Update (8/15/2023): Thanks to one of our readers who clicked on the link in the original report for one of the drives said to be used in the test we’ve become aware that this test may not actually be as it was titled in said report. According to Micron’s spec sheets, the drive used is actually a TLC drive. We have reached out to Scalability regarding confirmation of which drive was in fact used and will update upon notification.
A new report shows that the belief that SSDs are more power efficient than good old-fashioned platter drives may not be completely accurate. Solid-state drives began taking over as the go-to storage medium for consumers, and select hi-density enterprise solutions roughly a decade ago. Prices continue to drop as new technologies are introduced and older ones are more widely produced, and this is still true for both mediums. Now while the new report examines a very specific set of metrics it does expose a commonly misunderstood belief about the newer flash-based drives. The new report shows that specifically, hi-density QLC SDDs can use a significantly greater amount of power during read/write tasks.
QLC SSD vs. HDD
| Power Data Per Drive | SSD | HDD | HDD Advantage |
|---|---|---|---|
| Idle (watts) | 5 | 5.7 | -14% |
| Active read (watts) | 15 | 9.4 | 37% |
| Active write (watts) | 20 | 6.4 | 68% |
| Read-intensive workload (avg. watts) | 14.5 | 8.7 | 40% |
| Write-intensive workload (avg. watts) | 18 | 6.6 | 63% |
| Power-density read-intensive (TB/watt) | 2.1 | 2.5 | 19% |
| Power-density write-intensive (TB/watt) | 1.7 | 3.3 | 94% |
The above data was provided by Scality (via Tom’s Hardware) who was examining the power efficiency of hi-density drives which are more commonly found in enterprise or data center servers. Now as most consumers are unlikely to have such drives in their homes, except for the occasional enthusiast with a larger NAS or media server, the data may not be directly relatable but it’s still worth noting for those trying to keep their power usage to a minimum. Scality used a Micron 6500 ION 30.72 TB QLC SSD and a Seagate Exos X22 7200 RPM 22 TB HDD. These obviously are not like-for-like comparisons and as the SSD has a significantly larger storage capacity so it should be expected to consume more power than a similarly sized drive, but even then the numbers are a bit more extreme.
And the winner is. . .
Now the Seagate HDD clearly comes out on top as being more power efficient as it only loses out during idle operation but at that point, it’s only by 0.7 Watts, 5.7 W vs 5 W. Meanwhile, the Micron QLC SSD displays a bit more of an appetite for power when in use. Testing showed that the Seagate HDD could be anywhere from 19% to 94% more efficient in its power consumption than the QLC SSD. It was explained that two types of workloads were looked at, one more focused on read-intensive tasks and the other on write-intensive tasks: 10% idle/80% read/10% write and then 10% idle/10% read/80% write. The calculation methodology used for the data (which was rounded to the nearest tenth) in the above table was as follows.
Per Scalability
“Micron ION:
- Power consumption (read-intensive): (5*0.10 + 15*0.8 + 20*0.10) watts = 14.5 watts
- Power density (read-intensive): 30.72 TB / 14.5 watts = 2.1 TB / watt
- Power consumption (write-intensive): (5*0.10 + 15*0.10 + 20*0.80) watts = 18 watts
- Power density (write-intensive): 30.72 TB / 18 watts = 1.7 TB / watt
Seagate EXOS:
- Power consumption (read-intensive): (5.7*0.10 + 9.4*0.80 + 6.4*0.10) watts = 8.7 watts
- Power density (read-intensive): 22 TB / 8.7 watts = 2.5 TB / watt
- Power consumption (write-intensive): (5.7*0.10 + 9.4*0.10 + 6.4*0.80) watts = 6.6 watts
- Power density (write-intensive): 22 TB / 6.6 watts = 3.3 TB / watt”
The researchers do state that power consumption should not be the main metric for making a purchasing decision. In fact, there are many other details to be looked at and from consumers to prosumers the type of workload and each user’s unique needs ultimately should be the deciding factors but this report does provide a measure of myth-busting.
Discussion (19 replies)
Join Discussion →Since in regular use most of the time your drives sit idle this means SSDs are still far more power efficient in practice.
Well, that, and task efficiency should still be at play; i.e., how much energy does it take to read or write 1TB of data?
This depends on the energy used for the amount of time. Further, if a QLC or similar low-endurance NAND flash configuration drive is used constantly, it will almost certainly wear out sooner. Which means that these aren't catch-all solutions either.
