Samsung 840 120GB Endurance Testing

Corsair Force GT 240 GB: $199.99
Corsair Neutron GTX 240 GB: $214.29
OCZ Vector 150 240 GB: $234.99
OCZ Vector 450 256 GB: $206.60
SanDisk Extreme II 240 GB: $218.99

Now there are certainly instances where tiller is correct but they are the exception rather than the rule he proclaims.

SanDisk Ultra Plus 256 GB: $164.62
Mushkin Chronos Deluxe 240 GB: $159.99
Crucial M500 240 GB: $149.00

 

Most currently sold SSD will last as long as the system containing it is in use, and the typical failures prior to that point are controller and firmware problems, not exhaustion of program/erase cycles. However, a remaining concern for those who plan to use SSDs until the limit of their program/erase cycles was discovered during the pioneering reliability tests by folks at the Xtreme Systems forums. When SSDs do fail, they have a tendency to fail catastrophically. There are no chances to recover data off an SSD that has run out of cycles, even though according to the specifications, there should be. And this is why it's still important to back up information, even on SSDs.

 

@HTWingNut or others with a 840 EVO..

If you are using IRST, did you disable LPM (easily done with the latest IRST UI, no need to go in the Registry)? I didn't see what driver you were using in the OP.

Though I have not experienced issues in the short time with LPM enabled, I disabled it (saw those with 840 Pro had issues with it enabled)...

BTW, on my AW14, there was a jump in in performance from the stock AHCI driver to IRST v12.9

 

Techreport is back with an update to their SSD Endurance test.
The SSD Endurance Experiment: Casualties on the way to a petabyte - The Tech Report - Page 1

3 out of 6 drives have failed.
Kingston Hyper X 240GB - MLC: Died at 728TB
Intel 335 240GB - MLC: Died at 700TB
Samsung 840 240GB - TLC: Died at 900TB

3 drives made it to 1 petabyte (1000TB):
Kingston HyperX 3K 240GB - MLC
Samsung 840 Pro 240GB - MLC
Corsair Neutron 240GB - MLC

 

Wow. Even at 700 TB that's phenomenal.

Beamed from my G2 Tricorder

 

Those results suggest that most SSDs will outlive their owners.

John

 

John,

Not to me.


It's like the 1 Million mile engines run on a dyno... they produce some good numbers and 'scores'... but the real world is much harder than any man made testing in a lab can do.

 

Of course not.

I am surprised and impressed to see Sandforce-based drives exhibiting such durability.

 

Yup. It will outlive you, your CPU, I think most hardware in your computer will say bye bye before the SSDs does.

Like shown in the article, Samsung use protection on their TLC NANDs where they have a spare area in the NAND which they will reallocate to if you ever wear the cells off.

 

All SSD's have some spare nand chips for re-allocation. Just like HDD's before had spare blocks before (but when used; they would make the drive clunky slow).

Again; it's not the timespan of that life that counts... it's the quality of that lifespan.

And for SSD's, being able to 'survive' for 700TB+ is the equivalent of a 95 year old on life support for the next 100 years.

No point.

 

You didnt really read the previous update did you?
http://techreport.com/review/25889/the-ssd-endurance-experiment-500tb-update/2
They reached 500TB back then and the Samsung 840 with TLC showed absolutely no sign of performance degredation.

So its actually like a 95 year old being as healthy and energic as a 20 year old although that person will eventually die from ageing

I have reached 3TB write over almost a year of EVO usage. I think im the typical average user among us. 500TB/3 = 167 years.
167 years will it take me to reach the 500TB like techreport reached back in previous update, and it will still perform more or less like it was new (although with spare area being used)

 

Cloudfire,

From your link, the 840 Pro is slowing down... the consistency is not there... SE'ed after every milestone (sigh).

The 'rig' they used crashed (based on a Corsair Force GT 60GB) - log indicates 'unexpected loss of connection to drive'.


Enough said.


