Slashdot Mirror

With SSDs it is more like: Fast / Cheap / Reliable (pick two).

Samsung delivers the best reliability (for sample size of one anyway) and is acceptably priced. It also appeared to be one of the fastest but this bug has proved it otherwise.

So, if they can fix this performance bug they might just be all three.

Handling power off issues is a different problem. What the GP was referring to is how drives will fail spectacularly in the face of anything seen as corruption. You can see some examples in some longevity failure tests.

The problem in those cases was wearout, but the way that happens is scary. Let's say there's a bug in the firmware that causes a write to fail for no good reason. It's quite likely that the drive will kick into a mode where it doesn't trust itself anymore. And the way that will play out on most SSDs, the drive will shut itself down at the firmware level, so it isn't even picked up by the BIOS on boot anymore. What people would expect is getting read-only behavior there; instead they will find everything gone. And unlike most catastrophic spinning drive failures, you could easily hit the same bug that wipes out your data on both halves of a RAID-1 pair at the same time.

-_-, ok, this true (just looked it up).

http://techreport.com/review/2...

No other manufacture (that I know of) purposefully bricks hardware! Wow! Their enterprise products OTOH will put it in read-only. It will be real interesting to see a stress-test on the 750 series. If they start dropping like flies, there will be a huge community backlash; especially for a first-generation flagship product. Meaning, they can lose the foothold into the market as quickly as it could be gained.

It will be interesting if Samsung has something in the works.

I see your 900 TB and raise you to 2 freaking petabytes. Not quite a min raise, I realize...

http://techreport.com/news/279... Maybe nail it to the floor beams.

Some additional info from an earlier article:

According to Intel, this end-of-life behavior generally matches what's supposed to happen. The write errors suggest the 335 Series had entered read-only mode. When the power is cycled in this state, a sort of self-destruct mechanism is triggered, rendering the drive unresponsive. Intel really doesn't want its client SSDs to be used after the flash has exceeded its lifetime spec. The firm's enterprise drives are designed to remain in logical disable mode after the MWI bottoms out, regardless of whether the power is cycled. Those server-focused SSDs will still brick themselves if data integrity can't be verified, though.

SMART functionality is supposed to persist in logical disable mode, so it's unclear what happened to our test subject there. Intel says attempting writes in the read-only state could cause problems, so the fact that Anvil kept trying to push data onto the drive may have been a factor.

All things considered, the 335 Series died in a reasonably graceful, predictable manner. SMART warnings popped up long before write errors occurred, providing plenty of time—and additional write headroom—for users to prepare.

So, it sounds like this is the intended behavior for *enterprise* drives. It may not be the same for *consumer* drives, but that's a bit unclear.

While it may make you feel better if consumer SSD drives would go into a permanent read-only mode, it seems extremely unlikely that a typical consumer would ever actually reach this point in an SSD's life at all. So, I'm not really losing sleep that my own Intel SSD drives are going to brick themselves, when at a typical consumer write volume, this isn't going to happen anytime in the next century (seriously, look at the volume of data that was written). The drive will long be dead because of some electronic component failure long before I reach it's natural end of write life. Moreover, I'd appear to have plenty of warnings and could easily replace them long before that happened.

According to http://techreport.com/review/2... Intel TSX errata has been corrected

This can improve database performance (see http://www.anandtech.com/show/... )

I'm not a GPU designer.

But in the case of the ROP ..
(should I Google it? ROP = Raster Operator and here comes the URL: http://techreport.com/review/2...) .. supposedly doesn't matter (at all? Or for the number of pixels only?) because as the URL above says it:
"In an even crazier reality, that limit isn't even the primary fill rate constraint in this product, since the GTX 970's shader arrays can only send 52 pixels per clock onto the crossbar."

If it's the case that 56 is more than what is needed due to limits in other places on the card then whatever. I guess it's even fair to round of 1.75 MB to 2 MB. Heck, even 3.5 to 4 GB =P

As for how the slow access of the last 0.5 GB affect the other 3.5 GB I don't know and I won't look it up for this post. But sure, that's a real problem. And they should post the correct information and I guess being excessive on details is a good thing.

