Passmark as a comparative tool for old/low power CPUs?

[champ222]

Hi All

Short version:
I'm looking at CPU performance of old CPUs and LOW power CPUs. The only metric i can seem to find to compare some of these low power CPUs against older "regular" CPUs is Passmark. Is passmark a reasonable tool for this? The use case is a NAS with UNRAID.
Examples are:

CPU	Passmark CPU Mark	Passmark Thread Mark	Comments
N3350	1108	801	CPU from Zimablade 3760 (Dual core version)
N3450	1978	765	CPU from Zimablade 7700 (Quad core version)
J3455	2256	815	CPU from Zimablade 7700 (Quad core version)
E3950	2096	742	CPU from Zimablade 7700 (Quad core version)
J5040	3244	1346	CPU from my current NAS server (ASROCK J5040ITX)
6100U	2633	1333	CPU from NUC 6i3
5250U	2458	1526	CPU from NUC 5i5

the J5040 is an Asrock ITX motherboard with the CPU fixed, and passively cooled. Works really well, and does all i need from it. Its largely just NAS duty. I run a Plex server, but dont really use it to be honest, Plex does my head in a bit.

From the passmark results, it looks like its faster than a 6th Gen i3 NUC and a 5th Gen i5 NUC. But is that actually the case? i guess the different CPUs have different features, but does that even matter for my needs?

Additional info:
I created a new NAS a while back using the Asrock J5040 motherboard mentioned above. I now have a large drive "pool" that works pretty well. Previously i used an HP microserver with an ancient version of freenas and a bunch of HDDs that were all individual shares. Kodi could use multiple locations for movies etc so it didnt really matter that it wasnt one pool. I then backed up to a bunch of USB HDDS. This is why i didnt want one pool.

I'm looking into creating another NAS to create another big pool to backup to. it could also take over NAS duties in case of main NAS failure.

I was going to use an old NUC i already had (as above), with an M.2 to PCIE adapter, to an HBA, to a bunch of drives. but if a Zimablade can do the job, it seems like a better option, lower power and passively cooled. I'm leaning towards the 7700 quad core version (it can come with any of the CPUs in the list) but if i can do it with the 3760, then even better, but will it be too slow?

The main NAS is in a coolermaster NC100 case that i modified to fit the ITX board, with 4x HDDs in the place where the GPU would go. It was a good fun project actually, though i still need to finish it off. I got three NC100 cases for dirt cheap and they have an SFF PSU (that reviews well). If i used a NUC or Zimablade/zimaboard i would hardly need to modify the case at all and could fit 6HDDs in, plus and SSD or two. little bit of electronics to get it all working from the SFX PSU, and it should be a good project too.

 

[Armageus]

I'm looking at CPU performance of old CPUs and LOW power CPUs. The only metric i can seem to find to compare some of these low power CPUs against older "regular" CPUs is Passmark. Is passmark a reasonable tool for this?

Yes, it's as good a tool as any for comparing CPUs of different ages / makes / models.
However rather than just focusing on the overall Passmark, it's also worth factoring in the Single Thread Rating (as some tasks don't always scale well with cores and/or need good single thread performance). An older chip with lots of cores might have a similar overall mark, to a newer chip with fewer faster cores.

From the passmark results, it looks like its faster than a 6th Gen i3 NUC and a 5th Gen i5 NUC. But is that actually the case?

I'd almost certainly say so, it's got 4 actual cores (rather than 2 cores + HT), and is 5 years newer

 

[champ222]

Yes, it's as good a tool as any for comparing CPUs of different ages / makes / models.
However rather than just focusing on the overall Passmark, it's also worth factoring in the Single Thread Rating (as some tasks don't always scale well with cores and/or need good single thread performance). An older chip with lots of cores might have a similar overall mark, to a newer chip with fewer faster cores.


I'd almost certainly say so, it's got 4 actual cores (rather than 2 cores + HT), and is 5 years newer

i was convinced that the NUC5i5 was a quad core, but your right, they are both dual cores. Intels mobile CPU naming is stupid

Thanks, Ive added the "Thread mark" results to the table.

I'll try to do some research on whether Unraid is better with 4 cores over 2. Looking at the Zimablade options, there isnt much between them in the single thread test.

 

[Tetras]

i guess the different CPUs have different features, but does that even matter for my needs?

It depends, unfortunately you often only find this out when you realise you're missing something you needed.

E.g. there are IGP features like vPro and the display outputs, as well as what they can encode/decode.

There are extensions/instruction sets that certain apps might need to function, or make really good use of. I'd check the minimum requirements they publish, since that can reveal the baseline.

I'll try to do some research on whether Unraid is better with 4 cores over 2. Looking at the Zimablade options, there isnt much between them in the single thread test.

I'd always prioritise the single thread unless the app is very well threaded, since even when the app IS multithreaded, many of the tasks you do will still heavily rely on the single thread performance.

I was going to use an old NUC i already had (as above), with an M.2 to PCIE adapter, to an HBA, to a bunch of drives. but if a Zimablade can do the job, it seems like a better option, lower power and passively cooled.

If relevant to you, keep in mind with those M.2/PCIE adapters that some very old PCs can't boot SSDs from them (though there are workarounds).

 

[champ222]

It depends, unfortunately you often only find this out when you realise you're missing something you needed.

E.g. there are IGP features like vPro and the display outputs, as well as what they can encode/decode.

There are extensions/instruction sets that certain apps might need to function, or make really good use of. I'd check the minimum requirements they publish, since that can reveal the baseline.


I'd always prioritise the single thread unless the app is very well threaded, since even when the app IS multithreaded, many of the tasks you do will still heavily rely on the single thread performance.


If relevant to you, keep in mind with those M.2/PCIE adapters that some very old PCs can't boot SSDs from them (though there are workarounds).

Thanks for that. I think Unraid should work fairly well on any of these options. Its the additional plugins etc that can make the PC performance more of a concern. this is a simple file server so i think i'll be ok.

Unraid boots from a USB flash drive (it uses this for licensing) which is actually a challenge for the Zimaboard, as it defaults to the built in eMMC storage, and the bios isnt backed up with a battery. I'm waiting to hear back if the blade is the same. I suspect it can be modded for a battery somehow.

Thanks guys. this was really helpful.