Originally Posted by Fred B
It shows when boot from ssd or hd , the hd uses easy qd 10 and the ssd mostly 1 and a few 2/3 . The ssd is fast enough to answer a reqeust , while for hd it is a complet diferent story . When running in ide mode you could spread the load between two ssd/hd and have advantages of multiple iq request .
This is the end of the boot for xp when it reaches qd 3 , but it is a little piece on the end of the boot that uses for ms the higher qeue dept .
The iq time becomes faster when using ssd raid but that does not show in faster boottimes , so it think the os is a real bottleneck
The same end of the boot from fragmented hd , it is random and the iq time is terrible but hits qd10
Thanks for posting these tables! Nice example of queue depth, booting Windows is a situation even with SSDs, where it is easier to see how it works. Loading a complex game would likely show a queue depth (QD) above one with a SSD.
Any idea what recorded that data? Was that a feature in XP that we no longer have in Windows 7, etc? Interesting to see this with XP, since boards/SATA controllers and drivers in those days did not have AHCI capability, nor does XP have a native AHCI driver. Must have been later on in the XP SP3 time.
Also an example of the superiority of SSDs over HDDs, the SSD is so fast at some points the QD is zero, which means on that system the SSD actually had nothing to do while the CPU was working, I'm surprised. Plus the SSD is a 40GB Intel G2 SATA II SSD, a decent but hardly high performance SSD by today's standards.
I'd love to see similar data for booting Windows 7 and 8, which are larger than XP and likely have more services, etc, starting during boot. I wonder if the "IO Size" shown in the data, all relatively small files, is similar with Windows 7 and 8? Also note there are writes occurring during boot, not a surprise but sometimes forgotten. The tables are not showing the same point in the boot process, so my comments may not be entirely valid, I'd love to see all the data.
It is hard to argue that AHCI/NCQ is extremely important all the time for SSDs used in a single user PC environment, given the QD we see here. But I see no reason to go back to only single I/O requests being sent to a drive, unnecessary and just plain primitive.
If someone does not want to use AHCI with their SSDs, go for it. Personally, I'll keep it.