If you didn`t like Samsung SSDs before, you will now...

Well, here is one 'proof' that RAPID mode is not all it's cracked up to be...

See:
A closer look at RAPID DRAM caching on the Samsung 840 EVO SSD - The Tech Report - Page 1



Just like I predicted in post 4 of this thread.


See:
http://forum.notebookreview.com/sol...ung-ssds-before-you-will-now.html#post9302702



As the conclusion of the first link states:

RAPID mode is definitely a marketing ploy by Samsung to leverage the Nvelo buyout they sunk good money into... and using benchmarks (i.e. 'gigo') to substantiate their claims.


And while the first link also concludes somewhat positively about RAPID's future - I am still betting that no software solution will improve in real world use the performance of any SSD going forward (unless the performance of the SSD is sub-par to begin with - with RAPID disabled).


Seems like the Samsung lagginess lives...

 

...in your head.

 

The link Tiller supplied worked. I guess that wasn't in his head.

I thought the most interesting observation was this.

I'm still waiting to see a "Steady State" mixed Read/Write review to see if EVO improved over the 840 and 840 Pro's mediocre "real world" results. Get off your butts [H]ardOCP and review this drive already! Samsung emphasized low queue depth performance with the EVO maybe it will really perform outside of it's empty drive benchmarks.

 

No, not in my head and that quote, Bullrun, is where I mentioned the "And while the first link also concludes somewhat positively about RAPID's future ..."

The only caching software that made a real day to day difference in actual work with my workflows/machines was eBoostr.

See:
How to Make Your Computer Faster | eBoostr


As soon as real SSD's became available (Intel 510 Series 250GB) eBoostr became obsolete too.


Take care.

 

OR, TechReport didn`t test RAPID like they were suppose to.

You don`t just fire up an application and see the boost right there and then. You have to do the same task for a period of time before the software cache it in the RAM

 

Or, someone doesn't read the techreport fully:

Right on the first page...

 

Yup, the link works and actually shows that samsung's "lagginess" make their latest SSD top of the line in almost every test.
Unlike other contenders... except 840 pro ? They also have the best price/perf ratio if I'm not wrong, I think it should be noted.

So yes, samsung's lagginess definitely only lives in his head

About RAPID, except the DriveBench, it seems pretty decent. Let's just hope it will improve over time and become really useful without issue.

 

No, the trademark Samsung slow to react to input storage devices 'issues' are not in my head. But it's okay if you think they're making your system as fast as your system can be.

To get a modern version of eBoostr (and see RAPID done right) you should check this out:

SuperTalent USB 3.0 Express Dram Disk


See:
SuperTalent USB 3.0 Express Dram Disk Review - Simplicity With Ultra High Performance | The SSD Review

And benchmarks are not what to compare to... real world use is the only metric that counts (if you don't want to simply throw money away).

 

All SSDs slow down with data on them or in a used state referred to as "Steady State." If the 840 EVO performs like the previous 840 and 840 Pro in the "Steady State" metric the "lagginess" that Tiller experienced and talks about will continue. As far as I know, the SNIA guidelines "Steady State" metric that [H]ardOCP tests under is the closest to a "real world" performance test metric done by any review site for the average user.

To see how drives stack up, I included a few links to result pages.

840 Pro review:
HARDOCP - Iometer & Steady State Testing - Samsung 840 Pro SSD Review
840 500 TLC review:
HARDOCP - Iometer & Steady State Testing - Samsung 840 Series 500GB TLC SSD Review
M500 480
HARDOCP - Iometer & Steady State Testing - Crucial M500 480GB SSD Review
Sandisk Extreme II 240:
HARDOCP - Iometer & Steady State Testing - SanDisk 240 GB Extreme II SSD Review

I wish more sites did this test metric, as [H]ardOCP hasn't done a SSD review in almost three months. I think they are still looking for a reviewer. It seems most sites do the usual, ATTO, Vantage, Crystal. etc. TweakTown at least does Vantage with their "Drives With Data Testing" as well as empty. Why don't more sites do something the consumer can use to judge how the drive might really perform for them? All the empty drive benchmarks are useless for "real world" performance. I don't know anyone that will leave the drive empty. All the "real world" tests that are done on empty drives aren't done in "real world" conditions and are meaningless to average consumer usage.

