Hey, thanks for that... did read that Tom's Hardware article a couple weeks ago but lost it. I think it was probably my main reason for posting this in the first place, because of how hot the CPU ran when under load in their tests and how it throttled at stock clocking with compressor cooling.
Yes noise is a big concern, but equally is overheating. Live Audio processing is one of the most intensive tasks you can throw at a CPU, and I don't want my desire for portability to be at the expense of reliability. I've had my Mac CPU throttle when I'm working remotely with a client, and I have to say to them, 'look we have to bounce down things and then wait 20 minutes or so to start working again'. So it's about finding a balance which will give the most performance and the lowest likelihood of overheating + becoming too noisy.
The idea of the 8700k was that the single-thread performance bests even Intel's non-consumer CPUs and it is very affordable. This is one of my main reasons for coming back to Windows, for maximum single-thread performance that Apple doesn't offer hardware for. The audio applications I use, and moreso just audio processing in general, can't utilize multiple cores + threads as much as I believe other workstations tasks like rendering etc can, because the audio is summed in real time and that process takes place on a single thread. Other tasks within the application can be, but these tasks are quite minimal compared to the task of live processing, and that is a big thing I want to improve with this build over my Mac, so that I'm not having to bounce things to disk when they still need more work, which wastes time and clogs up my SSD, just to get a bit more performance.
The thing about single-threading not hitting the actual TDP is interesting - can't believe I never considered that before. Would be interesting to see how different it actually is in practice with the monstrous CPU draw of the applications I'm running... because I have no doubt that even without underclocking/volting I'll be running the first thread into the ground, but probably won't touch the others that heavily at all, so the fact that you are getting those temp. results even with an app like Handbrake is interesting. Might have to find someone locally with an 8700k to test it out.
Out of interest, what case are you using?
No worries; the article popped in my mind.
Er, wait: that article actually doesn't show stock clocks throttling, but it's also high-end water / below-ambient cooling. More so, their benchmarks stress most or all of the 12 threads simultaneously.
Got it, yes. This makes sense: the i7-8700K rips through single-threads faster than anything else at stock clocks. Unrelated: if it's mainly single-core, you do have a kind of time-intensive route of OC'ing an i3-8350K/i5-8600K on just Core 0 (you can set overclocks per core on Z370). It's time-intensive for testing for stability and if it's a production environment, I'm much less eager to take the time, but it's definitely
possible to reach i7-8700K stock performance on a single-thread with cheaper CPUs.
Exactly: with just a single-thread stressed, the chip won't get anywhere near its TDP. For shits and giggles, I set Handbrake to a single core affinity (i.e., only core 0 pegged @ 100%) for a ~5-minute 3Mbps 1080p H.264 clip: average temp was 55C with a peak of 58C and this is the stock Intel cooler with some higher-end TIM, but nothing like liquid metal. The Turbo Boost bounced between 4.1GHz and 4.3GHz as I had a few other programs opened; and, right: still on the stock clocks/volts i5-8600K.
For noise, I'd not choose the Intel stock cooler which definitely winds up on six-core workloads--eh, it doesn't even come with the K-series CPUs any more; I had a spare one from an old i5-4670K, so probably an easy decision for a new system haha.
Oh, right! I'm using a
Lan Gear Effy for the peculiar situation of wanting a single GPU slot in a pretty tight mini-ITX case that cost less than $80. It does have pretty decent cooling chops, too, though which helps the temperatures.
I would
imagine an i7-8700K would performs similarly, but likely a tad hotter as it's a single-core boost of 4.7GHz instead of 4.3GHz and a local test is always definitive.