Real world is nothing like lab experiments with excuses and remedies (SE'ing) to 'prove' an arbitrary and ultimately non-interesting point (re; nand can last... yawn).


Sure, these type of tests were valuable a few years ago - today, I know enough to safely ignore them. Not from the testing done; but from my own internal testing.

I can trust SSD's as far as I can throw them - and my pitching arm is not so good anymore.

But I can trust them long enough to complete a project (several days/hours) and after that, their 'reliability' is irrelevant anymore.

I take the performance they offer over a RAID0 VRaptor setup that was my storage subsystem standard a few short years ago, but other than that; all this endurance testing is like the 1 Million mile analogy I inferred a couple of posts up: totally out of touch with the real world.


Not to mention that the 840 has poor performance/productivity to begin with - so to be able to 'keep' that low performance is not so amazing...

 

How much spare area does the EVO have by default? I can't find that info anywhere. All Samsung says is some space is left open.

 

tiller, could you please describe your average workflow?

I'm fairly confident your usage scenario is atypical of the "average consumer", and if that's the case, you'd do well to put in a disclaimer to that effect instead of making these borderline misleading comments. 7 months into my 840 Pro now, and it's as fast as the first week I started using it.

 

You can`t get any closer to a heavy strain on the NAND than what techreport have shown. Its actually unrealistic high, much higher than what you would have gotten if you used it outside of a lab.
So if anything, their tests clearly shows that TLC can keep up even on a heavy workflow thanks to Samsung own precautions with spare area ready to take over IF you wear out a block.

Samsung 840 is somewhat slow. I agree here.
But EVO, which also use TLC is one of the top dogs. And techreport would have gotten the exact same TLC endurance result on a EVO as a 840, if not better, since I`m sure Samsung have introduced better wear monitoring and other safe guard protocols the controller does now and then thanks to the firmware, vs the Samsung 840.

 

I dont think he have ever used an SSD and just making up things

That Samsung 840 Pro shows some performance degradation according to tiller, with the horrible test techreport does to the drive, is because of firmware (garbage collection, not being able to deal with sick workload over a loooong period of time something Samsung of course dont code in to the firmware because it will never be a real scenario etc) or some other temporary stuff.

If it was the quality of the NAND, it would be an example why TLC is just as good as MLC because Samsung 840 with TLC doesnt have any performance degradation.
I don`t understand why tiller mentioned that as an example.

 

I've described my workflows many, many, many times.

'As fast as the first week I started using it' is not an indication of a fast drive (feels very laggy to me).


No misleading comments intended. If you have a question that might clarify sometime for you; ask.


I have a range of workflows; even though when I initially signed up at this forum it was only for my so called 'extreme' setups: converting hundreds and hundreds of GB's of RAW images to TB's of finished images on my many workstations (both desktops and notebooks). Multiple vRaptors in RAID0 crushed the early SSD's in those workflows - until I discovered OP'ing with an Intel 510 SSD (at 50% OP'ing).

The editing of those images requires fast SSD's used as Scratch Disks (at up to 70% OP'ing) for maximum speed, sustained over time (10+ hours and sometimes (depending on the project) several days).

Storage of the finished images are to multiple NAS units (the latest of which have SSD caching and 10GbE as available options).


The above workflow needs a productivity-oriented SSD like the SanDisk Extreme II and now the SanDisk Extreme Pro - capable of sustained performance over time without missing a beat. The 960GB capacity of the Pro also makes for a huge increase in productivity; more images can be converted/edited/worked on per batch, which means less time waiting while simply copying the old/new batches over the network and being able to start on the new images.


Another hat I wear is the billing of these finished projects: this usually involves consulting my timeline and ancillary cost spreadsheets for the creation of those images, opening/searching (emailed specifications of the images needed) and creating many PDF's , running accounting/billing software, searching online for similar images (if they exist) and ensuring mine are unique and finally printing a few hard copies of not only the final invoice - but also the decisions/comparisons/work involved to get to that final price (if it can't be justified; it isn't billable).


