Expected Temps for Ryzen 3950x and Threadripper 3960x Air Cooling Only

[dpriest]

I am thinking about upgrading my Threadripper 1950x build to either the 3950x or TR 3960x. I am currently cooling my 1950x very effectively with the Thermalright Silver Arrow TR4 with a 3.7GHz overclock, pushing all cores with Handbrake using very high settings. I'm not interested in water cooling at all. According to AMD the max safe temps for all Ryzen 3000 processors is 95C instead of 68C for 1st and 2nd gen processors. I'm looking for feedback from those who are air cooling these processors stock settings and overclocking if possible. I play it safe with temps. With that said, I am only interested in temps within the recommended safe range under very heavy cpu load with applications such as Handbrake. Thanks for your feedback in advance.

 

[doyll]

Subbed, interested in hearing owners' cooling feedback.

 

[Dan_D]

I am thinking about upgrading my Threadripper 1950x build to either the 3950x or TR 3960x. I am currently cooling my 1950x very effectively with the Thermalright Silver Arrow TR4 with a 3.7GHz overclock, pushing all cores with Handbrake using very high settings. I'm not interested in water cooling at all. According to AMD the max safe temps for all Ryzen 3000 processors is 95C instead of 68C for 1st and 2nd gen processors. I'm looking for feedback from those who are air cooling these processors stock settings and overclocking if possible. I play it safe with temps. With that said, I am only interested in temps within the recommended safe range under very heavy cpu load with applications such as Handbrake. Thanks for your feedback in advance.

Let's get something straight. Maximum safe temperature for Ryzen CPU's was never 68c. I know what it says on the website, but this figure is misunderstood. This temperature is the target CPU temperature threshold for the CPU to achieve its maximum boost clocks under load. At default "stock" settings, the CPU's won't go above 68c on their own. You have to enable auto-overclocking or PBO, PBO+AutoOC, or manually overclock them to override this limit. Once you do any of that, the CPU will get up to 95c depending on your cooling and what your trying to do with it. Simply enabling AutoOC will remove the 68c limit and you'll see the CPU hold its boost clocks for longer and they'll potentially spike higher.

I've run and own or owned every single generation of Ryzen CPU and some Threadripper CPU's as well. I've reviewed a bunch of them and I've tested them extensively. I've even run some of them on air cooling from time to time. The second you enable PBO, AutoOC, or manually overclock them that 68c figure, like the CPU's TDP goes out the window. It's also perfectly safe to run these CPU's at higher speeds and temps. I've got a 2700X that's been running at 4.2GHz on an AIO 24/7 for about a year now. Under load it's well past 68c and doing fine.

You aren't seeing temperatures greater than 68c on your Threadripper because you aren't pushing it. In fact, if you are manually setting it to 3.7GHz you are probably leaving performance on the table. Just let the thing boost on its own. It should be faster at default settings and it will still stay at 68c. Kyle and I both have tested these things quite a bit and often times, there was no benefit to traditional overclocking on them. This has been the case more and more with each generation.

 

[dpriest]

Let's get something straight. Maximum safe temperature for Ryzen CPU's was never 68c. I know what it says on the website, but this figure is misunderstood. This temperature is the target CPU temperature threshold for the CPU to achieve its maximum boost clocks under load. At default "stock" settings, the CPU's won't go above 68c on their own. You have to enable auto-overclocking or PBO, PBO+AutoOC, or manually overclock them to override this limit. Once you do any of that, the CPU will get up to 95c depending on your cooling and what your trying to do with it. Simply enabling AutoOC will remove the 68c limit and you'll see the CPU hold its boost clocks for longer and they'll potentially spike higher.

I've run and own or owned every single generation of Ryzen CPU and some Threadripper CPU's as well. I've reviewed a bunch of them and I've tested them extensively. I've even run some of them on air cooling from time to time. The second you enable PBO, AutoOC, or manually overclock them that 68c figure, like the CPU's TDP goes out the window. It's also perfectly safe to run these CPU's at higher speeds and temps. I've got a 2700X that's been running at 4.2GHz on an AIO 24/7 for about a year now. Under load it's well past 68c and doing fine.

