M.2 SSD help sought

[bhelmes]

Originally Posted by PhilF

How much data have you written to it?

Linux mint needs 12 Gbyte space on the disk, without any optimations.

I think you can add a pci-card for 2 or 4 M.2 cards for every pci-slot.

It may be interesting how long the M.2 card holds.

I bought 64 Gbyte ram (expensive) and a M.2 card (cheap) in order to run my programs in the first level cache.

[PhilF]

Originally Posted by bhelmes

I think you can add a pci-card for 2 or 4 M.2 cards for every pci-slot.

...

I bought 64 Gbyte ram (expensive) and a M.2 card (cheap) in order to run my programs in the first level cache.

Uhm, that's not quite how it works. For a M.2 card to operate anywhere close to level 3 cache speed, it has to be installed in a motherboard that has a special M.2 slot. That way it is attached directly to the processor's bus and can operate at full speed.

I'm not saying you can't get PCI cards that holds M.2 cards, but I am saying if you use them in that manner it makes accessing the data on them much slower, closer to standard SSD speed.

[chalsall]

Originally Posted by kriesel

Well there was that one time I formatted a 1TB rotating rust Seagate in an external USB enclosure, and that was the end of being able to access it.

Please don't get me started on SeaCrap...

Originally Posted by kriesel

Too late for me. Good advice for other potential buyers.

Indeed. I also learnt something here. Never would have thought of that!

Originally Posted by kriesel

The jury's out yet on the 2TB; no errors shown yet at over 190GB / 10%. Write rates are poor, ~4-8MB/sec oscillating at ~1Hz.

I was interested in just how bad that was, so quickly ran some experiments...

Burrow is my main workstation (M2.SSD attached to the MB and polished rust). ArchitectsCubed is a certain client's public-facing VPS with SSD.

[root@burrow tests]# dd if=/dev/zero of=sixteen_gig.zeros bs=1G count=16 oflag=dsync
16+0 records in
16+0 records out
17179869184 bytes (17 GB, 16 GiB) copied, 25.3828 s, 677 MB/s

[root@burrow tests]# dd if=sixteen_gig.zeros of=/dev/null bs=1G count=16 oflag=dsync
16+0 records in
16+0 records out
17179869184 bytes (17 GB, 16 GiB) copied, 7.79819 s, 2.2 GB/s

[root@burrow tests_rust]# dd if=/dev/zero of=sixteen_gig.zeros bs=1G count=16 oflag=dsync
16+0 records in
16+0 records out
17179869184 bytes (17 GB, 16 GiB) copied, 141.977 s, 121 MB/s

[root@burrow tests_rust]# dd if=sixteen_gig.zeros of=/dev/null bs=1G count=16 oflag=dsync
16+0 records in
16+0 records out
17179869184 bytes (17 GB, 16 GiB) copied, 122.485 s, 140 MB/s


root@architectscubed tests]# dd if=/dev/zero of=sixteen_gig.zeros bs=1G count=16 oflag=dsync
16+0 records in
16+0 records out
17179869184 bytes (17 GB) copied, 34.4538 s, 499 MB/s

[root@architectscubed tests]# dd if=sixteen_gig.zeros of=/dev/null bs=1G count=16 oflag=dsync
16+0 records in
16+0 records out
17179869184 bytes (17 GB) copied, 31.1031 s, 552 MB/s

Originally Posted by kriesel

If there's interest, I could follow up with a "2TB SSD" testing update later.

Please. We learn faster together!

[chalsall]

Originally Posted by chalsall

I was interested in just how bad that was, so quickly ran some experiments...

Sorry, everyone. I realized an assumed variable not given...

kriesel's low bandwidth /could/ be because of the USB interconnect, not the "SSD" actually provisioned...

If I may please say Ken... I can't wait to see you crack those products open and post the pictures here. Just to see what you were sold...

Learning opportunities abound. Add this to your reference material...

Seriously...