Sandforce SSDs: Less memory chips = less performance?

[Urbi]

Hello!
I want to get my brother a Sandforce SSD for Christmas but I wonder if I should go for a 60GB or a 80GB disk performance wise: The 60GB Corsair Force F60 for example has 16 NAND-chips, the F80 has 12. The F60 shares on PCB with the F40, that only uses 8 chips while the F80 (12 chips) has the same PCB as the F120 that features 16 chips.

I could get the F80 for roughly 30% more if both where the same speed it would be good deal because 80GB is also roughly 30% more space and the F80 features 96GB - that's more spare-space than the 64GB F60 can deliver.

Corsair states that the F40 does not have lesser channels than the F60 although it has less memory chips. This might as well apply to the F80/F120.

However, Hexus has benchmarked the F80 versus a 120GB Patriot Inferno and a 120GB OCZ Vertex 2 and the F80 seems to be clearly inferior regarding the Crystal Disk Mark and AS SSD Benchmark values.

So what do you think: Is a F60 faster than a F80? For which one should I go?

 

[sub.mesa]

Normally Sandforce = 8 channels; which can be done with any multiple of 8 chips. The older Intel 160GB G1 had double the chips of 80GB but the same number of channels; later they kept the same number of chips on the 80GB (10 chips) but used double density NAND chips to achieve higher capacity, and write speeds only increased by 25%.

It could be that 12 chips has 96GiB of physical NAND; formatted to 80 or 90GB of usable capacity. This might also mean it has only 6 channels instead of 8.

Personally, i think you should wait for Sandforce SF-2000 or Intel G3, coming in just a few months.

 

[spfite]

Personally, i think you should wait for Sandforce SF-2000 or Intel G3, coming in just a few months.

Hello!
I want to get my brother a Sandforce SSD for Christmas.

That about sums it up.

 

[sub.mesa]

Still, if i were his brother, i would want to wait. I'm just offering my insights; he has to consider his options on his own.

 

[ghost6303]

However, Hexus has benchmarked the F80 versus a 120GB Patriot Inferno and a 120GB OCZ Vertex 2 and seams to be clearly inferior regarding the Crystal Disk Mark and AS SSD Benchmark values.

i dont know where you are seeing "clearly inferior" because the 120gb vertex2 was equal or better in every single benchmark over the F80 in that review. perhapse you are looking at the RAID0 pair of F80's compared to a single vertex2? of course a RAID0 of SSD's will be faster than a single drive....

but also you should notice how the numbers on most of the drives get sequentially worse from the first test to the last? take a look at this explanation, which applies to all sandforce based drives. benchmarking, especially repeated benchmarking, does not reflect real world drive usage patterns, and actually causes the drives to slow down by design and on purpose to protect themselves. it doesnt give drives time to idle and recover and run their garbage collection routines.

almost no reviews state how much time was given between benchmarks, but it can take days for a sandforce drive to recover from heavy abuse, and i bet they dont wait a full week between benchmark runs.

 

[SicKlown42012]

Most decent review sites will run Secure Erase between test, which resets the drive back to factory condition. To anyone who's buying on of these drives, I'd suggest you do that same thing if you feel the need to run multiple benchmarks in a short amount of time.

 

[Urbi]

Normally Sandforce = 8 channels; which can be done with any multiple of 8 chips. The older Intel 160GB G1 had double the chips of 80GB but the same number of channels; later they kept the same number of chips on the 80GB (10 chips) but used double density NAND chips to achieve higher capacity, and write speeds only increased by 25%.

It could be that 12 chips has 96GiB of physical NAND; formatted to 80 or 90GB of usable capacity. This might also mean it has only 6 channels instead of 8.

Personally, i think you should wait for Sandforce SF-2000 or Intel G3, coming in just a few months.

