So it's the same underlying hardware as options from OCZ, Mushkin, Corsair, Kingston, and some others.  But because it's an Intel sticker on a SandForce drive, it costs more.  How nice.

Intel is promising better reliability.  Apparently they've tweaked the firmware a bit, supposedly for bug fixes.  Performance usually isn't changed as compared to other SandForce SSDs.

Admittedly, the second generation SandForce controller has had some reliability problems.  But even if you believe that the Intel version of SandForce has fixed those, and other versions of the same hardware haven't, why an Intel SSD 520 over a Samsung 830 or a Crucial M4?

Much faster write speeds on the smaller drives is one reason. Not something i'd spend the premium for though as there is almost nothing i do on my PC that would require the faster write speeds. The larger drives are a bit expensive but for some reason they priced the 60gb at £2 per gigabyte when other brands are starting to head towards £1/gb these days.

Intel do offer 5 year warranty opposed to 3 years which is nice i suppose.

I should probably back up and say, if you're an enterprise user and a drive failure will cost you tens of thousands of dollars, then I can understand why you'd go with Intel.  The price premium just isn't that important for that sort of user.  But for consumer use?

Its good to see improvements made to the sandforce controller.  But I'd still worry about its longevity especially coupled with 25nm nand.

Samsung or Crucial for sure. 

Originally posted by Quizzical

I should probably back up and say, if you're an enterprise user and a drive failure will cost you tens of thousands of dollars, then I can understand why you'd go with Intel.  The price premium just isn't that important for that sort of user.  But for consumer use?

Well, I do have 4 now old Intel X-25 but at the time they performed best in the benchtests I checked and I had at the time a lot of money so why not?

When Intel performs about the same as the rest I rather get a Samsung or Corsair instead but I think most the customers for this drive will be companies anyways. I guess they might lower the price if it sells badly.

Certainly, I could understand getting an Intel X25-M two or three years ago.  Three years ago, there wasn't any viable alternative to Intel SSDs.  Two years ago, the only serious alternative was Indilinx, and Intel was still better.  I bought Indilinx because it was cheaper and, in my view, good enough.  But I could understand buying Intel then.

But I'm talking about today, not two years ago.  Just look at the prices:

http://www.newegg.com/Product/Product.aspx?Item=N82E16820148525

128 GB Crucial M4 for $180

http://www.newegg.com/Product/Product.aspx?Item=N82E16820147134

128 GB Samsung 830 for $220.  And that's actually up, as it was more like $200 recently.

http://www.newegg.com/Product/Product.aspx?Item=N82E16820167095

120 GB Intel SSD 520 for $243.  And that's probably not coming down.  If you go with Intel, then you pay more than 1/3 more than Crucial for less capacity, comparable performance, and not necessarily better reliability, even.  The Crucial M4 has been out for about a year, and that means ample time to find and fix bugs as they show up in the real world.  Crucial found the 5000+ hour blue screen bug because a fair number of their drives have already been in use for over 5000 hours.  That's not the case for the Intel SSD 520.

so why does nobody mention the OCZ Vertex 3, or the Kingston HyperX?

Today is a weird day. I just read Cherryville in the thread title as Chernobyl.

Originally posted by eyelolled

so why does nobody mention the OCZ Vertex 3, or the Kingston HyperX?

The problem with SandForce SSDs is that they're less reliable than some of the competition.  One can argue as to whether or not a few percent difference in the chance of something going wrong is all that important--and "something going wrong" doesn't necessarily mean data loss.

My point was, even if you put a heavy weight on reliability, you can get reliability for much cheaper than what Intel charges, at least by the standards of reliability likely to be of interest to consumers.  Maybe not for some enterprise uses where a 99% chance of never having performance or reliability problems isn't good enough, and you need redundant everything.

Originally posted by Quizzical

Originally posted by eyelolled

so why does nobody mention the OCZ Vertex 3, or the Kingston HyperX?

The problem with SandForce SSDs is that they're less reliable than some of the competition.  One can argue as to whether or not a few percent difference in the chance of something going wrong is all that important--and "something going wrong" doesn't necessarily mean data loss.

My point was, even if you put a heavy weight on reliability, you can get reliability for much cheaper than what Intel charges, at least by the standards of reliability likely to be of interest to consumers.  Maybe not for some enterprise uses where a 99% chance of never having performance or reliability problems isn't good enough, and you need redundant everything.

I see. So the Sandforce controller is less reliable then the marvell controller?

Originally posted by eyelolled

I see. So the Sandforce controller is less reliable then the marvell controller?

Firmware, not hardware.  I think part of the issue is varying NAND flash.  Crucial wrote their own firmware for the Crucial M4.  They make their own NAND flash, so they knew exactly what flash memory they were going to use.  So it's exactly the same bill of materials for every single SSD they make, except for the varying capacity.  That makes validation (testing) a lot easier.  It also helps that Crucial is a big company with a lot of experience with validation, which makes them more likely to catch and fix bugs.  The same is true of Samsung and their SSD 830.  Intel uses a few different controllers, but also puts huge amounts of work into validation.

