Slashdot Mirror
Faulty Marvell Chips Delay SATA 6G Launch
Vigile writes "The SATA 6G standard offers more than simply a faster 6.0 Gb/s data throughput speed, to wit: improved NCQ support, better power management, and a new connector to support 1.8-inch drives. While modern-day, spindle-based hard drives struggle to keep up with SATA 3G speeds, modern SSDs are nearly saturating the existing standard, and a move to SATA 6G was welcome in the hardware community. It looks like that technology will be delayed, though. The only chip supporting the standard today, the Marvell 88SE9123, is having major issues. Motherboard vendors including ASUS and Gigabyte, which had planned on releasing SATA 6G technology using the chip on Intel Lynnfield platform motherboards later this summer, are having to remove the Marvell 88SE9123 and redesign their boards at the last minute due to significant speed and reliability issues."
... "modern SSDs are nearly saturating the existing standard" ...
æeee!
Faulty Marvell Chips...
Maybe they should stick to comic books.
When our name is on the back of your car, we're behind you all the way!
>> The only information we have gotten from anyone revolves around some hardware AND software issues that are preventing the SATA 6G speeds from actually reaching 6.0 Gb/s
Instead of delaying the launch, they could just rename the chip. SATA 5.5G for example.
I wonder if The Green Gobllin had anything to do with it...
Watch the Teaser Trailer for "The Lightning Thief" Her
Even the 6G standard won't hold for high-end SSDs (which seem to be raid striped in one unit, AFAIK). The long-term solution for those are ones that connect via PCIe, so this doesn't seem to be that big a deal, really.
The PCIe versions of SSDs I've see so far doesn't seem to support booting. That's a major crimp in their usefulness. Also, they're not all that useful in laptops.
To be bad enough that these low-end board shops would reject it this Marvell part must be truly heinous. Asus et al. usually don't hesitate to ship boards with badly flawed components. Cox talked about this a few years ago.
If you want good SATA avoid the third party chips these board makers integrate. Especially the RAID crap. Wait for Intel to build it into their regular chipsets.
Nitpick:
I'll bet they implemented striping a different way than RAID. Since all they need to do is act like a standard hard drive to the machine, they can implement something faster and more specific to flash memory (since every optimization counts).
*cough*Doesn't increase bandwidth*cough*
The PCIe versions of SSDs I've see so far doesn't seem to support booting. That's a major crimp in their usefulness.
True, but that's temporary. I've heard the next major crop of them coming out over the next quarter or so will be bootable. I certainly wouldn't buy one until they ARE bootable.
Also, they're not all that useful in laptops.
Laptops have PCIe connections available, too; that shouldn't be a problem.
TFA said "nearly saturating", but regardless...
http://www.anandtech.com/storage/showdoc.aspx?i=3531&p=24
250MB/s is just below the theoretical limit of SATA 3Gb/s, which is 300MB/s. It's possible that there are still other bottlenecks beside the hard drive.
I also have seen SSD RAID benchmarks somewhere, but I don't remember where.
Convert FLACs to a portable format with FlacSquisher
Intel's X25-M seems to perform 200 MB/s constant throughput. Granted that's "only" 2/3rds of the 3 Gbps that SATA 2 delievers, but the quote was "nearly saturate".
And if we're already at 2/3rds, that's a fairly compelling argument to upgrading. On laptops it can become an issue much quicker, as you usually only have 1 eSATA port, and port multipliers do not increase bandwidth. Hooking two X25-Ms onto a single eSATA 2 port can saturate it while doing non-random transfers easily but still have room left over on eSATA 3.
But if we're merely talking SSD in general, we can always point to Fusion-io ioDrive which bottoms out at 429 MB/s.
The Fusion-io ioDrive Duo could, but connected directly into a PCI-E slot rather than SATA, but it is definitely possible to make a drive that will.
upon the advice of my lawyer, i have no sig at this time
Fusion-IO has PCI-e cards that they claim sustain 570MB/s up to 1.5GBps
Can't boot from them, but there you go.