In the end it's still the card I would had bought.

nvidia did much more egregious deceptive marketing well over a decade ago in the GeForce4 MX, which didn't have programmable shaders like the GeForce3 series. I actually learned about it because the company where I worked was fooled into buying one of these for my computer, when the program I was using required one of the higher-spec devices. That misleading advertising actually did real financial damage, at least in my company's case, both in lost productivity and in replacing the hardware.

That really bothered me, and I started buying ATI cards for a time after that, since I didn't want to reward a company for underhanded behavior. I only started buying nvidia again quite a few years after that since the ATI drivers were giving me more trouble than they were worth.

This is not quite as bad, but it definitely smells a bit the same. A "mistake", really? I'm betting they wouldn't have made a "mistake" in underestimating their card's performance. Marketing at it's finest.

And if you, logically, don't trust the manufacturer's tests there are some others who have pushed over a terabyte to SSDs:
SSD endurance experiment

> How long can you sustain these kinds of I/O rates before burning the thing out?

If you were to sustain 1550 MByte/s write for 1 year, you'd write a total of 48 PB. (1550*60*60*24*365/1000/1000/1000), or 0.13 PB/day. In Techreport's endurance test, only two drives made it past 1.5 PB. So, if that is the bar, the drive would last only 11 days.

However, that would give you no time to read the data you'd written. Since you're not likely to write at max speed 24/7, the drive should last considerably longer.

http://techreport.com/review/2...

That's why SSDs keep a reserve of about 10%-30% of the logical storage as pre-TRIM'd. Some of the newer SSDs even reserve another bunch of the drive as a scratch pad for writes.

Looking at this benchmark, it seems reads are more likely to have a random long access time than writes. http://techreport.com/review/2...

Wear-leveling with SSDs isn't about reliability, more about longevity because a specific bit (really page) on a chip can only be rewritten so many times. Though once that point is hit, the data is still readable, just not writeable.

This sometimes happens in the best behaved drives, but in practice that's more of a myth than something that really happens. Most SSDs start forgetting data after they're powered down when they hit their end of wear life. See the SSD endurance at TechReport for some real world examples.

Samsung's 850 Pros have a 10-year warranty, although they are still quite expensive.

Also techreport.com has been running an endurance test, and a couple of the drives have reached 1.5 petabytes of writes without failing. I think they all lasted well beyond the manufacturers' expected write limits.

Basically, they've reached the point now that the average consumer can't wear out a drive.

You haven't done any math, if you had you'd realise they haven't been doing the test 24/7.

Look at the screengrab on this page:
http://techreport.com/review/2...

Here:
http://techreport.com/r.x/ssd-...

It shows that their test was rather slow, they were only writing at 208MB/s

At that speed it would take 58 Days 19 hours and 45 minutes to write 1 Petabyte.

I already did the math for a HDD - it would take about 100 days to write 1 Petabyte to a HDD.

Since they started the test in 2013, they could have done that easily.

You haven't done any math, if you had you'd realise they haven't been doing the test 24/7.

Look at the screengrab on this page:
http://techreport.com/review/2...

Here:
http://techreport.com/r.x/ssd-...

It shows that their test was rather slow, they were only writing at 208MB/s

At that speed it would take 58 Days 19 hours and 45 minutes to write 1 Petabyte.

I already did the math for a HDD - it would take about 100 days to write 1 Petabyte to a HDD.

Since they started the test in 2013, they could have done that easily.

Let's do some math here, shall we? At 200 MB/s, you can overwrite a 1 TB drive in an hour. 1 PB you can reach in a month. The hard drives are a few times larger than the SSDs, so you'd need ~ 10 TB instead of 2, which means 10 months.

Include all the actual variables, and you might get a usable answer. Just blowing data on the disk isn't the only thing this is doing (AFAIK). You've gotta detect errors, so you've gotta read back the data and validate it. This page goes through their full testing methodology (hint: they're using Anvil, a static file collection that includes a copy of a windows install, some applications, some movies, and some incompressible data, among other things, and every file has its md5sum checked after writing): http://techreport.com/review/2...