One problem with SandForce is that they wrote firmware without knowing what NAND flash their partners would use.  So a lot of different NAND flash gets used, and if some piece of firmware code works perfectly with all of the ones that SandForce tested, some partner might pair it with something that SandForce didn't test as much, and then an end user pairs it with a particular combination of a motherboard, processor, and memory that SandForce hadn't tested at all, and then maybe you get problems.  SandForce also doesn't have experience with validation (as it's a startup company that makes SSD controllers and nothing else), so they have to rely on their partners to catch a lot of things.  That might change now that LSI has bought SandForce, but it will take time.

One advantage to being able to use NAND flash from a lot of different sources is that you can buy whatever is the cheapest and meets your claimed specs.  If Micron is charging more than Hynix one week, and then prices are reversed the next, then you can buy whichever is cheapest each week.  That saves money, and lets you pass on the savings to customers.  Aggressively doing this is one reason that OCZ has been able to push SSD prices down.  It's also one reason why OCZ's SSDs are less reliable than some competitors.

It's also a reason why OCZ has a zillion different models.  For example, compare OCZ's lineup:

http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100008120+50001550&IsNodeId=1&name=OCZ+Technology&Page=1

to Samsung's:

http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100008120%2050001077&IsNodeId=1&name=SAMSUNG

OCZ has 61 models in stock.  Samsung has 7.  And Samsung's 7 is really only 4 different physical SSDs, with different packaging options for three of them.  And even those four are really only four different capacities for the same controller.

Meanwhile, OCZ has Vertex 2, Agility 2, Agility 3, Solid 3, Vertex 3, Nocti, Vertex 3 MAX IOPS, RevoDrive 3, RevoDrive 3 MAX IOPS, Colossus Plus, Vertex Plus, Vertex, Colossus LT, RevoDrive, Octane, Petrol, and Synapse.  And that's only counting 120-128 GB, so it's not even different capacities of the same thing.  They have at least five different SSD controllers, several different interfaces, and who knows how many different bins and brands of NAND flash.

If you want to buy one particular SSD model, guess who did more validation work to ensure that that one particular SSD model works.  Samsung only has one model to validate, so all of their testing goes into making sure that one model works.

Now, that's not to say that if you get an SSD from OCZ, it's going to fail.  It might be the difference between a 95% chance of working right and a 98% chance of working right.  The SSD that I personally use in my desktop is from OCZ, and it works.  But if I had a choice between buying an SSD from Crucial and buying one from OCZ for the same price, I'd go with Crucial.  If OCZ is $50 cheaper for the same capacity, then OCZ would be awfully tempting.  And indeed, that's basicaly why I bought from OCZ rather than Intel in the first place.  (At the time, the only good controllers on the market were Intel's and the the Indilinx Barefoot.)

I had heard that due to TRIM procedures, that the Samsung 830 can slow right down to less then standard HDD speeds. From my understanding that the Samsung controller leaves it's garbage collection for periods of low or no activity and that when the OS determines that an area needs to be overwritten it then forces to SSD to clear the contents and then write to it which slows things down considerably. I had also thought that the Crucial suffered from the same sort of design, while the Sandforce did not.

(sorry for deviating away from the original point of the thread, but I was thinking about starting a thread about preffered SSD's and this one was very close in what I wanted to discuss)

Some SSDs are more aggressive than others in their garbage cleanup.  There are trade-offs to be made, and there isn't a clear "right" solution.

If an SSD is too aggressive in its garbage collection, then when you want to access data on it, it might well be busy with garbage collection and make you wait.  These wait times usually won't be very long, and might not even be noticeable in the real world, but they're clearly visible in synthetic benchmarks.

On the other hand, if an SSD isn't aggressive enough in garbage collection, then things can get really jumbled up, and the SSD can get really slow as you continue to write, at least until you let the SSD idle for a while to recover.  For a while, OCZ recommended that if you want to manually trim your SSD, you go to the Windows login screen with no one logged in, and don't log in but just leave it there for a while.  That would make the SSD sufficiently idle to do all of the garbage collection that it could plausibly want.

Crucial philosophically leans toward less aggressive garbage collection than some competitors use.  That's not intrinsically good or bad, but it does mean that Crucial's SSDs are pretty much immune to the former problem, at the expense of being more prone to the latter.

You should realize that enough writes to cause problems is really a huge number of write operations.  If you do tens of thousands of random write operations per second for hours on end, it's a lot easier to get a Crucial M4 into trouble than an Intel SSD 510 (which uses the same Marvell controller).  But to do that, you're doing tens of millions if not hundreds of millions of random write operations.  You're taking the amount of stress on an SSD that even a heavy duty consumer might take months to do, and compressing it onto a couple of hours or so.  That might be a concern if you sit and run synthetic SSD benchmarks all day, but not for real world use.

