"AMD has signed up for it....."

I guess we can wave goodbye to any chance of driver fixes for both the shocking performance and the tendancy of the array rebuild/migrate routines to regularly hang the machine stone dead* courtesy of the Promise controllers they currently implement in the SB600 and SB750 then.

Hint: Hard disabling writethrough cacheing is not a bloody fix AMD, even if it does stop you losing data when the thing inevitably goes titsup.

Hint 2: If you can't write a reliable RAID driver and controller firmware to save your fucking lives, changing the controller ain't going to fix this.

*Not that it doesn't do this in normal operation too, just not every five bloody minutes.

How is this different from the "md" devices that have been included in the Linux Kernel since 2.4, and which consistently outperform proprietary (therefore, non-verifiable) software RAID?

"core logic chipset-based RAID solutions, which provide fewer features and are unable to meet the stringent demands of 24/7 server-class RAID technology."

Nevermind if anyone buying a low end server needs more than the "few" features, nor what they impose as the stringent demands those customers /must/ want.

I"m all for more features at lower cost, if implemented well and without introducing new bugs but this is putting the cart before the horse. Many would already want those low end $400 servers if they put LESS money into idle CPU cores and a little more into a beefy enough PSU to handle more drives, and a few more SATA ports onboard.