SF-2281 on SATA II - reliable?

[JonTom]

I'm getting ready to delve into the waters of SSDs. My PC (not too likely to upgrade anytime soon) has SATA II, not III. I understand that the problem with the latest SF controller'd SSDs is rooted in them running too fast. Am I likely to experience reliability issues/BSODs if I use one in my system (I understand that not everyone is having this problem...)?

TIA

 

[Emulex]

steer clear of sandforce then. Why not a good intel x25-m or 510 ? solid and cheap. or heck $1/gb all day long you can score kingston drives. I just picked up a pair of new 128gb SVNP325-2 for $180 (raid-0 with sata-2 will haul balls). I paid $160 for an hp pull (3hrs) x25-m G2 160gb. Kingston (all toshiba) 96gb drives are $96 if you search around.

fact is the latency of near zero will give you a boner and even an X25-V is so much faster than hard drive - don't have to spend big bucks on super-e-peen sata 3 ssd when you will always have a home for an ssd (resale values are extremely high). or throw it in that ancient i7 laptop you got with a 5400rpm drive.

I'd say no problems using SATA-3 drive (other than buggy sandforce drives). Intel has the lowest return rate across the industry - that says something.

 

[sub.mesa]

No reason to go Sandforce indeed. In my opinion, a Crucial M4 is a much better choice at the moment. Samsung 870 or Intel 320 might be other candidates, depending on your needs.

 

[Gigantopithecus]

Going to an SSD from an HDD is the major improvement. Going from any current (fast) SSD to another (even faster) SSD isn't worth risking the reliability issues of the current crop of Sandforce drives. I went from an 80GB X-25M to a 120GB SF-2xxx SSD and it always BSOD'd on wake from sleep. Sure it booted up 3-4 seconds faster and shut down a second faster, but I couldn't put the laptop into sleep mode. The X-25M is back in the laptop.

 

[sub.mesa]

Indeed, besides; a 60GB Agility 3 or Vertex 3 is not really faster than Crucial M4 64GB. I'd say most modern SSDs perform about the same. The differences are measurable but may not be noticeable.

 

[blackened23]

Indeed, besides; a 60GB Agility 3 or Vertex 3 is not really faster than Crucial M4 64GB. I'd say most modern SSDs perform about the same. The differences are measurable but may not be noticeable.

They're undoubtably faster than the m4 even with the new firmware in sequential writes and compressible data read/writes. I just would stay away from OCZ and get a force GT or kingston hyperx instead, if choosing a SF22xx drive.

 

[sub.mesa]

Faster in which regard? Since they are only faster when writing heavily compressible data, such as zeros which benchmarks like ATTO do. When using incompressible data, the Crucial M4 is about twice as fast as an Agility 3 in simple synthetic benchmarks. At least when comparing the 64GB versions.

 

[blackened23]

crucial firmware 0009. Don't be a fanboy - the new firmware is a 20% boost but not faster than SS2281 with sync NAND. The agility 3 is their cheap trash model with async NAND.

That said, the crucial m4 is a better choice than any ocz drive whether the OCZ is faster or not. OCZ has some major reliability issues to work through, i've not heard of many problems with other SF2281 drives.

Still, don't get me wrong - the m4 is a great buy, performs great and is reliable....its a great buy for anyone leery of the sandforce horror stories.

 

[sub.mesa]

As is usually the case, reviewers often test the large capacity versions of major SSD brands, but not the much more common 64/128GB sized variants.

When looking at the 64GB Crucial M4, you can see:
Sequential read: 530MB/s
Sequential write: 110MB/s
4K read QD32: ~200MB/s
4K write QD32: ~100MB/s

That is faster than the Sandforce variants. The Crucial M4 doesn't scale as much as Sandforce or Intel controllers, which is good if you're into 64GB or 128GB sized SSDs. For 256GB-class SSDs the Crucial M4 is not that attractive anymore. But I would argue that any such SSD basically wastes alot of performance. Imagine what a RAID0 of four 64GB Crucial M4 would do with perfect scaling:

Sequential read: 530MB/s -> 2120MB/s (needs additional read-ahead)
Sequential write: 110MB/s -> 440MB/s
4K read QD32: ~200MB/s -> ~800MB/s (needs a good RAID0 driver and plenty of queue depth)
4K write QD32: ~100MB/s -> ~400MB/s

Of course, casual Windows RAIDs won't see such performance scaling, but it's obvious the 256GB class SSDs are severely performance capped. That's also the reason I opted for multiple smaller SSDs, especially since I'm on UNIX sporting advanced RAID functionality and advanced threaded I/O backends which allows for excellent RAID0 performance scaling.

