Intel 510 not fast enough ?

[gaida]

Hi All,
I am having some trouble getting the best of my Intel SSD 510 120GB Drive.

Here is my config :

i7-860 @ stock
Asus P7P55D-E Pro
16GB (4x4GB) DDR3
1x Intel 510 120gb ( OS drive 30GB free )
1x Seagete 500GB ( temp files + pagefile )
1x 4GB Ram Drive ( Firefox cache )
Win7 Ultimate

Running the latest Intel RST in ACHI and latest Marvell Drivers ( From Station-Drivers.com ).
I originally had the SSD on the Marvell 6Gb controller.
On the Marvell Seq Read 328Mb / Write 166Mb ( tried a couple of drivers same result )
On the Intel I am getting Seq Read 256MB / Write 210Mb ( so write speed is spot on , but read is way off ).
I've read most of the reviews of the 120gb drive and people are getting in the 400s MB seq Read on Sandy Bridge.
The 512k random results for Read are about 50MB/s faster on the Marvell ( 276MB/s ) and 30MB/s slower.
The 4k random are pretty similar with the Marvell reading a bit faster and writing a bit slower.

Is the 256mb/s read on the Intel controller a limitation of the P55 platform ? Or can I do better?
I prefer the Intel controller as TRIM seems to work on it, whereas it is a bit debatable if it works on the Marvell.

All tests were done with Crystal Mark 3.0.1 64bit

Regards
Nick

[Pob255]

Is the 256mb/s read on the Intel controller a limitation of the P55 platform ? Or can I do better?

that's pretty close to the limits, remember the p55 chipset is only SATA 3GBit/s and there is always some overhead on communication that means it will never get to the full speed.
I don't know why the marvell controller is 50Mbit/s slower on write, my guess would be some latency limit of the motherboard, but I'm not sure.

Have you checked for a BIOS update, that might be able to help.

[gaida]

Thanks for your reply.
Running the latest bios for the mobo. I've also installed a "hacked" driver for the Marvell, but the same slowish write speed. It seems weird that the Intel controller tops at 256mb or 2^8 nice round number
Maybe my data drive is eating some of the bandwidth, I will move it to the JMicron controller to see if it makes any difference at all.
to be fair, TRIM is more important to me than 60-70mb/s read speed, but wouldn't it be nice to have both ? On Sandy bridge other ppl seem to get much faster speeds.
Does the fact that my drive is quite full affect the performance ? I know most reviews are using the SSD as a secondary drive, and it is nearly empty when they do the tests.

[blueball]

How did you setup the drive? Did you install Win 7 as a fresh install or did you clone the installation from a previous HDD?

If you cloned it then it may not be aligned properly. See this post.

[gaida]

I cloned my RAID 0 set-up to it
looking here http://www.sevenforums.com/tutorials...alignment.html it seems like my drive is aligned properly. I will investigate further and let you know.
I would imagine that if the drive was not aligned , it would be much much slower.

[blueball]

If you run AS SSD Benchmark (free download) it will tell you if your drive is aligned properly (top left of program main window, several lines of info one of which reports alignment state)

[gaida]

I didn't think to check with AS SSD ( I did run it the other day ).
The partition is already being aligned with Paragon Alignment tool, so we shall see

[gaida]

So partition has been aligned.
It did not make any difference to the sequential and random rates at all.
It did however gave me a 10% increase in READ when QD=32 and a massive 50% in WRITE

I am beginning to think 256mb/s is the max the 3GB/s interface can give me.
Time to start tinkering with the Marvell controller

[gaida]

After some more research it seems like the maximum out of SATA 3GB/s is around 280mb/s so close enough.
The Marvell was able to hit nearly 330mb/s by stealing PCI-E bandwidth from the PCI-E slots. ( possibly starving my 2nd GFX card ).
I shall leave it at that for the moment and not waste any more time on it

Thanks for all your help

[Pob255]

It doesn't take anything away from the gpu, on the non-E version it also has a pci-e x4 (slot is x16 size but is actually x4) they dropped that slot down to a x1 and used the additional lanes to supply the PLX chip which uses it to supply the USB3 and SATA3