Phison PS5016-E16 and PS5019-E19 PCIe 4.0 Client SSD Controllers - AMD Invested Heavily
Taiwanese Phison has been showcasing a range of PCIe 4.0 NAND storage solutions at the Computex, from the flagship PCIe 4.0 PS5016-E16 (E16) SSD controller for gaming, to storage solutions aim for AIoT, automotive, and edge computing applications.
One of the applications that can really benefit from the bandwidth of PCI Express 4.0 are SSDs, Phison has developed and introduced the "PS5016-E16" SSD host controller. Current high-end products boasts a read of 3400 MB/s and a write of 3200 MB/s. That number now can nearly double.
For the PS5016-E16 word is that AMD to invested $15 million to help develop the controller in 9 months. The first PCIe 4.0 compliant SSD controller PS5016-E16 introduced by Phison is manufactured with a 28 nm process. It supports flash memory connections up to 8 channels and can support up to 8TB. The DRAM cache has a capacity of 1/1,000 of the total NAND capacity. In order to support 8TB, two DRAMs (4GB x 2) need to be installed, and so on, so at this stage, 2TB is the maximum capacity of general installed products.
The performance has already been confirmed in several articles here at Guru3D, but the nominal value for such NVMe SSDS are roughly is 5000 MB/s for reads and ~4400MB / s for writes. That is about 40% faster than the high-end class SSD of PCIe 3.0 x4.
Galax to launch HoF Phison PS5012-E12 based M.2. and Add-in SSDs - 03/29/2019 10:28 AM
Galax has been working on SSDs for a while now, newly introduced is what you'll see below the fold. And all white and perhaps a bit overengineered heat pipe cooled Phison PS5012-E12 based M.2. SSD s...
Phison PS5012-E12 goes volume production: Over 20 new SSD pending - 09/10/2018 09:00 AM
Phison is a name that you probably have not heard off that much, but in fact, a lot of Value SSDs make use of their controllers. In fact the higher end Corsair models, for example, use them. The Phis...
Senior Member
Posts: 1239
Joined: 2014-07-22
AMD has been doing things like this for much of its history. As tunejunky says, it's leveraging every advantage against the financial might of Intel, not to mention Intel's ability to mount massive and often completely false/misleading ad campaigns. And it's just as advantageous for the partnering companies as it is for AMD.
Senior Member
Posts: 1239
Joined: 2014-07-22
You can do what I did, I just ditched my 512 970 Pro for 2 x XPG SX8200 Pros in Raid 0 on a mini-ITX. 4GB write, 6GB read. Many of the full-size boards have 3 NVMe slots now.
You can't go wrong either way. It's amazing how cheap storage has become with 3D stacking. DRAM prices to follow.
Yes, there is a lot of misleading information about RAID 0 still circulating--like the complaint that it has no fault tolerance, for instance. However, there is no fault tolerance with standard IDE/AHCI drives, either! You lose one drive in a two-drive RAID 0 configuration, you lose all your data--however, if you lose your single IDE/AHCI drive--you also lose all your data. Lots of old wives' tales about RAID 0, unfortunately. I used RAID 0 for many years without incident--but I will add the caveat that it's done better, I think, with a dedicated hardware controller--but that's just me thinking back. AMD's x370/x470 motherboard RAID 0 support seems very solid from all accounts.
Senior Member
Posts: 13295
Joined: 2004-05-16
...
People talk about fault tolerance in relation to raid 0 because the rate of failure is increased, not the same as a single drive. Also because if you have two drives it's worth noting that you could mirror those drives for a performance/storage tradeoff, vs having no fault tolerance. Which isn't even an option to talk about with a single drive. It's not a old wives tale, it's just something you mention because the user having two drives enables you to talk about it lol
Senior Member
Posts: 132
Joined: 2015-03-30
...
People talk about fault tolerance in relation to raid 0 because the rate of failure is increased, not the same as a single drive. Also because if you have two drives it's worth noting that you could mirror those drives for a performance/storage tradeoff, vs having no fault tolerance. Which isn't even an option to talk about with a single drive. It's not a old wives tale, it's just something you mention because the user having two drives enables you to talk about it lol
I don't know that the rate of failure is increased, but the potential definitely is doubled. I think fault tolerance is an enterprise issue as they're usually running 24/7 and there are still a lot of physical platters in the enterprise. That is slowly changing as the cost of enterprise ssds comes down. With physical disks and early 2.5" SSDs at the consumer level the rate of failure may have been high, but m.2, I think it's a non-issue nowadays. Having said that, Macrium reflect is on an automated schedule, it gives me a sense of security. It's awesome.
Senior Member
Posts: 132
Joined: 2015-03-30
You can do what I did, I just ditched my 512 970 Pro for 2 x XPG SX8200 Pros in Raid 0 on a mini-ITX. 4GB write, 6GB read. Many of the full-size boards have 3 NVMe slots now.
You can't go wrong either way. It's amazing how cheap storage has become with 3D stacking. DRAM prices to follow.