Solution
acer0169
Distinguished
- Apr 9, 2009
- 1,039
- 0
- 19,460
If you plug a SATA2 hard drive into a SATA3 port on your motherboard you wont see a performance gain. If your old motherboard was garbage and the new board is a good-spec one; you might get a little faster speeds but nothing noticable. It's a good move to buy SATA3 motherboards though - keeps your computer future proof!
sub mesa
Distinguished
- Apr 17, 2009
- 2,914
- 0
- 21,160
You would need an as-of-yet non-existent SATA6Gbps SSD that can actually use the bandwidth (unlike Micron C300).
HDDs for the next 8 years or so will be limited by SATA/300 so any SATA/600 HDD is mostly marketing, though some minor advantages in performance (propagation delay), power consumption (SATA link management) and other issues (expanders not using 6Gbps bandwidth when using non-6Gbps HDDs).
SATA harddrives do 150MB/s at most, which is about the limit of SATA/150 or the first SATA generation. If your board has 6Gbps ports, they may in fact be slower than your 3Gbps ports, especially in RAID. This occurs frequently for ICH10R 3Gbps SATA RAID and a cheap marvell SATA 6Gbps controller running on single lane PCIe 2.0 x1.
So the conclusion is that these kind of stuff is irrelevant for most consumers, as it doesn't really change anything about the speed; SATA 1, 2 or 3; your HDD is still the biggest bottleneck, and only the newest HDDs can break the SATA1 or SATA/150 barrier with sustained I/O.
HDDs for the next 8 years or so will be limited by SATA/300 so any SATA/600 HDD is mostly marketing, though some minor advantages in performance (propagation delay), power consumption (SATA link management) and other issues (expanders not using 6Gbps bandwidth when using non-6Gbps HDDs).
SATA harddrives do 150MB/s at most, which is about the limit of SATA/150 or the first SATA generation. If your board has 6Gbps ports, they may in fact be slower than your 3Gbps ports, especially in RAID. This occurs frequently for ICH10R 3Gbps SATA RAID and a cheap marvell SATA 6Gbps controller running on single lane PCIe 2.0 x1.
So the conclusion is that these kind of stuff is irrelevant for most consumers, as it doesn't really change anything about the speed; SATA 1, 2 or 3; your HDD is still the biggest bottleneck, and only the newest HDDs can break the SATA1 or SATA/150 barrier with sustained I/O.
SATA/6G hard drives make sense if they have a large cache e.g. 64MB
and they are configured in a RAID 0 for speed: under these conditions,
the caches are additive and buffer-to-host transmissions will be fast.
One short-lived secret is the sheer performance of the
Western Digital WD5003ABYX:
http://www.newegg.com/Product/Product.aspx?Item=N82E16822136697&Tpk=N82E16822136697
... ~150 MB/second with a single drive (see ATTO software).
That's probably why Newegg can't keep them in stock:
these 7200 rpm drives out-perform WD's famed VelociRaptors --
especially with a short-stroked partition --
probably because the former use PMR.
p.s. Putting 2 or 4 HDDs in a RAID 0 array does not increase
the probability that any one of those HDDs will fail.
Cf. conditional probability (a branch of Statistics).
Actually, there is little practical difference: if one member HDD fails,
you replace it, just as you would if it were a JBOD HDD (non-RAID).
MRFS
and they are configured in a RAID 0 for speed: under these conditions,
the caches are additive and buffer-to-host transmissions will be fast.
One short-lived secret is the sheer performance of the
Western Digital WD5003ABYX:
http://www.newegg.com/Product/Product.aspx?Item=N82E16822136697&Tpk=N82E16822136697
... ~150 MB/second with a single drive (see ATTO software).
That's probably why Newegg can't keep them in stock:
these 7200 rpm drives out-perform WD's famed VelociRaptors --
especially with a short-stroked partition --
probably because the former use PMR.
p.s. Putting 2 or 4 HDDs in a RAID 0 array does not increase
the probability that any one of those HDDs will fail.
