QLC Flash Wars Heating Up: QLC SSDs Headed To Desktop PCs This Year

Status
Not open for further replies.
This is great but to truly take HDD market share they need to get prices down a lot. Right now you can get a WD Blue 1TB for $45 bucks on Newegg, making it $0.045/GB. The 1TB NVMe WD Blue is $204 making it $0.204/GB. So SSDs are still about 5x the cost of an HDD.

For the same $200 you can get a Seagate SkyHawk 6TB which makes it $0.033333/GB, even cheaper overall than the 1TB.

Of course this depends on the brand and class you get and WDs are not top of the line but I was comparing like to like.

Personally I was to see a 4TB for $150-200 then I will probably switch out to an SSD exclusively and drop HDDs altogether except for external storage.

Then again I am tempted to wait till NVDIMMs become common and cheap enough.
 


Having a hypothetical 1 terabyte drive with 1 full Drive Writes Per Day is the same as having a 100 gigabyte drive with 10 full Drive Writes Per Day in terms of endurance.

The only difference is that with the 100 gigabyte drive you can't install much more than Grand Theft Auto 5.
(I don't still play GTA5. I just remember it was massive lol)

 


We will have to really see them in action but until 3D stacking becomes more viable upping the bits per cell is one of the best way to increase storage density per chip and thus lower the overall costs. Otherwise costs wouldn't drop.

 
So basically if you're primarily reading data and not writing as often, then QLC is fine. But if you're writing data back often, either stick with TLC or ensure you've over-provisioned the QLC drive with plenty of free space to ensure adequate wear leveling.
 
To anyone who compares SATA mechanical drives to NVMe drives are like comparing oil and water. SATA is a very aged technology, the interfaces are ubiquitous and the cost is in turn less than its NVMe also in unit sales in the consumer space most computers have 4-8 SATA connectors but only 1 or maybe 2 slots for NVMe. In time there will likely be a SSD technology that puts SSD and HDD at near identical price points but for now (and probably the next 5-10 years) it isn't happening. At that point there will be a new interface or NVMe 1.4 drives will be much cheaper than NVMe 2.0 drives... ETC. the market changes fast and a 1TB HDD has been around for many more years than a 1TB SSD and in turn a 1TB NVMe Drive.
 


1TB SATA SSDs can be found at $170 (4 times HDD). It's not entirely fair to compare only NVMe to Mechanical.

Also to me the SSDs are indeed 5 times better, not only the massively better random and parallel performance, but also lack of noise, vibration, safety of moving while running (for laptops), and space savings (especially for NVMe. I've removed all the HDD cages in all my cases for airflow or radiator/pump and rely on 1TB SSDs and NAS. Laptops gain even more from NVMe)
 
That slight increase in density is not worth the massive losses to performance and endurance. 3D QLC performs even worse than planar TLC.
Plus, a 33% reduction in the cost of materials is actually going to have almost no impact whatsoever to the retail price of SSDs. $300 drives are not going to go down to $200, it's going to be more like "The $300 drives will go on sale for $275 more often".

The hit to endurance is so bad that if you were to over-portion a QLC drive to offer a drive life comparable to TLC, then you actually end up with a lower capacity (and slower) drive than the TLC equivalent. That effectively makes the cost per GB higher in QLC compared to TLC in most workloads. The only exception is Write-Once Read-Many workloads where you are essentially using a drive as ROM. Desktop usage is not a Write-Once Read-Many workload. They are just banking on the idea that a lot of people don't actually use their PCs any more.

Where this will get particularly annoying is in laptops where they pre-install the cheapest possible SSD into the only nvme slot, because it is a pain to transfer data to something decent- especially if you have 2 years worth of data saved when your drive craps out on you.
 
@derekullo

But the density did not go up 10x, it went up 33%

So your comparison would be more in the vein of a 100GB drive with 10x drive writes per day vs a 130GB drive with 1x drive write per day.

Although there's some optimistic claims that 3D QLC is only half the endurance of 3D TLC - but you still lose significantly more in endurance than you gain in capacity.
 
I'm not so sure performance would be overall retrograded so long as there's enhanced parallelism on write-backs with QLC NAND. But that's going to be entirely dependent on free space / over provisioning slack and how manufactures implement this new technology. So at best, there's an improvement. At worst, it is in fact degraded. I'm thinking that's going to vary a lot between vendors.

As for endurance, agreed. It will be worse.

So long as the cost per GB goes down in accordance with the QLC, it might actually make sense to keep the OS drive with TLC, and relegate QLC for larger 1TB+ data drives where they're mainly read-only archive/media storage. But then again, if I'm going to use a separate secondary data volume, I'd probably go with a 6+ TB HDD with additional HDDs used for backing up all that data too. Of course, it's entirely situational to your needs.
 
GIRORO,
You are making some big assumptions since some of these QLC SSD's won't come out until 2019.

Write endurance isn't that important for things like storing games. I'll never even use 5% of the writes my 1TB Samsung TLC drive has for endurance since I rarely ever write data to it.

So there will definitely be a use-case for QLC SSD's.

Not sure why performance would necessarily change either. My TLC drive has about the same performance as my SLC drive.
 
"light-use applications, which includes desktop PCs. "

Have you ever wondered why that red light keeps on blinking continuously on your favorite Windows PC?
 
Status
Not open for further replies.