The NAND memory controller manufacturer advises that proper case air-flow and a large heatsink might not be enough past the upcoming generation of M.2 SSDs.
The NAND memory controller manufacturer advises that proper case air-flow and a large heatsink might not be enough past the upcoming generation of M.2 SSDs.
PC Enthusiasts and motherboard manufacturers both received a modest wake-up call last week as Phison, manufacturers of consumer and enterprise-class NAND memory controllers, warned that the next high-performance NVMe SSD generation might be the last to be supported by passive cooling alone. With Gen5 SSDs on the way and Gen6 models in early development, the age of fan-assisted SSD cooling may begin sooner than you think.
NVMe SSD cooling, particularly in the M.2 form factor, was a hot topic of conversation in the not too distant past. Flags were raised over the heat generated by first high performance/high capacity PCIe SSDs NVMe SSDs, particularly models with NAND chips present on both sides of the PCB. Elaborate heatsinks were criticised as a potential insulating layer, trapping heat between the motherboard and SSD PCB.
Discussion has cooled somewhat since then, but large aluminium heatsinks are an increasingly common sight amongst consumer PCIe Gen3 and Gen4 M.2 drives. Motherboard manufacturers have also begun to ship M.2 slot heatspreaders as standard with even mainstream designs. However, as storage performance continues to leap forward, the situation may become untenable with passive solutions.
As Phison discuss, PCIe Gen4 NVMe SSDs currently hit 5GB/s and early Gen5 testing indicates that the package could be rated at up to 12GB/s (or even beyond). With an industry rule of thumb sitting at 1 additional watt of heat generated for each GB/s of bandwidth, concern is mounting that passive solutions won’t keep up even with good case air flow.
Excess heat beyond the typical 70-85C of most NAND devices will cause thermal throttling, pushing your bandwidth down to a fraction of what you might expect. While some might think that will only be an issue for users that hammer storage I/O – through uncompressed video capture for instance – other applications designed to benefit from high speed storage may also be affected. That’s particularly pertinent to gaming given Microsoft’s introduction of Direct Storage into the Windows and Xbox API toolkit this month.
Shrinking the manufacturing node will tend to reduce power and voltage requirements, as will reducing the number of active NAND channels, but there’s a limit to the benefits these steps can provide. And as a result we may start seeing active (fan assisted) cooling on some Gen5 and Gen6 SSDs, i.e. M.2 heatsinks with a small fan bolted too them.
Such heatsinks won’t be huge and they’ll likely be required to remain within the M.2 standard footprint, so we’re not talking CPU-cooler size here. But they might be taller even than conventional VRM heatsinks, posing something of a challenge to system design and internal layouts.
As you might have guessed, there’s an additional wrinkle to this on most motherboards. The M.2 slot(s) is most often located adjacent to or underneath the main PCIe x16 slot; elaborate SSD cooling could easily interfere with these bulkier components. Furthermore, GPUs are only getting more power-hungry, dumping more heat into the case directly alongside the storage and necessitating even more efficient air-flow least it reach hot-box conditions.
According to Phison new consumer connector standards are in the works that may alleviate the issue somewhat by improving heat conduction through the motherboard. For now though, you might want to keep SSD cooling in mind – both case air flow and slot clearance – when buying your next motherboard if cutting-edge performance M.2 storage is part of your plans. Experts believe that heatsinks will be standard for the Gen5 SSDs due in 2022, and fans to augment air flow may be inevitable all too soon.
For a deeper discussion on the challenges involved check out Phison’s blog ‘Turn down the heat on SSDs‘.
