SSD Thread (Benchmarks, Brands, News, and Advice)

Maybe you should compare an 860 Pro SSD to the 2Tb Sabrent QLC drive, it's faster in 99% of everything. Again show me where a Sata SSD is faster than the sabrent QLC drive

 

For the average person probably won’t make a difference either way though. Still not sure it’s worth skimping on a little money as prices aren’t too far for a QLC Sabrent just for PCIE 4.0 vs a good 3.0 TLC nVME drive like a $115 1TB Inland Premium or MyDigital BPX nVME (Both are Phison based) which will have better sustained speeds for those who need it and greater endurance.

You also probably won’t have much difference in app loading or boot times between a QLC nVMe and good MLC/TLC SATA drive. I think what ole was alluding to was that on extremely large sustained transfers (since he’s a storage enthusiast) probably like doing video work or something of the sorts, sustained writes on a good TLC/MLC SATA will be higher than a QLC nVME (bigger the difference as drive fills and dynamic pseudo SLC caching size is reduced, not to mention much higher life for heavy use on MLC/TLC. Granted if you have that heavy sequential usage you probably want nVME TLC/MLC anyway. Your Q1T1 sequentials show you can achieve 500 MB/s on SATA, not uncommon.

 

Samsung 860 pro
https://www.anandtech.com/show/12348/the-samsung-860-pro-512gb-and-4tb-ssd-review

Sabrent QLC
https://www.anandtech.com/show/16136/qlc-8tb-ssd-review-samsung-870-qvo-sabrent-rocket-q


There is a SINGLE benchmark that sabrent is slower in.......

Sabrent



Samsung



Every other test there is no compairson

That's as close as I can find for benchmarks comparing a GOOD sata drive to a "crap" QLC drive

 

maybe you should write 400gb of files sequentially and tell me what speed you're getting. I understand from the way you reply it, seems like you dont read what you dislike, so I suggest reading past posts again to get accurate info of what I said.

edit: but damn, a NVMe PCIe drive, not 2.0 or 3.0 but PCIe 4.0 is slower than a SATA drive in 4k random read LOL.. shows all marketing masking everything with high sequential number and caching.

 

No you said QLC drives are worse than sata drives, and I'm telling you that NONE of my sata drives are better than my QLC drive and my QLC drive has performed better than any of my Sata drives. Do you use QLC and sata drives together in the real world and have real world experience with them?

 

Can you link to the post?

And do you have sata drives with MLC nand? BTW @ole!!! don't talk about sata drives with QLC or TLC nand.

TLC vs. QLC SSDs: What are the differences?

While you can generally follow that MLC will be faster than TLC, and TLC will be faster than QLC, new SSDs contain multiple optimization methods that help cover up or negate the shortcomings of slower NAND. A great example of this is “SLC-caching”, where unused areas of a drive will act as pseudo-SLC NAND. This allows for very good performance for shorter, burst-type workloads, as is often the case for most PCs and client-computing environments. This can be clearly seen in our previous consumer and enterprise SSD article.

 

https://www.guru3d.com/articles-pages/sabrent-rocket-q-1tb-nvme-ssd-m-2-review,10.html

This is the 1TB QLC drive not the 2TB drive I have, so again show me how any Sata SSD's are better at anything than this drive?

 

below are all of the posts since I started replying to you. now since you are too lazy to go back to read them, I do it for you. now tell me in which post I say such a "vague, without context" statement as you claimed.

people really need to read properly without jumping to conclusion. read without the proper context and reply right away soon as you see something you disagree with is no good. also you wanted him to provide graph so bad yet you ignore mine, what kind of immature discussion is this?

 

So in that nice fancy graph which of those drives are Sata drives?

