101–110 of 135 posts
With the introduction of QLC consumer SSDs, with worse write endurance and life time but the same price as TLC, I took a look at used enterprise SSDs and never looked back. The only issues is the connectors. You need 8x PCI-E lanes (4x might work) or a SAS connector that nearly no consumer board has.
Earlier quoted context omitted.
QLC is complete trash and should never be used outside of a WORM architecture in a larger pool. That they are trying to force it on consumers is criminal. Same with SMR.
>QLC is complete trash and should never be used outside of a WORM architecture in a larger pool. Not everyone is a creator editing/rendering 4k videos. For the typical surfer/office worker/gamer QLC drives are fine because they rarely need to dump 50+GB of writes at a time. What you're suggesting is basically "ban all atom processors because they're criminally slow!"
Does anyone have a list of SSD manufacturers that don't do this? Are SK Hynix and Intel still reliable?
All manufacturers multi-source parts. That will always have some very minor impact on power/reliability at least. The difference here is it makes a major difference. See also:. Apple sourcing 4g modems from both intel and Qualcomm, and the performance being quite different so they disabled features of qualcomms chip to make them similar.
No, not really. The whole point of buying Samsung SSDs is that they're guaranteed to contain a Samsung controller and Samsung NAND. Likewise with Hynix or WD/SanDisk.
Do we now have to place our SSDs in raid configuration to get reliability? The problem is that I don't have enough slots.
Earlier quoted context omitted.
QLC is complete trash and should never be used outside of a WORM architecture in a larger pool. That they are trying to force it on consumers is criminal. Same with SMR.
Current consumer QLC devices still do very well in read speed and especially latency. The latency often is better then on older enterprise SSDs. Only device managed SMR is terrible. Host managed SMR is fine for data-hording if the savings in size/price are forwarded.
Exactly. When you architect your tiered data storage solution to take advantage of technologies such as SMR/QLC it works great. This is absolutely not the case in consumer applications. Consumers will use QLC for their boot drives and fill their NAS boxes with SMR drives and have a horrible experience.
I used to work in their SSD vertical on the software side. For what it's worth, SSD/memory business unit is their pride. They never used to cut corners and do rigorous testing to ensure their customers aren't screwed over. Even when a bad batch of SSDs were shipped by mistake, they would recall/refund all of them proactively. When ever a new model is released, they would first ship limited units to a set of SSD enthu…
This is actually better for most consumers. The SLC cache was increased nearly three fold and the controller is a superior one (it now uses the same one as the 980 PRO). TechPowerUp[0] has a much better post on this, and you can clearly see there that the new one is better in most cases than the old one. The only ones disadvantaged by this change are people who constantly write >42GB which I would think are video edi…
>The SLC cache was increased nearly three fold How can they do that without reducing the overall capacity? My understanding is that part of the MLC storage in SSDs is used as an SLC cache so that it's faster, but can store only half, one third or one fourth of the data it otherwise would.
However, every SSD out there has unallocated space. This is for trim and drive write purposes.
At the mfg level, if you needed to get more space, you can. This will affect the drive rates per day (DWPD) rating of the drive.
In the server world, you can get “high endurance” drives that are 90% just under-provisioned for storage space.
Earlier quoted context omitted.
QLC is complete trash and should never be used outside of a WORM architecture in a larger pool. That they are trying to force it on consumers is criminal. Same with SMR.
Current consumer QLC devices still do very well in read speed and especially latency. The latency often is better then on older enterprise SSDs. Only device managed SMR is terrible. Host managed SMR is fine for data-hording if the savings in size/price are forwarded.
I'll believe it when I see it, though.
Unless you're trying to cram more data into a 2.5 inch drive, there seems to be zero upside to getting an SMR drive. They're priced about the same per byte despite "25% density improvements", and the high capacity 3.5 inch models on the market don't use it.
Earlier quoted context omitted.
>QLC is complete trash and should never be used outside of a WORM architecture in a larger pool. Not everyone is a creator editing/rendering 4k videos. For the typical surfer/office worker/gamer QLC drives are fine because they rarely need to dump 50+GB of writes at a time. What you're suggesting is basically "ban all atom processors because they're criminally slow!"
I think they were complaining about reliability. QLC cells can't handle as many writes before failing.
Earlier quoted context omitted.
I think they were complaining about reliability. QLC cells can't handle as many writes before failing.
QLC also has shitty slow write speed. (Slower then linear writes on spinning rust!) So consumer SSDs now depend on a limited dynamic part of flash being used as SLC to cache writes. Just that this dynamic part mostly gets smaller the fuller the device is. And as soon as you do anything with large writes you will hit a brick wall of slowness, on full devices sooner then on empty ones.
And I bet the vast majority of people basically never do >10GB writes at >200MB/s.