Nobody? OK no cream.
*cough*Regular mints are better if it's just a cough. No point in taking allergy medicine if you aren't allergic.
Take a long hard think about this.
What manufacturer is going to make SATA SSD's that can saturate the fastest SATA port?
Sounds to me like a bad idea. Surely there are some tradeoffs between speed and some other also important metric. Ex, faster might mean larger erase block sizes.
It is likely that Intel could have made their product much faster, but without any benefit at all to doing so, they wouldn't.
This new SATA, while important, it sort of too-little-too-late. We need a much higher ceiling, and we need it yesterday.
"His name was James Damore."
The only reason SSDs aren't capable of higher speeds is because the bus is not capable of more. There's no point making a controller capable of 2GB/s if you are only able to transfer at 300MB/s.
Most sequential throughput benches hit the drive with sequential requests, and are not multithreaded. You end up seeing throughput lower than the theoretical interface bandwidth because of the latency involved for each request. A queue depth of 2 or 4 will give an X25 enough heads up to truly saturate the bus. I've recorded as high as 285 MB/sec from an X25-M using an NCQ-capable benchmarking tool.
Also, don't forget SATA uses 8/10b encoding, so you have to account for that overhead as well when calculating theoretical maximum throughput. Doing the math, you'll find 285 comes to 95% of a 3 GB/sec SATA interface. That's way higher than 200 MB/sec and close enough to call saturated.
Allyn Malventano
Storage Editor, PC Perspective
this sig was brought to you by the letter
What manufacturer is going to make SATA SSD's that can saturate the fastest SATA port?
The same idiot manufacturers who deliver super-high sequential reads at the expense of everything else. There's a jumble of drives running off a particular J-Micron controller (or two :P), that deliver faster sequential reads than the Intel X25-E, but have random access times in the 50ms range (average). That's five times slower than a cheap 5400rpm notebook drive! They still sell like hotcakes because they're cheap, and we all know how much America loves cheap garbage. You could stick a SATA port out the side of a fresh turd, and if you price it low enough, some dumb fuck will buy it.
SATA 6G is too slow, yes, but the companies involved seem to be interested in controlling their steady income by keeping everyone on an upgrade treadmill. If they had given us 6G at the very start, they would not have been able to milk the industry over the last six years with these minor upgrades and feature enhancements. Look at Firewire, ignoring Apple's idiocy, it was a very fast external bus that was far ahead of its competitors in terms of performance, reliability and ease of use. It was so good that even the 800mbit upgrade is considered redundant, the old standard is "fast enough" for most uses.
Better get used to it though, SATA is the new bottleneck, and it will be for years to come. There's just no way around it. It took this long for people to finally ditch IDE, even if there were a new contender to leave SATA's generationally-challenged performance in the dust, the industry simply is not ready to change interfaces again.
-Billco, Fnarg.com
Honestly, Marvell chips have cost me more grief on Linux installs than all other vendors combined. If this gets mobo vendors to design out Marvell, then I say: "Grand!".
This lack of knowledge on /. is sickening. Not only a single SSD already almost saturates a SATA 3.0Gbps link (300MB/s with 8b-10b encoding), but even regular hard drives do. Transfers to/from the on-disk buffer chips are bottlenecked by the 300MB/s speed. And SATA enclosures placing multiple (3 or more) drives behind a SATA port multiplier also easily saturate SATA 3.0Gbps links (the sequential read speed of a 1TB Seagate 7200.11 is 120MB/s, so 3 of them do 360MB/s).
I have to agree there. Marvell chips have given me nothing but grief, both in Windows and Linux, both network and ATA. They made me buy a SATA DVD writer to replace the fully functional PATA device, as well as a Silicon Image-based SATA controller to attach harddisks (couldn't use the Intel controller, because MSI still insists on placing the "good" SATA connectors under the GPU cooler).
All in all, I'm happy about anything bad happening to Marvell.
Fight hunger. Filet a politician and send him to a 3rd world country of your choice.