An easier calculation would be to scale their timelines to the HDD stats. For example:
Samsung 840 Pro sequential read/write: 540MB/s / 520MB/s (390MB/s for 128GB)
WD Caviar Black: about 180MB/s read/write (ex. http://www.storagereview.com/w...)
Rough math: 520 / 180 = 2.89 = it'll take 2.89 times as long to do the test on the same size drive.

Samsung 840 Pro size in the article: 256GB
Assuming you use WD Caviar Black 1TB = 4x's the size.
2.89 * 4 = 11.55 = that many times as long to do the same operations they've done thus far.

Their test has been running for over a year. So it'd take (roughly) over 11.5 years to do the same on the WD Caviar Black. I understand that's a very very rough estimate, but I think it's MUCH closer to the ballpark than 10 months!

My bet: the WD will be dead long before that time. I've had drives last longer than that, but they got VERY VERY little use and were simply powered on all the time. I've had some that lasted longer than that and got a fair bit of use (ex. db servers), but they were never filled to capacity, they were enterprise drives, and some of their neighbors did die (RAID).

Hybrid drives do not use their meager flash to cache writes. The flash would wear out in an instant if they did that.

Hybrids that use some of their cache on writes have been around since the Desktop SSHD 2TB in 2013. They play around various flash types and methods for using it to keep that from wearing out too fast.

600TB total writes - http://techreport.com/review/2...
800TB total writes, and some of these consumer grade drives start to fail - http://techreport.com/review/2...

I know, it's amazing. :)

Actually that story seems to still continue with two of the disks approaching 1.5 PB written.

600TB total writes - http://techreport.com/review/2...
800TB total writes, and some of these consumer grade drives start to fail - http://techreport.com/review/2...

I know, it's amazing. :)

Actually that story seems to still continue with two of the disks approaching 1.5 PB written.

600TB total writes - http://techreport.com/review/2...
800TB total writes, and some of these consumer grade drives start to fail - http://techreport.com/review/2...

I know, it's amazing. :)

Actually that story seems to still continue with two of the disks approaching 1.5 PB written.

...but if you run that exact same app on a solid state drive you will, depending on the specific NANDs used and the speed of your code, soon begin to lose bits. As enough of those bits fail, you lose those sectors ...

Actually, there is no way to write to specific blocks of an SSD. The controller automatically stripes writes across multiple die, and the wear leveling algorithm makes sure that it is not always the same blocks. SSDs will wear out, no doubt, but not as fast as you think. For some actual data see: http://techreport.com/review/2... Even TLC drives like the Samsung 840EVO 250G got past 600 TB. When you add up all the hours you saved by not waiting for spinning rust to do it's thing, SSDs are a fantastic value.

The 'wear out too fast' concept is wildly overblown. You can listen to old rumors, or read actual test data.

600TB total writes - http://techreport.com/review/2...
800TB total writes, and some of these consumer grade drives start to fail - http://techreport.com/review/2...

"By far the most telling takeaway thus far is the fact that all the drives have endured 600TB of writes without dying. That's an awful lot of data—well over 300GB per day for five years—and far more than typical PC users are ever likely to write to their drives. Even the most demanding power users would have a hard time pushing the endurance limits of these SSDs."

By contrast, my main home machine (120GB Kingston SSD) has ~7GB total, in over 2 years of 24/7 use. I'll leave you to do the math on lifespan for that.

The 'wear out too fast' concept is wildly overblown. You can listen to old rumors, or read actual test data.

600TB total writes - http://techreport.com/review/2...
800TB total writes, and some of these consumer grade drives start to fail - http://techreport.com/review/2...

"By far the most telling takeaway thus far is the fact that all the drives have endured 600TB of writes without dying. That's an awful lot of data—well over 300GB per day for five years—and far more than typical PC users are ever likely to write to their drives. Even the most demanding power users would have a hard time pushing the endurance limits of these SSDs."