You aren't seeing temperatures greater than 68c on your Threadripper because you aren't pushing it. In fact, if you are manually setting it to 3.7GHz you are probably leaving performance on the table. Just let the thing boost on its own. It should be faster at default settings and it will still stay at 68c. Kyle and I both have tested these things quite a bit and often times, there was no benefit to traditional overclocking on them. This has been the case more and more with each generation.

I appreciate your feedback. I am manually overclocking the 1950X and when the ambient room temps are warm the processor will touch 72C. I can overclock it to 3.9GHz and with the volts required to make it stable I will probably being seeing temps around 78C. If that's safe to run for hours and hours straight, I will do it. I actually have not done a controlled test with Vidcoder. I am going to encode a video at stock settings and see what the avg. FPS is and then overclock it to 3.7 & 3.9GHz and see how much of an improvement there is. I never enabled PBO or PBO+AutoOC because I wasn't sure how to do it and just did it manually instead. Since you and Kyle have tested these extensively, what do you feel the max. safe temp is for this processor to run for hours and even around the clock for a couple of days at a time?

 

[Dan_D]

I appreciate your feedback. I am manually overclocking the 1950X and when the ambient room temps are warm the processor will touch 72C. I can overclock it to 3.9GHz and with the volts required to make it stable I will probably being seeing temps around 78C. If that's safe to run for hours and hours straight, I will do it. I actually have not done a controlled test with Vidcoder. I am going to encode a video at stock settings and see what the avg. FPS is and then overclock it to 3.7 & 3.9GHz and see how much of an improvement there is. I never enabled PBO or PBO+AutoOC because I wasn't sure how to do it and just did it manually instead. Since you and Kyle have tested these extensively, what do you feel the max. safe temp is for this processor to run for hours and even around the clock for a couple of days at a time?

It is perfectly safe to do. As I said, the 68c figure is just what the CPU targets by default for it's boost clock behavior. When you throw AutoOC, PBO or manual overclocking into the mix, that number is meaningless. 68c would be a ridiculously low safe threshold for most modern semi-conductors. Threadripper and by extension all Ryzen CPU's can handle at least 95c. They'll throttle after that which prevents them from being damaged. It wouldn't make sense to put your safe zone almost 30 degrees below your thermal throttling threshold. In fact, the safe temperature limits for the CPU are likely above 100C. There is buffer there as thermal throttling won't necessarily reduce the CPU's actual temperature before the CPU crosses the 95c line.

In other words, at 95c throttling gets triggered, but there is a small delay before that takes effect and your temperatures could spike beyond 95c before the throttling can successfully reduce temps This could occur due to your CPU fan falling off or some type of catastrophic failure. If throttling cannot reduce temperatures quickly enough, the CPU's actually shut themselves down or hard lock to prevent damage. It's been quite safe to push these things hard since the Athlon 64 days and earlier in the case of Intel CPU's. About the only way you can really damage one is to overvolt the crap out of it. Even then, you might have to make modifications to the board to go far enough to hurt the CPU. Most AM4 motherboards I've seen won't allow for more than 1.5v. You can disable most of the motherboards safeties, such as over current protection, but you can't actually disable the CPU's internal ones.

 

[Mav451]

Heh - AMD's come a long way since the pre-OCP days - i.e. see the Thunderbird roasting on the open test bed (old Tom's HW video).

 

[LurkerLito]

I keep my 1950x at 3.7ghz because I can hit 85 at load during the summer. It's my psychological temp throttle point and I find it is quite difficult to over come it . I hit 90c for a time during the winter after I got my 1950x with a new Enermax AIO at load when I OCed to 1.35v @ 4 ghz. It was nice but I couldn't stop worrying or staring at my temp readings the whole time so I dropped down to 3.8 to keep my machine usable with out my psychological throttle point stopping the process because it's a $1k CPU so don't break it lol. I finally settled for 3.7 so I wouldn't have to switch between a winter setting and summer setting .

