Comparison of NUC models, 10i7FNH vs. 7i5BNH vs. 8i5BEK


Due to a friend giving me one, I am in the lucky situation that I have three NUCs on hand, and need two of them for ROCKs, and I’m going to sell the third. I’m not picky about form factor (I have an external USB SSD enclosure), just about performance. And when I say performance, what I mean primarily is pure “snappiness of remote responsiveness on search”, then “track start / switch” speed / “playlist navigation speed”, and only finally DSP - which I use, but not extensively or to the point that I’ve taxed any of my systems; and I don’t care about library processing speed. Let’s just assume that they each have similarly speedy RAM / Roon DB SSDs.

I am guessing that the 10i7 is going to be the fastest according to followed by the 8i5, but I’d be interested if that’s so. I always see people saying things like “well, you actually don’t get any advantage with this model because the cores blah blah” where I don’t quite understand what blah blah means. I’m well intentioned, reasonably technical for a layperson, but this isn’t what I do for my day job.

Thanks for any help!

Actually the single core performance (which matters for Roon) for i5 vs i5 and i7 vs i7 is faster on the 8gen than the 10gen. 8i5 vs 10i7 I don’t know, I would guess same-y.

I would think the 8i5 is faster than the 7i5.
10gen generally uses considerably less power.

Without heavy DSP, multiroom or huge library I don’t think it matters. I have a 10i5, 50,000 tracks, one zone, only DSP is volume leveling, and my speed factor in the signal path is x100. (x1.2 would be still fine)

1 Like

That’s helpful. So your opinion is that for “search performance only” it’s just single-core performance, and that the single core performance of the 8i5 and 10i7 will be about the same?

Also, how are you (or I) paying attention with Spain/Germany on?

Yep, Roon is single core except when doing file analysis. Equivalence of 8i5 and 10i7 is a guesstimate based on 8i5 being faster than 10i5 on single core. And we have hit a bit of a wall for Intel single core performance generally, the differences are not very large anyway.

I am having a hard time getting into this world cup. TV is on, but watching only with half an eye.

wow, sidebar but among US friends who even marginally care about soccer, this is off the charts enthusiasm compared to previous years, and not because people are dreaming about US team, rather because so many ridiculous results - Japan d. Germany? SA d. Argentina? even Iran d. Wales was kind of a surprise. Tho I will admit that there is a lot of media hype about US v. Iran with a must-win, that’s predictable that they’re building hype over this esp given geopolitics, but this Iran team is fun to watch.

Germany (where I am) is being depressed because of FIA corruption, a world cup in winter, and Qatar human rights violations. The serious football fanclubs are in full-on boycott mode and are refusing watching.

1 Like

Ah, for some reason I thought you were elsewhere in Europe. I had heard that not just German players were protesting but people boycotting watching as well - amazing to hear that it’s going on on the ground. This whole Qatar/FIFA thing is absurd - the beer tents to homophobic repression to environmental consequences. Personally I’m glad the beautiful game is catching on more and more here but yes, agreed.

1 Like

And I hate the video referee. 3 times today I saw a goal, then drifted off and did something else, and when I started watching again it was a different score again. I am just now realizing that it’s 0:0 although I saw Rüdiger score earlier. It’s ridiculous

In league games for weeks before it started, the ultra clubs on the stands unfurled huge banners in protest. It’s quite heart-warming seeing this in football stadiums.

“We make everything clean for money” - Bayern München stadium, with pictures of Bayern leadership like Oliver Kahn:

“Air conditioned stadiums instead of climate protection. Prosecution of sexualities. (etc)”, in Berlin

Similar in Mainz

Borussia Dortmund. “Turn off Qatar”, “Boycott”, etc

“15000 dead for 5760 minutes of football. Be ashamed”. Bayern München again

And the TV stations are playing constant documentaries, reporting from Qatar, etc

1 Like

Noticing this as well with the US colleagues at the workplace, and it’s great. A beautiful game indeed, I love the free form: “Here’s a pitch, a ball and 90 minutes. Rest is up to you”. Sad that it has come to this

I think they will all be good.