By contrast, my main home machine (120GB Kingston SSD) has ~7GB total, in over 2 years of 24/7 use. I'll leave you to do the math on lifespan for that.

Correct me if I'm wrong, but don't SSD's have a point where they put on too many write's per bit?

Tech Reportchecked a bunch of SSDs for write durability and virtually all of them made it to 600 terabytes of data writes or better.

For an ordinary desktop user, write durability is not a problem. Now what about storage durability? With 3 bits per cell, how long before the data fades?

According to Techreport, Intel's three-dimensional NAND. will enable 10TB flash drives in servers in 2 years

not sure what scene that was, but the different was never THAT big that i saw in benchmarks.... citations: 1, 2, 3, 4

Anandtech disagrees. Techreport. So, in fact, do huge numbers of user reports which suggest that SSDs really do last a long time.

Further, multiplying this problem manyfold, is that when an SSD fails, it tends to fail totally.

I have seen this happen, but its not due to endurance of the flash cells but on the quality of the firmware / controller. The actual cell failures apparently cause reallocations (according to techreport's tests, and to common sense). And you create an interesting dichotomy; what does it look like for an SSD or HDD or CPU or RAM to fail "not totally"? You get most of your bits back? All tech generally tends to fail catastrophically.

And ironically specialized hardware is better than the CPU at raytracing and Intel might lose that battle as well after being its lone champion for so long.

http://techreport.com/news/261...

I realize that it is entirely anecdotal, but my miserable early experience with OCZ disks seems to match others, to the point that would never in a million years purchase another OCZ product again. Heck, it seems mighty telling that they're not even considered on the tech report longetivity test:

http://techreport.com/review/2...

According to http://techreport.com/news/268... :

- It can stream games from a PC
- It has a stylus to allow handwriting recognition
- It has processing power to emulate a nintendo wii (see https://www.youtube.com/watch?... )

Absolutely not true.

The Core 2 Duo is approximately 2x faster clock-for-clock versus the Pentium 4, and the current Haswell core is barely 40% faster than that (assume a 7% speedup per-clock for every core rev since). That gets you somewhere in the 2x-3x performance improvement range for Haswell, barring corner-cases that are embarrassingly easy to leverage AVX/FMA (most real-world use cases show small improvements).

Intel proved that they could do a whole lot better than the Pentium 4, but your performance improvement factor is off by half!

Have a 250GB Samsung 840 which so far has been reliable. Then again it has only been a year since installing it.

Have a look at this article: https://techreport.com/review/26523/the-ssd-endurance-experiment-casualties-on-the-way-to-a-petabyte

The Samsung did die an early death but the sample size is too low to be conclusive. Though, this does not worry me at all since my SSD is only for games. Plus I make backups :-)

They have always had issues with write endurance, since Day One. It has been their Achilles Heel.

Not according to real world testing.

Those are 240-256GB drives that have all had 600TB written to them, and many haven't even started to dip into the protected sectors. That's already up to 2400 P/E cycles on drives with 3000 or less on their specs. To put it into real-world terms, that's the equivalent of completely erasing the drive every day for 6.5 years. I think they'll last the same 7 to 8 years as your hard drives when used more realistically.

Nvidia PAYS for removal of features that work better on AMD

http://www.bit-tech.net/news/h...

Reading the link you posted above, it seems like a bit of a non-factual load of waffle. Nvidia deny paying, Ubisoft deny being paid, and the only sources mentioned are anonymous speculators we have no way of knowing are not just a few paid ATI shills.

Nvidia pays for insertion of USELESS features that work faster on their hardware

http://techreport.com/review/2...

Wow, another example of amazing journalism here.

Some guy moaning about Crysis having loads of detailing that is only used in the DirectX11 game. He give loads of examples of this, then posts a summary page of wild speculation with no sources quoted other than his own imagination. He never asks any of the companies involved, he just posts a bunch of stuff about why this might be the case.

I have another possible suggestion as to why this was the case: Crytek like making stuff look overly detailed and include graphics detailing that means their games continue to max out graphics cards long after they are released. They always make they games playable on the budget cards if you crank the detailing down, but they also like catering to people who buy a new graphics card then go back and play a few oldies that they had to crank the detail down on previously. Crytek also probably also quite like their games being used in hardware reviews because their games hammer the hardware.

