News Silicon Motion reportedly prepping SM8466 SSD controller with a PCIe 6.0 x4 — leak claims it will be unveiled at FMS 2025, sporting speeds of up to...

The article said:
Silicon Motion's MonTitan SM8466 SSD controller for enterprise SSD
...
As for client PCs, Kou expects SSDs with a PCIe 6.0 x4 to address PC OEMs only in 2030 or later.
Thanks Anton. Just wanted to emphasize these points, in case anyone missed them.

The article said:
the company expects its partner to start shipping their SM8466-based products sometime in late 2026 – early 2027
There won't even be PCIe 6.0-capable CPUs or servers until 2026.

BTW, I was interested to see no mention of CXL support.
 
Enterprise as oppose to consumers has no problems with many 16x PCIe channels. Consumers actually also can afford motherboards with 4-5 16x PCIe slots if they are really needed. I for example add 3 additional M.2 NVMe SSDs via PCIe enclosures on the motherboard. In these enclosures only 4 channels are used, others are empty. Or these not used lanes can be used for other 3 additional M.2 drives using bifurcation. So why restrict enterprise to just 4 channels on the SSD and not use other 12? This way speeds like 28GB/s would be available 5 years ago when PCIe4.0 appeared.

Besides such dumb restrictions to limit themselves just to 4 channels the problem always was in the lag in the development of advanced nodes controllers which were always hot like a solar surface because used obsolete technology.

Resolve these two restrictions and you will be able to think about not 28 but 128 GB per second drives on these PCIe 6.0 motherboards. I suppose the controller developers have the early access to the PCIe6 or even 7
 
Last edited:
Enterprise as oppose to consumers has no problems with many 16x PCIe channels.
That stat was actually in reference to NAND channels, not PCIe lanes. In general, the more channels a controller has, the more NAND chips you need to fully-saturate it. I think the current top-end consumer SSDs have only 8 channels. To feed a controller with 16 channels, you would probably need more NAND chips than you could fit on a M.2 drive.


As mentioned in the headline, the controller has only x4 PCIe lanes. The reason why they use PCIe 6.0 x4 instead of 5.0 x8 or 4.0 x16 is to optimize density, so that you can pack the highest number of SSDs per server.
 
ssd nvme 3.0 ok
ssd nvme 4.0 hot but you can place some metal to cooldown
ssd nvme 5.0 You need direct touch heat pipes to get to work
ssd nvme 6.0 You will need tower style heatsink or water cooler to keep at check levels... and don't forget your extra power cable.

Some one need to bring The PC Climate Change.
 
That stat was actually in reference to NAND channels, not PCIe lanes. In general, the more channels a controller has, the more NAND chips you need to fully-saturate it. I think the current top-end consumer SSDs have only 8 channels. To feed a controller with 16 channels, you would probably need more NAND chips than you could fit on a M.2 drive.

As mentioned in the headline, the controller has only x4 PCIe lanes. The reason why they use PCIe 6.0 x4 instead of 5.0 x8 or 4.0 x16 is to optimize density, so that you can pack the highest number of SSDs per server.
I have not seen the actual spec and what specific market they are targeting, but first reaction is that using 4 NAND channels would be detrimental for reaching largest bandwidth specifically with these rates. They mention about PCIex4 and I commented about this. As to chips, they can be made with any number of channels. SN850X for example uses 8 NAND channel controller but uses only 2 chips. It is among fastest on the charts
 
As to chips, they can be made with any number of channels. SN850X for example uses 8 NAND channel controller but uses only 2 chips. It is among fastest on the charts
I'd be surprised if there's any NAND on the market with more than 4 channels per package. The only 16-channel controllers seem to be for datacenter drives and target much higher capacity than you could fit on a M.2 drive.
 

TRENDING THREADS