Increasing processor usage

[Stonecold]

What temperatures were you getting when it was running full-blast? I don't know what was giving you the warnings, but it may have been overly cautious in its definition of "too hot." The consensus I've seen online is that an i7 can run up to 75-80 degrees C constantly without being at much risk of heat-related damage, and I've had my i7-930 running in that range for over a year and a half without any issues. Everyone has their own definition of "uncomfortably hot," so ultimately it's your call as to what temperature's you're okay with, but if some program was telling you that 55C was too hot it was really erring on the side of caution.

I run my i7 at almost 90C 24/7, and it has not been damaged. The TjMax for my processor (2630QMQ) is 100C.

[Zagen30]

Stonecold: I forgot to clarify the desktop/laptop difference, which I have since edited into post. Yeah, 90C for a laptop i7 is not unusual.

[Stonecold]

Stonecold: I forgot to clarify the desktop/laptop difference, which I have since edited into post. Yeah, 90C for a laptop i7 is not unusual.

Good point. I knew that laptop CPUs have higher heat tolerance, but I never thought it was that significant. How much more heat can they handle?

[Jesse_V]

Stonecold: I forgot to clarify the desktop/laptop difference, which I have since edited into post. Yeah, 90C for a laptop i7 is not unusual.

Good point. I knew that laptop CPUs have higher heat tolerance, but I never thought it was that significant. How much more heat can they handle?

I've seen mine hit 94C. I made sure it didn't stay there for long...

[Stonecold]

Stonecold: I forgot to clarify the desktop/laptop difference, which I have since edited into post. Yeah, 90C for a laptop i7 is not unusual.

Good point. I knew that laptop CPUs have higher heat tolerance, but I never thought it was that significant. How much more heat can they handle?

I've seen mine hit 93C. I made sure it didn't stay there for long...

On a desktop?

[Jesse_V]

On a desktop?

Laptop.

[Stonecold]

On a desktop?

Laptop.

93C isn't too high for a laptop to be at briefly, is it?

[Jesse_V]

On a desktop?

Laptop.

93C isn't too high for a laptop to be at briefly, is it?

It's pretty high, and I really don't like those temperatures since its only 10F below my max CPU temperature.

[k1wi]

"Too high" is a pretty relative term..... Increased temperature generally speaking lowers the lifespan of the processor, though only above a certain temperature would the decrease in lifespan be noticeable and I believe the relationship is not linear - the impact of a one degree increase from 60-61 degrees is much different and less than say 99-100... Given that there are many other factors in the lifespan of a CPU, 'Damage' from high CPU temperatures may not be easily apparent, but rather result in the MTBF (mean time between failure) decreasing, increasing the likelihood of a failure earlier than a cooler CPU.

However, that said, rapid temperature fluctuation/cycling may have an even greater effect on lifespan than a high, constant temperature.

Do note that it is important to differentiate between core temperature and CPU temperature, as the former tends to be around 10 degrees hotter than the latter...

[jrpower]

I restarted with no luck of increasing CPU usage, i have even tried to increase the priority in the task manager but nothing.

it was running at 78 and and my motherboard was hitting 80s and its a new HAF 932 case with a Geforce gtx560 ti, a sabertooth p67 motherboard, an i7 and a few other things.. its a great computer it normally runs at about 35c with the motherboard around 40-50 depending on where you read from the sensor (motherboard comes with build in sensors)

[Stonecold]

I restarted with no luck of increasing CPU usage, i have even tried to increase the priority in the task manager but nothing.

You should never increase the priority of FAH. FAH will gladly take up all of your CPUs time and lock the whole system up, preventing other processes from running. FAH by default has the lowest priority possible so that any other programs will be able to run.

[Zagen30]

I restarted with no luck of increasing CPU usage, i have even tried to increase the priority in the task manager but nothing.

I feel like we've omitted some information that has caused some confusion, so I'll try to be thorough.

FAH has 3 basic kinds of work for PCs: uniprocessor, SMP (Symmetric Multi-Processor), and GPU. GPU work runs on supported GPUs, but we can ignore it for the purposes of this discussion. Uniprocessor, as the name implies, uses one and only one CPU core no matter what you do to it. SMP uses multiple cores to process one work unit, as opposed to running a separate WU on each core.

