What are you looking for ?

SSDs Need Heat Management

As heat is bad, but extreme cold isn't great either.

Phison Jean 1648293461This blog was posted on March 21, 2022 by Phison Electronics Corp., written by Sebastien Jean, CTO.



Turn Down the Heat on SSDs
Why SSD Heat Management Is Key

When SSDs began to replace rotating HDDs across industries in the early 2000s, one of the touted side benefits was that SSDs used less power – and therefore generated less heat. At the time, heat management was virtually a non-issue for the latest drives.

SSDs have come a long way since then. They’ve evolved from ere MB/s of performance to 5GB/s and up, and early looks at Gen5 SSDs are reporting 10 to 12 GB/s and beyond. With greater performance and speed comes greater heat gen. Because today’s M.2 NVMe SSDs can overheat quickly, heat management is a necessity.

New SSD gens mean new heat challenges
In recent interviews with MSI Insider and StorageReview, Jean, CTO, discussed heat management.

He said: “As the speed continues to go up with each new gen, our challenge is to manage the heat. With Gen4, sometimes people need a cooler, sometimes they are okay with a metal foil label. If their board has enough airflow, it’s fine.”

SSD speed is measured in gigabytes per second, and he estimates that with each additional GB/s of speed, an SSD also requires approximately one more watt of power. And more power means more heat. While the power vs. performance relationship isn’t perfectly linear and changes in design or process can help, it’s a good simple benchmark.

Ignoring extreme heat in SSD can seriously affect its performance. The result is thermal throttling, which is when an SSD automatically slows down to stay below its maximum threshold temperature to prevent damage.

You know you need more cooling when you benchmark your SSD and the performance starts at 7GB/s and then [it suddenly slows down to] 500MB/s,” CTO said. “When you see that rapid drop, it’s very likely the SSD went into thermal throttling.”

Reducing power requirements can help reduce heat
At Phison, engineers are minimizing heat issues by managing growing power needs in the latest SSD gens.

We’re trying to stick within roughly the same power envelope as a 7GB/s SSD as we scale up to 14GB/s by making a lot of other changes,” said Jean.

One strategy is using a smaller process node – for instance, going down from 16nm to 7nm. Smaller process nodes can operate at higher frequencies with lower voltage. Also, less energy is needed to toggle the transistors, which in turn lowers the power used. Using less power means the SSD generates less heat.

Another way to decrease power requirements is reducing the number of NAND channels used by the SSD. That’s possible today, thanks to an improved ONFI bus speed, which is the rate at which data is moved from the NAND to the SSD controller.

CTO said: “In practical terms, you no longer need 8 channels to saturate the Gen4 and even Gen5 PCIe interface. You can potentially saturate the host interface with four NAND channels, and reducing the number of back-end channels reduces the total SSD power by typically 20 to 30%.”
What you can do to manage SSD heat?
There are lots of things that we’re doing to keep the SSD power within a reasonable envelope,” he said: “But for sure, the SSDs are going to be hotter, in the same way that CPU and GPU got hotter in the 1990s. As we move to Gen5 and Gen6, we may need to consider active cooling.”

He recommends a heatsink on Gen4 SSDs at a minimum and stresses the importance of good airflow. Most PC cases have decent airflow.

He said: “The M.2 motherboard connectors are near the PCIe connector, so you’ll get some airflow from your GPU. Just don’t neglect the heatsink on the SSD.

Once the SSD starts to get hot, he said: “There are 2 ways to pull that heat out. The first way is “the conduction path, which is, believe it or not, the little M.2 connector that plugs into the board. Not super efficient, but it still contributes.

There’s also a small screw at the back of the M.2 PCB that attaches the SSD to the motherboard. Thermal conductivity is based on the material and the volume. The small metal screw has more thermal mass than the gold fingers on the M.2 connector.

The screw represents 70% of the thermal conduction path from the SSD to the motherboard,” said Jean. Some vendors have started using a nylon screw on the motherboard to save money, which eliminates the most important conduction path.

The second way to dissipate heat is through convection, which occurs as heat comes off the SSD itself and warms the air around it. If a PC case has good airflow, it could eliminate the need for a heatsink.

The CTO likens the evolution of SSD heat management to what happened with gaming computers and GPUs over time. Early gaming computers typically had small metal heatsinks and eventually a small fan was added to manage heat. Today’s rigs have massive coolers and fans, and Jean sees more deliberate and robust cooling mechanisms down the line for SSDs.

“I would expect to see heatsinks for Gen5,” he said. “But eventually we’ll need to have a fan that’s pushing air right over the heatsink, too.”

When it comes to server-side form factors, he said: “The main thing is to have good airflow through the chassis itself, and the heatsinks essentially reduce the need for crazy, high-speed fans because it gives you a much larger dissipation surface. The EDSFF E1 and E3 specs have form-factor definitions that include heatsinks. Some hyperscalers are willing to trade off storage density in a chassis for a heatsink and a reduced need for high-speed fans.”

Temperature matters
So how hot is too hot? He said that NAND technology sets the temperature limits: “NAND, typically in the consumer space, operates from 0°C (32°F) to anywhere between 70 and 85°C (158 to 185°F), depending on the grade of the NAND. “And as heat goes up, retention of data in NAND goes down.”

As an SSD fills up, it becomes much more sensitive to heat. Jean recommends keeping and SSD under 50°C (122°F).

The controller and all the other components are good up to 125°C (257°F),” he said. “But the NAND isn’t, and the SSD will go into critical shutdown if it detects that the temperature of the NAND is above 80°C (176°F) or so.”

Heat is bad, but extreme cold isn’t great either.

If most of your data was written really hot and you read it really cold, you have a huge cross-temp swing,” said Jean. “The SSD is designed to handle that, but it translates into more error corrections. So lower maximum throughput. The sweet spot for an SSD is between 25 and 50°C (77 to 122°F).

Future of SSD heat management
As SSD manufacturers and their technology partners develop more and better heat management tools for SSDs, heat will become less of a concern for users overall.

Looking forward, Jean sees even more innovative solutions on the horizon: “If you look at the bigger question of where PCs are going, there’s an understanding that, for example, the M.2 PCIe Gen5 card, as it is today, has reached the limit of where it can go. The connector will become a bottleneck for future speed increases So new connectors are being developed and they’ll be available in the next few years. They will greatly increase both the signal integrity and the heat dissipation capability through conduction to the motherboard. These new connectors may allow us to avoid putting fans on SSDs.