This type of workflow needs 'snappiness' and a high enough sustained productivity in an SSD. And this is the only reason I kept the 1TB EVO's for. And admittedly, they excel here (mostly for snappiness). Even though I have seen the drives hiccup a few times when I'm trying to open many PDF's (about 20, 10MB and smaller) at once.

In almost any other workflow (even where capacity was king) they would be kicked to the curb; yeah, that bad when a hard core productivity scenario is needed.



Finally, I frequent a few dozen forums, reply to emails from (too many) accounts and am generally reading my corner of the web as a book (with a few dozens of tabs open in IE).

This too works extremely well for the 1TB EVO as snappiness is mostly what is required here too.

Consistency is just as important and although I can't prove (yet) that the EVO is at fault for pausing/stuttering (it could be the ISP after all); as soon as I get my hands on a SanDisk Extreme Pro 960GB and have it installed and use it with this workflow I can see if my equipment is beyond the capabilities of the services I subscribe to.



Storage subsystems are only part of the equation for me (small part; in the sense that their performance is dependent on everything that comes before them as indicated below).


Currently, all the above assumes I'm working with at least Windows 8.1 x64 Pro Update 1, an i7 QC IB or Haswell platform, 16GB RAM or more and O/S animations, hibernation, pagefile, System Restore and sleep timers disabled for the system I'm working with. Yeah; even for just browsing the web.


Again, I want to repeat: I have not used a Samsung SSD that I could do real work with (well, at least not one that I wouldn't be complaining about while I was using it anyways). Why?

They simply are not tuned to real world needs - they bench great (no arguments there...), but 'scores' are right up there with advertising claims: take them with a huge grain of salt and even then; until you have something to compare to (in a 'real' workload); what seems fast is not necessarily so (1TB EVO OP'ed to 650GB) and what seems slower/less snappy (480GB SanDisk Extreme II OP'ed to 312GB) may in fact be the workhorse of the two.


Fast is relative depending on your experience(s). Samsung SSD's are just beginning to 'get' what fast really is.

 

The other thing that irks me is that burst performance seems to be downplayed a lot. I understand the importance of sustained performance and how certain people's workflow depend on that, but once again I think it's fair to say for the average consumer or gamer, burst performance/"snap" carries a lot more weight. If SSD A has say 20% better burst performance but 50% worse sustained performance than SSD B, I'd still go for SSD A becauase it suits my workflow better. Yes I understand workflows can change and better to "have all grounds covered" but short of a drastic change in your life (career change perhaps?), I fail to see how that could happen overnight.

 

But as Anandech stated in the SanDisk Extreme Pro review; SSD's are within 10% of each other (even at burst performance...), so choosing a lower overall performing SSD is like buying the cheapest Porsche and thinking you can race with the 'real' GT's...

Sustained performance is important in things as mundane as monthly Windows Updates, virus scans, program installs and simply copying data back and forth.

While the 'snappiness' is seductive and sexy at first glance, it is the underlying beauty/power that makes for a satisfying and long lasting relationship... er. You know what I mean!

 

I'm not saying your comments are invalid, but please keep in mind that for the average consumer whose workflow (90% burst, 10%sustained) could be the polar opposite of yours, your comments when taken out of context could be very unsettling to them.

Reading through your comments, if I was an average consumer I'd have a sense of gloom and feel terribly about that Samsung SSD I just purchased, which in fact works just fine and as intended for the typical consumer workflow. This is what irks me the most about your comments. Hope you see where I'm coming from.


Also, I'd wait for Samsung's updates (850 Pro/Evo) before passing judgement on burst performance -- after all the Extreme Pro is technically a new release while 840 Evo is still yesteryear's tech. And I'd like to point out that in TweakTown's review of the Extreme Pro, they clearly point out the Extreme Pro has an overclocked controller, so perhaps SanDisk is resorting to the same tricks to boost burst performance.