Cf. conditional probability (a branch of Statistics).
Actually, there is little practical difference: if one member HDD fails,
you replace it, just as you would if it were a JBOD HDD (non-RAID).
MRFS
sub mesa
Distinguished
- Apr 17, 2009
- 2,914
- 0
- 21,160
My view on the subject..
The Samsung F4 2TB 5400rpm and WD EARS 3-platter ~5400rpm would also be able to beat the older velociraptor in sequential I/O; the F4 is capable of up to 140MB/s sequential reads. Being 5400rpm disks they approach the sequential performance of the newest 7200rpm and even 10.000rpm disks, because they have the highest data density of all. 7200rpm disks are still limited to 500GB platters while the F4 and newer EARS would use 666GB platters, leading to higher sequential throughput.
The higher rpm disks would still win from any 5400rpm disk regarding random IOps; though for that task any HDD sucks and an SSD would be the most logical choice; for example as system disk.
While i agree with MRFS, that RAID0 doesn't increase the chance of disk failure, is does increase the chance of failure of your storage, due to problems with the RAID engine itself. RAID engines kicking non-TLER disks out of the array on every bad sector causing the disks to perform recovery more than 10 seconds, is a prime example. So the disks don't die earlier, but non-RAIDed storage versus RAIDed storage has the advantage of not introducing another layer to your storage setup which can fail in itself; RAID is a single point of failure, just like your filesystem.
So using any RAID may increase the chance of at least the RAID layer failing or giving you trouble; RAID can improve reliability but can also decrease it; common onboard-RAID drivers available on Windows and virtually all Hardware RAID controllers would have problems with disks performing deep recovery. These kind of RAID implementations are less secure than modern software RAID under non-Windows OS (ZFS, MD-raid in Linux, geom raid in BSD).
Whatever you do, never trust your important data to just one thing. Have it backed up, and double backups for your most important data. If you do that, you can focus on performance and use RAID0 and don't have to be scared about failures, assuming it's no problem to restore from backup and have a bit of downtime.
Cheers
-sub
The Samsung F4 2TB 5400rpm and WD EARS 3-platter ~5400rpm would also be able to beat the older velociraptor in sequential I/O; the F4 is capable of up to 140MB/s sequential reads. Being 5400rpm disks they approach the sequential performance of the newest 7200rpm and even 10.000rpm disks, because they have the highest data density of all. 7200rpm disks are still limited to 500GB platters while the F4 and newer EARS would use 666GB platters, leading to higher sequential throughput.
The higher rpm disks would still win from any 5400rpm disk regarding random IOps; though for that task any HDD sucks and an SSD would be the most logical choice; for example as system disk.
While i agree with MRFS, that RAID0 doesn't increase the chance of disk failure, is does increase the chance of failure of your storage, due to problems with the RAID engine itself. RAID engines kicking non-TLER disks out of the array on every bad sector causing the disks to perform recovery more than 10 seconds, is a prime example. So the disks don't die earlier, but non-RAIDed storage versus RAIDed storage has the advantage of not introducing another layer to your storage setup which can fail in itself; RAID is a single point of failure, just like your filesystem.
So using any RAID may increase the chance of at least the RAID layer failing or giving you trouble; RAID can improve reliability but can also decrease it; common onboard-RAID drivers available on Windows and virtually all Hardware RAID controllers would have problems with disks performing deep recovery. These kind of RAID implementations are less secure than modern software RAID under non-Windows OS (ZFS, MD-raid in Linux, geom raid in BSD).
Whatever you do, never trust your important data to just one thing. Have it backed up, and double backups for your most important data. If you do that, you can focus on performance and use RAID0 and don't have to be scared about failures, assuming it's no problem to restore from backup and have a bit of downtime.
Cheers
-sub
Facebook
Twitter
Instagram