Enter the CPU model numbers on https://www.cpubenchmark.net/ and then you can compare their single thread performance.
But does it matter? I have a 10i5 and it’s easily fast enough for a convolution filter plus volume leveling with 196/24 material, showing the performance indicator at 23x. With convolution and leveling with 44.1/16 it is above 100 and hence disappears.
In turn it uses less power. Edit: Can’t complain about UI snappiness or other performance either.
Of course, if you want to do heavier processing then a faster CPU makes things possible that a slower one doesn’t, but as long as it is fast enough for what you want to do….
So I don’t really care about DSP. I’ve never come close to worrying about it - I don’t need “extra headroom” - I’ve never gotten below ~10-15x, and unless I’m doing something really silly it never gets below 30-70x. The only thing that matters to me is the responsiveness of the interface - search, page paint, etc. Some pages are just slow - like genres. Way slower than they should be. I imagine it’d be better if they were cached. But since they probably aren’t going to be cached unless Harman decides to send team Roon on a clean up mission, I just want my server / ROCK to deliver content as fast as is possible. Hence my question about my RAM speed and my SSD speed above.
This probably doesn’t warrant the higher price of the 10i7 (unless one uses many zones), but at least it doesn’t seem to be slower? Plus the turbo speed of the i7 is also a bit higher, and its L3 cache is larger
I did not do any scientific tests with roon but the single core/thread performance in standard benchmark rankings like the aforementioned Passmark is a good starting point. A good turbo mode ratio helps on top as a leverage (e.g. 2.4GHz > 4.2GHz for the aforementioned i5Gen11 or 2GHz > 2.9GHz for my surprisingly fast Celeron) as roon seems to hand over computing-heavy tasks like composition/recording list and coverflow from one core to the other. And I had the impression that more modern CPU like Gen11 and Gen 12 and even simpler models like current Celerons have a certain advantage. Should add that I did not test any Gen 10 NUCs.
I actually have a 7i5 lying around
Sounds like a good choice if you want a snappy usability experience but do not intend to do a crazy number of parallel heavy streams.
CPU Benchmarks is a good resource. However, the Single Thread Rating probably takes into account maximum possible Turbo Speed, which may matter little, if at all to Roon performance. For Roon, an ideal metric would seem to be Single Thread Rating based only on nominal Clockspeed (or, alternatively, whatever is the highest longterm sustainable Turbo Speed).
But you said “10i5 shows higher Roon Performance Index # than 10i7, thanks to higher single core speed”, and whatever the reason is, the single core speed doesn’t seem to be it.
The 8th gen single core speed is pretty good for their age and on par with 10th gen, comparing the CPU that Intel specs say is in the NUC8i5BEH:
Maybe, but as the turbo is 4.2 GHz for the i5 and 4.7 for the i7, this may account for the difference in the benchmark, but then the base speed would still be approx the same and the i7 would not be slower.
Why do you think the turbo speed is not important for Roon? I would have thought that it of course isn’t for DSP and other extended performance, but might have an effect on short-lived tasks like response time for quick database operations for instance
Turbo Speed, which may matter little, if at all to Roon performance
I had the impression it does help in terms of snappy experience as it is usually a matter of max. 10 or 20 seconds of high-intense computing of one core with other cores usually on pause (If you do not stream to 4 zones with heavy computing).
Ah, no I missed it I would have thought that there are many things in Roon that are not long-lived, like database retrievals, but maybe that’s not “real use” for Danny or overall it doesn’t have a noticeable effect on snappiness things.
Audio Analysis is one activity that probably can benefit from Turbo Boost. When I run Roon on a NUC8i7BEH, Audio Analysis modulates the fan quite a bit, ostensibly as the processor kicks in and out of Turbo Boost. On a NUC8i3BEH, the fan modulation is less, even zero, presumably because the 3.0 GHz nominal Clockspeed already is quite high.
Turbo speed is important for short-lived tasks like response time for quick database operations you mentioned. It is usually set in BIOS for 32sec, but the CPU may switch it off faster than that, when temp limit is exceeded for example (this is where passive cases show advantage, as they have much higher temp inertia).
Ok, well to try to get some closure, can we agree that the “interface / search snappiness” experience of ROCK (ie, not DSP where I think I’m well-covered with existing hardware) is some combination of the following factors:
CPUBenchmark single core performance
Turbo boost
RAM speed (and 2 banks vs. 1)
RoonOS SSD speed
If that’s right, then the relative importance is not that big a deal to me. It’s fairly easy to max out SSD and RAM speed, so that leaves the single core CPU Benchmark performance.
Not sure about SSD as I am under the impression that the database is completely in RAM during operation and ROCK does not swap so I suppose all of ROCKs processes are as well
I would have thought that there are many things in Roon that are not long-lived, like database retrievals, but maybe that’s not “real use” for Danny or overall it doesn’t have a noticeable effect on snappiness things.