[SOLVED] Safe temp testing without damage risk?

[dirk101]

I have a connundrum regarding the general PC stress testing process I've found among PC building enthusiasts.

I've just built my first true gaming system, and before I tie everything down with zip ties I wanted to check if it gets enough air cooling for my purposes. It's the first time I've considered doing this with a PC I've built. The trouble is that everything I've read about stress testing in the past week generally points to Prime95 as the best CPU test and FurMark as the best GPU test, but this is only if you're willing to take the risk of component damage.

I'm not keen on component damage. The alternative to using artificial workloads would be to actually use the PC for a long enough time in real-world scenarios. The PC is aimed at 1440p gaming at max settings, and at photo editing and graphics design with very large projects. The trouble with this is that I currently lack the time required to sit down and do these activities for long enough periods of time. There's the option of waiting until I can make time, but knowing I have an "incomplete" PC sitting on my desk, and seeing it every day, has birthed an irritated little gnome in the back of my head that just wants to know the build is finished and trouble free and cable managed and had the foil unwrapped off the glass panel.

My questions now is, is there a stress testing guide out there aimed at the non-enthusiast user, that has no interesest in burning their components to the ragged edge? I'm just looking to answer "Is it aircooled enough?". Could I, for example, achieve this by just running Prime95 and FurMark for 5 minutes each, with a twitchy finger on the stop button and bulging eyes on HWMonitor?

PS Are idle temps for a system representative of cooling potential?

PPS It's also my first time monitoring temperature. If I leave the PC running for an hour while I work around it, with HWMonitor on, whenever I look at it the CPU temps unde "Value" are constantly at around 33-35 C, yet the Max temps somehow record temp spikes when I'm not looking between 59-62 C, even though the acoustics never change, i.e. I never hear the CPU fan revving up for higher temperatures. Is this sort of temp spiking normal during idle?

PPPS After a heck of a lot of searching, the most thorough articles on the subject have actually been from the Tom's Hardware team, with How to Test Your Graphics Card and How to Stress-Test CPUs and PCs (Like We Do). If anyone from the team's reading this, many thanks for that.


System (no OC apart from the manufacturers') -----------------------
CPU: Intel Core i5-11600K 3.9 GHz
CPU Cooler: Noctua NH-U12S chromax.black
Motherboard: Gigabyte Z590 UD AC
Memory: Corsair Vengeance LPX 32 GB (2 x 16 GB) DDR4-3000 CL16
Video Card: Gigabyte GeForce RTX 3070 8 GB GAMING OC
Power Supply: Corsair HX Platinum 850 W 80+
Case: Corsair 4000D Airflow (front intake: 2x be quiet! Shadow Wings 3 120mm + 1 stock Corsair 120mm case fan, back exhaust: 1 stock Corsair 120mm case fan)

 

[Vic 40]

You can set Cinebench R23 to loop so can use that for temps without having to resort to prime95. Think myself i would put that Corsair fan at the front to the top, but that is me, maybe something to play with.
For gpu can you use some of the benchmark tools like the Unigine benchmarks.,

 

[mikewinddale]

Assuming you aren't overclocking, and assuming that you have a reasonable cooling setup, with everything installed correctly, the risk of damage is low.

Those softwares warn you of component damage because of (1) legal liability (covering their ass) and (2) some people are overclocking, which CAN kill a system. But if you aren't overclocking, and if you've installed everything correctly, your system should stand up to every stress test with no problem.

Still, what you can do is work your way up from less-stressful to more-stressful tests.

Cinebench is a good place to start because it is a real-world benchmark that doesn't stress your CPU in an artificial way. Under File, choose Advanced Settings, and set a minimum amount of time to run.

Monitor your temperatures and fan RPMs. I like to use HWiNFO64.

Also, look at your CPU's current frequency, which you can see in HWiNFO64 and also the Windows Task Manager. If your CPU frequency is above base clock, then you know you're safe. The CPU will boost above base as long as it has headroom. Better cooling means more headroom and more boosting, but any boosting at all means your cooling is at least minimally adequate.

Next, run 3DMark.

If your system passes those, then you can start ramping it up.

In Prime95, you can set the number of cores to use. Start with a low number of cores and see if your temperatures remain safe. If they are, start increasing the number of cores. Go as gradually as makes you feel safe. You can even start with just one core.

If you pass Prime95, the proceed to MSI Kombustor, which a nice GPU burn-in. Try running Prime95 and Kombustor at the same time to generate maximum heat. Prime95 will use all your CPU and make your system sluggish and somewhat unresponsive, so start Kombustor first, then start Prime95.

Finally, regarding your fans: check your motherboard manual for its fan configuration. You may have to adjust a fan curve profile in the BIOS. I just glanced briefly at your motherboard's manual, and indeed, there's a section called "Smart Fan 6". It looks like the default is to adjust fans to suit the temperature. But go into the BIOS to be sure. Make sure every fan uses the correct temperature sensor. For example, on my Gigabyte board, some fans were set to adjust themselves by the motherboard chipset temperature, so I changed that to CPU temperature, since my CPU gets hotter than my motherboard.

Also, in the BIOS fan control, you can set all fans to run at 100% max speed, all the time, regardless of temperature. Maybe turn that in during stage 1 of your stress-testing. If you pass, then turn automatic fan-adjustment back on (which Gigabyte calls "Normal" mode).