Samsung has emphasized low queue depth performance with the EVO. They know 32QD is pointless in consumer SSDs, as they are all too fast to let requests stack up like in an HDD or in server/enterprise environments. Hopefully, EVO will do well in "Steady State" when tested. Until then, I'll pass on the empty drive benchmark kool-aid that made so many purchase.

 

Your link is full of benchmarks, credibility just went down to 0. Oh, wait, it was already there, sorry.
As for samsung slow to react to its devices issues, that's fine with me since they don't seem to be flawed in the first place.
I used 2 intel SSD, a crucial SSD, and 2 samsung SSD, samsungs are definitely snappier.

FYI, I live in real world.

Well I don't see any lag in those HardOCP results.
Steady state Read : Samsung 840 pro first
Write: 3rd or 4th
Read/Write mix: 2nd ex-aequo.

Can you find one superior drive ? Oh yeah, samsung 840 evo.
And guess what, being the best won't prevent them to also get the best price point.

I think you guys don't know what "lag" means. In this context it would actually mean being (far) behind the others.
If you don't like samsung as a trademark, fine, there are valid reasons to this, no need to make up some.

Let me quote the linked techreport test for you :

"Some of our tests then put the SSDs into a used state before the workload begins, which better exposes each drive's long-term performance characteristics. In other tests, like DriveBench and FileBench, we induce a used state before testing."

 

Samsung EVO is the fastest drive out there along with Vector. It beats Samsung PRO. Period.
Samsung 840 EVO 750GB SSD Review - Real world tests | Myce.com

Also, I dont trust that TechReport review. We had a user here applying the RAPID and he saw real gains.

I will get my Samsung EVO next week, then I will do my own tests before concluding. But even without RAPID, its the most snappiest drive out there.


The 4K Reads are the fastest out there, and that is really important for the user, since that will translate in to snappiness.

 

Well, they didn't say there was no gain, they saw some too, especially in the write department. But drawbacks seem to outstand them as for now.
I couldn't see myself the difference, but I'm not picky. So I disabled it to gain that 1gig of RAM back, which I need more =)

 

I will do my own testing too.

I hope they increase the 1GB limit to 2 or 4GB if they use a percentage limit. So 16GB/32GB users can do 4GB

 

Second to four different drives really isn't second is it? The M5 Pro Xtreme does better here than the M5 Pro linked.

840 Pro: I think being down by roughly 30% is close enough to the definition of "lag." Never reaching 20K while others are in the upper 20Ks to low 30Ks in a mix of low queue depth Reads and Writes, that is where most users will be most of the time with their OS drive, is what I would call mediocre and that's not trashing the drive. Cloudfire talks about "snappiness" or "anti-lagginess" in 4K Reads being important and it is but not on an empty drive. That's only part of the equation that users will operate under. "Steady State" Read/Write Mix is the test metric where it's shown.

TechReport takes a drive secure erases, runs their test for fresh then runs a 30 minute IOMeter workload and calls this a tortured used state. The drive is, essentially, still empty. It's slightly better at predicting long-term performance characteristics than the fresh state. At least it's not an empty drive, quick burst benchmark. Again, no user that I know of will leave their drive empty. There are guidelines to get the drive into "Steady State", where all drives will eventually be, but it seems only [H]ardOCP uses them for their reviews.

I certainly understand defending your purchase decision, as I would defend the purchase of my Samsung 830 256GB, a good drive at a great price, based on what I knew at the time of purchase. But I won't ignore what I've learned since then. Take off your blinders and read what I wrote: "Samsung has emphasized low queue depth performance with the EVO.... Hopefully, EVO will do well in "Steady State" when tested..." I'm not damning the EVO... yet... or singing it's praises. At the heart of EVO is a small SLC layer, like the newer Sandisks. But the body is TLC. The 840 TLC was at the bottom in the results linked.

Price is definitely a factor in taking a purchase decision. EVO is competitive with the M500 comparing 480-500GB, depending on the day or sales, but at 960-1TB Crucial has the edge by $50-$70, again sales. I wouldn't consider a smaller M500, period. The M500 was designed to compete with the 840 in the value market. It also competes well in the enthusiast market as an OS drive because of it's excellent "Steady State" performance.

Maybe EVO will be a superior drive, other than it's superior idle power draw (important) and it's superior empty drive benchmarks (unimportant), The "Steady State" jury is still out.

