G Skill ECO & SSDs -- Any thoughts/impressions?

[andymcca]

I don't see G Skill SSDs recommended often here, despite having read wonderful things about them elsewhere. Is there a reason for this? I'm not too familiar with their customer service, QC, QA, etc. Any anecdotal information is welcome!
And are they available in European markets?

And while we are talking about G Skill, how do people feel about their 1.35V ECO series DDR3 modules? I've been keeping an eye on the DDR3 1600 CAS7, but have been a little annoyed that the 4x2GB is consistently more expensive than two 2x2GB sets. Not that I believe it makes a huge difference, but ..!

[HFat]

They don't have their own controllers and they're probably not the best at supporting them. I would rather go with OCZ which has made some efforts to get good firmwares to its customers in the past according to Anandtech and I would much rather go with a manufacturer who's developing their own controllers (so far as I know: Intel, Samsung and Marvell/Crucial).
Current G.Skill SSDs also seem to be using Sandforce controllers which I would rather avoid anyway (but maybe others would find them useful).

[andymcca]

They don't have their own controllers and they're probably not the best at supporting them. I would rather go with OCZ which has made some efforts to get good firmwares to its customers in the past according to Anandtech and I would much rather go with a manufacturer who's developing their own controllers (so far as I know: Intel, Samsung and Marvell/Crucial).
Current G.Skill SSDs also seem to be using Sandforce controllers which I would rather avoid anyway (but maybe others would find them useful).

I thought the top OCZ SSDs were also using sandforce. Is this not the case? The "vertex" series, or whatever, is what I am thinking of.

Edit: Also, what is so scary about sandforce? People seem very anti-sandforce, but from what I can tell it performs admirably. Is there something about it that I am not understanding?

Edit 2: Interesting:
http://www.guru3d.com/news/sandforce-sf ... -firmware/
This tidbit suggests that OCZ has access to firmware updates for sandforce chips before other companies? Not sure how true this is...

[quest_for_silence]

Is there something about it that I am not understanding?

You may give a worthy look at the "SSD Reliability" thread in the Silent Storage section: they seem suffering of a failure rate up to 26% (Corsair: or ~15% as a mean average, while Intel goes for 2% and Crucial around 4%): so, yes, they perform very well, but maybe performance isn't all about an SSD, I guess.

[andymcca]

You may give a worthy look at the "SSD Reliability" thread in the Silent Storage section: they seem suffering of a failure rate up to 26% (Corsair: or ~15% as a mean average, while Intel goes for 2% and Crucial around 4%): so, yes, they perform very well, but maybe performance isn't all about an SSD, I guess.

Good point! And a glance at the Newegg G Skill 60GB model has it in the same 10%-20% "failure rate" (in quotes because I think it is quite clear that user reviews will skew high), similar those discussed in the SSD Reliability thread.

However, I don't think this is actually out of line with the apparent failure rates of normal HDDs... I glanced around at a couple of different consumer-grade HDDs and the Newegg reviews seemed similar (including 5400rpm 1-platter drives, which I would have thought would have better dependability). Obviously this does not explain cross-controller differences, and may even mean that Intel SSDs are more reliable than magnetic HDDs... Then again, looking at Intel SSD prices, I wonder if price is not a significant confounding variable (in the nature of reviews, and who reviews).

Anyway, per all storage devices, I would not depend on it for data security, and would really just enjoy the spiffy startup times.

Definitely something to keep in mind, I guess.

[quest_for_silence]

This tidbit suggests that OCZ has access to firmware updates for sandforce chips before other companies? Not sure how true this is...

Enough true...

[HFat]

I don't trust (nor do I distrust) these reliability statistics. I guess they're better than nothing but the data stinks.
No SSD is better than all other across the board. But some SSD have serious issues or are plain mediocre.
My main problem with SandForce controllers is that they rely on compression. I don't know if they misrepresent the amount of flash on their drives or if they merely use compression to increase performance and to reduce wear but the bottom line is that the drives don't perform as well if you use encryption or if you have data that's already compressed. Plus it introduces an unecessary complication which might make the drives less reliable.
Generally, SandForce controllers might be better suited for servers than consumer use. They're good at randomly writing non-compressed, non-encrypted data.

OCZ uses different controllers for different drives (even when they have the same name). It's confusing but it looks like they're at least trying to deliver and support good products.

[andymcca]

My main problem with SandForce controllers is that they rely on compression.

Interesting! At the risk of making it very clear I've not done my homework: I was unaware of this. I assume this inflates performance figures on information-sparse data by lessening the impact of a flash i/o bottleneck?