 

That fancy graph was meant to show you the drop in sustained performance from the Rocket who contain QLC nand. Then re-read my post to finally understand what it means... TLC vs. QLC SSDs: What are the differences? The drop in sustained performance and the lower endurance is one of the reasons @ole!!! want to avoid them.
http://forum.notebookreview.com/thr...ews-and-advice.429972/page-1140#post-11072352

If you have other needs then you can freely pick what you want from the shelves. None force others to pick the same or from same brand. But be sure you know the pros vs cons. That's it.

QLC vs TLC vs MLC vs SLC - buying an SSD - Tech jack of all...

 

are you telling me you can't figure this out? the QLC drive at 200 MB/s which is slower than many sata drive on the market. it shows from this reply, you only want to pick out the error to win an argument, rather than focusing on what has been said.

 

Are there figures on how QLC drives behave in random read/write as they fill up? Genuinely curious since the article @Papusan just listed only shows sequential write jumping off the cliff into 'worse than hard drive' hell. If sequential read and random r/w stay ok then QLC might still be decent even at 1TB/2TB capacities for typical consumers (although tbh the price difference with a TLC drive is small enough at that point where I would probably go one of those instead)

Also, Golden award doesn't make the SSD fly and neither does 'winning' internet arguments, no need to get too upset if others don't like your stuff.

 

QLC is totally fine for average consumers, the controller nowadays are damn good and companies know what to do (including firmware) to make it fast. TLC had similar issue too when debut but QLC is simply having a more difficult time in seq write, they will need a lot more channels on the controller in order to have higher sustained sequential writes.

and as we all know, more channel = more latency so random read/write suffers, just like any other hardware (ssd raid, more cores vs less core, dual gpu configuration, quad channel ram vs dual channel etc). the higher latency is masked up with SLC cache, at least most software aren't capable of differentiate no more.

for your question on performance for filled drive, iirc only reviewer Jon Coulter from tweaktown and Chris R. (tweaktown, rwlab or tomshardware) test performance on filled drives but may only be iops rather than sustained workload.

 

Ok how about 109GB of MP3s and video files.

The HDD in my Stealth is a Rocket 4 Plus 1TB

Utilizing the thunderbolt 4 port on my MSI Stealth copying TO the Sabrent Rocket Q2 2TB
The Rocket Q drive in in a Vantec NVME/USB 3.1 Gen 2 Enclosure


Drive isn't even empty either

Here's the CD score when in the enclosure



Here is WRITING 109GB TO the drive



Utilizing the thunderbolt 4 port on my MSI Stealth copying TO the WD 2TB Blue Sata SSDdrive in in a Vantec Sata/USB 3.1 Gen 2 Enclosure


Drive isn't empty




That's about as close to real world as I can get, Sata Drive vs QLC drive and guess what the QLC drive takes less time and is faster and doesn't have the terrible write drops and its not an empty drive halo test either.

So what i'm saying is in MY use case scenarios MY QLC drive is not slower than Sata, and the Sata drive I used is no where near a top performing Sata drive.

 

BTW If I'm reading the write performance results posted on the previous page correctly, the write speed of the Sabrent Rocket QLC only collapsed after some 300GB written, so the good outcome of your 100GB test are consistent with that. Nice work though, always good to double check some synthetic results and this is reassuring that unless someone plans to regularly write 300GB+ in one go, that QLC drive will do a decent job at a very competitive price.

 

Here's the HD tune on the Rocket Q 2TB, a momentary blip which is not sustained, and the performance doesn't dip.



How come no one else wants to do some real world testing?

 

Again, if you want to confirm (or disprove), the result earlier, don't bother with crystal disk bemchmark using 1GB chunks, or HD tune. Would have to copy, say 500GB of data between a 3000MB/s source drive and your Sabrent, over a 4xPCIe link or using a full Thunderbolt enclosure. USB 3.2 is too slow. The reason is that you are transferring at 1/3 the max rate, the controller has that much more time to manage the writes so that the drive doesn't choke. If your primary use case is access via a USB 3.2 enclosure then no worries - you bought a great drive for this use case, (although you won't be able to use the drive to its full potential this way - get a true TB ebclosure) . You would probably have to write 900GB+ of data to see a slowdown, if it happened at all.