DSD upsampling is pretty demanding - esp if you turn off the setting to spread it across two cores.

So if you have an endpoint that supports this, try it.

You can check the multiplier in the Signal Path screen.

Just compare the ‘Single thread’ number in this post NUC 8th Gen still the best option? and best retailers? - #33 by simon_pepper
Higher the better for RoonOS

1 Like

It looks like the single threaded numbers are trading in a narrow band - they’re not improving that much. Which is what you said. So six of one?

Actually Simon, just to satisfy my desire to “get to the right answer”, can you show me where on the Passmark Benchmark site I can find the model specific numbers? I can find by CPU, but can’t find a searchable PC model page. Don’t want to put you to the trouble, so if you can point me at the right source I’ll definitely appreciate it.


They are likely all going to work fine.
the 10i7 has 6 cores, so if you wanted to do DSD upsampling on multiple zones it would have an advantage over the 7i7BN which has two cores and so might struggle supporting more than one or two zones doing DSD upsampling.

So many zones, and lots of DSD upsampling the 8i7 (four cores) or 10i7 might have an advantage.

But for most use they will all work fine, and you won’t see much difference.

(I should add that for most zone related processing, roon runs each zone on a single core, with the exception of DSD upsampling that can optionally be split across two cores).

Thanks. Don’t do DSD upsampling except on one zone and occasionally, and if I’m gonna do that I’m definitely not listening to multiple zones. To the extent that I do do DSP, I’ve never taxed any of these beyond 40x, so I’m safe in that arena - they’re all way more than adequate.

My real question is about “snappiness” and “non-lagginess” of remote response. Less jellyfish. Not that my jellyfish hang around that long, but I have them noticeably for a second or so on many actions.

So the real question is would any of them make for a better experience there. And I’m guessing that that is just as much a product of memory speed and Roon db SSD speed as it is single-threaded performance? But also that this way lies insanity because you can’t really tune to perfection?

Would love to squash me some jellyfish.

Hi, This get to this list, I had to look up the NUC type on the Intel specifications site e.g here for an NUC10i7FN and then follow the link to the CPU used in that NUC, which is the Processor number, i7-10710U. Then add that to the comparison table on PassMark, and then copied out the relevant details into my own spreadsheet, where I just want to record single-thread performance, # of cores & threads, TPW and overall performance.

However interesting results as in the NUC7i7 to NUC10i7 models single thread performance is now 2,165 to 2,800, it is just the overall that varies, which is more down to additional cores and Turbo frequencies, which both consume extra power.
The same was true for the NUC5i3 to NUC7i3 range as well.
Only with the NUC12 and the latest CPUs is anything significantly higher, but these CPU offer the 2 types of CPU cores, efficiency and performance types, however how is the workload moved from an efficiency core to a performance core and back again? Is this within the underlying BIOS kernel, the OS or application layer?
RoonOS seems to be built on a single-pipe processing model with each endpoint zone operating within a pipe. So will there be the ability to utilize the performance core, dynamically, with the Roon Core/RoonOS returning the workload back to the efficiency cores when dormant?
If not, then the NUC12 benchmark figures are somewhat meaningless, as the applications running the benchmarks will be behaving differently to RoonOS, at a CPU Core assigning and load-balancing across the CPU core to how the embedded operating system of RoonOS undertakes its workload management.

Personally, I moved from a NUC5i3MYHE, which had been perfectly adequate since 2017 with the launch of ROCK for the Library I was running, which at the time was circa 90k tracks, to a NUC7i7DNK (same board as in a Nucleus+) only because I was worried about supply issues as I approach the 100k track mark. Other than faster scanning of new material and any database updates, there has very little noticeable difference

1 Like

Has anyone on this thread extended the performance question to more recent models? Always wondering when my 10i7 gets a little sluggish if I should bite the bullet and get a 13th gen or an m2 Mac mini or something.