 

[doyll]

Dan_D is the expert here, but I perfer keeping temps well below throttle. I may be wrong, and I'm not sure if it's chip, or motherboard components that eventually fail and take other things out, but the harder we push them the more likely their life will be shortened. I still have an old i7 920 @ 4.2Ghz still working, but it never goes above 75c and rarely above 68c. I keep using it just to see how long it will last.

 

[Dan_D]

Heh - AMD's come a long way since the pre-OCP days - i.e. see the Thunderbird roasting on the open test bed (old Tom's HW video).

Who could forget that video? That's exactly what I was thinking of. By the time of the Athlon 64 (S754) that problem was a thing of the past.

I keep my 1950x at 3.7ghz because I can hit 85 at load during the summer. It's my psychological temp throttle point and I find it is quite difficult to over come it . I hit 90c for a time during the winter after I got my 1950x with a new Enermax AIO at load when I OCed to 1.35v @ 4 ghz. It was nice but I couldn't stop worrying or staring at my temp readings the whole time so I dropped down to 3.8 to keep my machine usable with out my psychological throttle point stopping the process because it's a $1k CPU so don't break it lol. I finally settled for 3.7 so I wouldn't have to switch between a winter setting and summer setting .

I understand wanting to stay within your personal comfort zone, but frankly, this is a technical matter and there is no reason not to let the chip perform as it was intended. It's highly probable that you guys running at 3.7GHz fixed are losing performance. You are literally gimping your CPU's and preventing them from performing as well as they could. In all the testing we've done with Ryzen and Threadripper specifically, we found that letting them boost on their own to 4.0GHz will yield the fastest results 9 times out of 10. The CPU will basically run balls out to 4.0GHz if it can run cool enough. When you let them run on Precision Boost, they'll increase their clocks and then lower them to hit the target range. This means that your CPU will fluctuate between it's 3.4GHz base clock and 4.0GHz boost clock. It will probably be around 3.8GHz to 3.9GHz most of the time. Basically, let it do it's thing. Seriously, just try it. You can set it back if it doesn't smash your 3.7GHz manual settings in any benchmarks of your choosing. I bet it will.

Again, you won't hurt the thing so long as you don't overvolt it. Voltage is what kills CPU's. Heat doesn't as there are way too many safeties between the motherboard and CPU to allow that to happen.

Dan_D is the expert here, but I perfer keeping temps well below throttle. I may be wrong, and I'm not sure if it's chip, or motherboard components that eventually fail and take other things out, but the harder we push them the more likely their life will be shortened. I still have an old i7 920 @ 4.2Ghz still working, but it never goes above 75c and rarely above 68c. I keep using it just to see how long it will last.

I prefer keeping my temps as low as possible. With boost clocking, the lower you can keep them the faster your CPU will run. It's that simple. This has been the case since turbo frequencies became a thing. With Intel's you can still clock them to their max boost clock for all cores, which is a nice boost, but with the AMD's you should really just let them do their thing. You can add auto-OC to PB2, or PBO, PBO+AutoOC into the mix and get even more performance. Again, the CPU will not let itself get hot enough to damage it and if you leave the voltage settings alone, it won't ever overvolt itself. Running like that, the system will probably outlast it's useful operational lifespan.

As for pushing things and shortening their life, it's only true when you push too hard. These things can be run pretty hard without dying on you. I've had CPU's overclocked every day of their service life for 5-7 years and I've never killed a board or CPU doing that. To date, I've only killed two CPU's to overclocking that I can recall. A Pentium III 600MHz and a Core i7 6950X. The former simply degraded until it had to run below stock speeds. The latter was known for degrading when overclocked, even on what seemed like reasonable voltage at the time. Beyond specific cases like that, I wouldn't worry about it. I've overclocked probably more than 100 CPU's, many of which were either abused on the test bench with little regard for longevity or left in systems for years at a time with little regard for longevity.

