Question 5900x Low Cinebench Scores & High Temps?

[invictus9407]

Hi everyone, I recently upgraded my CPU and GPU, and I have been doing some testing on performance & temps. Based on comparisons, it seems like my Cinebench scores with the R9 5900x are pretty low (more details below). I've also noticed that my temperatures seem quite high, even on idle. I know Zen 3 can run hot, but I just want to make sure that I'm getting the best out of my CPU while also keeping temperatures on the lower end if possible. I've explored a bit of undervolting & overclocking, but I don't have much experience with either so would love to get your thoughts and some advice/tips.

Below are my specs & different scenarios that I have explored so far along with Cinebench results.

Specs

Mobo: MSI B450M Mortar
CPU: Ryzen 9 5900x
CPU cooler: NZXT Kraken X62 AIO Water Cooling Unit - 280mm rad
GPU: MSI RTX 3090 Ventus 3X 24GB GDDR6X
RAM: Kingston Fury Renegade DDR4 3600MHz CL16 32GB Kit (2x16GB) - XMP Profile 1
PSU: Corsair RM 850 W 80+ Gold Fully Modular ATX PSU
Storage: WD SN750 1TB NVMe SSD with Heatsink
Case: NZXT h400i mATX

Performance & Temps

On idle/minimal use (i.e. Google Chrome only & some background apps like AIDA64), I have temps around 60-70C. I've also tested a few different configurations with Cinebench so far, and below are my results. Bear in mind room temperature is somewhere around 20-21C right now.

• Stock CPU settings - temps between 88-90C, clocks ~4.5GHz, Cinebench multi core score: 10407
• PBO limits off, curve optimizer set with all cores negative 20, temps between 87-89c, clocks ~4.7GHz, Cinebench multi core score: 10590
• PBO limits: PPT 165, TDC 115, EDC 160, curve optimizer set to all cores negative 20, temps between 89-90C, clocks ~4.7GHz, Cinebench multi core score: 10366
• PBO limits: "motherboard", boost +100MHz, curve optimizer all core negative 10, custom fan profiles on NZXT CAM, Cinebench multi core score: 11703

Temps in games are usually around 75-80C so far, but varies depending on settings.

I know there's probably a lot to explore here and the above settings may seem random, that's because they are, I have little experience with undervolting & overclocking and I've been following some guides here and there, but I'd love to know if my Cinebench scores are normal (other results suggest to me that they should be much higher for the 5900x), and if not, how I can improve them while also keeping temperatures in check.

Thanks in advance!

 

[Unolocogringo]

Take the side panel off and check temps again.
That is a lot of hot hardware in a low airflow case. Restricted front intake.

 

[Phaaze88]

NZXT Kraken X62 AIO Water Cooling Unit - 280mm rad

Let's see if you don't have the fans facing the wrong way - it does happen.
For reference:

Spoiler

^ You want the front fans' airflow direction pointing inside the PC. The rear and top fans should obviously be pointing out.
Also, the AIO fans should be as far away from the front panel as possible.

 

[invictus9407]

Take the side panel off and check temps again.
That is a lot of hot hardware in a low airflow case. Restricted front intake.

Thanks! I've tried with the side panel off but temps are pretty much the same. Do you have any thoughts on the cinebench scores and whether this is normal for a 5900x? The thing that I'm particularly concerned about is it seems like the CPU may be underperforming and still producing a lot of heat.

 

[invictus9407]

Let's see if you don't have the fans facing the wrong way - it does happen.
For reference:

Spoiler

^ You want the front fans' airflow direction pointing inside the PC. The rear and top fans should obviously be pointing out.
Also, the AIO fans should be as far away from the front panel as possible.

Thanks for sharing. The front fans/rad for the CPU cooler point inwards into the PC so based on the above seems correct - the rear and top fans point out. In total I have: 2 top exhaust, 1 back exhaust, 2 front intake (CPU cooler rad).

 

[Phaaze88]

Thanks for sharing. The front fans/rad for the CPU cooler point inwards into the PC so based on the above seems correct - the rear and top fans point out. In total I have: 2 top exhaust, 1 back exhaust, 2 front intake (CPU cooler rad).

I take it the fans are positioned behind the radiator and not in front of them?

There's a connection somewhere with either poor airflow through the radiator - because even AIOs need air.
Or, it's the paste application; might not be covering the entire IHS).

 

[Karadjgne]

With Ryzen at idle (that means no other stuff running like Chrome or Aida), all the cores except 1 shut down. That means the entire background load is on only 1 core, so if you only use a temp register of 1 temp, you'll always see 60+°C as that's the hottest core. Reality puts the cpu closer to 35°C or so for the rest of the cores.

