I have an Intel core i7 5820k over-clocked to 4.2ghz running along side an MSI GTX 970 video card.
The CPU is liquid cooled and can run at 100% at 42 degrees c. all day.
The video card however is not liquid cooled and quickly overheats and crashes.

So, how do i either configure the system to use 0% of the GPU, or limit its usage to a certain threshold, without also limiting the CPU?

---

In the preferences in BOINC (don’t remember off hand which drop down its in, maybe “Options”) there’s an option to select something like “don’t use GPU” or “never use GPU”.

You can also deselect the sending of GPU tasks on your preferences on the website.

Third, your GPU shouldn’t be overheating. Even on air cooling. Make sure it’s free of dust, maybe replace the thermal paste, and ensure that the case has adequate air flow.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

It's the 'Activity' tab. (Advanced view, if you're not using that already)

In the preferences in BOINC (don’t remember off hand which drop down its in, maybe “Options”) there’s an option to select something like “don’t use GPU” or “never use GPU”.

No, this is a project preference only, not a BOINC preference. For Seti it is found at https://setiathome.berkeley.edu/prefs.php?subset=project (but check the correct venue/location).

Mind that this will stop BOINC asking for work for the GPU, it won't affect any work for the GPU already in queue. That will have to be run or aborted.

---

The card shouldn't be overheating if it it working correctly. As the temperature increases the card's fan should also increase to compensate and drop the temp down to stay within the manufacturer specified range. You can also use any of the many Windows overclocking utilities to limit power and temp, such as Afterburner or EVGA PrecisionX. The TThrottle program by eFMer is specifically designed to control temps while running BOINC and probably is your best solution.
https://efmer.com/tthrottle/
But first verify that the fans are even spinning on the card. Are they caked with dust. Blow the card out with air.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

It's the 'Activity' tab. (Advanced view, if you're not using that already)

thanks again for the clarification

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

Indeed... this machine is an ancient Intel board with 3x970s stuffed into it, with a few mm between them, in a too-small case with only a single 80mm case fan, and they run the AIO 24x7 without overheating. I'm suspecting clogging with dust as well.

---

Probably preaching to the choir, just be glad I'm not singing Christmas carols...

Adequate airflow can be an issue even if the GPUs' fans are working OK. I tried cramming a Radeon RX 570 along with a Radeon RX560 in my Mac Pro. The airflow through the PCI & hard drive area just wasn't adequate for those two GPUs. It is adequate for two RX 560s.

GPUs, just like CPUs, use thermal compound to move heat from the silicon to the heat sink. This compound hardens with age and with heat (like we generate running CPUs at 100% load 24/7). As it ages and hardens the compound loses its thermal conductivity.

If the card is at least a couple of years old and you've blown all the dust out, it may be worth disassembling the card to replace the compound. I recommend Arctic Silver 5, it's worked fine for me for years. Be sure to clean both surfaces with isopropyl alcohol and blow them dry to ensure good contact. I use a piece of plastic wrap over my finger to apply the compound so I don't end up with a mess. OK, I still end up with a small mess, I blame old age...

I get at most three years from a laptop before it needs new compound. Not only do laptop processors run hot but the joint between the silicon and heatsink flexes from both heat extremes and mechanical motion. Desktops usually last longer before needing new compound.

---

Another possibility is faulty PSU rail. I actually had one PCIE rail from the PSU died after working fine for a few years. Windows booted without a problem but intense games would crash a few seconds in. Basically it can only work up until the GPU really needs to pull power from the PCIE pins.

Luckily my GPU had power indicator which made the issue very obvious once I start investigating. After I switched to a different socket for my PCIE cables, the power warning LED no longer lights up and everything is back in order.

---