Jump to content
Corsair Community

SSD question


Silvio

Recommended Posts

The situation is heavy.

I am running 2 force gt 120 on a dual lane 9182 marvell embedded onto a x58 RampageIII Black Edition.

 

the above attached are the bench.

o.s. w7/64

drivers used : marvell_91xx_1.2.0.1016

 

1) Considering that the system had got (for the mom) a real turbo boost from the time I installed my new gt ssd drives, - w7 loading time is really faster - ~ 15 secs against ~ 75 secs I had with my previous raptor 600 sata3 - I am quite happy - but not saddisfied.

 

2) My embedded 9182 shows on post 5Gb/s connection and not 6Gb/s (yes true at least they are honest lol) and here my consideration is - many ppl - like me - thrust on great brands - generally more expensive than others - and buy their products. Specially when the product is high end one, I believe they have not to fit controllers not native and with plenty of issues. (when I attached onto sataIII my raptor I had to pain a lot)

 

3) At present I could not understand if trim is available or not trought marvell controller. nobody tells me if trim command was able to reach the device. this Is umbelievable!!!!

 

4) Looking at bench specially write part of it, I got depressed!.

 

So the question - is better to buy a low profile for lot lot less cash and similar performances, or high end stuff as beta testers/Pilot customers (translated in rude words ppl paying development for others!) ?

 

really confused.

 

pls excuse my terrific english.

 

regards

 

Andrea

1928197254_Marvell9182.jpg.99ff81e81da2b9bbfeaf91104090411e.jpg

Link to comment
Share on other sites

the issue is not with the sata chip in that last post.. but with the incompressible data used in that particular benchmark.

 

The Sandforce controller used on these drives is going to show stronger speeds when using easily compressed data(such as ATTO would use) because that's how the controller works when leveraging it's internal data compression.

 

So, with that in mind.. I would recommend using ATTO to put your mind at ease here. The sequential transfer results will literally triple on the write side of the equation.

 

Also just a friendly FYI here.. don't run those CDM3 benchmarks at defualt 5 x 1000MB settings as it writes around 20 gigs of random test trash to your drives. This is the hardest data set to recover from and may even lower performance in fairly short timeframes as the drive is forced to GC those blocks back into the fresh pool for the next anticipated write load. 3 x 100MB is usually more than adequate to "see" the speed of these drives. Hope that helps.

 

PS. I just noticed that they were run with 9 x 1000MB test sizes. Gulp.. that's even worse for the amount of trash written/fresh blocks consumed.

Link to comment
Share on other sites

Wired, has Marvell confirmed that their drivers pass the TRIM command? Marvell provides zero end user support for their SATA chipsets, that is up to the mother board manufactures. Do I trust the board manufactures, that sell us "SATA 6Gb/s on the Marvell 91XX chipset...". to supply a honest or even knowledgeable answer to that question? No I don't.

 

Some are RAID drivers, others just AHCI. The RAID drivers sure don't, but I have never seen confirmation that the Marvell non-RAID drivers will pass TRIM. Or will the Marvell RAID drivers act like Intel IRST drivers used in RAID mode, and pass TRIM to drives that are not part of a RAID volume? Or will the msahci driver used on the Marvell chipset pass TRIM to drives?

 

With TRIM and chipset drivers, there are very few simple yes or no answers, IMO.

Link to comment
Share on other sites

the issue is not with the sata chip in that last post.. but with the incompressible data used in that particular benchmark.

 

The Sandforce controller used on these drives is going to show stronger speeds when using easily compressed data(such as ATTO would use) because that's how the controller works when leveraging it's internal data compression.

 

So, with that in mind.. I would recommend using ATTO to put your mind at ease here. The sequential transfer results will literally triple on the write side of the equation.

 

Also just a friendly FYI here.. don't run those CDM3 benchmarks at defualt 5 x 1000MB settings as it writes around 20 gigs of random test trash to your drives. This is the hardest data set to recover from and may even lower performance in fairly short timeframes as the drive is forced to GC those blocks back into the fresh pool for the next anticipated write load. 3 x 100MB is usually more than adequate to "see" the speed of these drives. Hope that helps.

 

PS. I just noticed that they were run with 9 x 1000MB test sizes. Gulp.. that's even worse for the amount of trash written/fresh blocks consumed.

Ty groberts for your informations. lol now I am more confused. I tested my ssd at least 4 times with different benchmarks and all at maximum strongness. (included ATTO) I hope I have not damaged my SSD!

In any case even ATTO in write shows really slow speed.

I wont repete anymore benches on my ssd lol. (....just thin********. what the *** very delicate devices.) I hope they will work at least a couple of years then Corsair 4sure will have some new technology faster and more reliable.... :) )

 

take care

 

Yes TRIM should be working for you.

 

Ty so much Wired for reply.

Link to comment
Share on other sites

Silvio, You have not damaged your SSD, just put a little more wear on it than necessary by testing with large writes.

 

SSDs do not read or write at one speed with every type of file and data. When they say "550 MB/s", that is the best it can do with very large files, but any drive is actually slower when reading small files, and many small files one after another. Your test results are normal and generally the same as everyone else.

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...