 

"Real "World" tests but not in "Real World" conditions. The drive is empty. Do you know anyone that won't put anything on their drive? I don't. Myce.com doesn't.

They should have put data on the drive and re-run these tests. Your link would carry more weight. They could have easily done this because they do a "Sustained Performance Test" with data on the drives. But they choose to run benchmarks again, with burst results.

TweakTown runs PCMark Vantage for their real world tests like most sites, on an empty drive. But they also run Vantage in their "Drives With Data Testing."

The results here are great for EVO, better than 840 Pro.

http://www.tweaktown.com/reviews/5630/samsung-840-evo-750gb-ssd-review/index10.html

That's part of the equation for "snappiness." But it's not 4K Reads on an empty drive alone, without writes and in a "Steady State" Samsung Pro (don't know about EVO yet) has excellent read results in "Steady State." It doesn't really change from it's "Fresh State." Writes on the other hand aren't great. And the most important, a mix of reads and writes, OS like performance, the results aren't impressive.

HARDOCP - Iometer & Steady State Testing - Crucial M500 480GB SSD Review

The fastest runner in a 100M race may not be the quickest (snappiest) to 50M but wins the race. An OS SSD needs to be quick, the race is over at 50M. So, to say that Samsung SSDs suffer "lagginess" is fair and accurate based on these results, It's empty drive benchmarks show it as one of the fastest and quickest. But not it's "Steady State" results. All SSDs slow down with data on the drive and eventually get to a "Steady State." EVO has excellent empty drive benchmarks and does very well with data on the drive in heavy read related tests. But it has not been tested yet in "Steady State", that I know of.

Are you serious, Cloudfire? You would trust the word of a user over a review site? You can, at least, challenge the methods they use if you think they got it wrong. Just as I think MyCE could have done their "real world" test better or TechReport tested in their "used state." If you think they are biased show why. I may agree with you. If you don't trust them because you don't like their results you lose all credibility and you're just a fanboy. An end user has no bias to justify their purchase in their testing.

I think you already have concluded.
If you really want to test it, how about putting it in "Steady State" following SNIA guidelines? Of course, that's if it's possible with your system and you determine it won't introduce unnecessary writes to your drive. You can be the first as [H]ardOCP hasn't reviewed it yet.

[H]ardOCP

SNIA Rev 1.1
http://snia.org/sites/default/files/SSS%20PTS%20Client%20-%20v1.1.pdf

 

Ok, that is where you're wrong.
That is called normal operation, as you explain it yourself:

As you can see, there's nothing wrong with samsung's steady state results.
Are you all too young and never used HDD ? That kind of performance I would call lagging nowadays.
A drive isn't lagging just because it's not first in the results.

I'll leave that discussion now, no need to debate on such futile arguments, enjoy your SSD everyone, whatever it is

 

Man what has this guy against Sammy?? Lagginess? I don't see it with my 470 and that is a few years old now....

 

Bullrun and tiller throws around scenarios and technical terms that mostly doesnt matter to anyone.

First they kept telling everyone to buy Intel 520 yet it was painfully slow compared to the rest. Why? Because it could do QD32 and sick iops datastreams better than other consumer SSDs due to the fact it use an enterprise controller. Now it is the Extreme 2 they have their eye on amd steady state which apparantly the drive is better there.

It will always be some corners and aspects of any SSD where it is not entirely top notch. And therefor people who are fanboy of some sort of brand always spin and spin to lure people in.

Just try to read reviews that have real world scenarios (like MyCE I listed earlier) or read workbench (like Anandtech's light workload)) where they throw real programs at the SSDs. And make up your own mind.

 

The best idea is always to look for tests that use the drive in similar ways you will. Also of note is that most people do not notice differences between SSDs in actual use.

 

Exactly. And you dont go recommend a SSD based on your use if it differs from the guy you are trying to help....

As for most people do not notice difference between SSDs, maybe not on its own with nothing to compare against, but take this as an example: Installing Office.
There is 11 seconds between Intel 520 and Samsung EVO.

 

As a prospective buyer I appreciate Bullrun's critique.

The speed increases for an empty drive are not a big deal to me. Any SSD will be significantly faster than my HDDs. I'm much more interested in durability and consistency.

It seems to me as though most of the "features" on many of these drives are avoiding the real technical advances which would be updating the actual hardware of the storage system itself. e.g for RAPID I don't really care to sacrifice my total RAM if storage is no longer my bottleneck. Sure faster is better, but I don't want to see faster storage at the expense of either durability or other system components.