The F80 definitely has 12 chips and a total of 96GB. But your explanation does not respect Corsair's statement, that the F40 supports all channels despite having only 12 memory chips (a total of 48GB). I don't think Corsair would make false pretenses or at least none like this or would they?

i dont know where you are seeing "clearly inferior" because the 120gb vertex2 was equal or better in every single benchmark over the F80 in that review. perhapse you are looking at the RAID0 pair of F80's compared to a single vertex2? of course a RAID0 of SSD's will be faster than a single drive....

but also you should notice how the numbers on most of the drives get sequentially worse from the first test to the last? take a look at this explanation, which applies to all sandforce based drives. benchmarking, especially repeated benchmarking, does not reflect real world drive usage patterns, and actually causes the drives to slow down by design and on purpose to protect themselves. it doesnt give drives time to idle and recover and run their garbage collection routines.

almost no reviews state how much time was given between benchmarks, but it can take days for a sandforce drive to recover from heavy abuse, and i bet they dont wait a full week between benchmark runs.

I meant that the F80 was inferior. Sorry. I corrected my original post.

Still, if i were his brother, i would want to wait. I'm just offering my insights; he has to consider his options on his own.

I guess you can always wait for new technology to come out. The new Intel SSDs do not seem to be faster that SF-1200-SSDs and the SF-2000 is regarded as an Enterprise/ Industrial Chipset - the successor of the SF-1500. If it is, it may as well be rather pricy...

 

[john4200]

I'd avoid Sandforce and go with either an 80GB Intel X25-M, or a 64GB Crucial C300.

 

[Old Hippie]

I'd avoid Sandforce and go with either an 80GB Intel X25-M, or a 64GB Crucial C300.

QFT.

Many SSD novices are hung-up on the SF drives inflated "non-real world specs".

Do yourself a favor, ignore the factory published specs and look at ASSSDs final scores on the drives you are considering and go from there.

You may have to do some research for yourself but if your serious about the purchase, make the time.

 

[Megalomaniac]

QFT.
You may have to do some research for yourself but if your serious about the purchase, make the time.

This is ultimately key. Dont get swayed by the SSD tag, not all SSDs are created equal. Intel and Crucial 300 are two of the best as john4200 said.

 

[xeonMP]

QFT.

Many SSD novices are hung-up on the SF drives inflated "non-real world specs".

Do yourself a favor, ignore the factory published specs and look at ASSSDs final scores on the drives you are considering and go from there.

You may have to do some research for yourself but if your serious about the purchase, make the time.

no comments from the peanut gallery

 

[SicKlown42012]

QFT.

Many SSD novices are hung-up on the SF drives inflated "non-real world specs".

Do yourself a favor, ignore the factory published specs and look at ASSSDs final scores on the drives you are considering and go from there.

You may have to do some research for yourself but if your serious about the purchase, make the time.

AS-SSD use only non-compressable data. In real usage, just about everything can be compressed to a point. The few exceptions are mp3's, avi's, jpg's, and etc. The truth is that Sandforce drives perform somewhere between the ATTO score and AS-SSD score. I have a Vertex 2 here now, and have used an Intel 160GB in the past. At the very least, the drives are equal when actually using the computer. But there are situations where the Sandforce is noticeably faster than Intel(can't comment on Crucial). The ony way I can't recommend a Sandforce is if you do nothing but move compressed files around, or run benchmarks all day.

 

[john4200]

AS-SSD use only non-compressable data. In real usage, just about everything can be compressed to a point.

For general data, that point not-very-much, especially when the requirements are that the compression be done at 270 MB/s with 1W of power consumption.

The fact is that Sandforce cannot obtain signficant compression on most data. Most data will be nearly incompressible to the Sandforce controller.

So, while it is true that the performance you see with a Sandforce drive will be somewhere between AS-SSD and the inflated Sandforce specification, the truth is that the performance will be very close to AS-SSD, and far from the inflated specification.

And what is worse is that the performance of Sandforce drives will vary, depending on how heavily the drive has been used. The sequential write speed can drop to 80 MB/s or below after heavy use, and stay that way for days. That is less than one third of Sandforce's inflated spec. I have seen some claims that this is a feature, not a bug, but whether Sandforce implemented it deliberately, or whether they are just trying to cover up a bad, buggy implementation, the fact is that Sandforce drives are the only ones that have this problem. I do not need my SSD deciding that it is going to throttle my performance -- I am perfectly capable of deciding how much I want to write to my SSD and whether that will significantly reduce the lifetime (which is an overblown fear anyway).