Not all SSD use is the same for causing trouble like this, either.  Reads aren't a problem at all.  Sequential writes really aren't a problem, either.  (Quick and dirty method of trimming an SSD:  write a single enormous file to fill up all the free space, and then delete that file.)  Large files aren't a problem.  It's small file random writes that can cause trouble.  If you're running a heavy-duty database on your SSD, then maybe you should worry about this.  But for general consumer use, it's completely irrelevant.  And furthermore, even if an SSD does get into trouble this way, all you really have to do is to let it idle for a while and it will correct itself.

Again, there are trade-offs, and one trade-off to consider is spare area.  It's easier to do garbage collection well if you have more spare area.  The problem is that more spare area means less usable capacity.  There are a lot of SSDs on the market with 128 GB of physical NAND flash, but they range from about 107-119 GB of usable capacity.  Reducing the usable capacity by expanding the spare area doesn't affect the cost of making the SSD at all, but it does reduce the amount of capacity that an end user can take advantage of.

Intel and Crucial make SSDs based on the Marvell controller with 128 GB of physical NAND flash, but Crucial's leaves more capacity available to the user (119 GB versus 112 GB).  For most consumers, more usable capacity is nice, but the added spare area does let Intel do better garbage collection.  In the extreme case, an SSD designed for cache (which means enormous amounts of small random writes) may have half of the physical NAND flash set aside as spare area.  That allows you to make garbage collection that is nearly bulletproof.

well that is a prime example of how more information just leads to more questions. I guess the one thing I can  take away is that if there is no truly right answer, then either choice is not exactly wrong either.

I will probably continue to use Sandforce SSD's as OS drives as I think a little more up front garbage removal is better, while I will consider using the samsung 830 for secondary/gaming drives as there really shouldn't ever be much worry about Trim.

thanks for the info

---

Originally posted by drbaltazar
Performance?the old hhd will come back very soon with a lazer.for terabyte per second speed.

---

Pretty sure they already did that, and it was called the CD, then the DVD, and now the Blu-Ray.

Intel has a strong brand of reliability, which they in the SSD 520 Series back up with 5 years warranty.

It's not aimed for the consumer mass market, but instead to companies and prosumers who need to keep their data safe and running at all times.

Intel is making a metric crap ton of money on selling the NAND chips to other manufacturers. Their own SSD lineup is just an icing to the cake.

Originally posted by tom_gore

It's not aimed for the consumer mass market, but instead to companies and prosumers who need to keep their data safe and running at all times.

That's why you make backups of your backups and have redundant everything.

Do you really trust an Intel SSD that launched just yesterday to be more reliable than a Crucial M4 that has been out in the hands of consumers for a year?  The latter has had hundreds of thousands if not millions of people searching for problems (i.e., if something goes wrong, contact tech support to complain).  An SSD that launched just yesterday doesn't have that.

And warranty length isn't necessarily indicative of reliability.  Many products give you the option to pay extra for a longer warranty, even though it's exactly the same underlying product as before.  For a while, Patriot was offering a 10 year warranty on their Torqx SSDs.  Are those really the most reliable SSDs ever?

Honestly, as someone who purchases a lot of business hardware:

For critical machines, you just go with RAID 5 and a serious and rigorous backup plan with off-site storage. If anything happens to a drive, RAID saves you, and if anything happens to the computer as a whole (say, fire), then your offsite backups have your data protected. It takes a lot of work to set these up and get them going smoothly - the price of a single hard drive is inconsequential, and the reliability benefit of a RAID 5 array mitigates any small differences in drive quality.

That being said, for less critical applications, we just rely on a routine backup plan: usually an external USB drive the employee can carry around with them and use at their leisure, combined with some network storage in the office.

As long as the drives are comparable in quality/warranty (i.e. no DeskStar), then it really doesn't matter, even for the critical machines. We haven't been using SSDs in our RAID arrays (yet) because we need a lot of storage, but we have been moving all of our laptops over to SSDs - it used to be Intel, but now we are using Crucials. Part of it has to do with marketing as well: if we can tell our clients we are using XXX brand, and they recognize it positively, that imparts a message that our company cares about quality - even if it is essentially the same drive under the hood, the brand recognition (and even the least savvy of our customers usually recognizes Intel) goes a long way. Intel hasn't really had a "great" product since the X25's, so we started migrating to Crucial. I don't know that we'll go back, as we've had pretty good results with the C300's and M4's.

The "reliability" benefit of these Intel drives is nearly negligible at this point, as the 520 hasn't been around long enough to prove any reliability numbers: no one is seriously buying these in the corporate world with any notion of reliability. Anyone who is really needing that critical of drive application (say, stand-alone server in a remote location, we have a few of those) - you need proven reliability, not something just off the line with no history. Now, Sandforce does have some history, but not with an Intel sticker and "tweaked" firmware.

Now we usually don't go out of our way to advertise what brand hard drives we use, but occasionally a client will ask what hardware we select, and we give them the full run-down. The more parts they can positively identify, the better message that sends.

That, and in corporate setting, we can write off the cost of hardware on taxes as part of capital expenses.

But that's corporate. For home use, it's a good deal of money to pay for little benefit. The only thing going for it is brand recognition/loyalty.