That's ok if others would prefer that (i.e. some may have heaps of RAM) but I think it's important that some knowledgable users explain to people like me what the differences and potential drawbacks are.

 

Potential drawbacks? For the RAPID RAM cache, I'm not aware of any other than RAM usage. Actually my Linux box (with 32G RAM) uses all unallocated RAM space as cache. For the EVO fake SLC thing, 4X write amplification on TLC NAND isn't fun if you care about durability.

SSD brands sell products with those functions because they sell the drives as independent items, not integrated into systems. They want you to believe that the product is superior on its own.

Emulated SLC does make sense if you can't have enough RAM though, as long as durability isn't a problem.

 

I guess simply challenging the "fastest SSD" or "best SSD" claims that are based on empty drive benchmarks, gets me accused of having something against Samsung, (more than once already in this thread). I, certainly, understand people defending their purchase decision. Blind loyalty, not so much. I've learned a whole lot more after reading what knowledgeable members shared here than I knew before joining. I wanted to learn about Clevo, I started here. Again, I've got nothing against Samsung. As I wrote, I'm happy with my 830 purchase. And all my other Samsung products for that matter. Highly recommended :thumbsup:

If the difference of your definition of "lag" is SSD versus HDD, then yes, I'm wrong. I thought this thread was comparing Samsung SSDs versus the competition (it's earlier models too). Buy the cheapest asynchronous NAND SSD you can, it won't lag like an HDD.

I will assume that my use of the word "down" and not "slower" was a poor choice and led to a misunderstanding, writing "down 30%" meaning 30% slower.

Yes, all SSDs slow down with data on the drive. They don't all slow the same amount. They use different controllers, firmware, NAND type and size, page size, number of die on NAND, ...etc. Look at the M500, it's "Steady State Read" is about even at the bottom. It's "S.S Write" is only better than the TLC. But it's "S.S. Mix" is excellent. Crucial did something right here with their firmware, I suspect.
.
The "Steady State" graph posted has the Extreme II at about 23K at QD6 840 Pro 18K that's about 22% down or slower. Of the links I posted that was the closest. Then there is 28%, 35% and 40%. Saying it is down (slower) roughly 30% wasn't unfair but maybe generous. The king of the empty drive benchmarks behaves differently when you buy it and use it (get it to "Steady State" naturally) as an OS drive. It is not in the top 5. That doesn't mean I wouldn't buy it before other drives that are ahead of it in the "Steady State" metric. More factors than this one should lead to a buying decision that is best for the individual.

 

Really, Cloudfire? How a SSD performs the way most people will use it, with data on the drive and in "Steady State", as an OS drive, doesn't matter to anyone? I guess "Steady State" is the technical term to which you refer? Here's [H]ardOCP's explanation:

How SSDs perform here should matter to everyone that is interested in performance or getting what they paid for, value. These results should matter to manufacturers and more review sites. A consumer that buys a SSD based on the unrealistic empty drive benchmarks, that they are flooded with, could easily be disappointed or unhappy with their purchase shortly after buying. A happy customer is a return customer. This industry is trying to get consumers to switch over to their tech and to pay a much higher premium per GB. Truth in advertising will go a long way.

You must have me confused with someone else. I've never used or owned an Intel 520. I would like to though. It's too expensive right now to justify a purchase. When I bought my Samsung 830 256GB, I chose it over the Intel 520. I bought it based on what I knew at the time and at a great price. I also own an Intel 335 240GB, value drive on sale, in a light resource system. And I own a Crucial M500 960GB in a class by itself at the time of purchase. So, I'm a fanboy of what brand?

You get what I've written about QD32 exactly wrong. QD32 is meaningless for the average consumer. SSDs are too fast to let requests stack up. High queue depth performance is important in Server/Enterprise SSD environments and seen in HDDs.

2.5" I recommend Extreme II in 240GB performance M500 480GB and 960GB price/performance. If EVO does well in "Steady State", when tested, I could easily recommend it.
mSATA Plextor 256GB mSATA No "Steady State" results yet. but price/performance with reviews available it's a worthy choice. Hopefully, the newly released but untested M500 480GB mSATA will perform like the 2.5".

Don't buy any asynchronous NAND SSD. I don't recommend 120-240GB M500 2.5" or mSATA big performance hit from architectural changes. 120-128GB Samsung 840, 840 Pro, PM841. PM851, 830 or EVO big latency issues in this size.