Might be great for the ridiculous ASCII I/O software we use at work Nothing like writing 5GB of ASCII to the disk when it compresses to less than 10kB... Oh, and then read it 10-20 times in order to modify a few lines (but definitely read it once per modification).

So I guess the hardcore gamers avoid sandforce drives, then, since their textures are likely all very information-dense storage types? Are there figures floating around of bit-random performance? (/me puts on his google hat)

[HFat]

Performance is still decent with non-compressible data. Anandtech has some numbers.

[dhanson865]

From http://images.anandtech.com/graphs/graph4031/34317.png and http://images.anandtech.com/graphs/graph4031/34315.png you get these numbers

Writes
Random
Sandforce 120GB compressible data 164 MB/s
Sandforce 120GB random data 120 MB/s
Crucial C300 128GB any data 110 MB/s

Sequential
Sandforce 120GB compressible data 214 MB/s
Sandforce 120GB random data 136 MB/s
Crucial C300 128GB any data 131 MB/s

and from http://images.anandtech.com/graphs/graph4031/34318.png and http://images.anandtech.com/graphs/graph4031/34316.png

Reads
Random
Crucial C300 128GB any data 78 MB/s
Sandforce 120GB compressible data 58 MB/s
Sandforce 120GB random data 50 MB/s

Sequential
Crucial C300 128GB any data 226 MB/s
Sandforce 120GB compressible data 212 MB/s
Sandforce 120GB random data 170 MB/s

Why buy the drive with 15%+ failure rates when the drive with 4% failure rates is basically just as fast in most tests and faster in others? It's a wash on writes with random data and the C300 wins outright on sequential data (significantly more so when you use random data).

Oh and what is the deal with random data you might ask? Random data prevents compression and simulates how the sandforce controller will handle files that are encrypted or already compressed including MP3, video files, etcetera...

[frostedflakes]

You may give a worthy look at the "SSD Reliability" thread in the Silent Storage section: they seem suffering of a failure rate up to 26% (Corsair: or ~15% as a mean average, while Intel goes for 2% and Crucial around 4%): so, yes, they perform very well, but maybe performance isn't all about an SSD, I guess.

Good point! And a glance at the Newegg G Skill 60GB model has it in the same 10%-20% "failure rate" (in quotes because I think it is quite clear that user reviews will skew high), similar those discussed in the SSD Reliability thread.

However, I don't think this is actually out of line with the apparent failure rates of normal HDDs... I glanced around at a couple of different consumer-grade HDDs and the Newegg reviews seemed similar (including 5400rpm 1-platter drives, which I would have thought would have better dependability). Obviously this does not explain cross-controller differences, and may even mean that Intel SSDs are more reliable than magnetic HDDs... Then again, looking at Intel SSD prices, I wonder if price is not a significant confounding variable (in the nature of reviews, and who reviews).

Anyway, per all storage devices, I would not depend on it for data security, and would really just enjoy the spiffy startup times.

Definitely something to keep in mind, I guess.

It makes no sense because NewEgg reviews pretty poor indicator of reliability IMO. I'd take any trends gleaned from them with a grain of salt.

[quest_for_silence]

NewEgg reviews pretty poor indicator of reliability IMO.

So which could be, with a grain of salt (and as far as you know), any meaningful yet practical indicator for SSD dependability?

[andymcca]

NewEgg reviews pretty poor indicator of reliability IMO.

So which could be, with a grain of salt (and as far as you know), any meaningful yet practical indicator for SSD dependability?

I agree. The most straightforward conclusion is that there is some real difference in rate of failure.

I suppose there is a good argument for getting a C300 over g skill et al. Plus the read speeds supposedly improve with an upgrade to 6Gbps SATA. (according to newegg specs)
http://www.newegg.com/Product/Product.a ... 6820148357

Edit: Of course, if for some reason I cared about spongy data write speeds there would be some argument for the sf-1200... But as a boot drive I don't think I'll be doing too much writing. Still debating where to keep /var/log. On my old pen-drive installation I kept them in tmpfs, but this does not seem great for a "real" system.

[frostedflakes]

NewEgg reviews pretty poor indicator of reliability IMO.

So which could be, with a grain of salt (and as far as you know), any meaningful yet practical indicator for SSD dependability?

NewEgg and manufacturer RMA rates. Of course since we don't have access to this information, NewEgg reviews are probably the next best thing. Just saying that it's not really an ideal indicator of failure rate.