 

You missed the point of my previous reply re: workflows. No problem if you don't want to admit that I can take into account 'average' workflows even if my most punishing storage subsystem workflow is above 99.999999% of the people on the planet. Your choice.

But all my workflows are subjectively worse on a Samsung SSD, and Anand himself has effectively stated the same thing; without OP'ing the 840 Pro is a sad example of an SSD (consistency-wise which i feel as a laggy/stuttering storage subsystem and not worthy of the money I've spent on a few examples of them).



However, overclocking the controller is not a trick; it is better hardware.


The 'trick' is that EVO pseudo SLC nand is vastly inferior to SanDisk's nCache scheme. And here's why:

Pseudo SLC in the EVO is max 12GB in their 1TB models (which is one reason I only bought those capacities of the EVO's). The actual firmware though runs out of steam when that 12GB 'fast nand' is used up. I'll let you do the math of how long you can use an EVO at 'full speed' before it hits a wall. Hint: seconds.

Not only that... but the way the pseudo SLC nand is used is also slowing down the drive and making it hit sub HDD speeds when used. This is in contrast to the SanDisk Extreme Pro which uses less than a GB of nCache... because it's firmware is much more efficient at getting the data to the nand it belongs to.


See:
AnandTech | SanDisk Extreme Pro SSD (240GB, 480GB & 960GB) Review: The Fastest Just Got Faster

I agree that a year old model is yesteryear's tech - but I also see Samsung shuffling along the same 'performance' roadway for a very long time now.

If they can change their spots with the new models, I'll be the first to support them here and with my wallet.


As things stand now; your stating that it feels fast to you only strengthens my guess that your experience and/or sensitivity with storage subsystems is not at the same level as mine (and that is okay too).

 

So apparently you can take them into account. You just choose not to, which is both unhelpful and disingenuous.

 

My EVO mSATA is currently at 2.7TB writes after about 4 months.

If my usage pattern changes and storage writes increase 10-fold then it's still 6 years to reach 500TB.

John

 

So apparently you can't read?

 

I've seen your posts and whenever you speak disparagingly of any SSD you always mention it doesn't suit your workflow, but you never quite mention what your actual workflow is. This mostly happens in threads when people are asking for advice or validation, most of the time you give a vague "mundane workload like file copying and transfer". Well, when it's hundreds of GB at a time bordering on the TB, that's very different. Also misleading and disingenious.

To your credit you have mentioned a few times recently that the 1TB 840 Evo is the king of snappiness, so at least that's a start.

But when Samsung does it, it's a trick.

Once again, vastly irrelevant for the typical consumer's average workflow.

Well, my workflow and experience is typical of the average consumer-gamer type (at least 80% if not more of the market), and more directly relevant to theirs. So I don't have to qualify my statements and they can be taken at face value. Your workflow is something entirely different, and none of your comments can be taken at face value without some kind of qualifying statement regarding your workflow.

And you're right, there's no competing with a man who claims he can feel microseconds of difference.

EDIT: Let me just reiterate this. I really don't care that you're disparaging towards certain SSDs, you're certainly entitled to your own opinion. What I find really inexcusable is this: your comments when taken at face value and without any qualification, gives the average Joe a sense of gloom and makes them wonder if their spanking new SSD that they spent a few hunderd dollars on is total crap, will end up slower than an HDD over time and fail within the year. All of that is patently false given an average consumer's workflow. But if omeone who's not tech savvy strolled by and read your posts, they would become unsettled based on wrong conclusions generated from misleading premises. THAT is the major issue.

 

You can twist what I say however you like.

The people that need the information I provide get it. You don't have to.

The face value of your statements are that you're possibly not experienced enough to have even a qualified opinion that can be applied to someone else - unless they are just as inexperienced as you?

And you have to read the links I provide - and keep reading the things you don't understand in those links to be able to say something is irrelevant. No reason to prove you wrong here. You do that yourself admirably.


Finally, I support/validate products that actually work better in my varied uses; not just because some websites once gave them a 'good/great' rating.


