K8s on CWWK X86-P5 & X86-P6 mini-pcs

[naBs]

Hi I've just seen your post and was wondering what the storage limitations are on the p6 devices? I just recently purchased a p6 n355 and I'm going to get 48GB vengance SODIMM DDR5 and was looking to try and kit it out with 4* 4TB m.2 but I've heard mixed things regarding limitations on the storage?

Would appreciate if you can provide some insight before I go ahead and buy some drives!

 

[alkersan]

The systems I bought are:
...
N305 X86-P5 "Super Mini Router"

I just recently purchased a p6 n355

Hello @bhigh and @naBs. I find it very fortunate that found you both in one place here .

I'm planing to build a similar toy cluster and trying to decide between P5 and P6 boards. Leaving the BIOS aside, the difference is supposedly only in the CPU - an N305 (AlderLake) vs a slightly boosted N355 (TwinLake). Given that today there are literally zero 1-to-1 comparisons of these CPUs on the internet, I wanted to ask you for a favor - could you please run for me a couple of CPU tests? Even a couple of runs of GeekBench would be very much appreciated.

Thank you

 

[amalurk]

Hello @bhigh and @naBs. I find it very fortunate that found you both in one place here .

I'm planing to build a similar toy cluster and trying to decide between P5 and P6 boards. Leaving the BIOS aside, the difference is supposedly only in the CPU - an N305 (AlderLake) vs a slightly boosted N355 (TwinLake). Given that today there are literally zero 1-to-1 comparisons of these CPUs on the internet, I wanted to ask you for a favor - could you please run for me a couple of CPU tests? Even a couple of runs of GeekBench would be very much appreciated.

Thank you

CPU Mark

Intel 3 N355 vs i3-N305 [cpubenchmark.net] by PassMark Software

 

[alkersan]

Thanks @amalurk , I saw that website but It's very similar to the infamous userbenchmarks. The thing is that it's an aggregation of dozens of arbitrary user submissions, using different hardware platforms, e.g. some of those ~400 results for N305 are very likely done from an itx nas board, from different OEMs, bios settings, maybe power limits, etc.
Here, I believe, we could have a more apples-to-apples comparison, as both mentioned users have in possession the exact boards I was considering for the lab build. Well, maybe I'm overcomplicating things... dunno.

But thanks for suggestion anyway. A 7% difference for the 20% of price is already something to take into consideration

 

[SnJ9MX]

Thanks @amalurk , I saw that website but It's very similar to the infamous userbenchmarks. The thing is that it's an aggregation of dozens of arbitrary user submissions, using different hardware platforms, e.g. some of those ~400 results for N305 are very likely done from an itx nas board, from different OEMs, bios settings, maybe power limits, etc.
Here, I believe, we could have a more apples-to-apples comparison, as both mentioned users have in possession the exact boards I was considering for the lab build. Well, maybe I'm overcomplicating things... dunno.

But thanks for suggestion anyway. A 7% difference for the 20% of price is already something to take into consideration

I trust cpubenchmark.net far more than userbenchmarks, which is total garbage. Sure there are a lot of variables that go into performance (PL, the other things you mentioned, etc), but at least the numbers mean something for cpubenchmark.

Infamously, userbenchmarks writes things like this where they say the AM of AMD is "Advanced Marketing" -

http://imgur.com/a/userbenchmark

. They also constantly tweak their metrics so that Intel is faster. The current CPU speed metric is "effective speed" which is of course meaningless.

https://www.reddit.com/r/pcmasterrace/comments/1i7xx2t

 

[SlowmoDK]

Just found this tread ... Im running a "fleet" of p5 n100 and p6 n305

Main Issues with these boards is ram compatibility .... the first month I tried every possible combination of psu, nvme, and the ram I had onhand at the time, without getting systems stable. they would run for anything between 10min and couple of days but would eventually crash.

turns out all my problems stemmed from compatibility with my 16gb crucial sticks .... once i finally tried some g.skill ram suddenly all my boxed stopped randomly crashing like magic

With the right ram the psu that comes with boxes seems fine, but have changed for meanwells

Edit : I even got the crucial 16gb sticks RMA'ed and exchanged for new with the exact same issues ... and the same sticks runs just dandy in my AMD 7745 system, so clearly not the ramssticks fault

 

[alkersan]

Just found this tread ... Im running a "fleet" of p5 n100 and p6 n305

CWWK P6 with n305 cpu? Mind sharing the link? Cause I've just ordered a P6 board and it's only available with n355 CPU.