Nvidia cripples their own middleware to disadwantage competitors

http://arstechnica.com/gaming/...

Ok, congratulations on actually posting an article that was real journalism, with quote sources and not just made up of the authors own conjecture.

The issue here though seems to be that there was an optimisation, moving from x87 to SSE that they did not do on a bunch of legacy code. Instead they rewrote it from scratch, which took slightly longer to use SSE.

This was not them intentionally doing something to hobble a competitor, this was them not doing anything to help them quickly. That is very different.

They did however ultimately fix it:

"PhysX SDK 3.0 was released in May 2011 and represented a significant rewrite of the SDK, bringing improvements such as more efficient multithreading and a unified code base for all supported platforms"

Intel did the same, but FTC put a stop to it
http://www.osnews.com/story/22...

There is a massive difference here, Intel's were intentionally hobbling the code their complier created based on finding a competing vendor name in the product string. They did not say "wait for version 3" like the PhysX case, they just did something then just sat their tight lipped until it went to court and they were forced to change it.

This is something FTC should weight in just like in Intels case.

As I said earlier, Nvidia made the all important change to use SSE when running PhysX on the CPU without the FTC being involved.

http://www.bit-tech.net/news/h... - What? It's just story speculation here

http://techreport.com/review/2... - the article doesn't state that Nvidia pays anyone, it's a statement you made up yourself.

At this point I decided not to waste anymore of my time after looking up the first 2 links

Nvidia PAYS for removal of features that work better on AMD

http://www.bit-tech.net/news/h...

Nvidia pays for insertion of USELESS features that work faster on their hardware

http://techreport.com/review/2...

Nvidia cripples their own middleware to disadwantage competitors

http://arstechnica.com/gaming/...

Intel did the same, but FTC put a stop to it
http://www.osnews.com/story/22...

so how exactly is that not Nvidias doing??

Nvidia is evil and plays dirty. They dont want your games to be good, they want them to be fast on Nvidia, any means necessary. They use "means to be played" program to lure developers in, pay them off and hijack their games to further nvidias goal.

For example how come Watch Dogs, a console title build from the grounds up with AMD GPU/CPU optimizations to run good on both current gen consoles, is crippled on PC when played on AMD hardware? How does this shit happen?

This is something FTC should weight in just like in Intels case.

Netflix? House of Cards and all of their new original series are shot and displayed in 4K if your device and display support it.
 
Also, there's a much higher quality Samsung 4K display for $50 more, that is probably the model you want.

they pull shit like this world's greatest virtual concrete slab:

http://techreport.com/review/2...

or this
http://www.bit-tech.net/news/h...

>improvement was to the tune of 20 percent as a result of DirectX 10.1 on ATI cards.
>there was some co-marketing between Nvidia and Ubisoft

All just to make AMD look slower.

You're in luck, as that time is right now!
http://techreport.com/review/2...

tl;dr - the Samsung 840 series is the only drive to really suffer problems but that's strictly relatively speaking; it's allocating from reserve capacity and to reach the point it's at now you'd have to have 150 gb of writes per day for 10 years, which is probably at least an order of magnitude higher than even a heavy standard user. And that's the consumer version -- the Intel ssd, aimed more at production / business environments, fares even better. Which mechanical hard drives do you use that support 150 gb of daily writes for 10 years?

For an "AMAZING" product like this is supposed to be, those are some awfully curated results right there. The product is only compared with a Core i3 (outdated Ivy Bridge) when it's actually going to win the test, and there's a plethora of multi-threaded tests that nobody on earth actually uses to put AMD's quad-core in the best light. Also, there are NO BATTERY LIFE TESTS to speak of, just "trust us" quoted TDP figures, and no pricing information.

And while it is MUCH faster at games than Bay Trail, it's not fast enough to play ANY modern games, even on the lowest setting possible. This leaves it firmly parked somewhere between tablet and ultraportable processing capabilities, so there's the question about product positioning.

