Why do SandForce SSD controllers have a bad reputation?

[hojnikb]

Which reviews have you been reading?! Even anandtech alludes to the great similarities in their review and the huge slowdown with TRIM enabled very much akin to Sandforce II.

Are we even talking about the same drive buddy ?

http://www.anandtech.com/show/7803/intel-ssd-730-480gb-review

[patrickjp93]

And going back to your bashing of the 840 EVO in sustained writes, I'm afraid a huge user base has something of a very different experience. http://ssd.userbenchmark.com/Comparison

 

And the Mx100 does not fair well.

[Askew]

This is the part in question?

 

[patrickjp93]

Are we even talking about the same drive buddy ?

http://www.anandtech.com/show/7803/intel-ssd-730-480gb-review

Did you read that review? It's in exactly that one.

[hojnikb]

Did you read that review? It's in exactly that one.

Please give me ONE quote from that review, that shows similarity with sandforce. ONE

[hojnikb]

And going back to your bashing of the 840 EVO in sustained writes, I'm afraid a huge user base has something of a very different experience. http://ssd.userbenchmark.com/Comparison

 

And the Mx100 does not fair well.

Its a known fact, that EVOs write speeds drops significantly (specially on 120 and 250 versions) once you fill up turbowrite cache.

120 drops to around ~150-170 and 250gb is good for around 250MB/s

[patrickjp93]

Its a known fact, that EVOs write speeds drops significantly (specially on 120 and 250 versions) once you fill up turbowrite cache.

120 drops to around ~150-170 and 250gb is good for around 250MB/s

No one buys a 120 except to house an OS and a couple games. It's a moot point. At 500GB+ this simply isn't true and the endurance of the 840 EVO outshines the MX 100 because Marvell sucks at wear leveling.

 

You can kill the 512 GB version in less than 300TB of writes. The 500GB 840 EVO lasts 1.1 Petabytes.

[hojnikb]

No one buys a 120 except to house an OS and a couple games. It's a moot point. At 500GB+ this simply isn't true and the endurance of the 840 EVO outshines the MX 100 because Marvell sucks at wear leveling.

 

No it doesn't. And even if it sux, samsung can't make up 3x loss in endurace due to TLC.

So you're bullshiting again.

[hojnikb]

PS:

I'm still waiting on that quote

[patrickjp93]

No it doesn't. And even if it sux, samsung can't make up 3x loss in endurace due to TLC.

So you're bullshiting again.

there is no 3x loss due to TLC. That's a fabricated myth and the 840 EVO has proven so. The estimates on TLC endurance have proven totally false. The mass enterprise market is picking them up now as an alternative to higher-cost MLC drives of lesser capacities.

 

This also has partially to do with the fact Samsung provides many extra blocks on the 840 series to use if and when cells actually die.

 

P.S. I gave it.

[hojnikb]

there is no 3x loss due to TLC. That's a fabricated myth and the 840 EVO has proven so. The estimates on TLC endurance have proven totally false. The mass enterprise market is picking them up now as an alternative to the Intel 730.

 

P.S. I gave it.

 

Its not a fabricated myth, but its common knowlege for everyone, that knows a little bit about flash tech.

If you stuff more states into a single cell, its bound to have lower endurance.

 

http://www.anandtech.com/show/7173/samsung-ssd-840-evo-review-120gb-250gb-500gb-750gb-1tb-models-tested/3

~1100 cyles vs 3000+ for typical MLC drives

 

 

But that obviously doesn't mean, you're EVER gonna wear it out. Its just lower endurace, that all. Still plenty for 99% of people.

[hojnikb]

PS

Still waiting on that quote

[patrickjp93]

Its not a fabricated myth, but its common knowlege for everyone, that knows a little bit about flash tech.

If you stuff more states into a single cell, its bound to have lower endurance.

 

http://www.anandtech.com/show/7173/samsung-ssd-840-evo-review-120gb-250gb-500gb-750gb-1tb-models-tested/3

~1100 cyles vs 3000+ for typical MLC drives

 

 

But that obviously doesn't mean, you're EVER gonna wear it out. Its just lower endurace, that all. Still plenty for 99% of people.

Anandtech got their asses handed to them by independent reviewers. TLC has shown near-equal write endurance to MLC. http://www.tweaktown.com/articles/5196/samsung-840-ssd-storage-endurance-testing-tlc-to-the-end/index4.html

[hojnikb]

Anandtech got their asses handed to them by independent reviewers. TLC has shown near-equal write endurance to MLC. http://www.tweaktown.com/articles/5196/samsung-840-ssd-storage-endurance-testing-tlc-to-the-end/index4.html

It doesn't. Samsung cleverly upped the spare area to accomodate more program fails, when flash gets abused like that.

Here is a better review

http://techreport.com/review/26523/the-ssd-endurance-experiment-casualties-on-the-way-to-a-petabyte

 

As you can see, samsung started to show bad blocks way sooner than any other mlc drive. While other MLC drives died aswell, they went because set limit was reached and they failed. Not because they burned all the spare cells.

