Kingston HyperX Predator M.2 SSD On Ubuntu: Linux Might Have Problems With It

[gigaplex]

I can't even see the graphs properly since the mobile site prohibits horizontal scrolling and zoom. All the useful information is cropped out. At the very least, allow clicking on the graphs and spec chart to open as an image in a new tab where I can pan and zoom.

[molecule-eye]

I can't even see the graphs properly since the mobile site prohibits horizontal scrolling and zoom. All the useful information is cropped out. At the very least, allow clicking on the graphs and spec chart to open as an image in a new tab where I can pan and zoom.

I've had this issue for ages too. You can sometimes put your device in landscape mode to fit it in. It's funny, because the whole site seems to have been designed with being response in mind, but then all the juicy data is stuck in these non-responsive tables.

https://techblog.livingsocial.com/blog/2015/04/06/responsive-tables-in-pure-css/

[NotMine999]

Check the Linux kernel configuration for: CONFIG_BLK_DEV_NVME That is a block device driver that can be compiled into the kernel or compiled as a module. I can see it as an option in Gentoo Linux 4.0.5 My guess is this: Without this driver loaded by the kernel, this type of SATA-like storage device is accessed via SATA/AHCI legacy mode. Intel contributed the Linux kernel code: http://lxr.free-electrons.com/source...ck/nvme-core.c

[s_j_newbury]

NVMe capable drives are still quite rare(I think the Samsung 950 Pro might actually be the first consumer targeted NVMe drive). This particular Kingston drive is PCIE x2 + AHCI according to the web

Even if so, then normally the SSD may well be able to operate as either legacy SATA directly through the edge connector and the AHCI controller on the adapter will be responsible for interfacing like with mSATA or with a logical AHCI controller on the SSD card interfaced through the PCIe interface.

From wikipedia.org on M.2:
PCB of an M.2 module provides a 75-position edge connector; depending on the type of module, certain pin positions are removed to present one or more keying notches. Host-side M.2 connectors (sockets) may populate one or more mating key positions, determining the type of modules accepted by the host; as of April 2014, host-side connectors are available with only one mating key position populated (either B or M).^[4][11][16] Furthermore, M.2 sockets keyed to support SATA or two PCI Express lanes (PCIe ?2) are referred to as "socket 2 configuration" or "socket 2", while the sockets keyed for four PCI Express lanes (PCIe ?4) are referred to as "socket 3 configuration" or "socket 3".^[1]:15[17]
For example, M.2 modules with two notches in B and M positions use up to two PCI Express lanes and provide broader compatibility at the same time, while the M.2 modules with only one notch in the M position use up to four PCI Express lanes; both examples also support SATA storage devices. Similar keying applies to M.2 modules that utilize provided USB 3.0 connectivity.^[4][16][18]

[nils_]

I can't even see the graphs properly since the mobile site prohibits horizontal scrolling and zoom. All the useful information is cropped out. At the very least, allow clicking on the graphs and spec chart to open as an image in a new tab where I can pan and zoom.

Chrome on android can be set to force enable zoom, so you can zoom on mobile "optimized" sites as well.

https://support.google.com/chrome/answer/2633135?hl=en

[gamaral]

Ran into the same issues, in fact the Kingston support dudes thought it was the drive and RMAd it. Unfortunately the new drive also has the same problems. I've noticed DIO is actually worse on it than on my Samsung SSDs. I did find it does not handle queued read commands correctly which could cause mayor slowdowns if read from it constantly (kernel patch in the works as I write this). Disappointing to say the least.

[gamaral]

Just confirmed that the Linux SATA driver makes the drive and/or link go wonky. If you boot straight into Windows from power-off it works as advertised.

[jwegner]

An update from my previous post. I tried using the Fedora 23 Beta, and my issues with the drive being recognized at boot have (so far) been resolved. I have not run any benchmarks yet, but I haven't noticed any terrible performance issues thus far. Once I get the test suite up and running I may be able to give another update.

[supertzar]

Holy thread revival!

I believe I've run into this issue as well, with similar symptoms to what gamaral reported. I purchased a new HyperX Predator 960G with the M.2/PCIe form factor. I get system freezes with it, see the type of errors below logged in syslog when it happens, and then eventually responsiveness returns once the SATA link resets. Granted this is an older SSD design, but I bought it since it has the built-in Option ROM, and my system won't boot with a NVME drive. It looks like this SSD is fubar for linux use, unless someone else has found a resolution. I can't find much else on the Internet other than this thread.

Feb 27 13:58:40 ganymede-work kernel: [ 109.940155] ata7.00: exception Emask 0x0 SAct 0x2 SErr 0x0 action 0x6 frozen
Feb 27 13:58:40 ganymede-work kernel: [ 109.940162] ata7.00: failed command: WRITE FPDMA QUEUED
Feb 27 13:58:40 ganymede-work kernel: [ 109.940167] ata7.00: cmd 61/08:08:30:b4:0b/00:00:67:00:00/40 tag 1 ncq 4096 out
Feb 27 13:58:40 ganymede-work kernel: [ 109.940167] res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
Feb 27 13:58:40 ganymede-work kernel: [ 109.940170] ata7.00: status: { DRDY }
Feb 27 13:58:40 ganymede-work kernel: [ 109.940174] ata7: hard resetting link
Feb 27 13:58:40 ganymede-work kernel: [ 110.268280] ata7: SATA link up 6.0 Gbps (SStatus 133 SControl 300)
Feb 27 13:58:40 ganymede-work kernel: [ 110.269559] ata7.00: configured for UDMA/133
Feb 27 13:58:40 ganymede-work kernel: [ 110.269562] ata7.00: device reported invalid CHS sector 0
Feb 27 13:58:40 ganymede-work kernel: [ 110.269566] ata7: EH complete
Feb 27 13:59:11 ganymede-work kernel: [ 140.931818] ata7.00: exception Emask 0x0 SAct 0x2 SErr 0x0 action 0x6 frozen
Feb 27 13:59:11 ganymede-work kernel: [ 140.931827] ata7.00: failed command: READ FPDMA QUEUED
Feb 27 13:59:11 ganymede-work kernel: [ 140.931836] ata7.00: cmd 60/08:08:00:40:47/00:00:6d:00:00/40 tag 1 ncq 4096 in
Feb 27 13:59:11 ganymede-work kernel: [ 140.931836] res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
Feb 27 13:59:11 ganymede-work kernel: [ 140.931840] ata7.00: status: { DRDY }
Feb 27 13:59:11 ganymede-work kernel: [ 140.931847] ata7: hard resetting link
Feb 27 13:59:11 ganymede-work kernel: [ 141.259953] ata7: SATA link up 6.0 Gbps (SStatus 133 SControl 300)
Feb 27 13:59:11 ganymede-work kernel: [ 141.261310] ata7.00: configured for UDMA/133
Feb 27 13:59:11 ganymede-work kernel: [ 141.261319] ata7.00: device reported invalid CHS sector 0
Feb 27 13:59:11 ganymede-work kernel: [ 141.261330] ata7: EH complete

[gamaral]

Just to update the thread, during my original post I talked with HyperX support and they managed to reproduce it (that was a year and a half ago). They replied with "It looks like our engineers have duplicated the issue and have reported it to our controller vendor for analysis."

No update from them, I ended up replacing the drive with a Samsung M.2 950 (now 960) EVO.