 

[alkersan]

Probably I've got too optimistic after seeing this one reply and immediately imagined running an 8 GB/Core workload on these boards

 

[SlowmoDK]

CWWK P6 with n305 cpu? Mind sharing the link? Cause I've just ordered a P6 board and it's only available with n355 CPU.

Sorry P6 MAX not P6, I really like this formfactor.
I'm cooling my 4 unit mini-cluster with a single almost inaudible 20cm Noctua USB fan

Probably I've got too optimistic after seeing this one reply and immediately imagined running an 8 GB/Core workload on these boards

Dunno about 64GB but 48GB seems verified from many sources

 

[SlowmoDK]

Edit : I even got the crucial 16gb sticks RMA'ed and exchanged for new with the exact same issues ... and the same sticks runs just dandy in my AMD 7745 system, so clearly not the ramssticks fault

Another "funny" is that while 16GB Crucial sticks caused all sorts of trouble for me, The 32GB version with the exact same chips onboard works without issues

 

[alkersan]

Another "funny" is that while 16GB Crucial sticks caused all sorts of trouble for me, The 32GB version with the exact same chips onboard works without issues

Alright, wish me luck. Just got a bare bones P6 (although, I was ordering the dev board model without enclosure, and they sent me this metal brick). Going to try it with 64GB stick next week.

Spoiler: Photo

Update: it posts. Measuring stability will take some time

Spoiler: memtest

 

[wistak]

Brilliant! Would you be open to posting your files? I would like to modify it so that it has two vertical 2.5" drive slots so that the drives could be slid in from the top. That way, I can avoid destroying the enclosures of the SATA drives, make them easy to swap, and make sure nothing rattles inside. Always wanted to jump on this 3D printing bandwagon, but never had a pretext.

Hey there
I'm thinking about something similar, have you managed to try it?

 

[wistak]

Here's a number of files, in STEP format - you should be able to pull them in to any cad program or 3d printer. There are various options. All have a common mounting plate that slips inside the case snug down against the SODIM socket with the rear sticker removed.

The sticker can be eased away without damage and then re-attached to the side - mine haven't pealed away. I push from the inside starting at one end and letting it peal itself off.

The inner mounting plate needs to have 4 m3 thermal bushes installed - available from amazon, easy to install with a soldering iron. On the outside you can then mount the simple antenna plate which first needs to have the antenna a leads attached. With just wifi mounted its a two part assembly. It will be all very obvious how it assembles once you either import to CAD or 3D print.

If you want to use the mounting for the sata drive its a three part assembly as shown in the earlier post. There are two versions, depending on the size of the sata guts you use. I found the 4tb drives need a larger housing which is a few mm taller than a 2tb drive, which is a more 'elegant' solution - if you can call it that.

I print with ABS since it has a higher resistance to heat. But I haven't found there to be any heat issues - the drives I have used stay pretty cool - they don't generate the type of heat you see with an nvme. I have seen ssd's getting to 50 degree, but then others sticking at a max of 35. So that is probably worthwhile thinking about. ABS is good up to close to boiling point.

The sata assembly is in three parts since I first wanted to get as good a mount to the computer case as I could and then build up from there. The sata cable and the antenna leads have holes to thread through. While you 'could' change the sata drive once everything has been assembled it makes for a neater assembly if you assemble from the outside-in. You will figure out what I mean once you put it together. All fasteners are m3 countersunk.

I Had to upload as a zip file as that's the only format that would work for the step files on this msg board.

Btw, you don't need to destroy any sata drive. All the ones I've played around with, with the exception of things like the WD external drives, all come apart pretty easily - either plastic cases that can be unclipped or metal cases that have either an exposed set of screws or screws under a label. If you are going to go the full drive route you should take a hard look at if the design is strong enough - it's a different proposition having a small, static circuit board (ssd) in a small enclosure hanging off of parts of a case that doesn't have a lot of metal exposed to mount to, vs a larger assembly where there will be the loads from insertions. Not saying it won't work, or it will fail, just I designed with different assumptions and application. If I wanted to build the application you are thinking of, I 'may' look for a beefier solution - the challenge is not doing actual surgery on the case.

Hey, so which of the backplates i want? I see that i want faceplate sth not 0.1 since i want to use sdd, and i'll use 4tb version. But not sure about backplate. I think v2? Because those look like space for screws from the 4nvme expansion board that i can see in my device?