on the CPU for a couple of minute I got the temparature on package at a high 95.3 C, on the CCD#0 98.1 C, by the way what it means?, and the core max temp 92.4 C.

First question: are the temperatures actually ok and I'm worrying for nothing?

If not, what I could do? I was thinking about inverting the upper fans, or try to repaste the CPU. Any other ideas? It's possible in some BIOS setting telly to do a less aggressive ramp up of the clock?

PS: there's a pgoto of the build, if you'r einterested.

[BigAl01]

I think as you approach 100 C that things are getting way too hot.  The current fan orientation looks good, with three intake fans on the front, and one year exhaust fan with two top exhaust fans.  I believe you should add a second fan to your CPU cooler, and orient them so the fan towards the front of the CPU cooler is pulling air into the CPU cooler, while the new fan on the rear of the CPU cooler is pulling air out of the CPU cooler.  If that doesn't improve things then consider an AIO setup, with the radiator mounted on the top.

[Sam_AMD]

Hi, @mike71 (nice to e-meet you).

I have some information for you.

CCD stands for Core Chiplet Dies (CCDs). A CCD, sometimes also called a "chiplet", is a cluster of eight CPU cores that share access to a common L3 cache. Some AMD CPUs, like the Ryzen 7 7700X and Ryzen 7 7800X3D, have one active CCD per physical processor. Higher-end chips in the Ryzen 9 Series have two active CCDs, typically referred to as CCD0 and CCD1.

Your temps are a bit too high. I suggest you start by checking to see if there's enough thermal paste on the CPU itself. It can get a little messy, but a great place to start troubleshooting. Make sure it's spread consistently across the surface of the CPU.

Your Noctua NHD15-S has plenty of heatsinks (the larger the surface area the higher the heat dissipation), so I would worry much about this device, as long as it is resting on the CPU, and, again, there's enough thermal paste. By the way, is the 140mm fan working/spinning?

About the airflow, check out this picture, it clearly shows how the air should flow in and out of your rig:

Simply put, you're looking for the Push-Pull effect (air being pulled into your case from the front panel and pushed out from the rear and top of your case).

You've done a great job in cable management, the inside of your rig looks pretty good to me, but be sure that the rear fan is actually sucking the heat from the CPU cooler out of the case. Just double-check the orientation of the fan.

Lastly, as @BigAl01 mentioned, consider going with an AIO...

 

[mike71]

Thank you for the answer, so the CCD£0 is the master temperature of the die.

As a stopgap measure, before damaging something I disable the boost and the processor runs at 4500 MHz.

The fans are spinning and are oriented in the same direction.

The thermal paste it's another thing to check, and actually I'm not sure that I spread it correctly. In case it didn't cover all the heat sink surface it's sufficient to add only a bit in the missing points, or clean it the CPU and heat sink and reapply from zero?

Another thing I'm going to try is to use a 140 mm Noctua fan as second one instead of the Fractal Design one that is a 120 mm one with mas speed of 1200 rpm, Noctua is 1500, and using the extra one to put in series with one of the top fan and getting more air on the top part of the case.

[Sam_AMD]

Hello, @mike71 

Great idea! Disabling AMD Precision Boost can help with troubleshooting. 

About the thermal paste. You should remove it completely and re-apply it evenly across the entire surface of the CPU. Don't put too much, it should be a thin layer.

About the fans (including the 140 mm Noctua fan), be sure they are pulling the (hot) air out of your case. Place your hand over them once you're done with the installation/re-configuration. 

And if you don't mind, please let us know how it goes.

[elpuertorro]

If i might chime in, I noticed you mention you added a Fractal fan to the Noctua to help in push pull, Here is the thing, in that case that both fans should be working in unison for push/pull, 1:1, it is better to have both fans of same brand and model, why you may ask?
They can both run at different rpm as different specs from manufacturing and can actually damage each other, if one is pulling more, the other runs faster not as programed and vice versa, so if possible try to get both fans on the Noctua of same brand and model.
ps: As Sam suggested make 100% sure those 2 are going the same direction pushing towards the back of case.
Sometimes it happens to the best of us, so just double check.
Report back how it went.

[mike71]

I'm back after a week but I bought some stuff and had some problem to make progress.

I bought some new thermal paste from Noctua (because I lost somewhere the old syringe and the offset bar that Noctua suggest to use with AM5 Ryzen CPUS.

Now: this was the CPU when I detached the cooler:

Thermal paste before

Then after mounting the offset bars:

Thermal paste applied with AM5 bars

And while I was remounting all the stuff I noticed that on the top of the intake fan on the front the Pop XL Air Silent has a gap. I closed with some electric tape to make sure that the upper fan isn't sucking air from the inside.

Electric tape to the rescue

Now the results:

Without the front panel:

Without the grille

With the front panel:

with the grille

Having the front panel on means an 1.5 increasing on core temp but more or less same peak CCD#0 temps.

Themerature are a bit lowe and I newve seen peaks of 98 C as before, so for now it's a meh result. What I noticed is that the peak on CCD#0 happens when I try a screensot of HWMonitor with Cinebench running, as if something interferes between those programs.

The other suggestion I have to follow is to get a Noctua fan and use it, to having more cooling rather than with the Fractal design fan that runs at 1200 rpm insted of 1500 and it's smaller.

[ulieq]

literally designed to run that hot.

[mike71]

From what I've gathered, the Ryzen 7 7700X in normal operation has a thermal throttling at 95 C, and sometimes it overshoots.

I' ve found that it's possible to enable the ECO mode, that power limits the CPU. Unfortunately in Gigabyte BIOS it's hidden under a submenu,  AMD CBS->SMU Common, that is quite counterintuitive. With that limitation the maximum temperature it's 81 C but the performance hit meauserd with Cinebench isn't too big.

At the end, using the offset Noctua bars for the AM5 socket helped a bit, and maybe also the repasting. But using the ECO mode helped a lot more to sta at lower temperatures.

[eebiii]

I've really never been a fan of the P method. Leaves to many gaps. A nice AIO would bring those temps down.

[mkr24ac]

What do you mean by "P method?"

[-SABRE-]

Hello! I don't really know if it is relevant anymore, but I personally recommend to either undervolt CPU by a curve a little or just lock-in the maximum voltage that Motherboard can use in the BIOS on some optimal level. The thing is in recent years manufacturers of Motherboards are going insane with their power plans and often enough overvoltage the CPUs which can even lead to a fried ones and it is not an AMD thing, Intel ones are frying left and right too because of it. So to be safe it is better to manually tune it, especially with such a high-end and pretty hot chip. 

I personally lock my voltage on 1.2V on my R7 3700X and lock the clock on the 4.4Ghz(which is max boost clock for my CPU). My temps are really low(Around 70 C max in stress tests and 60-62 C in gaming with very cheap cooling system) and the performance is even better than with dynamic clocks. But you need to find this voltage for each CPU by changing and then running some test over and over so it would be stable and cool(even if it is just a 5 degreec C it is still a win isnt it?)

And regarding the CCD I just see it as a "hotspot" thing in GPUs, maybe it is not correct by I just view it like this and don't really care about it unless it is too hot.