Only getting 200MB/s in Sata 3 in Raid 0

[refriedfood]

Hello, maybe someone here can help me.

I have a setup as follows:
Motherboard: GIGABYTE GA-Z68X-UD7-B3
Hard Drives: Seagate Barracuda ST3500413AS 500GB x2 in RAID 0
OS: win 7 x64

These are Sata 3 (6Gb/s) drives, and they are detected as such in Intel Storage Utility. However, when I benchmark them, they max out at roughly 200MB a second.

I'm using the Z68 Intel chipset raid controller in Sata RAID mode. They are in the correct sata3 ports on the mother board.

I'm completely at a loss and have searched for the answer to no avail. At raid 0 in sata 3 I should have a much faster read/write rate than I do. Windows index scores the volume at only 5.9.

Anyone have any suggestions?

 

[Dereck47]

Read/Write speeds for your drive is 156MB/s. So with 2 drives in RAID-0 the maximum theoretical speeds would be 312MB/s.

AS-SSD and CrystalDiskMark use highly incompressible data to test Read/Write speeds.
If you are using either software to benchmark, that could be the reason you're getting 200MB/s Read/Write speeds.

Benchmark with ATTO, which uses highly compressible data to test Read/Write speeds.

 

[Pinhedd]

RAID-0 does not double your read/write performance, it typically adds about 33-50% 200MBps is right on the mark and SATA III will not add anything to a platter drive

 

[Dereck47]

RAID-0 does not double your read/write performance, it typically adds about 33-50%

Yes, due to overhead it's not double, but it's way more than 33-50%.

I have 2 Vertex 2's in RAID-0 and I get 91% using ATTO and 88% using AS-SSD; and my drives are SATA II.
If I had 2 Samsung 830s or 2 Intel 520s or 2 Vertex 4s I'm sure my percentages would be even higher.

 

[refriedfood]

Why exactly would a Sata 3 @ 6Gbps in raid 0 only do 1.6Gb (200MB)... That is piss poor.

I have a SSD sata2 and it does 220MBps max, so you're telling me because they're "platters" (disks) and in raid, sata 3 essentially doesn't mean a thing and is worse than a ssd in sata2?

If that is the case, isn't that blatant false advertising on Seagate's part?

 

[Pinhedd]

Yes, due to overhead it's not double, but it's way more than 33-50%.

I have 2 Vertex 2's in RAID-0 and I get 91% using ATTO and 88% using AS-SSD; and my drives are SATA II.
If I had 2 Samsung 830s or 2 Intel 520s or 2 Vertex 4s I'm sure my percentages would be even higher.

You have two SSDs (SSDs should not be put in RAID-0 because there is no TRIM support for SSDs that are in RAID) and are running a synethic benchmark which uses compressable data which the Sandforce controllers love. This is not reflective of a real world scenario in which data cannot be compressed and HDD access times come into play. In non-synthetic applications (that is, all the applications that count, Raid-0 adds between 33 and 50% depending on the size and fragmentation of the dataset. In fact, RAID can actually decrease performance for small datasets due to added latency

 

[Pinhedd]

Why exactly would a Sata 3 @ 6Gbps in raid 0 only do 1.6Gb (200MB)... That is piss poor.

I have a SSD sata2 and it does 220MBps max, so you're telling me because they're "platters" (disks) and in raid, sata 3 essentially doesn't mean a thing and is worse than a ssd in sata2?

If that is the case, isn't that blatant false advertising on Seagate's part?

There are two interfaces for any storage device

There's the interface between the initiator and the controller which will negotiate a link speed of either 1.5, 3, or 6gbps.

Then, there's the interface between the controller and the storage medium, which operates at whatever speeds the device is capable of operating at internally.

HDDs controllers have a disk cache that acts as a buffer for reading and writing data. This has both a high speed interface connection with the initiator and a burst connection with the storage medium.

Communication over the SATA bus with the HDD controller can read and write to the HDD controller's buffer at extremely high speeds but the buffer is quite small, topping out at 64MiB even on top end models. Once it's full, or empty, the speed of the initiator-controller interface becomes immaterial because the entire system becomes constrained by the controller-storage interface which has significant read and write delays (on average 10-20ms for a 7200 RPM hard drive).

This is why SATA-III does nothing for platter drives, the technology backing it can't read or write anywhere near that speed

 

[Dereck47]

You have two SSDs (SSDs should not be put in RAID-0 because there is no TRIM support for SSDs that are in RAID) and are running a synethic benchmark which uses compressable data which the Sandforce controllers love. This is not reflective of a real world scenario in which data cannot be compressed and HDD access times come into play.

Yes, you're correct with regards to compressible data; that's why I included my AS-SSD results which does not use compressible data.

Also, modern SSDs have very good Garbage Collection algorithms to maintain drive performance. Apple OS X has only supported TRIM for less than a year, yet Apple users have had SSDs in their system for many years.

 

[refriedfood]

Best answer selected by refriedfood.