No competition here with you or anyone else either... I'm sharing what I know to be true and trying to help others with that knowledge. You along with a few others are keen to prove how much I don't know (and when that fails) how much of what I know doesn't apply to others in similar situations.


If what I was recommending cost double, triple or even 10x what other's suggest; yeah, I could agree with a lot of the points made against me and my recommendations with regards to not only storage subsystems, but also RAM, platforms and CPU's as well.

But when the SSD's I recommend are the same price or cheaper than the 'highly rated' Samsung offerings, I think maybe listening a little is in order? At least it warrants a little further research anyways (and I've given enough hints where to do that research too).


So once again, disregard my input. That's your choice.

The people that want real performance (in all workflows) for similar $$$ or less can consider what I offer in these forums instead.

 

Pretty much.

 

Ad-hominem attacks look desperate in an argument over technical subject matter. If you continue to throw insults you will be thrown out the door.

 

Sorry, that is not what I was intending to do. But I admit I may have been getting a little desperate in my need to clarify my position. I'm not insulting anyone, rather I feel I'm being insulted by constantly having my statements twisted into something I haven't said.

But your warning has been noted. Thank you.


To put a technical twist on this:


n=1,

See:
The SSD Endurance Experiment: Casualties on the way to a petabyte - The Tech Report - Page 4


The last two graphs show (relatively) how slow and inconsistent (my words: laggy/stuttering) the 840 Pro is and how weak the 840 TLC is at actually writing files (the main job of a storage subsystem).

I don't feel a 'microseconds' of difference and never claimed to be able to. But a those same microseconds of difference can be felt when that action is played out over and over again as in a real world workload on a storage subsystem (even a simple one like navigating the O/S) and that is what I do report on.


I don't disparage certain SSD's for no reason. And I don't intentionally try to mislead anyone either. The information I provide is available on the 'net; I simply condense it a little.

The people that did become concerned and asked further questions were also answered in a straightforward fashion too, instead of simply assuming the worst (of their SSD or my intentions).

I have also not stated that any specific SSD will fail within a year. Have I? Again, please stop putting words in my mouth that I have not said.


A direct question will bridge an understanding between us better than stating what you think I'm doing wrong on these forums.

 

Its hard to take most of what tilleroftheearth says about SSDs seriously.

You keep repeating the same boxed in arguments over and over despite being proven wrong countless times.

You seem to have zero experience using any SSDs because you repeat theoretical arguments and try to apply them real world experience, but lack the experience to understand that they have no real impact on real world. Which you would know if you had experience.

When real world testing is presented, you are quick to defend yourself with that the drives have been run as secondary disks. But doesnt seem to grasp the fact that this is the closest to real world we can get due to testing protocols of this magnitude. So you keep bringing up your workflow and your experience, yet you have zero experience with SSDs.

You apply your "workflow" to people asking for recommendation for SSDs that will be used with normal usage. The workflow you mention time after time is very heavy, totally on the other side of a normal usage. But yet you seem to think that its better to go for a drive with MLC NAND instead of speed, which have been proven countless times, even in this thread, to not be any better with normal to semi heavy usage.

You twist and turn and defend every single test posted here, like you just did on the post above this, where you look at two single graphs from a totally unrealistic horror stress test, and try to persuade people that the Samsung 840 PRO is therefor bad at real world usage.

There wasn`t long ago when you had your love for Intel 520 and it was the "greatest drive out there" because Intel had used a controller thats good with enterprise workloads. So the drive did well with really really heavy workloads with high queue. You said it was better than Samsung 840 PRO because of this, but when presented with graphs and results from real world scenarios where the 840 PRO blew 520 out of the water, you started with your typical BS arguments that because the controller and firmware on the 520 was better with enterprise workloads, it was a better drive for the average joe.
Now you have shifted to Sandisk Extreme II and continues the same arguments.
You are like an open book, easy to read, and easy to foresee where the story is going.


