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

[TheLostSwede]

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 of IT Home in the PRC, we now have details of Silicon Motion's SM2504XT PCIe 5.0 NVMe SSD controller, which appears to be one of the first, if not the first, to be made on a 7 nm node.

Although it's not mentioned specifically who Silicon Motion's manufacturing partner is, it's most likely going to be TSMC, as Silicon Motion is a Taiwanese company. The SM2504XT will be a mainstream 4-channel controller, but it'll support NAND flash rated at up to 3600 MT/s, which is a step up from any other currently announced NVMe SSD controller. It also supports the NVMe 2.0 protocol and will have a PCIe 5.0 x4 host system interface. The new controller is expected to have hit the engineering sample stage some time in September this year.



View at TechPowerUp Main Site | Source

 

[TumbleGeorge]

There is suggestion of what would be speeds of SSDs with this controller: 14/12 R/W. Or it is possible better than Phison E26.

 

[TheLostSwede]

There is suggestion of what would be speeds of SSDs with this controller: 14/12 R/W. Or it is possible better than Phison E26.

A lot will depend on the type of NAND they can get hold of. The issue right now, is that at least Micron has slowed down the production of NAND and alledgely its most dense product, so we might be stuck with the first generation of PCIe 5.0 topping out at around 10k R/W.

 

[Wirko]

A lot will depend on the type of NAND they can get hold of. The issue right now, is that at least Micron has slowed down the production of NAND and alledgely its most dense product, so we might be stuck with the first generation of PCIe 5.0 topping out at around 10k R/W.

The NAND with the fastest data bus currently achieves 2400 MT/s, and four channels of that yield 9600 MB/s, right? Is there anything faster already in sight?

 

[TheLostSwede]

The NAND with the fastest data bus currently achieves 2400 MT/s, and four channels of that yield 9600 MB/s, right? Is there anything faster already in sight?

More recent NAND is already at 3200 MT/s, which is why we've already seen drives that can do just above 10 GB/s.
However, to hit 12 GB/s+ the 3600 MT/s NAND is needed and it seems to have been put on hold for now.

 

[Wirko]

More recent NAND is already at 3200 MT/s, which is why we've already seen drives that can do just above 10 GB/s.
However, to hit 12 GB/s+ the 3600 MT/s NAND is needed and it seems to have been put on hold for now.

For those who seek top performance there will be 8-channel controllers anyway, and 2400 MT/s probably won't be a bottleneck for those.

 

[TheLostSwede]

For those who seek top performance there will be 8-channel controllers anyway, and 2400 MT/s probably won't be a bottleneck for those.

Sure, but those controllers might not appear in consumer drives. They'll most likely also run hotter.

 

[bonehead123]

SilMo + TSMC = a match made in heaven....

Bring it !