MyCE and AnandTech do their "real world" tests or light workload on empty SSDs. Running "real world" tests or programs on empty drives are not "real world" conditions, I gave you a link to a better argument for EVO, "real world" tests with data on the drive and yet you insist on using your poor example and then you add another poor example.

 

Hello. Is there a big difference between Samsung 830 and 840 Pro in terms of speed ?

 

Nop, that was you saying (or approving tiller, which is quite the same) the samsung SSDs are lagging.
They're not "lagging" more than other SSDs once they're filled with data. Just because they're not the fastest anymore in a few benchmarks does not mean they're lagging, that's all I said.

Apparently you never used those. Try an eeePC with those first gen SSD in a steady state, then you'll understand what lagging really means. Seems like samsung evo with rapid enabled is actually lagging in some particular tests.

If I take your description of "lagging", then almost any brand's SSD is lagging since it won't be at the top for some tests. And that would be just plain ridiculous.
You will never see a reviewer make such a statement and for a valid reason : they actually know what they're talking about.
You can still point out that steady state is not samsung specialty and you would probably be right (at least for some of their drives). You can say they're not the best in that condition, no problem. But saying samsung SSDs are lagging (which implicitly states "more than others"), is completely wrong.

 

If to speak in a SSD type difference then yeah (one of the first vs one in the middle of the list which include newest SSD), but if to speak in general, then no, not a big difference.
For usual using you better buy cheaper one. Only if you store data base, SQL etc then you may want to use 840 Pro.

 

I'm quoting lastnikita because his remark makes a lot of sense.

I'm sure the more vocal forum members in this thread have already read this AnandTech article but i'd like to bring it into the attention of less documented readers. People should know this "steady state performance" is quite relative, one can control how much performance degrades by simply using more or less over-provisioning. It's not ideal, but it's part of your choice, as a consumer.

 

So much to comment on here.
As for the "Steady state" some people keep throwing around here, there is a bunch of fill tests available on the internet for everyone to see.

Does the EVO drop down in speed when you fill it? Yes, pretty much all SSDs does that. You can see yourself that the EVO is running faster than PRO when filled 75% with RAPID enabled. 56000 vs 48000. And PRO is bloody fast in real world.



Sandisk Extreme 2 is slower than Samsung PRO when filled, so that should tell you something...
http://www.rwlabs.com/images/articles/sandisk/Extreme_II_480GB/55.png

TLC is bad at durability is a beaten horse which have been proven wrong so many times. It doesnt matter for a user if he choose SLC or TLC. Take whats cheapest. SSD with TLC NAND will outlast ALL your other components in your computer, so it is a moot point. Technology progress, in 5 years the TLC drive will still have really good health, but SsD technology have progressed to the point it where you would want to replace it.


The whole Samsung bashing that goes around in this forum, with spewing nonsense like Samsung is lagging when they clearly are one of the fastest and most snappy drives out there, speaks volume about those people.

 

It would be very helpful if a review site over-provisioned drives at different levels(percentages) then put it in "Steady State" and different amounts of data on the drive. [H]ardOCP chose 75% data for their "Steady State" test. This would be much more beneficial for the consumer in the buying decision and helping them choose how best to OP for their use, if at all. Until that time, users have to go by what information is available, use what's important to their usage, and throw out the rest.

When I look to buy for myself or if I offer a recommendation, it's pretty obvious by now, I give a lot of weight to "Steady State" results and throw out empty drive benchmarks, (that is for an OS drive). There really is not much "Steady State" information available. So, then fill tests with "real world" programs would have possibly the most importance. AnandTech's performance consistency would gain more importance and workbenches. User experiences, I want to read the 1 and 2 star or egg reviews: why the bad review verified owner first and foremost, performance, reliability, biased.

 

Just to make things clear, [H]ardOCP did not overprovision to 25%, in fact they did not overprovision at all (and they did the right thing, they reviewed the product at nominal specs, as given by manufacturer). They filled 75% of the drive and then ran drive tests until performance numbers reached a steady state. There is a clear difference between free partition space and over-provisioned space.

• A drive with 25% free partitioned space may or may not contain 25% cells ready to be written to, depending on how many of them have been checked as clear through trim/garbage collection.
• A drive with 25% over-provisioned space will always have 25% cells ready to be written to.