I rarely pay for CPU's. I've run the damn things hard because I could either afford to replace them or because I could get another one for free and never worried about it. I've still rarely killed any abusing them. You aren't going to hurt your 1950X by letting it govern its own voltages and clock speeds. You also wouldn't hurt it running it at 4.1GHz if you can manage to keep it from throttling.

 

[dpriest]

Who could forget that video? That's exactly what I was thinking of. By the time of the Athlon 64 (S754) that problem was a thing of the past.



I understand wanting to stay within your personal comfort zone, but frankly, this is a technical matter and there is no reason not to let the chip perform as it was intended. It's highly probable that you guys running at 3.7GHz fixed are losing performance. You are literally gimping your CPU's and preventing them from performing as well as they could. In all the testing we've done with Ryzen and Threadripper specifically, we found that letting them boost on their own to 4.0GHz will yield the fastest results 9 times out of 10. The CPU will basically run balls out to 4.0GHz if it can run cool enough. When you let them run on Precision Boost, they'll increase their clocks and then lower them to hit the target range. This means that your CPU will fluctuate between it's 3.4GHz base clock and 4.0GHz boost clock. It will probably be around 3.8GHz to 3.9GHz most of the time. Basically, let it do it's thing. Seriously, just try it. You can set it back if it doesn't smash your 3.7GHz manual settings in any benchmarks of your choosing. I bet it will.

Again, you won't hurt the thing so long as you don't overvolt it. Voltage is what kills CPU's. Heat doesn't as there are way too many safeties between the motherboard and CPU to allow that to happen.



I prefer keeping my temps as low as possible. With boost clocking, the lower you can keep them the faster your CPU will run. It's that simple. This has been the case since turbo frequencies became a thing. With Intel's you can still clock them to their max boost clock for all cores, which is a nice boost, but with the AMD's you should really just let them do their thing. You can add auto-OC to PB2, or PBO, PBO+AutoOC into the mix and get even more performance. Again, the CPU will not let itself get hot enough to damage it and if you leave the voltage settings alone, it won't ever overvolt itself. Running like that, the system will probably outlast it's useful operational lifespan.

As for pushing things and shortening their life, it's only true when you push too hard. These things can be run pretty hard without dying on you. I've had CPU's overclocked every day of their service life for 5-7 years and I've never killed a board or CPU doing that. To date, I've only killed two CPU's to overclocking that I can recall. A Pentium III 600MHz and a Core i7 6950X. The former simply degraded until it had to run below stock speeds. The latter was known for degrading when overclocked, even on what seemed like reasonable voltage at the time. Beyond specific cases like that, I wouldn't worry about it. I've overclocked probably more than 100 CPU's, many of which were either abused on the test bench with little regard for longevity or left in systems for years at a time with little regard for longevity.

I rarely pay for CPU's. I've run the damn things hard because I could either afford to replace them or because I could get another one for free and never worried about it. I've still rarely killed any abusing them. You aren't going to hurt your 1950X by letting it govern its own voltages and clock speeds. You also wouldn't hurt it running it at 4.1GHz if you can manage to keep it from throttling.