 

I still don't think it's an issue at all, the only time I copy >100GB is maybe when I'm moving steam files or something. Other than that my QLC drives perform just fine. Again everyone's usage is different, but most people would be well off with a QLC drive...

Having said that, I have SLC transcoding drives in my server, not needed but you can't beat the endurance, lol...

 

SLC as in some of the older models or new ones? Mind sharing the models, just of curiosity not that I will ever need it lol.

 

Well i'm not spending 100s of dollars for a TB enclosure I used HD tune because I was told to, but as you can see i'm not really being limited by the interface as I posted the CD mark to show what the max transfer rates could be, again someone wanted to tell me QLC drives were terrible and slower than Sata drives so rather than just say no they aren't I showed real world tests of how I use the drives and again i'm not getting these slow downs every one talks about.

 

Funnily enough, found a Sabrent TB3 enclosure right away - will probably cost around $80 in the US

https://www.amazon.co.uk/dp/B08FT59SB6

Of course, only needed if you want to use the drive to its full potential though. Good to have such an option.

Edit: arrgh, this one only uses 2 lanes, so speeds up to 1600 MB/s. A good improvement over USB, but still half the max speed.

Here is a full speed one, 50% more but not "hundreds": https://www.amazon.co.uk/dp/B08DNLWCK7/

 

Sadly nothing exciting, they are old but work well for what I use them for:

HGST Ultrastar SSD400S.B HUSSL4040BSS600 400GB SAS drives. Nothing that would remotely work in a mobile platform, lol.

 

Three+ minutes to transfer 109GB?

You need to upgrade @Silvr6.

The hardware below should get you to 2021 levels of storage performance for little cost (sell what you have! Give it away! But don't keep using it and wasting 10x your time each time you do a copy to/from other fast storage).

See:
Sabrent Rocket 4 Plus review: This SSD is faster than anything available right now | Windows Central

See:
[Intel Certified] ORICO Thunderbolt 3 (40Gbps) M.2 NVMe PCIe SSD Enclosure, Aluminum External M2 2280 PCIe M-Key Adapter Case with 0.5m Thunderbolt 3 Cable (SSD NOT Included): Amazon.co.uk: Computers & Accessories

Above per link supplied by @etern4l.

 

If his use case is only to access the drive externally, then his current PCIe 3.0 Rocket will do, he just needs a faster enclosure. He would need to put that Rocket Plus in a PCIe 4.0 laptop to fully utilise the additional bandwidth (I understand).

 

If you put a 4.0 drive in a 3.0 slot or 3.0 drive in 2.0 slot, your sequential read/write speeds will be constrained by the interface, but how about random accesses?

 

It plummets too.

 

This is a Rocket 4 500GB in a PCIE 3.0 slot instead of PCIE 4.0 you could say that everything plummets compared to a PCIE 4.0 system but it's still respectable.



This is the PM961 in my Inspiron which is PCIE 3.0




It does better than my Samsung PM961 drive in my Inspiron but now an overly fair comparison because the PM961 drive is quite old now.

 

I do 3 minutes is too long!

So I would totally get a TB3 enclosure however the issue is that the TB4 port on my laptop is being used by a TB3 to dual DisplayPort adapter for my 4K monitors.

One other interesting bit of information regarding enclosures.

This is with the Rocket Q 2TB using the TB4 port as i've posted before



Now this is using the USB port on my MSI stealth



Its a fair bit slower and I do notice the file transfers (large ones being about 150-200MBps slower).

 

Looks like your MSI stealth has USB 3.0, limited to 5gbps.