Running anything set by the user wakes up all the cores, you aren't at idle, so if even with minimal use, and the clocks are set high or static, all the cores will be reflecting that shared load, and you'll see 70's instead, as that's the hottest core.

With that case, fans should be behind the rad in pull config, gives the greatest amount of efficiency and the most space for air input from vacuum.

Check your Cam settings. With my old X61 I just set it for Silent mode and let it do its thing. You'll want to see how the pump is responding since that's a variable speed pump and if the settings are off, the temps will reflect that.

I'd only run HWInfo64, look at all the cores speeds and temps, at idle 1 core will be active and it rotates. The other cores will be minimal speeds. If all the cores are at active boost, your cpu is set for high performance (either by bios or motherboard) or some other program such as MSI Dragon Center etc.

Power plan in Windows should be Balanced/Microsoft Balanced/Amd Balanced.

 

[Unolocogringo]

Have you tried "Curve Optimizer" in Ryzen master?
It can reduce voltage and keep boosts up by lowering heat.
My 5600x went from 1.41-1.45v for 4.65ghz to 1.21-1.36v for the same 4.65ghz speed reducing temps and increasing all core load boosts to 4.5ghz.
I had already did the laborious job of manually setting it core by core through bios.
And ran the software to check its results.
It was close but not exact,but it gives you a baseline starting point for your processor.

 

[invictus9407]

With Ryzen at idle (that means no other stuff running like Chrome or Aida), all the cores except 1 shut down. That means the entire background load is on only 1 core, so if you only use a temp register of 1 temp, you'll always see 60+°C as that's the hottest core. Reality puts the cpu closer to 35°C or so for the rest of the cores.

Running anything set by the user wakes up all the cores, you aren't at idle, so if even with minimal use, and the clocks are set high or static, all the cores will be reflecting that shared load, and you'll see 70's instead, as that's the hottest core.

With that case, fans should be behind the rad in pull config, gives the greatest amount of efficiency and the most space for air input from vacuum.

Check your Cam settings. With my old X61 I just set it for Silent mode and let it do its thing. You'll want to see how the pump is responding since that's a variable speed pump and if the settings are off, the temps will reflect that.

I'd only run HWInfo64, look at all the cores speeds and temps, at idle 1 core will be active and it rotates. The other cores will be minimal speeds. If all the cores are at active boost, your cpu is set for high performance (either by bios or motherboard) or some other program such as MSI Dragon Center etc.

Power plan in Windows should be Balanced/Microsoft Balanced/Amd Balanced.

That's great context and good to know. Based on HWInfo64, it looks like all my cores are at active, as they all seem to be running at anywhere between 3.6 - 4.9GHz, despite my Windows power plan being set to Balanced. I am guessing this probably is related to the bios settings which I currently have set within AMD Overclocking section as follows: PBO limits: "motherboard", boost +100MHz, curve optimizer all core negative 10.

That being said I do see that core temperatures vary quite a bit and some of them are as low as 33C, so you're right that the average temperatures seems a fair bit lower than the highest temperature I'm seeing.

In terms of the fans, I can confirm that they are behind the rad in pull config. I currently have CAM set for the Kraken X62 fan to run at 100% when the liquid temperature is 34c or above, and the pump to run at 90% when liquid temp is 30c, going up gradually to 100% when liquid temp is 34c. These were custom profiles that I was playing around with but of course I can adjust further.

EDIT: can I just confirm if I should be looking at the core clocks section of HWInfo64, or Core Effective Clocks? Because under core clocks it looks like these don't drop below 3.6GHz, but under core effective clocks it does seem like several cores are maybe idling?

 

[invictus9407]

Have you tried "Curve Optimizer" in Ryzen master?
It can reduce voltage and keep boosts up by lowering heat.
My 5600x went from 1.41-1.45v for 4.65ghz to 1.21-1.36v for the same 4.65ghz speed reducing temps and increasing all core load boosts to 4.5ghz.
I had already did the laborious job of manually setting it core by core through bios.
And ran the software to check its results.
It was close but not exact,but it gives you a baseline starting point for your processor.

I was actually using the curve optimizer within the BIOS itself, as I've read in several forums and also heard from several folks that Ryzen Master is not great. So currently I have a curve optimizer set in the BIOS for all cores with a negative 10 offset. As mentioned above though, it seems like all my cores remain active and boosting even with minimal activity such as browsing on chrome or just running something like Aida64 in the background...

 

[Karadjgne]

Idle should be somewhere around 5-8% cpu usage. If your background tasking is using more than that you aren't actually in idle, you are one step into active, so the power plans as far as lowering idle speeds, and voltages, don't apply. They are just relegated to non-boosted requirements. Only your foreground active apps will be actively boosted for performance.