This is very informative Dan_D. After looking into this further, my understanding with the TR 1950X and the 1000 series Ryzen chips is that manual overclocking is more effective because Precision Boost is an inherent feature at stock settings and does not push the cpu the way Precision Boost Overdrive does. When yoy mention Auto OC, do you mean leaving the overclock settings at the default where it says AUTO under Overclock mode? Aren't you talking about the 2nd and 3rd gen Ryzen/Theadrippper chips with PBO and PBO 2 since this didn't exist with the 1000 series chips? From what I understand, it doesn't make sense to manually overclock the 3000 series chips because it is more efficient to let PBO do its thing. Check this out:

Since I have the 1950x I will adjust voltages and frequencies, including stock settings to see what real performance gain I will notice in Handbrake and I will measure this with Avg. FPS. Right now, my 3.7GHz is adjusting voltage at 1.24 (1.23750). Again my temps average 62C and will creep up to 66C and during warm days in the room will creep just over 70C. To get a stable 4GHz overclock I believe I need to up my voltage to 1.3. This will up the temps in the high 70s or low to mid 80s. From my understanding, you are saying voltage and not temps is what shortens the life of the processor, motherboard, I should be fine with 1.3-1.35 volts.

Here is what I am also adjusting in the BIOS settings:

XMP Profile: XMP 2.0 Profile 1
CPU Load-Line Calibration: Level 1
VDDCR_SOC Voltage: Fixed Mode (Fixed Voltage=1.1)
DRAM Voltage: 1.45
DRAM_CD Voltage: 1.45
VDDCR_SOC_S5: Changed from Auto To 1.1

Any feedback, recommendations based on what I am doing is greatly appreciated.

 

[doyll]

I prefer keeping my temps as low as possible. With boost clocking, the lower you can keep them the faster your CPU will run. It's that simple. This has been the case since turbo frequencies became a thing. With Intel's you can still clock them to their max boost clock for all cores, which is a nice boost, but with the AMD's you should really just let them do their thing. You can add auto-OC to PB2, or PBO, PBO+AutoOC into the mix and get even more performance. Again, the CPU will not let itself get hot enough to damage it and if you leave the voltage settings alone, it won't ever overvolt itself. Running like that, the system will probably outlast it's useful operational lifespan.

As for pushing things and shortening their life, it's only true when you push too hard. These things can be run pretty hard without dying on you. I've had CPU's overclocked every day of their service life for 5-7 years and I've never killed a board or CPU doing that. To date, I've only killed two CPU's to overclocking that I can recall. A Pentium III 600MHz and a Core i7 6950X. The former simply degraded until it had to run below stock speeds. The latter was known for degrading when overclocked, even on what seemed like reasonable voltage at the time. Beyond specific cases like that, I wouldn't worry about it. I've overclocked probably more than 100 CPU's, many of which were either abused on the test bench with little regard for longevity or left in systems for years at a time with little regard for longevity.

I rarely pay for CPU's. I've run the damn things hard because I could either afford to replace them or because I could get another one for free and never worried about it. I've still rarely killed any abusing them. You aren't going to hurt your 1950X by letting it govern its own voltages and clock speeds. You also wouldn't hurt it running it at 4.1GHz if you can manage to keep it from throttling.

Indeed, keeping temps down definitely improves performance. I think it's more true on new AMD than ever before.

Indeed, reasonable voltage is key to life and performance. Keep temps safely below throttle point so chip can use it's full potential without thottling.

I'm sure having free chips (assume mobo & RAM too) makes a difference. I get plenty of cases and coolers and a fiar few fans, but not much in the way of electronics.

 

[Dan_D]

This is very informative Dan_D. After looking into this further, my understanding with the TR 1950X and the 1000 series Ryzen chips is that manual overclocking is more effective because Precision Boost is an inherent feature at stock settings and does not push the cpu the way Precision Boost Overdrive does. When yoy mention Auto OC, do you mean leaving the overclock settings at the default where it says AUTO under Overclock mode? Aren't you talking about the 2nd and 3rd gen Ryzen/Theadrippper chips with PBO and PBO 2 since this didn't exist with the 1000 series chips? From what I understand, it doesn't make sense to manually overclock the 3000 series chips because it is more efficient to let PBO do its thing. Check this out:

Since I have the 1950x I will adjust voltages and frequencies, including stock settings to see what real performance gain I will notice in Handbrake and I will measure this with Avg. FPS. Right now, my 3.7GHz is adjusting voltage at 1.24 (1.23750). Again my temps average 62C and will creep up to 66C and during warm days in the room will creep just over 70C. To get a stable 4GHz overclock I believe I need to up my voltage to 1.3. This will up the temps in the high 70s or low to mid 80s. From my understanding, you are saying voltage and not temps is what shortens the life of the processor, motherboard, I should be fine with 1.3-1.35 volts.