As a person who have " Wisdom listens quietly..." in your avatar, you come off as a pretty ironic figure especially when you usually sit there on your high horse and try to apply a squared piece to a round hole in the threads where people ask advice.
I can`t even count how many times threads like this become a tiller vs the rest. When will you get the picture? A start will be to become more humble and try to apply advice and arguments regarding other people`s usage, instead of your own.

 

Cool, thanks, I've learned something today, didn't know what "Ad-hominem" referred to, but google was my friend:
Ad hominem - Wikipedia, the free encyclopedia

 

Not to mention having one of the higher posts per day ratios of any active, non-moderator member on this board.

 

Unfortunately something that`s the reason why some people listen to him.
Im rarely active in this subforum and whenever I am, I see his typical posts.

But enough of tiller. I think most people who have been around in this forum know who he is.

Homing advertisement? Advertisement that is sure to work?
Am I close?

 

Back more to the topic at hand, Amazon has the 1 TB 840 EVO on sale for $419.99.

You can get the same drive for $414.99 from Newegg, but it requires purchasing the drive through the company's mobile app and using Google Checkout.

 

Cloudfire, only you and a couple of other posters don't take me seriously. I have answered your 'concerns' in the posts above; I won't repeat myself.

And please stop acting like you know me. You don't.

As ALLurGroceries said to me; let's discuss the technical details if you want. But attacking me personally is not productive nor beneficial to this forum.


I can't tell if you're trying to be funny with your post #133; all I can see is that you continue to make statements without even trying to research the facts, even when the link is right there for you.

You also ignore all the comments I have made above... but you still just come posting your own story/agenda.

I don't see the point of most of your posts other than 'tiller sucks. Maybe it's time to stop?

And the (real) competition starts.

 

tiller: ad-hom attacks were never intended, if there were any, I'm sorry. I will just say the most ironic part about everything is, I actually did listen to your advice and got the SanDisk Extreme II 480GB to try out, but sadly it was just inferior to the 840 Pro/840 Evo when it came down to my workflow -- boot times were longer, and the system just felt less snappy. Gave it a fair shake but after 2 weeks I switched it out and now it's just collecting dust. I'll be the first to admit that sustained performance is not part of my normal workflow, and there's probably multiple reasons why it felt slower, I'm just sharing my own personal experience so just take it for what it is k? And this is definitely not an endorsement of Samsung SSDs!

Also, if you will please read the comment thread started by fackamoto in that AnandTech article you linked to. I think it's very helpful to understanding the perspectives from both sides.

 

Cloudfire and saturnotaku, the same rules apply to you. Why do you think you're any different?

Let's move on from the drama in this thread, or I will close it and you will all become very sad and bored.

 

Did I say $419.99 for the 1 TB 840 EVO? Because I totally meant $399.99.

And to the mods and everyone else, you are right. The rules apply to me just as much as everyone else. So to tiller and everyone else, I humbly and respectfully apologize.

 

I do not apologize. I meant every single word I said.
But will honour this thread and let it run its course.

Not sure there are much more to be said now that we found out that TLC reached 900TB, some MLC crashed at 700TB while 3 other MLC drives reached 1PB. I mean, can you expect to reach even 50% of that with normal usage? Heck even for hard usage, 700TB, that must be projects over many years right? Say you buy an SSD for $500. Use it like crazy for 3-5 years with sick workload.
I say $100/year is really not that bad.

 

Where's the face palm emoticon? I'm starting to feel very sad and bored. It would be a shame to close HTWingNut's thread of hard work and a willingness to sacrifice an SSD for the sake of learning.

I think it's perfectly legitimate to attack a factual argument or method of testing. Both AnandTech and TweakTown test consistency differently. TweakTown believes their file size choice, random sizes, compared to AnandTech's 4K file size, is the better choice to test with. Who's right?

Attacking a person and not their argument is an instant loss in my eyes no matter if I believed their argument had more merit. I only quoted what I did because the rest lost any value it might otherwise have.

 

So, you couldn't tell the difference or just felt less snappy?

