Systemboard decision (part 2): SSD performance

Some of the board's SATA 6Gb/s ports should be Intel ports. Move your drive to one of them and retest.

Unfortunately, as you've discovered, spec sheets can mislead. You'll probably need to rely on reviews and/or benchmarks. The Intel controllers have, time after time, been shown to be a lot closer to their advertised specs (and there is always overhead) than the Marvell controllers.

Despite advertising to the contrary performance is NOT a good reason to invest in an SSD. The specs are all inflated and also reflect idealized conditions. The only time the average SSD user will notice a performance increase is during booting and program launch. After a few days/weeks you won't even notice that any more. It just becomes normal. The better reason for investing in an SSD is reliability. No moving parts. If that doesn't impress you then you should consider a HDD.

I've had more SSDs fail in use (2) over the last five years than I've had HDDs fail in use in at least ten years (none, but I did have a couple of DOA mechanical drives that were promptly replaced under warranty).

SSD controllers will fail electrically just like a HDD however most HDD failures are mechanical. I believe what you say but surely you must realize that you are an exception to the rule. BTW, what brand failed?

If you handle huge files frequently such as video files then you will certainly notice the difference however that means you are not the average SSD user that I was referring to in my last post. Eventually even you will become accustomed to the extra speed. Remember the first 1000mz processor you owned?

The issue is not so much your SSD/drivers it's your chipset and marvell controller connected to the system with a pcie 1x this will limit you to sub v2.x: 500 MB/s (5 GT/s)
Then your south bridge conntect to northbridge with 2GB/s sharing bandwidth with everything.

http://en.wikipedia.org/wiki/File:X58_Block_Diagram.png
http://sterlingdesktops.com/88se91xxa.png

wiki:
"Like 1.x, PCIe 2.0 uses an 8b/10b encoding scheme, therefore delivering, per-lane, an effective 4 Gbit/s max transfer rate from its 5 GT/s raw data rate."

sata drive -> sata controller -> southbridge -> northbridge -> cpu -> mem
That's a lot of stages to further increase overhead/latencies, and potentially put a ceiling on your peak transfers. ie: burst rates from the SSDs dram
Also if you have another sata or pata drive connected they'll share that pcie 1x connection.

I noticed you mentioned linux.
Things you can do: Verify that AHCI is enabled, update your kernel, update filesystem utilities, you might want to update your bios, update your SSD firmware, and run fstrim on the drive. Also i'm not sure if partition alignment has any effect on the samsung SSDs, you might wanna check into that.

My crucial M4 on a marvell 9128 card matches it's specs in linux.
I didn't see what programs you were using to bench either.

Others above mention specs not always matching up to realworld (ie: degredation).
Also what filesystem are you using ?
There are some newer marvell 4 port cards with pcie 2x connections (might be some asmedia ones too).
I doubt you'll get much faster buying a new card however.

The first one that failed was a 1st generation Corsair; I forget the controller. The second one that failed was an A-Data Sandforce drive. The one I got back from RMA seemed a little flaky too, but probably okay. I believe it's on a shelf right now; I replaced it with a Mushkin (also Sandforce) iirc. I'd probably try using it again, but the machine was for my wife, so I wanted to be extra-careful; she hates PC problems!

1. Your motherboard is an older Intel socket LGA 1366 model that is no longer in production. It does not properly support modern 3rd generation SATA 3 6Gb/s solid state drives. The Marvell controller was not that great.

2. The performance charts you refer to are for newer computer configurations that properly support modern 3rd generation SATA 3, 6Gb/s ssd's. However, there are problems with the synthetic benchmarks. They are called synthetic benchmarks because they are not real. The benchmarks were expressly designed to grossly exaggerate very minor differences in ssd performance. In addition, companies manipulate the settings and test conditions to present their ssd's in the most favorable light possible. For all practical purposes over 90% of the synthetic benchmarks do not accurately represent real world situations. Consider them an advertising gimmick.

3. If you are considering purchasing a new motherboard, then look at motherboards with newer Intel chipsets and drivers. The newest are the Intel LGA 1150 socket motherboards with Intel Z87 chipsets. Unfortunately that also would mean purchasing a new cpu and possibly new memory. Here is a link to some examples:

http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100007627%20600438202%20600315497%20600009016&IsNodeId=1&name=ATX&Order=RATING

4. I do not recommend PCI-e cards at this time, especially the inexpensive ones as they produce mixed results. However, there may be a viable alternative - affordable consumer oriented PCI-e based solid state drives. The new SATA Express standard has been adopted. The adoption signals the beginning of the migration to PCI-e ssd's for consumers. Samsung is already mass producing and shipping their new XP941 PCI-e ssd's to OEM's. That particular model is designed for use in ultrabooks and tablets. Consumers will have to wait at least until the end of the year for desktop versions.

5. You mentioned video capture and video editing. Rendering processes can slow down performance. When it comes to rendering processes there are performance differences among ssd's. The typical solution for consumers is to use one ssd for the OS, software applications, and utilities; a second ssd as a scratch disk; and a hard disk drive for data storage.

I maintain the ssd database listed in a sticky at the very top of this forum section. The database includes links to technical reviews for each ssd brand and model. I distinctly remember a handful of reviews that included testing video editing applications and rendering. Unfortunately I am a senior citizen who is growing old disgracefully and I forget things. If anything comes to me I'll post the information. On the positive side the technical reviews usually have a page with a chart that identifies the motherboard, cpu, memory, and other components used for the testing. Here is the link to the ssd database:

http://www.johnnylucky.org/data-storage/ssd-database.html

http://en.wikipedia.org/wiki/PCI_Express
"Capacity
Per lane (each direction):
v2.x: 500 MB/s (5 GT/s) "

The 9128 marvell chip only provides pcie 1x (configurable as v2.0 or v1.0), there is no other option. The png diagram you apparently can't view is me screenshotting the marvell flyer of the 9128's layout.

Constantly writing 1GB to an SSD is something that can degrade performance.

I guess you'll figure things out. Good luck.