Finally... QLC isn't cheap enough to really justify replacing 3D NAND and its ilk if performance is at all a concern.
From an enterprise standpoint even in flash arrays you can have a mix of the different flash types. Once vendors get on board with that idea for tiering then the cost could go down and durability be maintained.
And honestly spinning disk drives in an array setup even if they are 'idle' are still spinning and drives like that still fail. Happens all the time, I've seen raid's loose drives that literally had zero data on them. Because of the health check monitoring that the host is doing for that SAN.
And honestly if they are talking high density drives, but not enterprise drives... ok sure... But I do hope anyone actually running high density storage has a backup or redundancy solution built in. That's a lot of data to risk even for a consumer/prosumer.
But I do like the idea of tiering and modified maintenance based on durability of the hardware. High durability high speed flash as your fast cache. Tier 1 being frequently accessed data with high read write durability. Tier 2 being your less often accessed data on less durable but larger denser storage. Tier 3 being some chunky spinning disks... maybe.
I'd prefer to stay all flash. At least for those the health checks don't hit the cells and are more logic than physical block testing. And even some of the slower less durable flash is still going to be more immediately responsive than spinning disks.
That is listed in the article, where the HDD still "wins" albeit only by a small margin. But if you have any kind of random access performance demand HDDs are out of the question anyway.
For home use I think it is. The performance hit is non-existent in daily use, and I'm never going to get anywhere near the stated endurance.
I agree in general, though under sustained writes QLC will throttle excessively. It will continue to write of course, just slowly; I'd observed <100MB/s on my 660p when imaging the drive.
According to Micron's 6500 ION SSD Series Technical Product Specification, the SSD is equipped with Micron 3D TLC NAND Flash.
Main product page: https://www.micron.com/products/ssd/product-lines/6500-ion
A review of the drive also lists 3D TLC NAND in the specs:
[URL unfurl="true"]https://www.storagereview.com/review/micron-6500-ion-ssd-review[/URL]
I could be missing something as I'm half-asleep and just skimmed the first results from my search.
Regarding the acceptability of QLC in general, it may have its uses, but not in my personal machines. I want 3D TLC NAND at minimum. High-speed writes are of little use if the drive doesn't have the endurance to take advantage of them.
Scality even linked to Micron's product brief, which clearly states "Micron 232-layer 3D TLC NAND" in the specifications. That makes it hard to take their article seriously.
Those researchers biffed it. You got me curious and I clicked on the hyperlink in their own article.
https://media-www.micron.com/-/media/client/global/documents/products/product-flyer/6500_ion_nvme_ssd_product_brief.pdf?la=en&rev=0551c45ada82469aa6162c87473e2b06
I think whoever was doing their write-up got hung up on all the QLC mentioning that's all over the place (16x according to search), compared to the 2x of TLC.
I'm at my job right now but when I get home tonight I'll do an update to this. Funny, Tom's missed it too (at least as of this morning). Good catch!
Lol! A comedy of errors. I re-checked Tom's post to see if they've updated on this and they even put a link for that drive that goes back to a previous post on it that states it's a TLC drive while still calling a QLC in this post.
This:
which goes to this:
https://www.tomshardware.com/news/micron-launches-ultra-high-capacity-30tb-6500-ion-ssd-and-high-endurance-xtr
TH News Guy: "TLC and QLC are the same thing, right?"
Not gonna lie. I kept thinking to myself when doing this one why would anyone want to use QLC for a hi-density enterprise or data solution? Thanks to the other replies in this thread some of that question has been answered. I used some extremely low-cost Sandisk SATA III 256 GB QLC drives in our old workstations and laptops, and it worked out well but they were never really pushed to their limits since those stations were only SATA II. Out of around 50+ I only ever had may 2 or 3 fail in about 6 six years but I still would never use QLC for anything like a server or hi capacity storage.
I have three in rotation; two are in my wife's computers (laptop and desktop). Not an issue, though they're larger 2TB and 4TB drives, which helps mitigate potential issues.
Overall, they're basically slow PCIe 3.0 drives, but they do work without issue unless you're writing large files or just straight up imaging them. For things like media or even slightly older games, they're just as fast as anything.
Updated. I also sent Scalability a message and will update again if they respond. I'll check again in the morning.
I stopped reading at Tom's Hardware... :D:p;):rolleyes: