01-19-2010 01:26 AM - edited 01-19-2010 01:29 AM
One topic I haven't seen a lot of discussion on is whether it is better to run an SSD in IDE or AHCI mode. In fact, I don't think I've ever seen it mentioned in any of the professional reviews which mode the drives are tested in.
The benchmarks on my system show that AHCI gives approx. an 18% improvement in read speeds compared with IDE but that may be down to the fact that I was using Intel drivers under IDE and now I'm using Microsoft's under AHCI. In fact, I did test both drivers and the Intel ones did return a slightly lower read/write speed than the default ones Windows 7 installs.
However, I've noticed that when I boot up Windows 7 now, rather that it skip so quickly through the Welcome screen with IDE that if I blinked I'd miss it (I'm the only user on my PC so I don't log on), with AHCI is takes a second or two then the desktop appears. That seems to imply that when faced with lots of access such as during the loading of Windows AHCI is perhaps not as fast as IDE. I've tested the load times with IDE and AHCI and the latter are a couple of seconds longer typically. Other than Windows loading though I can't say I've honestly noticed any performance differences between IDE and AHCI while actually using Windows.
So which should I be using: IDE or AHCI? AHCI gives you access to hot-plugging and NCQ but I've read that NCQ is great for mechanical drives in servers but not so hot for SSDs.
01-19-2010 03:18 AM
You are coming up with some insightful questions just lately aren't you! Im sure targetbsp will be along with his 2p worth shortly, but I've always gone by the benchmarks the drive gives which are clearly about 10% better in my case with AHCI enabled (Microsoft driver) as for random access times you could be right, although I would dare say that the increased boot time is simply to do with the different drivers required for AHCI to be operational on the system. Im sure that once windows is actually booted its faster in all respects using AHCI over IDE.
01-19-2010 03:31 AM - edited 01-19-2010 03:43 AM
Here I am! Man I gotta get me a job
Djinnuk makes a good point there. The reason you can't switch between IDE and AHCI is Microsoft determined long ago in earlier versions of Windows that loading hard disk drivers slows down the boot quite a lot so it only loads the one you use during setup. So when you change your registry you're making it load more than it ought too. So it could be that.
As for NCQ I doubt it has any downsides I just also doubt it has any benefits based on the theory behind it. It's purpose is to give the impression of reducing seek times on drives. It does this by if the OS requests files 1 and 2, it works out if based on the position of the read head on the data on the disk whether or not it's faster to instead read 2 then 1. This shouldn't have any effect on an SSD as you don't move anything to read so assuming 1 and 2 are the same file size the time to read them should be the same. I'm not aware of any other feature in AHCI that's there to improve performance.
I would imagine, though I don't know, that any performance differences come down to driver differences or possibly caching option differences between the 2 modes/drivers. I can't test this on the grounds that even with the registry changes my Windows dies if I switch transfer mode. And as for the caching mode differences, consecutive benchmarks under 1819 varied so wildly that the margin of error was too large to test any minor performance improvements. I've not carried out consecutive tests under 1916 to see if that's better now. Coupled with the fact that benchmarking absolutely canes the drive life.
Having said all the above, it seems both you 2 have had a performance increase. So hopefully people that have tested it will post in here so we can get a picture of which is better.
01-19-2010 05:57 AM
Thanks both of you for the replies.
I think I'll stick with AHCI then as I don't think I'll be losing much sleep over Windows taking an extra second or two to load over IDE. I've not come across any issues with using AHCI - my DVD-Rewriter for example still loads and burns discs - and performance seems pretty much the same to me from general usage, i.e. Windows is still uber-fast and responsive the split-second the desktop appears post-load. The benchmarks suggest better performance with AHCI and the Microsoft drivers so who am I to argue?
Incidentally, if I did want to switch back to IDE is it just a case of changing the Start registry entry for Msachi back to 3, rebooting, switching the BIOS back to IDE then letting Windows load and reinstall the IDE drivers?
