Question ideal OC temp for R9 3900x ?

[Chrexolia112]

Hello, I got my R9 3900x to 4.3ghz all core stable at 1.34v.

I use dark rock pro 4

(Ambient Temp = 25c )

CPU is usually sitting at 40-47c when I do some light works ( adobe xd, browsers etc )

50-55c during gaming
and 71-75c on cinebench

Also I noticed the voltage drop to around 1.27v or something in cpu-z something during cinbench. ( is that normal ? )

I am new to AMD,

Is this a safe temp for long term use ? ( I don't plan on upgrading anything in the next 5 year )
Thank you

Edit :

10min in Prime95 hit me at around 83-86c
I turned everything down to stock speed with cpu offset 0.18 ( ASUS motherboard use 1.46v on CPU which is waaay too high )

Now I am at 35-45c idle and 70-73c on prime95 test. It boosted my core to 4.2 on cinebench and 4150 on prime95.

And according to ryzen master, I am using peak core voltage of 1.49v which is weird. it shows 1.30v in cpu-z

and 4.6ghz boost even at single core is non-existent.

 

[Roland Of Gilead]

In general terms, most folks would advise keeping your CPU (Intel or AMd) under 80c at full stress load, OC'ing.

Cinebench (the key is in the name - Bench)is a benchmark and not a true indicator of max temp for your CPU at a given voltage. You should test with Prime 95, small fft's and find out what temps your CPU is hitting on a real stress load. If you are hitting 75c on Cinebench, you will hit much higher on Prime. Prime will also show you if your CPU voltage is stable. If you have an errors while running prime, then that will mean your voltage needs adjusting, and your current settings are not stable.

Your temps are a tad on the high side, but it would be the stress load temps I'd want to see.

Otherwise 1.34 isn't to shabby. But if temps are higher with Prime, you'll either need to back down the OC, reduce voltage or get a better cooler.

Edit, the difference between your load voltage and set voltage, is called vdroop. This is normal and occurs. It's different from different levels of mobo, and sometimes isn't as bad on one mobo v another. You can manage vdroop by using LLC in your bios options. But if you are not familiar with it, then I'd suggest looking into it further or asking more questions

 

[Chrexolia112]

In general terms, most folks would advise keeping your CPU (Intel or AMd) under 80c at full stress load, OC'ing.

Cinebench (the key is in the name - Bench)is a benchmark and not a true indicator of max temp for your CPU at a given voltage. You should test with Prime 95, small fft's and find out what temps your CPU is hitting on a real stress load. If you are hitting 75c on Cinebench, you will hit much higher on Prime. Prime will also show you if your CPU voltage is stable. If you have an errors while running prime, then that will mean your voltage needs adjusting, and your current settings are not stable.

Your temps are a tad on the high side, but it would be the stress load temps I'd want to see.

Otherwise 1.34 isn't to shabby. But if temps are higher with Prime, you'll either need to back down the OC, reduce voltage or get a better cooler.

Edit, the difference between your load voltage and set voltage, is called vdroop. This is normal and occurs. It's different from different levels of mobo, and sometimes isn't as bad on one mobo v another. You can manage vdroop by using LLC in your bios options. But if you are not familiar with it, then I'd suggest looking into it further or asking more questions

Thank you, I will try prime95 today. Will update the result here.

 

[Roland Of Gilead]

Thank you, I will try prime95 today. Will update the result here.

Cool, let us know how you get on.

 

[Chrexolia112]

Cool, let us know how you get on.

10min in Prime95 hit me at around 83-86c
I turned everything down to stock speed with cpu offset 0.18 ( ASUS motherboard use 1.46v on CPU which is waaay too high )

Now I am at 35-45c idle and 70-73c on prime95 test. It boosted my core to 4.2 on cinebench and 4150 on prime95.

And according to ryzen master, I am using peak core voltage of 1.49v which is weird. it shows 1.30v in cpu-z

and 4.6ghz boost even at single core is non-existent.

 

[Karadjgne]

You can set Prime95 to use just 1 worker. That'll equate to a single thread. You'll probably then be able to see single thread boost clocks reliably. Very, very few games actually only use 1 thread anymore, most are a minimum of 2, many are 3-5, but anything multi-player online is going to hit 6-8 threads easily, maybe more if there's available threads.

Gotta figure that you get more work done with 2 slower threads (like 2.8GHz) than a single faster thread at 5.0GHz. So you'll very seldom, if ever, see heavy single thread usage on anything complex as even a really slow pc will get better performance with higher than single thread usage.

Game developers have to take into consideration that not everybody has 8+ threads to play with, many are still using older i3's and i5's, and they want to play too. So will pay for the game, if it'll work on their pc, even at low settings and less than 60fps.

4.2GHz across that many threads on a 3900x? That's not going to hurt anything. When using far less threads gaming, it'll probably boost higher as temps and voltages allow.

 

[Chrexolia112]

You can set Prime95 to use just 1 worker. That'll equate to a single thread. You'll probably then be able to see single thread boost clocks reliably. Very, very few games actually only use 1 thread anymore, most are a minimum of 2, many are 3-5, but anything multi-player online is going to hit 6-8 threads easily, maybe more if there's available threads.

Gotta figure that you get more work done with 2 slower threads (like 2.8GHz) than a single faster thread at 5.0GHz. So you'll very seldom, if ever, see heavy single thread usage on anything complex as even a really slow pc will get better performance with higher than single thread usage.

Game developers have to take into consideration that not everybody has 8+ threads to play with, many are still using older i3's and i5's, and they want to play too. So will pay for the game, if it'll work on their pc, even at low settings and less than 60fps.

4.2GHz across that many threads on a 3900x? That's not going to hurt anything. When using far less threads gaming, it'll probably boost higher as temps and voltages allow.

I read something about bios version currently I am at chipset 1.0.0.3ABBA which capped my boost at 4225 ghz. ( no idea why, maybe someone can explain me on this )
And gaming is not my first priority

 

[Roland Of Gilead]

10min in Prime95 hit me at around 83-86c
I turned everything down to stock speed with cpu offset 0.18 ( ASUS motherboard use 1.46v on CPU which is waaay too high )

Now I am at 35-45c idle and 70-73c on prime95 test. It boosted my core to 4.2 on cinebench and 4150 on prime95.

And according to ryzen master, I am using peak core voltage of 1.49v which is weird. it shows 1.30v in cpu-z

and 4.6ghz boost even at single core is non-existent.

Thats where you'd expect your temps to be.

If you allowed prime to run longer, the temps will continue to creep for about 25-35 minutes after initialising Prime.

 

[CountMike]

I read something about bios version currently I am at chipset 1.0.0.3ABBA which capped my boost at 4225 ghz. ( no idea why, maybe someone can explain me on this )
And gaming is not my first priority

You would have to activate/set PBO in BIOS.

 

[Chrexolia112]

You would have to activate/set PBO in BIOS.

I already activated and deactivated. It didn't do anything. it still boost at 4.2ghz

 

[Chrexolia112]

Thats where you'd expect your temps to be.

If you allowed prime to run longer, the temps will continue to creep for about 25-35 minutes after initialising Prime.

It's still weird tho, Ryzen Master shows idling temp around 45-55c but as soon as I restart it , it drops to 32c and will continue to goes up after several minutes. Also the peak voltage is 1.46v but shows in cpuz as 1.297-1.304v.

 

[Roland Of Gilead]

Yeah peak/boost voltage will be higher. It's not unusual at all to see spikes of 1.4+ this is normal.
It spikes mostly for a split second and drops back.