 

[dirk101]

Thank you both, I've taken your suggestions, together with what I've managed to find on web articles, and created myself a "safety first" list of bechmarks to go through in order. I haven't included a combined stress test that covers CPU and GPU at the same time because if temps are fine after going through the list below, that's good enough for me.

CPU

• Cinebench R23
• PCMark 10
• Prime95 Small FFTs
  • AVX disabled 1 thread
  • AVX disabled 50% threads
  • AVX disabled all threads
  • AVX enabled 1 thread
  • AVX enabled 50% threads
  • AVX enabled all threads

GPU

• 3DMark, Time Spy and Fire Strike
• Unigine Superposition
• Red Dead Redemption 2 in-game benchmark
• Metro Exodus in-game benchmark
• FurMark

Think myself i would put that Corsair fan at the front to the top, but that is me, maybe something to play with.

Testing has shown me that my GPU reaches 71 C at most, which is great given that ambient temperature can also get hot in the summer now where I live. But The CPU seems to reach 80 fairly easily, and it passes it without actually pushing the CPU with all of the tests. To try and improve this, I'll see if a top of the line thermal paste help bring it down by a few degrees, and I'll move the 2x Shadow Wings 2 to the back of the case to get hot air away from the CPU, while replacing the front with 2x Silent Wings 3 140.

Finally, regarding your fans: check your motherboard manual for its fan configuration. You may have to adjust a fan curve profile in the BIOS. I just glanced briefly at your motherboard's manual, and indeed, there's a section called "Smart Fan 6". It looks like the default is to adjust fans to suit the temperature. But go into the BIOS to be sure. Make sure every fan uses the correct temperature sensor. For example, on my Gigabyte board, some fans were set to adjust themselves by the motherboard chipset temperature, so I changed that to CPU temperature, since my CPU gets hotter than my motherboard.

Also, in the BIOS fan control, you can set all fans to run at 100% max speed, all the time, regardless of temperature. Maybe turn that in during stage 1 of your stress-testing. If you pass, then turn automatic fan-adjustment back on (which Gigabyte calls "Normal" mode).

I've also changed the fans to work off of the CPU, to ensure it gets as much cool air as possible, given the temps I'm seeing. Your BIOS fan speed suggestion actually helped me isolate and figure out which fan was making a lot of noise during full load, turns out it's the NH-U12S.

 

[Vic 40]

I don't see much use in Furmark which is quite unreal in temps, unless you really want to see max temps of the gpu and if it can handle those, think ingame benchmarks and the others will do fine.

For the cpu temps can you try and add a second fan to the cooler, see if this wil help a bit, but don't expect dramatic changes from it.

 

[mikewinddale]

For the cpu temps can you try and add a second fan to the cooler, see if this wil help a bit, but don't expect dramatic changes from it.

I second this suggestion. Note what Noctua says on their page:

Anti-vibration pads and fan-clips for second NF-F12 (optional)
For users who want to achieve even better performance by adding a second, optional NF-F12 fan to create a push/pull setup, the NH-U12S includes fan clips and an extra set of custom-designed anti-vibration pads that allow to off-set the rear fan by 5mm to improve acoustics in dual-fan mode.

and I'll move the 2x Shadow Wings 2 to the back of the case to get hot air away from the CPU, while replacing the front with 2x Silent Wings 3 140.

This YouTube channel has a series of videos testing different fan configurations, with each video being built around a different CPU cooler configuration. Here is a screenshot from the video about your CPU configuration. I think that fan configuration #2 makes the most sense. It offers great cooling and have more intake than exhaust, which will help reduce dust. (By having more intake than exhaust, you can choose to intake air only where there is a filter. Any excess air pressure will be exhausted through any vents in the case. By contrast, if you have more exhaust than intake, then the vacuum will draw air in through any vents, which might not have filters.)

https://imgur.com/a/VnViqXP

 

[dirk101]

I don't see much use in Furmark which is quite unreal in temps, unless you really want to see max temps of the gpu and if it can handle those, think ingame benchmarks and the others will do fine.

If temps behave fine up to FurMark, I'll use FurMark just to get a complete picture of how good my aircooling is, and to compare with other stress test temp results from case and GPU reviewers.

For the cpu temps can you try and add a second fan to the cooler, see if this wil help a bit, but don't expect dramatic changes from it.

I already found other U12S users who tried this and reported either no or negligible improvement. I'm assuming it's because the heatsink is fairly thin and the existing NF-F12 has enough performance to aircool all it can.

I second this suggestion. Note what Noctua says on their page:

I appreciate the heads up, I noticed this in the U12S manual while installing.

I think that fan configuration #2 makes the most sense.

Since I don't have a hub controller, I'll only use the 4 fans I mentioned, the 2 be quiet 120s in the back to get air out and 2 be quiet 140s in the front to get air in. I looked at the manufacturer's technical data between different models before deciding on changing fans, specifically to ensure I'll keep positive pressure in the case.

 

[mikewinddale]

I like having a second fan on my CPU cooler because it provides a little extra insurance in case one fan suddenly fails for some reason. It's not likely, but it's just a little bit of insurance.

And since they're Noctua PWM fans, the second fan will usually be quiet and spin at low RPM anyway, so it won't add much noise.