01-19-2010 06:39 AM
If you left your pciide registry entry set to 0 you can switch backwards and forwards at will without any further changes as you're loading both drivers. Setting the registry entry to 0 doesn't choose it as a mode it simply says load this driver whereas 3 means don't load it.
01-19-2010 07:45 AM - edited 01-19-2010 07:47 AM
I found that tweak across lunchtime while I was browsing more info on AHCI and IDE, targetbsp, but I knew I should have just waited for you to reply though... would have saved me an hour!!!
Does leaving both those AHCI and PCIIDE registry settings at 0 mean the both sets of controller drivers are being loaded or is it the BIOS settng that determines which one is loaded with Windows? I'll try the tweak only if one set of drivers are loaded.
By the way, I came across an interesting thread on another rival SSD manufacturer's website where someone else was asking about AHCI and IDE. The reply from the manufacturer was that IDE mode is best for SSDs because AHCI results in slower performance due to the extra features it has for hot-plugging and the fact that NCQ isn't needed (which you explained anyway). That certainly isn't supported by my benchmark results but then Windows 7 does seem to load a little slower...
The reason I switched to AHCI in the first place was because my benchmark scores were lower than others I'd seen for the same 128 GB model. With AHCI they are the same but I've seen people running the SSD in IDE mode over on the Overclockers forum post the same scores so I'm wondering if it's not the controller mode that is making that difference but the drivers. I did test Intel and Microsoft's drivers with AHCI and the former are definitely slower on my system. However, I only had the Intel drivers installed under IDE mode, I've never actually benchmarked with the default Microsoft ones. Maybe they'll give better performance?
I'm going to switch back to IDE, uninstall the Intel controller driver then benchmark one last time. If the scores are the same or comparable to the ones I'm getting with AHCI then I'm going to leave it set to IDE otherwise I'll switch back to AHCI. I do feel like I'm spending so much time worrying about getting the best performance from my drive that I'm not actually spending any time enjoying my PC!!!
01-19-2010 01:12 PM
As it turned out, the registry setting for PCIIDE was already 0 so it was very easy to switch back to IDE mode (a simple BIOS setting change). I changed the drivers for the controller from the Intel to the Microsoft ones but when I benchmarked with ATTO I got exactly the same 190 MB/sec sequential write and 225 MB/sec sequential read speeds as I did with the Intel controller driver under IDE. Since AHCI gives me faster random and sequential read speeds I've decided to go back to ACHI and just leave it at that.
I also noticed that the health of my SSD has gone down from 99% to 98% according to CrystalDiskInfo, no doubt a result of all this benchmarking so I'm *definitely* not running any more of those!!! Instead I'm going to enjoy using my PC... that is what I got it for!
03-11-2010 05:22 PM
Interesting discussion guys...wonder if you can shed any light on this.
I installed Windows 7 using AHCI and found I had a windows performance score of 5.9 for the ssd mentioned in this post...
I then tried IDE mode (just changed the bios setting back to IDE and booted) and got 7.4 which sounds about right.
But I benchmarked using ATTO on both occasions and under IDE Atto scores are on average 20% down on AHCI.
Anyone know what is going on?
And more importantly which is better a Windows performance score or a high Atto reading?
Many thanks for any light you can shed,
03-12-2010 01:45 AM - edited 03-12-2010 01:49 AM
I'd imagine you can't tell the difference in real life performance? I'd also think that AHCI is better from the sounds of it. WEI scoring is a bit strange. It'll run several tests that could put you in the mid 7's but if one of them is below a certain threshold it will cap your score at 5.9 no matter how well it did at the rest.
Are you using Microsofts AHCI drivers?
I see from the other thread that you can tell the difference. In that case I'd use IDE then.
03-21-2010 05:41 AM - edited 03-21-2010 06:35 AM
I'm seeing Stutter here on my C300 running AHCI (Marvell AHCI Drivers - V1001027_XpVistaWin7) on a 9123 SATA 6G controller.
Still playing around I'll report back.
The Stutter problem was found to be using a SATA 6 device in an icybox removable caddy.
My other SATA3 SSD runs fine in it; not so the SATA6 C300, it wanted to be connected straight to the port.