 

I'm guessing the lied in their specs then

 

Ok thanks, good to know. Looks like the 970 Evo plus will be staying in my 3.0 slots for now. Got two 2TB ones on the cheap during black Friday, so all set until I get a new system. They definitely run hot, but adding heatsinks from Aliexpress (my notebook has clearance for the 4mm thick ones) drops the temps from 80 on the controller to a comfortable 60 under synthetic and about 50 with typical use.

wrt the specs you might want to check your cable/enclosure, since I find it hard to believe they would lie about specs

 

I expect this is the manual for your Stealth 15m
https://download.msi.com/archive/mnu_exe/nb/1562_v1.0_English.pdf

Listed with USB 3.2 Gen 1

USB 3.2 Gen 1: originally known as USB 3.0, and previously renamed to USB 3.1 Gen 1. It’s the original USB 3.0 specification, and it can transfer data at up to 5Gbps.
https://www.theverge.com/circuitbre... 3.2 Gen 1: originally,speeds at up to 10Gbps.

 

Oof that is super misleading, but we learn something new everyday.

 

because that is the real world result we'd expect. if you go back to my earlier posts, you'll notice that I posted two graph, both are sabrient drive, one is at the 200MB/s in sequential write, the other is 400 or 500 MB/s sequential write during sustained workload.

in this graph I believe HD tune blue = read, orange = write. if it's read, the result would worry me if I have the drive and CDM is showing over 5000MB/s while HD tune only showing you 400.. further more we don't see the settings you choose for HD tune such as block size, or if you do it full disk span etc.

also, window transfer test is flawed. consumer window store files into ram before going into disks and I would not trust them one bit, too much variables involved and result will be skewed.

from my experience, it is my recommendation to try not to purchase a pre-enclosed external storage device, especially hard drives due to them able to mask the speed with USB bus or enclosure pcb controller if device has issues. unless you know the internal components its best to grab an enclosure with high performance and put in the SSD yourself.

though if it's drives that run on usb c or 3.2, or thunderbolt then those problem are of no concern, review will show full extend of what sequential is capable off assuming its not QLC.

 

So basically because the results I posted in the real world showing the my QLC drive is faster than any Sata drive I have, we are going to say the windows transfer speeds are flawed?

Again my QLC drive in my uses has always been faster than any Sata drive I have

 

thats what you think and assume. im simply laying out the facts and you don't have to like it, that is not my problem. when you copy files using native window OS, you get those massive several GB/s speed, then drop, then back up again, this applies to either NVMe or sata and more evident with SATA. this should already tell you something is wrong, but since you are so stubborn, I will no longer share my knowledge with you, you can use your money to continue to buy those lower quality SSD if you wish.

your use case will probably never fall outside of those number anyway so you will never notice the difference, become a sheep like majority of consumer if you will, up to you!

@Papusan @tilleroftheearth on the other note, some good news https://www.tomshardware.com/news/n...CP0WvfUvoreAdoJcWc1v1NnQDOVNdg9awlELjqMsuSIEk and shows I have been behind way too much.

some promising tech coming up could give us "proper" mass QLC storage, 16TB NVMe QLC im waiting.

 

Where are the massive drops when i'm transferring 100+gb of data like you asked me to do i'm seeing sustained 400-500Mps. I've made videos showing exactly how the setup works on my system and again there is no massive drop and my QLC drive is faster by a few minutes vs my Sata drive. You never did answer if YOU personally have any experience with QLC and sata drives in the real word, i'm guessing you don't and just keep rehashing benchmarks. I've proven to you and others than in my REAL WORLD usage scenarios there is ZERO performance degradation that some QLC have. I don't want your knowledge and I never did.

 

You're the one imply QLC is "better" even when I pointed out the disadvantage and weakness of QLC flash. I argue the sequential performance drop during a long sequential write workload, yet you continue to use graph that is irrelevant to argue.

continue with your way of arguing. people that is knowledgeable can tell in an instant you lack knowledge. maybe you should start thinking why company would market and design their product the way it is, to look "good" and hide other issue.

 

I was under the impression that the links I posted were to enclosures, not enclosed drives.

