Yeah, that's another truly great thing about Slackware... Pat's notes in the default config files. I've also learned a lot that way.
As does Slackware!

OK I'm now pretty convinced to buy a SAMSUNG 970 EVO PLUS M.2 2280 1TB PCIe Gen 3.0 x4, NVMe 1.3 V-NAND 3-bit MLC Internal Solid State Drive (SSD) MZ-V7S1T0B/AM although I could easily be convinced to spend the extra for the Pro version.

I do have a bit of a problem though in that...

1) Asrock Z77 Extreme 4 has no M.2 or U.2 slots but does have spare PCIe slots

2) Asrock Z77 Extreme beta BIOS supports NVME but many adapters for PCIe installation say they require Z97 or newer chipset.

I'm currently looking at this adapter that notes it is Linux compatible but it, too, says Z97 or newer.
https://www.newegg.com/silverstone-m...quicklink=true or https://tinyurl.com/yc5mfwrf

It appears Amazon has a wider selection of adapter cards and some don't mention chipset restrictions but a) I'd rather avoid Amazon if possible, and b) I don't know if the restriction exists and just isn't stated.

Anyone using a PCIe adapter for m.2 drives on older chipsets please jump in!

EDIT - Further research seems to say this may not be a chipset issue but a BIOS issue and my beta BIOS may have fixed it. It may boil down to UEFI boot vs/ MBR boot compatibility.

Let me assure I have been running NVMe board in a PCIe slot for years as bootable device on a Asrock 77 Extreme 4 Motherboard as I previously stated.
partitioned as follows:
HTH
John

3 members found this post helpful.

Even if it wasn't supported (which it seems from AlleyTrotter's post that it is), you could just have your /boot/ on another device that is bootable from your motherboard while keeping the rest of your filesystem on the NVMe.

And the 970 EVO Plus is a great one. I put a 250GB one in my HTPC (built that 2 years after the desktop). I've only written 596GB to it.

(But looking at this one confirms my suspicions that the Power On Hours readout isn't accurate. This is saying I only have 509 hours of on time, which is 21 days. I've had this machine running 24/7 since it was built almost a year ago.)

EDIT: Stupid work internet had my computer post it twice.

Solid choice. I have had several SSD's crap out over the years but never a Samsung. I currently have two 970EVO nvme's (256 & 512) in my Slackbox for the past year or two without issue. They really are a cut above the rest in my somewhat anecdotal experience.

I don't think there's any benefit for the above average home user to get the Pro. If you look at the charts on this article (which pits the regular 970 EVO vs 970 Pro, not the 970 EVO plus), you'll see the Pro offers 200MB/s faster write speeds, which when compared to SATA drives (including SSDs) seems like a lot, but it's just an increase from 2.5GB/s with the EVO to 2.7GB/s on the Pro. Read speeds are the same at about 3.5GB/s. But then when we start talking about the EVO Plus, the write speeds actually surpass the Pro by 600MB/s and are at 3.3GB/s compared to 2.7GB/s (read speeds stay the same at 3.5GB/s).

The only benefit the Pro has is the rating for TBW (TB written). The Pro has double the TBW warranty/rating as the EVO and EVO Plus, with the 1TB being rated for 600TBs on the EVO/EVO Plus and 1.2PBs on the Pro. Considering in 2.5 years I have only written 10TBs of data, I would never need the TBW rating provided on the Pro and I imagine almost all home users (even major power users) would have no need for the Pro.

1 members found this post helpful.

You're probably right, but there isn't that much of a price difference and the "peace of mind" factor is priceless.

In reality OP can't go wrong with either. They're both excellent. I use the Pro version in my office server and have an EVO in my laptop and another one in my home desktop.

But he's looking at the EVO Plus, which has 600MB/s faster write speed than the Pro. The only benefit with the Pro is having a TBW rating of 1.2PB vs 600TB. But for even someone like me who makes no changes in how I use my system with NAND compared to HDDs, it would take me 50-100 years to hit that limit. By then, 1TB NVMe drives would probably be the equivalent capacity and speed of a 3.5" floppy today. However, the warranty period remains the same for both at 5 years (or hitting the TBW, whichever happens first).