Coolant temps generally range from ambient to @ 45°C at most, so 33ush to 45ish. If you exceed that max, you seriously have a cooling issue somewhere. Coolant temps are not representative of cpu temps, but of cooling efficiency. A 280mm aio has a 300w± capacity so after an hour at 70°C cpu, you should be closer to 40°C coolant.

PBO doesn't change what a cpu does, it changes what a cpu can do, there's a difference. Stock has wattage/amperage limits in order to maintain TDP ratings. What PBO does is raise those limits. So if you push your cpu stock to the max, you'll run into those limitations, which changes the limits of the cpu core boosts. Enable PBO and that resets the limits higher, so it's possible to get more performance from the cpu, at the cost of higher power consumption, higher cooling requirements. The +200MHz boost is another limit to core max frequency. Doesn't affect anything lower.

 

[KyaraM]

Not connected to temps directly... Which Cinebench did you run, r20 or r23? In r20, that result would actually be really good. In r23, it's almost impossibly bad.

 

[invictus9407]

Idle should be somewhere around 5-8% cpu usage. If your background tasking is using more than that you aren't actually in idle, you are one step into active, so the power plans as far as lowering idle speeds, and voltages, don't apply. They are just relegated to non-boosted requirements. Only your foreground active apps will be actively boosted for performance.

Coolant temps generally range from ambient to @ 45°C at most, so 33ush to 45ish. If you exceed that max, you seriously have a cooling issue somewhere. Coolant temps are not representative of cpu temps, but of cooling efficiency. A 280mm aio has a 300w± capacity so after an hour at 70°C cpu, you should be closer to 40°C coolant.

PBO doesn't change what a cpu does, it changes what a cpu can do, there's a difference. Stock has wattage/amperage limits in order to maintain TDP ratings. What PBO does is raise those limits. So if you push your cpu stock to the max, you'll run into those limitations, which changes the limits of the cpu core boosts. Enable PBO and that resets the limits higher, so it's possible to get more performance from the cpu, at the cost of higher power consumption, higher cooling requirements. The +200MHz boost is another limit to core max frequency. Doesn't affect anything lower.

Assuming that coolant here is referring to the liquid temp, then that sounds about right. I haven't seen the liquid temp go about 38C or so, so it sounds like there's nothing abnormal happening there?

Not connected to temps directly... Which Cinebench did you run, r20 or r23? In r20, that result would actually be really good. In r23, it's almost impossibly bad.

These results are from Cinebench R23... so I guess my assumption that the CPU is underperforming big time is right?

 

[Karadjgne]

There's no 'all in one' test. Cinebench is good for stability and performance, not temps. TimeSpy is excellent for performance, Aida64 is excellent for stability. Prime95 is excellent for temps. You should pick and choose and swap tests repeatedly depending on exactly which criteria you are wanting to adjust for. Then make compromises with the results. So if Prime gets a temp you can live with but CB performance drops too much as a result, you'll need to adjust cooling or it's sources to enable CB to score higher but remain in acceptable Prime range.

It's why occt has a standard 80°C temp cap, pointless to continue the test since the user will not accept cpu temps over 80°C as being acceptable.

So if Prime gets you 70°, and CB gets you 11000, but you done pretty much everything possible with your setup and settings to get those results, you'll have to decide if a bump in voltages to allow a 100MHz boost to clocks and an extra 5°-10°C on temps is really worth changing the CB to 11400 etc.

 

[invictus9407]

There's no 'all in one' test. Cinebench is good for stability and performance, not temps. TimeSpy is excellent for performance, Aida64 is excellent for stability. Prime95 is excellent for temps. You should pick and choose and swap tests repeatedly depending on exactly which criteria you are wanting to adjust for. Then make compromises with the results. So if Prime gets a temp you can live with but CB performance drops too much as a result, you'll need to adjust cooling or it's sources to enable CB to score higher but remain in acceptable Prime range.

It's why occt has a standard 80°C temp cap, pointless to continue the test since the user will not accept cpu temps over 80°C as being acceptable.

So if Prime gets you 70°, and CB gets you 11000, but you done pretty much everything possible with your setup and settings to get those results, you'll have to decide if a bump in voltages to allow a 100MHz boost to clocks and an extra 5°-10°C on temps is really worth changing the CB to 11400 etc.

Thanks for the insights. This makes sense, I will also test with TimeSpy and a few others. I think my main concern is that, despite temps being fairly high, comparing other Cinebench results from folks who have the 5900x, my scores seem significantly lower (i.e. 11k vs 20-22k), so I'm particularly concerned about whether I am getting the best out of my CPU if that makes sense. I can live with higher temps but only if the performance warrants it which based on what I've read and heard so far, I am not sure that it does.