The reliability issues are not inherent to OCZ, they are inherent to the Sandforce controller. Unlike other controllers, Sandforce does not actually store your data but rather references to datablocks. This technique is called de-duplication and it adds a great amount of complexity not found in other controllers. It is quite likely at least a decent amount of bugs are inherent to this feature. Thus, other Sandforce drives should have similar failure rates; OCZ just happens to sell much more and thus gets mentioned alot more frequent than other brands.

 

[blackened23]

You're still comparing your m4 with the 140$ (for 120gb) agility 3 which uses async NAND.

m4 0009 firmware

kingston hyperx 120gb sync nand

 

[sub.mesa]

Blackened23, you're comparing a 3Gbps controller versus 6Gbps for the Kingston.

Also, since you've not commented on any of my arguments and instead revert to childish behavior by calling someone you don't agree with a fanboy -- twice -- I'll leave this discussion now and hope people can make their own determination.

 

[Emulex]

so basically any will do for a good best experience. skip sandforce due to open-issues that may or may not hit you. $1/gb is a good point of reference for a first experience. you can score 96gb kingston drives all day lone for $96 - that is a great starting point - not the fastest, safe (osx/windows) - built by toshiba using toshiba nand.

or intel or crucial who builds their own controllers (320/x25-m) and their own NAND flash.

Avoid: People slapping together other peoples controllers and other people's nand.

There is not enough NAND to go around. If they don't make their own controller - you have issues like sandforce where the company has moved on from the SF-1281 to better products and ditched millions of users with that controller with open issues. Their solution: buy our newer product. SF-22xx - lol. no thanks. burn me once.


Numbers are great - but reliability is greater. Intel 320 with 1 extra NAND chip to sustain a single chip failure = win. Intel 320 with parallel capacitors rather than 1 big one so give you battery backup in event of failure of the capacitor = win. X25-M not using write cache at all = win by simplicity.

So what if your first experience is 7 seconds to boot versus 4? you will find an outlet for that device (HTPC,neighbor,laptop) when you decide to spend more to go faster - but pick your first experience on a SOLID product that is affordable. Otherwise you will end up one of those bitter folk that think ssd is evil because their first super-e-peen product let them down when they were flashing it around

 

[nanaki333]

the new SF drives are only faster in benchmarks. they're all the same as far as an end user is concerned.

 

[groberts101]

the new SF drives are only faster in benchmarks. they're all the same as far as an end user is concerned.

that would probably be true for simple Windows GUI navigation or small tasks.. but you've obviously never tested them during heavy multitasking. Don't get it twisted and think that Sandforce 6G drives using toggle-nand(from any mfgr) aren't perceivably faster because they certainly are when you push them(especially with incompressible data like vid, pics, and music.

Just copying/writing a large 10GB file and going about your business will easily bog some controllers down(especially smaller capacity points). Faster drives with quicker writes(regardless of the data types), along with slightly lower latency can make all the difference in the world for more demanding users like myself.

Hell.. I've tested the P' outta 5 different controllers and there surely is a difference in performance when you really push them hard with simultanious tasks. Anand's heavy-test will weed out the true winner's almost every time.

Course.. as mentioned, the differences in the top-contenders is so small it's not worth splitting hairs unless you running larger/wider raided SSD arrays and need that last little bit x (the number of drives used). Or you have fast enough storage to see it in transfers between volumes.

Also consider that the biggest bottleneck on any system using a fast 6G SSD is going to be the storage volumes max speed. Can't really make use of these 6G speeds too much unless you're willing to wait for the slow-ass transfers from the HDD to the faster SSD to get that data natively stored on the SSD for best results. This is also why there aren't huge percievable gains when moving from a fast 3G SSD to a much faster 6G SSD.

IOW,.. having a fast raided SSD based OS will not show you the full potential of a balanced system until you match it up to a raided HDD volume of 4-8 drives in R0. This is why many don't think there is any gain to be had with raided SSD besides benchmarks. Most just don't have fast enough storage to really see any time savings in R/W between the volumes, is all. Same thing could be said about the perceptible gains between single 3G and 6G SSD's as well. Using a ramdisk as a temporary volume to copy/paste from/to the SSD will give you the real story.

 

[nanaki333]

ok.. for your TYPICAL end-user, there's no difference as far as they're concerned. most people are browsing the web, watching movies, or playing games. they're not doing straining tests or using their SSDs as storage to copy large files to

 

[Compddd]

Didn't Anand write that Samsung is just as reliable as Intel in the SSD sector?

I'd go with a Samsung or Intel drive.