100C+ is too hot for VRM for long-term stability. That being said, Prime 95 can end up being an unrealistic work-load for most use-cases. I'd try getting a genuinely cpu-limited game out (battlefield something?) and throwing lots of AI on the screen until it's a slide show and leave it run that for a while, then check temps on the VRM. Consider undervolting and / or underclocking AND then undervolting the CPU. Just disabling extended turbo durations - or altogether - may help alot. CPU COOLER: Use a downdraft cooler to help keep the VRM cool (same fan orientation as stock junk cooler, just a better cooler capable of actually cooling). Downdraft (vs tower coolers) will push air through the heatsink out onto the surrounding components. VRM Cooling: They make heatsinks and thermal adhesive/tape specifically for this type of upgrade. You'll definitely want to put some type of heatsinks on there for some insurance against heat. Otherwise, if you're not willing to do that and/or try to under-volt the CPU, send the board back as it's just going to be the point of contention for reliability's sake. Also, if you have a dead motherboard or know someone with one, go get it and rob the heatsinks off. Don't break a working board though - if it comes to that just send the new board back and get another one even if it's an ASRock or MSI model. If on a tight budget ASRock and MSI will make a more serviceable / better board for you at a low cost, as that's their appeal. Games will NOT create the type of load in 98% of use-cases that the Prime95 / Furmark will. I believe those are AVX heavy workloads and those do quite a number on any intel CPU. What would I do? TL DR Well I bought a Ryzen vs intel due to the heat issues with processors, and the anemic intel heatsink I got years before (never-mind absolutely dismal overclocking!). I would entirely send back the gigabyte board as it's obviously not making you very happy. Get a better board with better VRM, otherwise, read below. Buy a downdraft style cooler, noctua make a good one you should consider, but there's many choices there, do read reviews. If that alone doesn't solve it, affix some heatsinks to the VRM componentry and be careful not to short anything out down there. Making any permanent mods to the board will void warranty, so make sure you can undo any changes if and when it does eventually fail. --Good luck!
It's likely that the board has a 95W power limit set in the BIOS, you could check for that. What that means is that the CPU will boost to something like 1.25xTDP (or more at times, look up PL1 and PL2 for Intel CPUs, it all depends on what the motherboard manufacturer decides to set) for a short while, and then return to the TDP of 95W. By the way, Turbo is definitely not an OC. MCE is. For a 9600KF it's not that relevant since without HT, the CPU manages to stay below the 95W TDP most often. If you want to improve the VRM temps: - fan over the VRMs - undervolt the CPU - the lower the voltage, the lower the power draw, and the lower the CPU and VRM temps under load. You will have to stresstest each step as undervolting will eventually lead to instability. For example, my average 8700k silicon can run at about 1.096V vcore under full load, fully stable, default clocks, drawing around 110W. Better CPUs can go to 1.050V or even lower. On default voltage, the CPU would draw 135W+ and run at 1.2V or so. - unsure for Gigabyte, but on MSI you have settings such as CPU Lite Load which you can mess with to help with undervolting, also the package C-State limit does have an effect in lowering power draw (hence lowering VRM temps which is your goal), if I set mine manually to C8 it decreases power consumption with 8W at same voltage for some weird reason. - get a good ventilated case. When you game, most aftermarket GPUs will dump heat inside the case, and it will contribute to VRMs getting hotter. The fan over VRM should have the most visible effect. If you don't want to bother with it, just undervolt and be done with it, or simply do nothing, there's not much of a reason to stability/stresstest on a platform that does not offer overclocking. If it's stable, it's stable, just leave it at that. If it's not stable, it might as well be the RAM or some bad driver/Windows install, it's not necessarily the CPU.