 

Isn't the main competitor against QLC drives TLC drives? Why are we even comparing them with SATA here, when (other than the niche M.2 SATA) they aren't competing for a slot.

At least for me (and the market in Singapore), QLC drives aren't that much cheaper compared to TLC ones, so the 5% or so savings just aren't enough to justify the overall worse performance.

 

it is simply to compare the issue with QLC when during sustained workload. cant even use it as backup storage if speed was to copy at 400MB/s on a pcie 4x4 bus. if I want to copy 8 to 16TB at that speed.. i'd rather not.

 

Not so sure. I never got any answers on one of my previous posts. Regardless, QLC drives is cheap, and this for a reason. And this is already explained well in the thread + links.

Maybe we should move on...
Best SSDs 2021: From Budget SATA to Blazing-Fast NVMe tomshardware.com

 

Anandtech released a new benchmark suite that looks pretty interesting. Looking forward to more ssds being tested - the hynix p31 performs really well compared to some of the pcie 4.0 drives. I also liked how they did a lot of testing at 80% filled:

https://www.anandtech.com/show/16458/2021-ssd-benchmark-suite

 

From the link in the post above.

So, not real-world.

So, not real-world.

So, more real-world.


In the end, they admit what they're doing isn't for us to know what drive is better at what real-world task; it is for their sake. Ease of completing the bm. Less babysitting required. More graphs to obfuscate the actual performance differences (if and when they exist in real-world workflows where queue depths are arbitrarily inflated, idle times are truncated and other shortcuts to getting a 'story' out are made for the 'click-ability' factor to drive in more traffic to the site, and not with the usefulness or applicableness in mind to consumer or workstation workloads, of the graphs they strive so hard to provide.

Sorry, I really miss Anand... and the real reasons to want one SSD drive over another.

 

IMHO, it is really difficult to figure out how my 'real world workloads' will run on different hardware based on benchmarks, especially on CPU and SSD. Some of my time consuming tasks include:

1.) Compiling *very* large codebases
2.) Creating an Ubuntu VM in vmware from iso
3.) Saving and loading snapshots of VMs (which essentially flushes the RAM to/from disk)
4.) Compiling codebases *in* a VM

and the more and more specialized the tasks get, and the more such tasks I perform at the same time, the more unlikely I am to find a set of benchmarks that correlate well with my use cases, not to mention IO frequency (how long does the SSD get to flush SLC?). I end up trawling the mountains of data and guessing some tenuous relationship between metrics like 'QD4 random write' and performance.

At least with CPUs its pretty much just AMD vs Intel - testing one processor gives you a pretty good idea how the rest of the generation will perform, so that's just two tests to perform. With SSDs things get way more complicated due to the SLC/controller/etc.

Maybe the right answer is to stop reading benchmarks and just buy a Samsung MLC...

 

some of their chart is showing alright. like the 16QD 1MB sequential test, and still remained on 1QD for random read/write, though i'd prefer 2-4QD for sequential tests.

though they mention 80% filled I am hoping that 80% filled is applying to all graphs not just one test, they don't clarify on those and I am quite picky on it.

maybe someone to release a monitoring software showing which software read/write at what type of blocksize, queue depth/threads, what/which file name being read/written all at what sequential or random speed, what percentage drive is filled at and native ntfs block size, if theres raid or no raid etc.

edit: im still wondering if its because of windows natively cache copied files, why aren't reviewers include small file transfers. lots of tiny files hit hard even on some of the best SSDs including optane. it would be nice for them to explore on those areas.

edit 2:

optane!

 

Yeah, Optane!

 

Unfortunately, the price tag makes it unoptane-able for me (harhar), but hopefully in time it'll become standard consumer tech.

 

Sadly it won't:
"Intel has announced it will discontinue all of its Optane drives in the consumer space, even the top-end enthusiast-oriented products."

 

It will continue. Just not under an Intel/Optane banner.