Should you intentionally adjust it to reach 95c?
Should you intentionally adjust it to reach 95c?
I have some concerns about my 7950x after getting it. I know it's meant to run at 95c under load, but my temperatures stay between 70-80c when it's running hard. The fans start to speed up and get louder in that range. Should I set the fans to turn on only when it reaches about 90-95c? Would that change how long it lasts? Also, I noticed the ECO mode on 105w only cuts performance by about 5%. Is it worth changing that?
The expected lifespan at 95°C is likely to decrease, yet no one can determine the exact improvement or whether a cooler temperature would be more advantageous for higher frequencies during operation
No, keep the fans as they are. I've adjusted my NH-D15 fans to increase speed at 60C on my 7950X (the noise isn't an issue).
During long video renders with Topaz Video AI, all cores stayed near 90C, with a few reaching 95C. The exact temperature varies based on the chosen algorithm.
If you wish to challenge your CPU, consider using a Handbrake H265 conversion or compressing a large folder with WinZIP/WinRAR (several GB). These tasks simulate heavy loads and can stress your CPU, or you could perform a stress test to see if all cores maintain 95C.
if you examine the maker's design, the expected lifespan ends when the warranty expires
😉
the highest temperature is 95°C, beyond which it will reduce its performance
other parts such as the motherboard, GPU, or M.2 will also heat up, so keeping them cooler is advisable
I frequently collaborate with Topaz Video AI and sometimes Adobe Premiere Pro. During rendering tasks, these applications can significantly increase the load on both the CPU and GPU, bringing their usage to 100%.
For the 7950X model, the standard TDP is listed at 170W, and thermal throttling begins when temperatures reach 95°C. Likewise, if your application demands additional GPU power during rendering, the GPU may also operate at full capacity, affecting its own temperature regulation.
You achieve optimal performance when the CPU and GPU are pushed to their thermal limits. On the 7950X, this occurs when all cores reach 95°C. For an Intel 14900K, the threshold is at 100°C.
To extract more processing power from a modern CPU, you must enable higher boost settings while maintaining the same throttling level. Contemporary CPUs are engineered to push their limits until they hit their thermal cap, which is when you get the best results. To boost performance further or reduce render times slightly, a more efficient cooling solution is necessary to remove excess heat.
On a large air cooler setup, the 7950X can manage around 170W at 4.5GHz across all cores. With a big 420mm triple fan AIO or custom loop, the same unit might handle up to 200W at 4.65GHz all cores. These numbers are approximate, but they help convey the situation. As cooling improves, the maximum boost frequency also rises slightly.
In my experience, I ran 36-hour long renders with the 7950X operating between 90 and 95°C, drawing about 170W using a Noctua NH-D15 cooler. At the same time, my RTX 3060 ran at full power and 170W. I assembled the system in 2022 and am comfortable using its full potential. It’s a powerful tool, and I see no reason to limit its performance by restricting power output for better temperatures. Whether it lasts another decade or not is a matter of personal choice. By 2027, I plan to upgrade to an even faster configuration.
I estimated that using an RTX 4090 would cut render times by roughly one-third—from 36 hours to about 12 hours compared to my RTX 3060.
Under similar conditions, a cooler setup is preferable.
I've performed numerous Handbrake encodes on 3950X, 5950X and now 7950X.
Occasionally, I faced queues where my system maintained over 90% CPU usage for weeks.
My goal is to keep temperatures below 80°C. As mentioned, prolonged high heat can raise the temps of the motherboard, VRM and other components. Heat combined with power/voltage damage electronics over time.
For my Handbrake CPU encoding, especially at 4K, an all-core overclock with under-voltage gives solid results and usually stays under 75°C. My current setting is 4.85 GHz all-core for the 75°C aim. Going to 5.2 GHz would push me into the 85-90°C range.
I feel more at ease running the system unattended for extended periods at 75°C, accepting a slight performance drop compared to pushing it to 90-95°C.
By the way, I use a 420mm AIO cooler and this rig is located in my basement utility room, which is relatively cool.
The CPU doesn't improve simply because it hits its maximum temperature. In fact, it tends to reduce performance to avoid exceeding that limit. The goal is to maintain some margin, ensuring you never hit the throttle. If your system makes a lot of noise, consider using a 360mm AIO and adjusting the fan curve based on liquid temperature.
This suggests that when the CPU begins to slow down, the BIOS lowers its speed considerably, leading to much longer rendering times. It seems to me that throttling serves as a finer control, preventing the CPU from surpassing its rated limits—such as 90/95°C for AMD or 100°C for Intel. This approach keeps the processor temperature stable at those safe levels.
The outcome appears to be optimal performance from a system standpoint. That's why contemporary CPUs tend to increase their speed automatically during demanding tasks, provided cooling is sufficient. The quicker the cooling system responds, the more effectively the CPU can ramp up.
If I replaced my NH-D15 with a 420mm AIO, I'd anticipate slightly reduced rendering times, thanks to enhanced cooling at the same throttling level.
I compare CPU throttling to a vehicle's rev limiter. When you push the throttle fully, the engine reaches its maximum speed but stays below it. Similarly, CPU throttling acts as a protective measure to avoid exceeding manufacturer-set limits.
If concerned about electromigration, this piece of information could ease your concerns (or confirm them):
https://www.anandtech.com/show/15839/ele...l-your-cpu
As processor designers and semiconductor producers refine logic gates in CMOS and FinFET chips, they've implemented strategies to minimize electromigration effects. With smaller process nodes and lower voltages, the problem diminishes—though wires shrinking also has a counteracting impact. Manufacturers are now actively working to mitigate electromigration risks within processors.
Electromigration hasn't been a concern for most consumer electronics for a long time. The only instances I've experienced it were with a Sandy Bridge-based 2011 Core i7-2600K I used for overclocking at 5.1 GHz under harsh cooling conditions. Eventually, it required higher voltage to run normally after a few years.
Note: This Anandtech article focuses on standard CPU throttling up to the safe temperature threshold, not on pushing CPUs beyond their limits for extra performance. I believe an overclocked CPU will still throttle at the same temperature as a non-overclocked one.
If you plan to keep your CPU until 2045, consider capping the temperature at 75°C for added confidence, but be aware you might sacrifice rendering speed or achieve higher frame rates.
I don't use PBO on my 7950X and 3800X, but I do allow them up to 95°C. I've pushed two i7-4770K and a G3258 Pentium to their limits, along with an Athlon 955 and 965 from 2006. My older CPUs—pre-2000 models—still function, though motherboards often fail first due to burnout caps.
Ultimately, the choice of how you manage your computer is personal, and it may stay that way for a long time.