Version 6.23 is a uniprocessor-only client. As you saw when you were running it, it can only use so much CPU time since it's limited to a single thread. In the case of an i7, this will show up in task manager as roughly 13% usage; as has been mentioned, most i7s only have 4 physical cores but, due to HyperThreading, are presented to the OS as having 8 cores, which are often referred to as threads. Windows doesn't monitor the actual usage of each hardware module in the cores, so one of its 8 threads being used all the time will be noted as 12-13% usage based on minor fluctuations and rounding even though in reality roughly 25% of the total floating-point unit (the piece of hardware that FAH uses the most) processing power was being used (there's one FPU per physical core). I'm not entirely sure why you were seeing 8%, but it's still indicative of the fact that only one thread was being utilized.

Version 6.34 is capable of running in both SMP and uniprocessor work, though not at the same time. It's listed on the High Performance page because it may not be the best option for new contributors to start with; it takes some additional configuring to get it running in SMP mode, and SMP work has rather short deadlines, so the thought process has been that new donors should start with the uniprocessor client and later figure out if they want to contribute more. If you don't add the -smp flag to 6.34, it will run just like 6.23 and only use one core. It will get basically the same work that 6.23 can, and in that case will be no higher performance than 6.23.

In one of your posts you said "I think ill just stay with the high performance at least until i water cool my processor," which suggests that after trying out v7 you installed 6.34. Now, v7 and v6 clients are capable of getting the same work, barring a few exceptions; basically all of the uniprocessor projects can be obtained by 6.23, 6.34, and 7.1.43, and basically all of the SMP projects can be obtained by 6.34 and 7.1.43. The client is only responsible for downloading/uploading work and starting the the folding process. Scientific software called cores are what actually do the computations, so the client that's starting those cores has no effect on how hard your CPU is being taxed.

If you switched from 7.1.43 to 6.34 and saw a notable drop in temperatures, then it's almost certainly because 6.34 is not doing SMP work at this time (one core running full blast generates a lot less heat than all of them). You could have switched the v7 slot from SMP to uniprocessor and gotten the same results. When 6.34 is running SMP work, it will raise your temperatures about as much as you were seeing with v7. No amount of fiddling will get 6.34 running in uniprocessor mode to use up more resources since the scientific cores it's running will only use one core*. As Stonecold said, you really shouldn't raise the priority since it will use all unused resources to which it has access and setting it too high can interfere with your ability to use the computer at all, and increasing the priority will never cause a uniprocessor work unit to suddenly be worked on by other cores.

78C is on the upper end of what many people would consider the safe range for a desktop i7. You could try restricting the number of threads to which it has access; running SMP on 4 or 6 threads will reduce how much the CPU is taxed and should bring down the operating temperature. See here for instructions on how to add flags to a console client (which is the only version 6.34 comes in), and note the third picture that shows restricting the number of threads via -smp X. You'd want X to be 4 or 6.

Out of curiosity, what heatsink/fan did you use in that build? If you're not shooting for very high overclocks you don't usually need watercooling to keep your temperatures manageable, but you do usually need a decent aftermarket cooler since the stock HSFs are often not very good. By all means buy a watercooling setup, but a good air cooling solution can usually suffice and is not nearly as expensive (though those closed-loop CPU-only WC systems are more affordable than the whole-system solutions).

*To the vets, yes, I'm aware that core a4 can run on multiple threads, but I'm pretty sure if it's called in uniprocessor mode it will only run on one core.

[7im]

You don't have to increase fah's priority in task manager (and we generally recomend against it). Fah will always use all of a single core (cpu client) or all of the many cores (smp client) as long as nothing else is running.

*technically correct, a4 on one core on uni client.

[Jesse_V]

Also, its my understanding that when a WU finishes, the core quits so the process ends, and then when the next WU starts up the core is again started as another process. Consequently any priority change you make to the core will only apply to that WU, and then then your setting will be reset when the next WU starts up since the process you changed had died.

[Stonecold]

Also, its my understanding that when a WU finishes, the core quits so the process ends, and then when the next WU starts up the core is again started as another process. Consequently any priority change you make to the core will only apply to that WU, and then then your setting will be reset when the next WU starts up since the process you changed had died.

Yes, that's what happens. I tried making a batch file which monitored the core's CPU usage over time, but each time the core ended and a new one started the log was cut off (even though the processes had the same name). The process's PID changes for each core.