Silicon Motion Launches Third Generation PCIe Gen 4 SSD Controller for Future TLC and QLC 3D NAND Flash

[TheLostSwede]

Silicon Motion Technology Corporation ("Silicon Motion"), a global leader in designing and marketing NAND flash controllers for solid-state storage devices, today announced the SM2268XT, its latest high-performance PCIe Gen 4 SSD controller solution optimized for higher speed NAND transfer rates. The SM2268XT's superior performance and robust reliability allows customers to accelerate development of next-generation SSDs using current and future TLC and QLC 3D NAND flash with comprehensive data integrity and correction without compromising throughput and latency.

The SM2268XT features a dual-core ARM R8 CPU with four lanes of 16 Gb/s PCIe data flow and supports four NAND channels with up to 3,200 MT/s per channel, enabling designers to take advantage of higher throughput next-generation high-speed TLC and QLC 3D NAND flash. Its multi-core design automatically balances the compute load to deliver industry-leading sequential read and write speeds of 7,400 MB/s & 6,500 MB/s, and random read & write speed of 1,200K IOPS. In addition, its advanced architecture enables lower power consumption and rigorous data protection, providing high performance and reliability in a cost-effective DRAM-less PCIe Gen 4 NVMe SSD solution.




The SM2268XT offers a new system bus architecture, Host Memory Buffer (HMB) function and Silicon Motion's state-of-the-art 8th generation NANDXtend ECC technology with a performance-optimized 4 KB LDPC engine and RAID to maximize error correction capability. The SM2268XT is designed for next generation, cost effective TLC and QLC 3D NAND flash to enable high-value SSDs for a broader range of laptop PCs from performance to mainstream to value.

"We are proud to introduce the SM2268XT for Client SSDs," said Wallace Kou, President and CEO of Silicon Motion. "We have been scaling our SSD controller design wins with leading PC OEMs, and this latest evolution of our PCIe Gen 4 controllers demonstrates our commitment to developing products these customers want and need to support future higher speed NAND."

"We are excited about this high-performance PCIe Gen 4 controller from Silicon Motion," said Atsushi Inoue, Senior Director, Memory Division, KIOXIA Corporation, "We believe that SM2268XT will upgrade the standard of high-performance SSDs to the next level together with our leading BiCS FLASH technology."

"PCIe Gen 4 SSDs are experiencing explosive growth with shipments almost doubling in 2022 and will continue to be the mainstay storage solution in PCs for the next few years," says Greg Wong, President of Forward Insights. "Silicon Motion's latest SM2268XT DRAM-less controller enables cost-effective, high-performance storage in PCs supporting the most advanced 3D NAND technologies."

Silicon Motion is currently sampling SM2268XT to key customers.

View at TechPowerUp Main Site | Source

 

[Wirko]

Hm, the same product appeared as a 2022 product on the roadmap in your recent news post.

Silicon Motion's SM2504XT PCIe 5.0 NVMe SSD Controller Specs Leak

One of the major issues with M.2 based NVMe SSDs today is the excessive heat they output, part of the reason most of them run so hot is because of the older fabrication node they're made on. Most current PCIe 4.0 NVMe SSD controllers are built on a 12 nm node of some kind, but based on data out...

www.techpowerup.com

 

[TheLostSwede]

Hm, the same product appeared as a 2022 product on the roadmap in your recent news post.

Silicon Motion's SM2504XT PCIe 5.0 NVMe SSD Controller Specs Leak

One of the major issues with M.2 based NVMe SSDs today is the excessive heat they output, part of the reason most of them run so hot is because of the older fabrication node they're made on. Most current PCIe 4.0 NVMe SSD controllers are built on a 12 nm node of some kind, but based on data out...

www.techpowerup.com

CS: Dec'22 as in they were sampling customers in December last year, we're in mid Februrary now, so I guess they're ready to share it with everyone that the product has gone into MP (mass production) and is now shipping to customers so we can expect drives based around it, in a few months.

You made me correct a mistake in that post though.

 

[AnarchoPrimitiv]

More of a general question, but will this newer flash with higher MT/s finally make a substantive improvement to low queue depth random reads and writes? Because the random performance between an NVMe PCE 3.0 and 4.0 drive is basically the same and I'm hoping to finally see some improvements on something other than sequential performance.