To give an example of how monstrous 1.2PB is, if you were to take that 1TB drive and write 100GB each and every day (which is a ridiculous amount to do daily), it would take 12,000 days to hit 1.2PB, which is almost 33 years! If you go down to *only* 600TB, you'd be looking at 16.5 years until you hit that limit! And if this article is any indication, a lot of drives are able to go well past their TBW ratings (all the drives tested passed the 700TB mark and this hardware is from 2014 or earlier). The longest lasting, a Samsung 840 Pro that was rated for 73 TBW hit 2.4PB of data written (33x its rated amount)! If this is an indication of what the 970 Pro can do, you're looking at 19.7EB (19,700PB), although, I doubt that endurance has increased evenly over the years.

Totally agree here. Whether you have the "slower" 2.7GB/s of the Pro or the "faster" 3.3GB/s of the EVO Plus, both will be EXTREMELY fast compared to anything SATA based (IIRC they top out at 600MB/s).

All good points sir!
Yeah. It's pretty frustrating going back to a non-SSD computer now. Everything takes so long! HDDs were the Achilles' heel of computer performance.

I've converted all of my computers except for my home NAS (which runs 4 x 2TB WD Reds in a RAID... still going strong after 12 years and not keen to replace it yet) and two of the computers at the office, one of which runs a service we're transitioning away from and will be decommissioned soon.

Thanks bunches! You guys have convinced me I can choose a bit of bandwidth over guaranteed longevity without loss of peace of mind. As for speed, isn't it a laugh riot how quickly we adapt to new levels? Improvements in specific areas ebb and flow but I recall my first 386 and how fast it seemed even compared to a server level 286, and then, just a few years later after a 486DX2 knocked my socks off, the 386 was agony. LOL.

I watched a video of a guy copying a 2.7GB file from one part of a Samsung EVO Plus to another that was nearly instantaneous !!! So today I pulled the trigger on a 1TB Plus and a Silverstone PCIe adapter. I'm so pumped! Thanks everyone.

My only platters are for mass storage. I currently have about 46TB of HDD space on my desktop/server. I couldn't afford to do that in NAND. But I don't need fast speeds for that and the system itself is on NVMe.

@enorbet, you won't be disappointed!

I totally agree with the other posters here..

NVME drives are as much faster than SSDs as SSDs are than spinning rust, because it uses the PCI-X bus directly for data transfer, rather than depending on the SATA protocol, which is better optimized for spinning rust than flash memory.

UPDATE - Just Got It !!...

but have some issues. I practiced with hard drive UEFI installs for a week. It took that long because the old Intel system I practiced on has a buggy/damaged BIOS and possibly an older version of UEFI but did get a workaround success. The very first time practicing on my Asrock Z77 Extreme 4 went right in and booted perfectly. I was able to hit the F11 Boot Order key to continue to boot all my older MBR stuff. I'm not adept at all with either elilo.conf or refind.conf yet, though I now have some familiarity with efibootmgr from all the hoop jumping with the Intel boxen. Hard drive EFI worked. NVME EFI did not.

After the nvme boot failed I again booted my DVD Slackware install disk as UEFI and chrooted in to the 970EVO nvme install and installed rEFInd and looked for anything remotely obvious in the .conf files. I am pretty dismayed at the poor quality of error messages. Often it will just say "Going interactive due to config error" with not a clue as to what or where that error is. An additional issue is that since the GPT ef00 Boot partition is mounted by fstab as "/boot/efi" it can be very confusing as to what are correct paths before a file system is even loaded. It also often made more difficult as some configurations are written to nvram and those don't autoupdate. Frankly it is far easier to remember to run "lilo" after a new kernel is installed than to deal with stuff burned into nvram.

Right now I'm thinking I may have to set the ext4 2nd partition that I installed -Current on to MBR and just use lilo as the drive is only 1TB, I assume it should work fine though I cant yet tell how the Asrock bios sees nvme at boot time.

Well at least I'm learning new stuff and the performance is insane. From my old MBR booted Current usding hdparm I'm getting buffered and direct reads around 2600 MB/sec. Cached reads are 13058 MB/sec ! 0_0 I surely do want to boot from that. Maybe later today.....

Thanks again for all the comments and help.