Larger SSD = Faster SSD?

[JMFruitSalad]

Is it true that SSD performance increases with capacity? I am considering the 60gb corsair FORCE, but if there are some good reasons to go bigger then I might.. Also, why 60gb and not 64gb? lol seems strange.. anyway thanks forum

 

[Zap]

Partially true with some controllers. Basically just read the specifications for your answer.

 

[masteryoda34]

Realize that two SSD's in RAID0 will almost always be faster than a single drive of double the capacity. (Within the same product line.)

 

[Athadeus]

The Corsair Force drives are all SandForce controller based believe, which does not scale speed based on capacity. The Intel G1, G2, and the Crucial controller based drives all do, and the difference is significant (there may be others that do that I'm not familiar with). The 60GB Force has 64GB of NAND memory, which leaves ~6.8% as spare space which is typically the minimum used in SF based drives.

While RAID0 of say 2x 80GB Intel G2 vs 1x 160GB Intel G2 would definitely offer better sequential performance, real world performance (especially non-seq) would probably be worse after some use since TRIM provides a significant improvement to the G2s other garbage collection techniques.

 

[Athadeus]

Well, I looked at some more threads, and while I couldn't actually find a benchmark of the situation I described, mastery's opinion is probably more accurate based on what I saw. The G2 doesn't suffer as badly as I thought from performance degradation without TRIM as I thought, and there are various third party supplemental GC solutions.

I also didn't know that the difference between transfer rates on the 80GB and 160GB were so close. The 80GB G2 is capped at 70MB/sec write, while the 120GB and 160GB G2s have 100MB/sec write caps. 2x 120GB G2s on sale and in RAID 0 is still a really good SSD solution (perf/$) if you need that much space.

 

[taltamir]

RAID0 is not too good an idea with only 2 SSDs because of lack of TRIM.

A larger SSD with an identical controller, firmware, and NAND chip type will be faster than a smaller one.
But there are plenty of smaller SSDs that beat larger ones due to superior controller, firmware, and/or NAND chip type.

the reason to use 60 and not 64 is that 4GB is being "hidden" from the user on the 60GB model. This is because once you fill over a certain percent of the drive, you see a performance drop as well as accelerated loss of lifespan (arguable what it is exactly, around 80% to 90% full). By reserving space they prevent said drop and lifespan loss. Most drives have a certain amount of space reserved in such a manner.

 

[sticks435]

the reason to use 60 and not 64 is that 4GB is being "hidden" from the user on the 60GB model. This is because once you fill over a certain percent of the drive, you see a performance drop as well as accelerated loss of lifespan (arguable what it is exactly, around 80% to 90% full). By reserving space they prevent said drop and lifespan loss. Most drives have a certain amount of space reserved in such a manner.

isn't that only true if you not using win 7? I thought with TRIM and win 7, no space was reserved.

 

[llee]

Realize that two SSD's in RAID0 will almost always be faster than a single drive of double the capacity. (Within the same product line.)

just make sure that they are identical ssds or use the same controller chip

 

[JMFruitSalad]

the reason to use 60 and not 64 is that 4GB is being "hidden" from the user on the 60GB model. This is because once you fill over a certain percent of the drive, you see a performance drop as well as accelerated loss of lifespan (arguable what it is exactly, around 80% to 90% full). By reserving space they prevent said drop and lifespan loss. Most drives have a certain amount of space reserved in such a manner.

Interesting, I guess the same doesnt hold true for normal HDD's? I looked at the specs for various corsair force drives and they all seem to have the same write speed regardless of size.. Is SandForce still the best controller or is there something newer and better now?

 

[taltamir]

isn't that only true if you not using win 7? I thought with TRIM and win 7, no space was reserved.

1. reserved space is done in firmware and has nothing to do with the OS.
2. the way TRIM works is actually the same as making all free space equivalent to reserve space. Without TRIM things are much worse, but TRIM can't help you if you filled up the drive and there is no reserve.

The space on an SSD can be classified as one of those four:
1. Used space
2. Free space (space that at one point contained data, that has since been deleted without the knowledge of the SSD)
3. Reserve space (space that the SSD knows to be free)
4. Neverused space (space that never contained data, since last secure erase; this only exists for a very very short time, a few day to weeks after purchase)

However, SSDs cannot handle what I described to be "free space".
Without trim, the drive thinks all free space is used space.
With trim, the drive considers all free space to be reserve space.
GC scans the drive when idle for a known file system (NTFS and FAT) and identifies free space so it can be used as reserve.

So, if you have a 60GB drive (that is actually 64GB with 4GB reserved), you format it, put files on it, and it is now got 20GB of used space and 40GB of "free space". With TRIM that is considered 20GB of used space and 44GB of reserve. Without TRIM this is considered 60GB of "used space" and 4GB of reserve. A huge difference.

If you actually had 0 reserved and filling it up, TRIM will be unable to help, at all.

Interesting, I guess the same doesnt hold true for normal HDD's? I looked at the specs for various corsair force drives and they all seem to have the same write speed regardless of size.. Is SandForce still the best controller or is there something newer and better now?

Actually, a normal HDD that gets above 85% full cannot be defragged, suffers from severe fragmentation and subsequent performance loss, and suffers from performance loss due to writing at the inner rings of the drive (which are about 1/3 the speed of the outer edge of the drive). Spindle disks are written (typically) from outer edge inwards to maximize their speed.

So its a similar issue, for completely different reasons... oh and Spindle HDDs DO have reserved space... if sectors fail they are remapped to the reserves.

Amusingly, the causes for such similarities are completely unrelated and different. its just a total coincidence

 

[JMFruitSalad]

very informative post taltamir