 

No, that is not true. A drive with 25% over provisioning may have 25% of it's cells ready to be written to (just like the drive with simply free space), or, it may not - but it will depend on what the workload was like until that write request is executed.


The benefits of leaving 'unallocated' space on as SSD vs. simply leaving free space is that the SSD's controller will 'know' without O/S intervention (TRIM) that the data it contains is throw away (and therefore doesn't rely on TRIM to tell it first (before it gets deleted).

 

Enough of this HardOCP crap already.

A) They havent tested the EVO. Assumptions are being made and presented as truth but people here forget that
B) Over provisioning area on EVO compared to PRO have increased from 7% -13%
C) HardOCP tested the 256GB PRO against 480GB versions of other brands (unfair due to over provisioning)
D) The controller on the EVO is clocked higher than PRO that might result in an entirely different result
E) Data posted here shows that EVO is faster than Sandisk Extreme 2 when filled, included when used in real world scenarios. The only time Sandisk is ahead is on 4K QD32 workloads, where its much more consistent.
I posted 4K results, where EVO was over 11MB/s faster on 4K read than all other brands.

Seriously, stop and think for yourself instead of eating what other people say, raw.

Truth is that there are not just one but several SSDs out on the market that is really good. They all have their pros and cons. None of them are perfect.
Read reviews, look at scenarios you will encounter most, and pick your choice from that.

Bah, im out.

 

Thanks coercitiv, I can see how readers might conclude I inferred it was 25% OPed the way I wrote it. :thumbsup:

 

You can buy a first generation asynchronous NAND SSD today? Did you read my reply to Cloudfire? I wrote "Don't buy any asynchronous NAND SSD."

Is this "(which implicitly states "more than others")" from a definition of the verb lag (to lag)?
It is completely accurate because I linked 840 Pro lagging behind more than four others. It doesn't have to lag behind more than all others to use the word to describe it's performance as lagging. The definition of the verb "lag" (to lag, lagging) does not mean to finish last. I will use a marathon analogy. In many races different groups of runners are separated. It's accurate to say one group lags behind another by time or distance. If the winner has an easy victory, it's accurate to say the field lagged behind. But it wouldn't be heard as often as "he blew away the field" because the victor is talked about. In reviews it's "Samsung dominates" or words to that affect. I think being roughly 30% slower is enough to justify using the word lag. You disagree. We will probably have to agree to disagree.

 

Cloudfire you made me smile when I scrolled down and saw the chart. I thought, briefly, you're not so closed minded. ;p You're welcome, by the way. But then you turned my smile upside down. How is pointing out that 840 Pro's excellent empty drive benchmarks don't translate to it's "Steady State" results, bashing it or spewing nonsense? I at least, point to results. Don't shoot the messenger.

 

[H]ardOCP, [H]ardOCP, [H]ardOCP Sorry, I couldn't resist.

A. From, I believe, my first post in this thread: "I'm still waiting to see a "Steady State" mixed Read/Write review to see if EVO improved over the 840 and 840 Pro's mediocre "real world" results. Get off your butts [H]ardOCP and review this drive already! Samsung emphasized low queue depth performance with the EVO maybe it will really perform outside of it's empty drive benchmarks."
B. No dispute
C. (unfair due to over provisioning) What?! You can't really mean it's unfair because one brand OPs more than Pro. Who decided how much to OP the Pro, Samsung or it's competitors? There just aren't a lot of reviews to compare all drive sizes. Samsung 256GB and Vector 256GB are widely considered the top two drives and their results are used since they were first reviewed. Results for 256GB M5 Pro and 240GB Extreme II not just 480GB M500.
D. Who is speculating now on the untested EVO on [H]ardOCP?
E. You're welcome

The best thing you've written in this thread, a moment of unbiased honesty. :thumbsup: Remember though, pointing out a con is not bashing if you support the argument with references.

Now go read the whole Samsung 840 Pro review from beginning to end, Cloudfire You will start using [H]ardOCP

 

You already have the test you need done by Anandtech. The Storage Bench 2013 a.k.a the Destroyer writes about 900GB and reads about 1500GB of data in a 10 hour long test simulating all kinds of disk activities (from real world programs).

