SATA 6Gb/s worthless with Marvell controller?

synce

Distinguished
Apr 27, 2011
42
0
18,540
I bought an Asus Sabertooth x58 last year because I thought I'd be future proof with 6Gb, and then I hear about some people not being able to go above 400MB/s using the Marvell controller. Is it like this for every board out there? Any possible way to fix the issue?
 
Solution
1) It is true that Marvell has issue with getting over 400 MB/s, outside of the Marvell 9182 (newest version on highend boards).

2) This is true with all Marvell 912x boards but not true with boards running off of the Intel RST for SATA 6.0 or Marvell 9182.

3) Only way to fix is to either A) get a new motherboard or B) look at a PCIe based SSD like the OCZ RevoDrive X2... IMO

tecmo34

Administrator
Moderator
1) It is true that Marvell has issue with getting over 400 MB/s, outside of the Marvell 9182 (newest version on highend boards).

2) This is true with all Marvell 912x boards but not true with boards running off of the Intel RST for SATA 6.0 or Marvell 9182.

3) Only way to fix is to either A) get a new motherboard or B) look at a PCIe based SSD like the OCZ RevoDrive X2... IMO
 
Solution

kire

Distinguished
Apr 16, 2010
48
0
18,540
I have also read this and everything tecmo34 said is true to the best of my knowledge. My question is, would a 6GB/s SSD still perform better on a SATA port using the Marvell 9182 controller than on a SATA 3GB/s port?

(I also bought the ASUS Sabertooth x58 board on the same premise that synce did)
 

tecmo34

Administrator
Moderator

Yes... The Marvell 91xx (9120, 9128, 9182, etc.) SATA 6GB/s port would perform better than an Intel SATA 3GB/s for a SATA 6GB/s SSD.
 
Everybody looks at the "Advertized" Nrs. These are normally for Sequencial read/writes.
The Seqs are much better for sata III SSDs on SATA III controls. However, for the more important 4K randoms and IOPS there is not a big diff.

Most use an SSD for Operating system and programs. If you check you will find that approx 50% of your files are 16K (4 clusters) and a good number under 4K (only one cluster) and are stuck all over the disk.

The High Seq numbers are GREAT when working with DVD and Blu-ray video files. These files range from 1 Gig to 40 gigs in size - OPPS not many of these are going on that SSD. Editting large JPEG photos and large CAD/CAM drawings and large spreadsheet.

Bottom Line, yes the SATA III drives are better on SATA III, Just how much better than placing it on SATA II real life (Not benchmark) will depend on usage. After Booting and opening a program, as the OLd saying goes - HOW fast can you type.
 

kire

Distinguished
Apr 16, 2010
48
0
18,540
Just to make sure we're all on the same page as I just re-read what I wrote and I worry I wasn't specific enough.


Say for example I buy the OCZ Vertex 3 and use it on my ASUS Sabertooth x58. Because of the issue with the Marvell 912x, the Vertex 3 will not be able to reach its full speeds. So my question is, would the Vertex 3 still perform significantly better on the SATA III port than the SATA II port or would the gain be insignificant that I should just save my money and buy a SATA II SSD such as the Intel x25-M, 320 or Vertex 2?






I understand that random R/W and IOPS are the metrics that the "common folk" should reference and not sequential R/W. But are you saying that the difference between SATA II and SATA III ssd, with respect to random and IOPS, are not significant enough to be noticed?
 

synce

Distinguished
Apr 27, 2011
42
0
18,540
^ I believe it was written in a recent article here that the difference in performance between any two SSDs is insignificant to typical users. In my situation I would've benefited from higher seq. speeds (lots of WinRAR use) but seeing as there's no cheap and easy fix for the Marvell issue I've decided to go with an Intel 320. Thanks guys for the input
 

jerman885

Distinguished
May 26, 2011
1
0
18,510
I have an Asus P6X58D Premium Mobo, with at SATA III controller in it as well, its a Marvell controller, not sure what model though, I recently bought a Corsair Performance 3 SSD SATA III, hooked up to the SATA III controller on my mobo WITH a SATA III cable, and I am still only getting SATA II seq r/w speeds. I've been doing some research and can't seem to find another cause for this but the Marvell controller. So I don't think you should waste your money on a SATA III SSD, unless you are going to upgrade your system to a SandyBrigde build or something, I am probably going to return my SSD, too bad, I really like build, disappointed to find out that SATA III controller on this mobo is crap!


Win7 64bit Ultimate
Intel Core i7 930 @ 4.2Ghz
Corsair Dominator RAM DDR3 @ 1600 Mhz
Asus P6X58D Premium Motherboard
Seagate 500GB HDD
Corsair Performance 3 SATA III SSD
ATI HD 4890 Graphics card
Antec 850W PSU
 

kire

Distinguished
Apr 16, 2010
48
0
18,540
From what I can ascertain, the Marvell controller doesn't support TRIM, which reduces its usefulness for SSD. Well, what else is a 6Gb/s controller useful for anyway? &*##$%%@

That is something I wish I had not overlooked -_- but thank you for bring that up... now I know not to stick and SSDs on those ports
 

groberts101

Distinguished
Feb 20, 2010
363
0
18,810
The Marvell 91xx series is just fine for older gen sata3 SSD's like the C300 and will run it to max speeds(close anyways). The loss of trim on that controller may bite you over time though. If using a Sandforce controlled Vertex 3? then all you gain is sequential as Retired Chief already mentioned. Trim loss with that controller is a non-issue due to the way it works like none other. I run my V3 on it without issue or slowdowns.

Personally, I like the Intel as latency is a tad lower and write speed(small, random, and sequential) is stronger with the tradeoff of read speeds by about 100MB's.

The Marvell chip to have on non-native 6G supported mobo's is the 9182 due to the extra PCI-e lane availability. Otherwise the regular Marvell 6G chip works well enough for it's originally intended purpose which was 6G HDD's and slower first gen 6G SSD like the C300.