Is AHCI needed with SSD drives?
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
Is AHCI needed with SSD drives?
Since there is no increase in performance and it may decrease performance--see quote below--why use AHCI unless one needs hot swapping? Also, do these new SSD drives need AHCI for any reason in a desktop PC?
"Just a quick reminder for everyone on NCQ. It provides no performance benefit whatsoever in a typical desktop PC. In fact having it enabled usually imposes a small performance penalty, although it too is pretty small. It is only in enterprise environments, in which access patterns are far more random than in windows, and spread out all over the surface of a drive, that any real-word benefit is derived."
http://www.intelforums.net/showpost.php ... stcount=11
"Just a quick reminder for everyone on NCQ. It provides no performance benefit whatsoever in a typical desktop PC. In fact having it enabled usually imposes a small performance penalty, although it too is pretty small. It is only in enterprise environments, in which access patterns are far more random than in windows, and spread out all over the surface of a drive, that any real-word benefit is derived."
http://www.intelforums.net/showpost.php ... stcount=11
-
- Posts: 1386
- Joined: Fri Jun 18, 2004 6:53 pm
You do not need AHCI for SSD to work. They will work just fine in IDE mode. However, SSD drives, at least intel drives perform almost twice as fast at small read/writes with AHCI enabled compared to IDE mode. You just need decent controller, for example early nVidia controllers had problems with AHCI enabled.
-
- Posts: 404
- Joined: Sun Jul 04, 2004 9:18 pm
- Location: Illinois, USA
My long-in-the-tooth motherboard does not work correctly in AHCI/NCQ mode with the new Samsung F2 HDD, so I had to disable it. My X25-M SSD did not slow noticeably, although benchmarks indicate 240 MB/s vs the 260 quoted above.
The SSD is so much faster than an HDD as a system disk (about 8x for disk pigs like Acrobat Pro), that the further boost of NCQ isn't very important, and in my system, not noticeable.
The SSD is so much faster than an HDD as a system disk (about 8x for disk pigs like Acrobat Pro), that the further boost of NCQ isn't very important, and in my system, not noticeable.
Interesting the 4K Read and Write specs are better with AHCI off. The only significant advantage with AHCI in that test is Read 4K-64Thrd, which I know nothing about. I am interested in typical desktop usage.swivelguy2 wrote:That was written 2 years ago.
Here's some X25-M benchmarks, first with AHCI off, then with it on:
When you have many apps/threads hitting the disk randomly at the same time (which is why you got an SSD in the first place), performance is much better with AHCI on.
Are you sure about that? I think it was in Anand's review that TRIM was for IDE/AHCI but would not work for RAID until they do some kind of fix. So, I don't know why you say TRIM requires AHCI. Do you have a source of that information?Blue_Sky wrote:TRIM requires AHCI. Unless you have a SSD that will never have TRIM support, you want AHCI.
-
- Posts: 404
- Joined: Sun Jul 04, 2004 9:18 pm
- Location: Illinois, USA
NCQ makes handling multiple simultaneous requests more efficient, but adds a little bit of overhead in general.JVM wrote:Interesting the 4K Read and Write specs are better with AHCI off. The only significant advantage with AHCI in that test is Read 4K-64Thrd, which I know nothing about. I am interested in typical desktop usage.
In the large sequential test (the first one), NCQ helps a little bit, because it makes the drive "read ahead" a little more efficiently. In the small read/write test, the additional overhead of NCQ shows up. There's only 1 read or write request active at a time, so the benefits of NCQ for simultaneous stuff don't apply here. With 64 threads all making 4k read or write requests at once, NCQ really shines, providing a huge boost in performance.
Real desktop usage would be described by a combination of all 3 of these behaviors. The case of 64 threads making tiny little reads all over the disk is a little extreme, so it's not like you'll ever see that seven-fold performance increase by turning on ACHI in real life. A more realistic scenario would be something like 2-5 threads hitting the disk simultaneously, with a mix of small random and large sequential reads and writes.