The large capacity EVOs do quite well in the test, while any Samsung drive with a capacity equal or lower than 512GB shows some fatigue. There is a clear need for some over-provisioning in such a hard hitting test, but keep in mind this kind of load is unrealistic for 99% of users on this forum (almost 900 GB of data written in less than 10 hours).

 

What would you guys recommend for a swap (on-disk virtual memory) drive? 900GB within 10 hours is nothing for me when playing with numerical models, and I need a good SSD to handle the job since RAM size is limited to 32GB on notebooks.

The swap data is moved around in large pages so I don't care about sub-4k performance.

While wear isn't a problem for average consumers, it does matter to me. I'd like to avoid TLC unless someone can prove it's actually cheaper from a total endurance/cost perspective, which I highly doubt.

I don't need a large drive, ~256GB should be big enough. I do worry about cost so going all-out for marginal benefit isn't what I'm after.

 

If you plan to write more than 1TB per day and expect a decent lifetime for the drive... you are going straight into SLC territory, and budget hell.

 

But which is cheaper, (same capacity) cheap MLC/TLC SSD replaced more often, or expensive SLC SSD replaced less often?

 

The cheaper option in the long run is usually the more expensive option 'now'.

Quality is not found at bargain basement prices. Neither is TLC nand aimed at anything other than a way for consumers to have a 'checklist' item that yeah, they too have an SSD...

If you have a real and heavy workload SLC is the most reliable, most durable and most sustained performance over time you can buy.

However, don't ignore MLC based SSD just yet - you say you don't want to over spend (on capacity) - but by buying a 480/512GB model and only using ~200GB or so, you have effectively achieved SLC durability and performance levels - at less than half the cost (and up to 10x less in some cases).


Good luck.

 

The "unrealistic for 99%" is why I prefer [H]ardOCP's take.

If I used the Destroyer and the "Steady State R/W Mix" results to critique the 840 Pro 256GB I would be accused of being a "hater"

 

Let's take a look at what Intel has to offer and compare price/endurance of their DC 3700 and DC 3500 drives.

The High Endurance MLC based - 100GB Intel DC 3700 costs roughly 220Eur and is guaranteed to endure at least 1.825 PB of writes.
The MLC based - 120GB Intel DC 3500 costs roughly 170Eur and is guaranteed for 70 TB of writes.

So the 3700 is 30% more expensive and lasts 26x times longer. Need more data?

 

size guaranteed doesn't mean actual possible size.
SSD Write Endurance 25nm Vs 34nm

 

Sorry, but your math is wrong.

Guaranteed is not equivalent to "lasts 26x longer"...

As usual, it depends on the workload presented to the drive (and the DC3700 is for data centers - not workstation class workloads...).

 

The man asked for a swap drive with a daily write load greater than 1TB. What workstation drive would you recommend?

No comment.

 

You probably meant "write endurance" instead of "size". Now, can you please show some "actual possible" write endurance data for Intel 25nm eMLC?

 

This one might be worth a look. In the 240-256GB class, for the endurance/price, $250 for a 240GB PNY Prevail Elite, it uses 10K P/E cycles, Intel eMLC 25nm NAND flash, in a consumer drive. 3 year warranty but you get 5 years to simply register online. pny.com I would check the warranty to see if they have a TB written to NAND limit.
PNY Prevail Elite SSD9SC240GEDA-PB 2.5" 240GB SATA III Internal Solid State Drive (SSD) 10K Endurance - Newegg.com

Performance is fairly close to the big boys. Linked to specs page.
PNY Prevail Elite 240GB SSD Review - Specifications, Pricing and Availability | TweakTown

EDIT Oops not in stock. They have a 480GB model too. Found it for $433 at CompSource. I've never shopped there but they have a 4.6 out of 5 stars 218 reviews on google shopping Not sure about International shipping.
Pny Technologies SSD7SC480GCDAPB Prevail Elite Solid State Drive :

Found the 480GB here for $464 at CTI ComputechInternational Same story with this vendor, I've never shopped there, but 4.8 out of 5 stars 261 reviews on google shopping, I would buy from them. Again, not sure about International shipping.
PNY Prevail Elite 480 GB 2.5" Internal Solid State Drive

In the last generation, 240GB was the sweet spot for 25nm SandForce performance. So, the 480GB could be slightly behind.

 

Ummm I still don't like Samsung SSD's, supposedly as one of top manufacturers of SSD's they still haven't come out with a consumer based mSATA SSD while other manufacturers have been in the game for over 2 years now.