Pretty much all of them as we move into 3D NAND... I know samsung has one out already as the 850 EVO... everyone should be right behind them with similar products.
Newer 3D NAND is using a charge trap design which basically solves the electron leakage issue found with the older floating gate NAND...
Also, the move to the newer 3D NAND brings us back up to 40nm processes vs the 10nm gates we are currently working with, allowing for much better reliability.
Disclaimer: I've been selling enterprise flash storage for the last 6 years.
I get hit with 1-2 job opportunities every day or two from LinkedIn alone...
Some are good, some are cruft... it all becomes noise since I'm not looking for a job right now.
You can't read game files faster because the process that reads the game file is single threaded vs multi-threaded... so it can only read as fast as a single thread can read.
A fully saturated CPU has enough lanes to do about 26-28GB/sec, but a single I/O thread might only be able to do 100-200MB/sec.
There was never any reason in the OS before today to make that any faster because the spinning disks that fed data to the CPUs couldn't do more than 100MB/sec.
Now that we have all this great flash, code needs to be re-written to be able to use other idle CPU time to spin up more threads and read the data faster.
It's worth remembering that 98k IOPS will be at a very small block size and will rapidly drop as you increase block size to 4K-1MB as the larger transfer size will directly equate to less I/O.
The real problem is that apps are not written for multi-threaded I/O which is what you really need in order to take advantage of the throughput provided by PCI-e flash.
Booting from RAID is more supported, and the support is baked into nearly every BIOS out there... booting PCI-e over NVMe or UEFI is brand new and very few things support it and all the code is new.
Booting through a raid card that has it's own BIOS is nothing like booting off of a native PCI-e device.
Booting from pci-e uses either a UEFI driver or NVMe today, two technologies that are kinda in their infancy.
The code is not yet fully optimized/etc and you may see reduced speeds at least until you can get into an OS layer and load up a more feature-full driver.
The PCI-e native SSDs are indeed faster, the problem is, the code reading data off of them (your application/os) isn't written to take advantage of the increased speeds. Single threaded reads cap out at the read speed of a single thread, and that isn't that fast. This is especially true if they are 4K reads vs 1M reads, as you aren't going to saturate anything until you get up into larger read sizes.
To really take advantage of the bandwidth SSDs enable, you need to be running multiple parallel apps running multiple reads, or you need an app that can do multi-threaded reads.
This is because you get stuck leaving the CPU to handle all the context switching between virtual block storage in DRAM and memory. The CPU has to copy data out of block and into memory before it can actually use it, so by making a ram disk you end up giving the CPU 2-4x the amount of work to do for what should be a DMA read/write, which would normally be offloaded.
Also, your reads from your game are going to be single-threaded, and a single read/write thread is going to be pretty slow.
That's how I do it at least...
There's no reason to use SLC these days really... Once you start writing in large density with intent on retaining data for some period of time, you'll be striping that data across 10-100 SSDs... The combined wear-life even with cheaper MLC drives still puts you up over 100 years for most products.
It's pretty easy to take the Drive Writes Per Day (DWPD) or PetaBytes Written (PBW) for the drives and add them all up... most any install will 10+ drives will outlast any standard 5 year hardware refresh cycle.
Disclaimer: I work for a large flash company and have been selling this stuff for the last 5 years.
You don't want to just cut fuel with the same air intake... then you just run lean and ruin the motor with detonation.
The car simply closes the butterfly valve, cutting air, and thus cutting fuel as a result of less metered air in the plenum.
...but I appreciate what their marketing team has done for the rest of the headphone industry...
...is that in a properly-designed SSD, there is no such thing as data fragmentation. You lay out the nand as a circular log and write to every bit of it once before you overwrite, and maintain a set of pointers that translates LBA to memory addresses.
Pretty much every SSD vendor out there has figured this out a few years ago.
I have a PDF scan of all important IDs/health cards/etc on a drive in my safe deposit box. It's also where I store my long term email/document archives.
I keep a mirror at home, which is what I update most frequently and any time I go to the bank, I just swap the external home drive with the one in the safe deposit box, go home and rsync the current data to it.
My safe deposit box key lives in a floor safe in my home which should survive even a gas leak explosion/tornado/etc.