 

My apologies, it should've been a 480GB Crucial m500 in the first post.

 

The Samsung 840 Pro is the king of burst performance. Nothing can come close to what it can achieve in short span workloads. Its a function of Samsung's incredible vertical integration from NAND->DRAM->Controller->Firmware allowing basically perfect latency control.
The reason you mentioned is exactly why the Intel 730 was so underwhelming. That drive was the king of performance consistency and was priced actually more expensive than the enterprise DC 3500 drive.
Thats why I use 2 SSDs in my desktop, a Sandisk Extreme II for Boot/OS and a Samsung 840 500gb for bulk games storage. Best of both worlds, good consistency for the OS drive and good burst performance for loading + installing games.

Anyway
@HtWingnut
It seems like TLC NAND might actually outlive the MLC Sandforce parts. I can honestly say I would've never seen that coming.

 

That's some good points you bring up about the importance of burst performance to a lot of people, makes sense to me. What surprised me was the way round you configured your OS & gaming SSD's. Wouldn't you want better burst performance for your OS drive, and then better consistency for your drive that has lots of data & games installed on it? I say that because I could imagine queue depths being longer on your gaming/data drive than on your OS drive. Hence making more sense to have the high performance low latency burst drive for the OS to keep it snappy, and then the more consistent performer for the game & data drive (most of your writes & reads in terms of GB would be from the data drive right and more likely to be in a 'sustained' manner too). (I don't profess to be an expert on this, but from the stuff I've picked up, this makes more sense to me).

 

So many replies from tiller and he still can't describe his work flow or what his actual work is lol.

Sent from my Amazon Kindle Fire using Tapatalk

 

http://forum.notebookreview.com/sol...40-120gb-endurance-testing-3.html#post9694656


So many posts from people that don't follow a conversation but still feel the need to belittle at any chance they (think) they see.

 

My reasoning was that the OS is continually accessing the SSD for background operations like event logging and such. When I was researching the replacement for my M5S, I did some disk access profiling on my C drive. You'd be surprised at idle, Windows 7 basically continuously issues Random 4k Writes (about 50% of idle windows traffic) and a smattering of Random 4k reads, the number of 4k reads obviously increases with light to medium desktop browsing but the continuous write stream is ever present. Therefore the boot SSD is technically always in a multi tasking, albeit very light, so trading burst performance might mean a snappier interface but it will likely be a snappier interface with random hiccups or slowdowns due to the reduced consistency (the plextor m5s was notorious for this).
Personally I would trade 2 seconds of boot time for a smooth consistent desktop.

As the the game storage. I anticipated the workload to be most monotasking (e.g. dealing with game assets) logging mostly 4k sequential reads and writes with relatively large time gaps in between, thus I don't expect performance consistency to play a big role. The main workload is already executed at game loading and level loading.

 

Nope I read that and the multitude of other vague posts you've posted thru the years.

But feel free to act like an expert, whatever floats your boat.

 

And you're not alone in that approach...:thumbsup:

 

Some well made points I think you made. I've briefly looked at OS drive usage using Resource Manager, and it is a VERY light workload albeit consistent like you mentioned. It's so light with queue depths being virtually consistently 0.00 occasionally flipping up to 0.01 when just browsing the net, or flipping up very briefly a little higher when opening programs, that I don't see 'sustained' performance of any SSD being tested when used as a boot drive, that's why I still think an SSD drive with best burst performance would give a better experience. Moving big files around and installing new games on your SSD (the data drive) is going to be a lot more 'sustained' acitivity than the data drive, hence why I think 'sustained' performance would be a more important factor in an SSD for a data/gaming drive. I suppose I'm nitpicking though (even if there is truth in my words), as the perceived difference between having your Samsung 840 Pro or Sandisk Extreme II as the OS drive is probably close to zero; the same could probably be said for your gaming/data drive. Although, it is kind of fun to know that you've optimised everything to within an inch (or more aptly a mm) of it's life!