Alright here's a two questions post.

I've been running BOINC and participating in these projects since 1997 but with always a cheap computer replaced by another cheap computer and so forth so never been able to participate seriously.

Now, I got a I7-8700 in a case with 5 120mm fans (2 in and 3 out), with a Stryx RX-580 GPU... Oh! And the CPU has a aftermarket cooler not the crappy original CPU fan.

It's not the highend of highend of computer but it is quite good and can go up to 4.4 ghz and I did put 16 gigs of RAM with heatsink RAMs.

So, there I was thinking stuff would be more serious with BOINC as I could run not only more than 2 analysis at the same time but run many projects at full speed........

Big error on my part. The first try, my computer shut itself down 3 times when running BOINC since it heated my RX580 to dangerous levels so fast I didn't really have that much time to try and react since most of everything happened under a minute.

Now I even tried to use TThrottle but that thing turn out to be trash trying to maintain the GPU at more acceptable temperature scale than 90 degrees which it actually reached at shutdown.

After much work tweaking my GPU fans, I was able to run GPU analysis again at a acceptable 65-70 degrees max for 1 running GPU project and only 1 at a time.

The other problem I ran is at Turbo mode (Intel own UEFI option), reaching 4.4 ghz instead of stock 3.2, I ran into another problem where CPU temps came close to 90 degrees so I stopped everything when I would see temps of 86.. 87... 88... and I stopped.

Now that one, unable to resolve it at all, I had just 2 choices, either I stop using BOINC altogether and run my new computer permanently at 4.4 ghz without any problem or I continue to use BOINC but go back to 3.2 since it runs fine at 3.2, temps are always below 65.

I'm lead to wondering a few things, why after all this time we still do not have a clear control of GPU settings directly in BOINC yet when we are able to control with a fair degree of precision how intensive your CPU will work.

Also, why in the world am I hitting those crazy temperature scales? Obviously I could not run BOINC at 100/100 (100% CPU and 100% all cores) at 4.4 it would blow my computer.

Keep in mind it's AIR cooled but aftermarket aircooling is very much at par with closed watercooling. And the tower is a full tower with excellent airflow with all fans running at full speed.

P.S. Just to be clear, the 4.4 is obviously overclocked and would produce more heat but I'm talking about stock overclocking by using UEFI Intel Turbo boost so I'm not talking about very serious overclocking here, you either chose you are going to use your cpu at turbo or not.

From what I have read ATI/AMD cards are fairly heavy power users when running compute loads. There are ways to undervolt and underclock the RX cards so that they run cooler. Projects that make heavy use of ATI cards are Milkyway and Einstein. I bet you could easily find forum posts showing you how to optimally set up your RX580 card to both run cool and produce a lot of work.

Now, I got a I7-8700 in a case with 5 120mm fans (2 in and 3 out), with a Stryx RX-580 GPU... Oh! And the CPU has a aftermarket cooler not the crappy original CPU fan.

Pick a more aggressive curve for fan control and make sure there is nothing blocking the 580's fans intake. On a mining system with several rx570 I had to set all their fans at %100 for 24/7 crunching.

It's funny because those two I run often, both Einstein and Milkyway.

I did try to under clock it but I think mine is already sold overclocked albeit I was not totally able to confirm that. All my tries at under clocking the GPU were failed but then again I was pretty much in the dark when it comes to under clocking.

Also, under clocking is not exactly what I was looking for because it has a permanent impact on my system and I don't want BOINC to be the center of my decision making.

Ideally, what I want is to run both CPU and GPU in acceptable temps without affecting gaming as this remains a multitask platform for me, I loan CPU/GPU times but I do not at all use my computer just for BOINC in my case.

Yeah, I had to really tweak the fans and it did work for the GPU but it was a massive fail for the CPU and it forced me to move out of Turbo mode.

Turn off Turbo which puts a lot of excess voltage to the cpu which causes exponential power consumption and a lot more heat.
Instead pick a fixed cpu clock for all cores that you are comfortable with, say 4Ghz and lock the cores to that value. Only run the cpu at enough Vcpu voltage that is stable with that clock. Easily could drop 200mV off the Vcpu voltage that Turbo tries to run the couple of cores at 4.4Ghz. Should reduce the cpu temps greatly and what your cpu cooling can keep up with.

[Edit] Try the solutions and suggestions from this Reddit on the i7-8700
https://www.reddit.com/r/intel/comments/7rgqqr/got_my_new_i78700_nonk_and_the_temperatures_and/

Just in case you run any work for CPDN on CPU (It doesn't support GPU) overclocking increases the risk of tasks crashing with climate prediction. Not enough have done it and written about it to be sure of what you can get away with and what you can't.

It's so much complicated to try and find a solution where I can up the CPU about halfway from Intel Turbo provides when ENABLED because the main problem is that my CPU is 8700 not 8700K which means my options are limited.

I did try XTU and tried and tried to underclock when Turbo ON but... running BOINC showed temps going fairly fast to 92 degrees which is way beyond acceptable for me and it was probably going to get even hotter.

Running with Turbo OFF means even more complexity because you have even less possibility of moving a 8700 UP so Turbo has to be ON to be able to try to move down and trying to VCore Offset totally fails in lowering the temps with BOINC (but I do clearly see a decrease in temp in XTU).

VCore itself is even more complicated for me as I don't want to mess too much with VCore on XTU, safe limits are hard to fine on the internet for the 8700... plenty for the 8700K.

All in all I may just settle with safe temps with my computer being slower which I do not really see much impact when I do gaming at 3.2Ghz instead of 4.2 or 4.4Ghz. That way, BOINC never stresses the CPU more than 65C which I am very happy with and yes, I know that temps can be safe above that but a cooler CPU is always better for longer life as far as I am concerned. Also, this is still free involvement on our part so I do not want to age both my GPU and my CPU for donating my computer to these projects (been paid would allow people to change hardware when occurring fails or premature end of life).

So, currently, temps are under control:

CPU 60-65 max with all 6 cores in use by Einstein, SETI, Milkyway, collatz, etc.
GPU 60-70 max with aggressive fan curve control (you do hear the fan kicking in quite a bit the moment GPU projects starts computing) but that I could also stop all GPU computing if I ever feel it demands a bit too much from my brand new system.

The maximum is not the optimum.

CPU: Consumer CPUs are not optimized for constant loads, and running BOINC at turbo frequencies 24x7 is inefficient. Underclock and undervolt where possible. If you are running Windows, it is trivially easy to restrict CPU frequency using a power profile, which will make the biggest immediate difference. You can easily automate switching profiles so that you have full speed when you are using the box for something else.

GPU: the same applies - underclock and undervolt where possible. How you do this will be vendor-specific. With less contended RAM you may be able to run a comparatively higher compute frequency than with the CPU, but the physics of voltage and power is even more brutal because of the larger die area.

Faster RAM is often the best performance return on electrons employed. The architecture makes a bigger difference than increments in speed. With a fat L3 cache or more memory channels, or a generation jump (DDR3 to 4), you will get a better marginal return on more compute cycles. But with consumer gear, faster compute is often just waiting frantically for the RAM to hurry up, sucking power in the process.

---

But with consumer gear, faster compute is often just waiting frantically for the RAM to hurry up, sucking power in the process.

Certainly true with some of the current CPDN work where some users have seen greater throughput of work by not using all cores.