Call me back when AMD is willing to let reviewers just have at it. If your product does not suck, then it does not need to be coddled.

So for a given number of dollars the smaller drive will have better performance.

First, this is a red herring, since the price you pay for an SSD in a given size class won't buy you any significantly larger drive. So, a 60GB dog of an SSD for $60 is still far faster than the zero IOPS you get from a $60 120GB SSD. What you really need to compare is the cost per GB, because then you can compare things like the performance of a pair of 60GB drives in RAID-0 vs. a single 120GB.

That said, the primary factor in SSD speed is the number of controller channels that can be connected to the flash chips. For an example, see pretty much any review (like this one). Because of this, smaller drives always have lower performance. Even crossing manufacturers/lines can only rarely make this untrue, as a doubling of size doubles the channels, so the flash on a smaller drive would have to be more than twice as fast to make up the difference. And although you are correct that SLC is faster than MLC, it's not twice as fast.

So, if you can find a larger drive that costs less in total dollars than a smaller drive (and it is possible...there are a few 120GB drives that cost less than 60GB drives), in every case you will get astoundingly more for your money, as you get more storage and more channels used on the controller, which gives you more performance.

MANY simple ways to deal with this, first off, buy as many Windows 8.1 licenses as you require, then:

install Windows 7
OR

install a 3rd-party desktop application on to recover the classic Windows shell
OR

download the recently released Windows 8.1 Update which restores a Win7-like start menu, and allows Metro Apps to run in a Window. Win 8.1 Start Menu demonstration

Seriously, the INCESSANT whining and whinging on here about Win8 has been plenty aggravating (almost as annoying as the original Win8 UI), but seriously, with these modifications, Windows 8.1 is PERFECTLY usable, and has a substantially better kernel than even Win7's... With this capitulation, there's really no GOOD reason to avoid Win 8.1 anymore...

-AC

Look at page 4

http://techreport.com/review/2...

The 730 drive is in the middle of the pack for sequential reads. None of the drives reach 550MB/s.

Recent reliability testing has been downright horrifying for the TLC based drives. I predict a whole lot of people buying Samsung 840 drives because they're cheap are going to regret that.

SSDs have their own longevity problems. In the race to get prices down, some of the new models are using some pretty shady flash in their designs, with the Samsung models being the worst right now.

I tried to find some hard data on either statement. It looks like the model number in TFS is a typo, and the test I found that showed results with BF4 neglected to explain what the medium settings are. It does, however, show us an average of 28fps, which would support your definitely not 30fps by a hairsbreadth. Now if only there were some technology to make that difference from 20fps count...

I'm not going to cram my current desktop rig into my home theater, because it's a powerful machine that is capable of doing far more than spitting out a movie or playing a Steam library.

Well for you there's In-home Streaming

Per Valve:

You can play all your Windows and Mac games on your SteamOS machine, too. Just turn on your existing computer and run Steam as you always have - then your SteamOS machine can stream those games over your home network straight to your TV!

The streaming function doesn't require a beefy machine ( even Tegra 3/4 devices can handle it).

Talk about NSA. I came across http://techreport.com/news/25828/report-nsa-intercepts-computer-shipments-plants-malware

The write cache on the Crucial M500 survived my power plug testing with similar properties to this article, but it doesn't have SMART data for drive wear. Can't take that one seriously for business use. I've been very happy with the figures I've seen from Intel drives on their internal lifespan tracking; see my Intel SSD lifespan for example. I'm not the only one who noticed this flaw in the M500, the Tech Report review has another complaint.

None of the cheap and easy to buy Samsung drives (840 and 840 Pro) claim any power loss protection, and they all fail this sort of test. They have enterprise models that might work, but those wouldn't fit within the budget parameters here. I find it hard to take those seriously when the 840 drive they share design features with are so terrible handling lifespan failures.

I'm not aware of any Sandisk models with that feature, but I haven't gone looking for them either.

The only drive really missing here that might have passed are the Seagate 600 Pro models. Those haven't been shipping long enough for me to recommend them yet.