 

So generally speaking, TLC is inferior to MLC when dealing with endurance. But real life results vary (as not all flash is made equal) and this certanly doesn't affect typical users, as they will NEVER write so much data in drive's userfull lifespan.

But when dealing with write heavy enterprise applications, there is always MLC and HE-MLC and not TLC. Not that long ago, even MLC was considered inferior and only SLC was used.

[hojnikb]

PS

Still waiting on that quote

[patrickjp93]

It doesn't. Samsung cleverly upped the spare area to accomodate more program fails, when flash gets abused like that.

Here is a better review

http://techreport.com/review/26523/the-ssd-endurance-experiment-casualties-on-the-way-to-a-petabyte

 

As you can see, samsung started to show bad blocks way sooner than any other mlc drive. While other MLC drives died aswell, they went because set limit was reached and they failed. Not because they burned all the spare cells.

 

So generally speaking, TLC is inferior to MLC when dealing with endurance. But real life results vary (as not all flash is made equal) and this certanly doen't affect typical users, as they will NEVER write so much data in drive's userfull lifespan.

But when dealing with write heavy enterprise, there is alway MLC and HE-MLC and not TLC.

techreport didn't test the cells themselves. Techreport has been losing quality for years. tweaktown may write bogus articles ala wccftech every once in a while when it comes to specs or "the next big thing" but their tech reviews are spot on and detailed on a level techreport hasn't reached since 09.

[hojnikb]

techreport didn't test the cells themselves. Techreport has been losing quality for years. tweaktown may write bogus articles ala wccftech every once in a while when it comes to specs or "the next big thing" but their tech reviews are spot on and detailed on a level techreport hasn't reached since 09.

Well, thats a load of bollocks.

What were they testing then, if they wrote a shitload of data then ? How long is the sticker on the ssd gonna last ? :lol: :lol: :lol: :lol:

[patrickjp93]

Well, thats a load of bollocks.

What were they testing then, if they wrote a shitload of data then ? How long is the sticker on the ssd gonna last ? :lol: :lol: :lol: :lol:

using software to check for "bad blocks" is not accurate. Actually testing the individual chips and seeing what could be written and read after power cycles is much more trustworthy. The TLC flash in Samsung's chips holds up at about 2600 writes to MLC's average of 3000, but this may be due, in part, to Samsung's vastly superior silicon processes to the industry average, the only other high-quality producers being Intel and OCZ(when they aren't blowing up a portion of their company).

[hojnikb]

using software to check for "bad blocks" is not accurate. Actually testing the individual chips and seeing what could be written and read after power cycles is much more trustworthy.

 

Yes it. Its not like you can test any other way.

And techreport did data retention aswell

http://techreport.com/review/26058/the-ssd-endurance-experiment-data-retention-after-600tb

[hojnikb]

 

This pretty much says it all. If TLC was not inferior, than we wouldn't see so many relocated sectors

[patrickjp93]

Yes it. Its not like you can test any other way.

And techreport did data retention aswell

http://techreport.com/review/26058/the-ssd-endurance-experiment-data-retention-after-600tb

their data retention tests are outdated and tweaktown used decidedly more punishing periods of being offline. My guess is, given the wide discrepancy, techreport picked the worst performing of the 840 evo they tested, because at least tweaktown published results from all 10 drives tested, and only 1 was close to being the same as techreport's poor results.

[patrickjp93]

 

This pretty much says it all. If TLC was not inferior, than we wouldn't see so many relocated sectors

wear-leveling, duh. Having the space available gives more life to the drive, independent of the endurance of any one cell. Do you even think before posting?

[hojnikb]

their data retention tests are outdated and tweaktown used decidedly more punishing periods of being offline. My guess is, given the wide discrepancy, techreport picked the worst performing of the 840 evo they tested, because at least tweaktown published results from all 10 drives tested.

Excuses excuses. Why don't you just admit, that you're a big fat Samsung fanboy.

TLCs endurace is worse and every flash savvy person is gonna confirm that.

[hojnikb]

wear-leveling, duh. Having the space available gives more life to the drive, independent of the endurance of any one cell. Do you even think before posting?

Yeah it gives more life, but thats not my point. If TLC is equal to MLC, then why is 840 getting so many bad blocks then ? And why is samsung setting more space aside for spare area ?

Explain this shit.

[patrickjp93]

Excuses excuses. Why don't you just admit, that you're a big fat Samsung fanboy.

TLCs endurace is worse and every flash savvy person is gonna confirm that.

worse, yes, 3x worse? Oh hell no. It's a 12% difference. Just because Samsung can make it cheaply and make it perform drastically better than some nimrod speculators expected doesn't make them liars or TLC flash a bad technology. Why do you think samsung is moving to 4 bits per cell in their enterprise drives?