Here is what I am also adjusting in the BIOS settings:

XMP Profile: XMP 2.0 Profile 1
CPU Load-Line Calibration: Level 1
VDDCR_SOC Voltage: Fixed Mode (Fixed Voltage=1.1)
DRAM Voltage: 1.45
DRAM_CD Voltage: 1.45
VDDCR_SOC_S5: Changed from Auto To 1.1

Any feedback, recommendations based on what I am doing is greatly appreciated.

I am generally speaking about the 2000 and 3000 series Ryzen's and Threadripper CPU's. With the first gens, you would still likely see better performance letting them boost on their own versus manually clocking them super low at 3.7GHz. This would be especially true in games and anything that isn't heavily multi-threaded. With the 1920X and 1950X's, we generally locked them in at 4.1GHz. It was the last more or less traditional overclocking we got from AMD. With the 2000 series, PB2, PBO, and automatic overclocking largely worked the same as it does with the 3000 series. The main difference being that PBO is pretty much worthless for the 3000 series and sometimes retards performance. With the 2000 series, PBO is what you wanted.

When I refer to auto-OC, I'm refering to automatic overclocking on top of PB2 or PBO. Although, this more applies to the 3000 series if I recall correctly. I haven't looked at the 2000 series in awhile. I got rid of my 2920X awhile back.

 

[dpriest]

OK. Did some testing and want to make sure the cost/benefit is worth it as I am trying to squeeze every ounce of performance out of the Threadripper 1950X. Would love thoughts/opinions on this. Using the same settings for Vidcoder here are the results for one video h.264 1080P High custom settings:

1.) At stock /default settings: Max temp = 46c Avg. FPS = 45.1 Encode Time = 1:00:22

2.) 4GHz Overclock @ 1.29375v: Max temp = 72c (Avg. temp: 68c - 70c) Avg. FPS = 49.9 Encode Time = 54:34

This was one video. More than likely with a Queue of several videos that could run for hours and perhaps 1-2 days, the temps will increase. What would be the maximum voltage and temps that the 1950X could run at without shortening its life?

 

[Dan_D]

OK. Did some testing and want to make sure the cost/benefit is worth it as I am trying to squeeze every ounce of performance out of the Threadripper 1950X. Would love thoughts/opinions on this. Using the same settings for Vidcoder here are the results for one video h.264 1080P High custom settings:

1.) At stock /default settings: Max temp = 46c Avg. FPS = 45.1 Encode Time = 1:00:22

2.) 4GHz Overclock @ 1.29375v: Max temp = 72c (Avg. temp: 68c - 70c) Avg. FPS = 49.9 Encode Time = 54:34

This was one video. More than likely with a Queue of several videos that could run for hours and perhaps 1-2 days, the temps will increase. What would be the maximum voltage and temps that the 1950X could run at without shortening its life?

Any temps below throttling are fine. 1.4v is perfectly safe for these CPUs.

 

[Hakaba]

Heh - AMD's come a long way since the pre-OCP days - i.e. see the Thunderbird roasting on the open test bed (old Tom's HW video).

Are we referring to the Intel vs. AMD? Where Intel chips survives because it slowed down/turned off and the AMD chip started puffing smoke 3 seconds after the heatsink was removed?

 

[Dan_D]

Are we referring to the Intel vs. AMD? Where Intel chips survives because it slowed down/turned off and the AMD chip started puffing smoke 3 seconds after the heatsink was removed?

Indeed. That's the one.