 

[TheLostSwede]

More of a general question, but will this newer flash with higher MT/s finally make a substantive improvement to low queue depth random reads and writes? Because the random performance between an NVMe PCE 3.0 and 4.0 drive is basically the same and I'm hoping to finally see some improvements on something other than sequential performance.

I doubt it'll make a substantial difference, which is what I presume you meant?

This drive is according to the company that sells it, using Micron's 3D TLC B58R NAND, which is 232 layers and 2400 MT/s.

First Consumer PCIe 5.0 NVMe SSD Gets Tested, Makes a Lot of Noise

In Japan, the first consumer focused PCIe 5.0 NVMe SSDs have gone on sale and one of these drives has been put through some quick synthetic benchmarks by @momomo_us on Twitter. We're not familiar with the CFD Gaming brand which the drive is sold under, but the CSSD-M2M2TPG5NFZ—as the drive is...

www.techpowerup.com

 

[thewan]

I doubt it'll make a substantial difference, which is what I presume you meant?

First Consumer PCIe 5.0 NVMe SSD Gets Tested, Makes a Lot of Noise

In Japan, the first consumer focused PCIe 5.0 NVMe SSDs have gone on sale and one of these drives has been put through some quick synthetic benchmarks by @momomo_us on Twitter. We're not familiar with the CFD Gaming brand which the drive is sold under, but the CSSD-M2M2TPG5NFZ—as the drive is...

www.techpowerup.com

The SSD in that article is a reference based Phison E26 SSD, which only uses 1600MT flash. Since he is asking about higher MT flash, not PCIE Gen 5, that is not what he is talking about. Also here, have a better article:

Phison E26 B58R 1,600MT 2TB SSD Preview - PCIe Gen5 storage is here

Just as they were with PCIe Gen4, Phison is once again the first to bring PCIe Gen5 SSDs for mass consumption. 10,000 MB/s, oh yeah!

www.tweaktown.com

More of a general question, but will this newer flash with higher MT/s finally make a substantive improvement to low queue depth random reads and writes? Because the random performance between an NVMe PCE 3.0 and 4.0 drive is basically the same and I'm hoping to finally see some improvements on something other than sequential performance.

Yes and no. Higher MT does make some difference, but we need more controller and firmware optimizations for it to be a substantial difference. For example older Gen 4 Phison E18 SSDs with 1200MT vs newer ones with the same controller with 1600MT, which we do see some difference, but is it substantial? The Samsung 990 Pro for example uses 2000MT flash, and yet the "older" SK Hynix P41 is right up there with it despite using older flash (I can't find the MT its rated at but its 176 layer only, same as other 1600MT nand that most Gen 4 SSD rated for 7GBps seq speeds use.). This proves firmware optimization and vertical integration plus superior controller is needed to be "up there" not just raw flash performance.

Here's one of the first 2400MT flash SSDs:

Acer Predator GM7 1TB SSD Review - High Speed Flash

7,450 MB/s throughput from a 4-channel DRAMless SSD? Yep, that's what we've got in the lab for today's testing with the Acer Predator GM7 1TB.

www.tweaktown.com

But as we can see being DRAMLESS and being based on a budget 4 CH controller seems to hold it back, and yet it performs so well. So we also need moar better DRAM.

 

[Wirko]

More of a general question, but will this newer flash with higher MT/s finally make a substantive improvement to low queue depth random reads and writes? Because the random performance between an NVMe PCE 3.0 and 4.0 drive is basically the same and I'm hoping to finally see some improvements on something other than sequential performance.

No.

I know little about NAND technology but it's obvious by now that the ~40 us latency is the technological limit of flash itself. Much like the ~50 ns latency of DRAM. Everything that's faster is achieved by sequential access with minimum latency, massively parallel access, and queueing, which both technologies are highly optimised for.

(Yes, ironically the Dynamic Random Access Memory is good at many things but random access is not one of them.)

 

[Minus Infinity]

More of a general question, but will this newer flash with higher MT/s finally make a substantive improvement to low queue depth random reads and writes? Because the random performance between an NVMe PCE 3.0 and 4.0 drive is basically the same and I'm hoping to finally see some improvements on something other than sequential performance.

In short NO!. Scamsung were making noises about their new 230+ layer nand improving random io's, but so far the first gen 5 ssd with 200+ layers showed only small improvements, admittedly not Scamsung.