Read more.And Samsung has sent out invites to Unpacked 2021 (14th Jan) for the Galaxy S21 reveal.
Read more.And Samsung has sent out invites to Unpacked 2021 (14th Jan) for the Galaxy S21 reveal.
To be fair SATA SSD's are rather old hat these days. nvme are much faster and convenient too - a decent gen4 nvme drive is 10x faster easily
Old puter - still good enuff till I save some pennies!
Eh? And? They also use less power. Each to their own but gaming load is just one part of the difference an ssd makes - sata v nvme I did a quick test when upgraded last time and boot times reduced by 20 seconds, and loading apps/programs was much much faster and the whole system far more responsive. Heck even video rendering times was like 30% faster because of the faster io
Old puter - still good enuff till I save some pennies!
Last edited by LSG501; 04-01-2021 at 02:30 PM.
How is the 870 evo for write endurance and lifespan compared to previous ones like 860 evo, 850 evo, etc as well as the PRO versions?
the 870 uses QLC which is suppose to be more fragile and easier to wear out than MLC, but has the same write endurance as the 860?
the 860 uses MLC
not sure what the 850 used
the 840 used TLC
MLC has a higher bit error rate to SLC
Last edited by liquidflower; 04-01-2021 at 08:31 PM.
Benchmarks vs applications. Granted, some stuff may start taking advantage down the line, but as it stands there just isn't much difference unless you're regularly transferring lots of big files to a similarly fast drive or doing some very specific workloads (talking about desktop use here of course - workstation/server loads are different).
Regarding your observations - I'm doubtful the boot time would really be 20s faster on the same system unless something was broken with the SATA initialisation/POST or you're comparing against a particularly bad SSD. Plenty of places have tested this. Even if you loaded a PC with junkware I find it hard to believe there would be that much difference.
"Feeling" faster and more responsive? Again, could be confirmation bias, and actual blinded tests tend to show it's very hard to reliably tell the difference.
Video rendering. Yeah, that's pretty unlikely unless you were ripping raw footage through a VERY fast encoder or something. You will very rarely come anywhere close to even HDD speeds when doing that sort of work. Like much of the above, even a SATA SSD ceases to be the limiting factor in most IO workloads.
And lastly regarding power - overall SATA and PCIe drives tend to fare similarly, though some NVMe drives may naturally prove more efficient during transfers due to being faster, but on laptops drives will spend much of their time in deep sleep modes, and both types of drive are capable of tens of milliwatts. There's far more variance between individual drives than there is between interfaces.
Not that there's anything wrong with future-proofing, and entry-level NVMe drives are coming down in price, but they do still have a couple of disadvantages vs SATA drives. Cost is one but this is shrinking, also practicality is another - many systems will take one drive at full speed and a second drive at half speed (mine is one of them). On the other hand, I have lots of SATA ports should I want to add more storage.
Additionally, it's worth remembering that there are some particularly awful QLC NVMe drives which sink to sub-HDD speeds when you throw more than a few GB of writes at them, couple with woeful endurance figures. I'd take a decent SATA SSD over one of those any day of the week.
The line matters as well as the generation; Samsung may use MLC, TLC or QLC in the same generation. E.g. the 860 PRO uses MLC, the 860 EVO uses TLV, and the 860 QVO uses QLC. Those PRO, EVO and QVO lines persisted up until the 980 PRO which now uses TLC.
Likewise, the 870 QVO already exists and uses QLC. This 870 EVO will most likely use TLC.
The English is fantastic...could be confirmation bias
It was a basic clone drive and see....
poor old ssd was slow, new one is pretty speedy. It's subjective, and peoples mileage might vary but the change was huge...
Like I said it was just a quick test, not scientific but overall yes things were faster. The video encoding was a surprise but I guess many might not render to ssd but a spinny drive.
Old puter - still good enuff till I save some pennies!
The price difference is tiny these days, only ~£10 for 1TB drives (and SATA M.2 drives are equal price to NVMe ones)
I'd prefer a QLC NVMe drive to a DRAMless SATA one - going back over SATA for the drive map isn't great. QLC on SATA could work better, by limiting you to SATA speeds they could have sufficient time for the controller to handle the internal QLC-related rearranging in the background.
The suitability of QLC drives is workload dependent, of course, and I'd point out that multi-GB writes for home use tend to involve getting stuff over broadband (which is also sub HDD speed). If the only SSD in the system is a QLC one, then where is all this high-speed many GB data coming from?
MX500 1TB is frequently sold for £80. The cheapest NVMe +DRAM drive I'd touch is £120 - that's 50% more expensive. The MX500 is a well-regarded TLC based SATA SSD with DRAM. Comparing to QLC SATA drives is kinda academic since they're no cheaper and frankly pointless from what I've seen, and they would need to be substantially cheaper to be worth even considering IMO. And I don't see that as particularly likely given going from TLC to QLC only gains you 33% more raw storage per die - minus potential losses for stronger ECC, then factor in the substantially lower endurance and performance.
If you're not bothered about multi-GB transfer speeds then perhaps an NVMe drive isn't that important anyway? I know what you're saying (I promise I'm not just being contrary for the sake of it ) but their horrid sustained write speed once the cache is full just makes me grimace (we really are talking sub-HDD speed). And besides, it's not just sustained write speeds, their performance tends to suffer right across the board outside of some synthetic micro-benchmarks that make the SLC cache performance look OK. There are some workloads where they will fall behind SATA drives, especially when they get full and aren't able to hide the woeful QLC performance behind a huge dynamic SLC cache. This in particular is quite important since many benchmarks are carried out on an empty or nearly-empty drive which means a large dynamic SLC cache. This will be much smaller on a fuller drive.
Horses for courses, but I know what I'd choose, any day of the week. I do transfer things like game files or videos on occasion, and I don't buy a fancy NVMe SSD to be laughing at 80MB/s transfer speeds. I would consider some of the DRAM-less or HMB NVMe drives as some of them are quite good nowadays, but QLC can stay on the shelf. To me, QLC just seems like a toxic race-to-the-bottom for very little benefit. It's objectively worse in pretty much every way, and *substantially* worse in many, for a very small gain in cost per GB. The fact some QLC drives are no cheaper is just ridiculous IMO.
As I've said I'm not arguing against the potential benefits of NVMe drives, but they're frankly an unnecessary cost for many (if not most) real use cases. And that includes current games. My overall point being that SATA SSDs still very much have their place on the market.
I very much agree with the headline of this review: https://www.techpowerup.com/review/s...-870-qvo-1-tb/
Edit: BTW do have a look through that review - quite a few reviews of the QVO stop at synthetics and draw the conclusion that it's good but a bit overpriced. Things are different when real-world scenarios are tested, and it's not just a case of exhausting the SLC cache with bulk sequential writes either.
Last edited by watercooled; 04-01-2021 at 07:24 PM.
CAT-THE-FIFTH (05-01-2021)
I back-to-back tested an 850 Evo SATA3 500GB and 970 Evo Plus NVMe 1TB and there was very little in it. Load times were maybe a second or two better but for my use even getting a twice as fast PCIe4 drive probably wouldn't improve it much further. According to benchmarking the SATA drive has half the random IOPS and twice the non-random sequential throughput, but that's just due to caching.
SATA SSDs are trivial to install in my system compared to NVMe (and can be done hot) so going forwards as and when I replace my mechanical HDDs with further SSDs I'd prefer SATA - I'll keep my OS/caches on NVMe and use SATA for the rest.
There are currently 1 users browsing this thread. (0 members and 1 guests)