Mac SSD performance and TRIM in OSX | Hacker News Reader

Mac SSD performance and TRIM in OSX

bit-tech.net

24 pointssoundsop17 comments

Does the SSD really treat overwriting everything with zeroes as a low level format? I'm guessing "no", and that the entire test was performed on dirty disks.

that struck me too. he pretty much said early on they were doing the tests on a used computer given to them by the vendor. a box that had previously been given to other reviews to test/review. therefore, I don't see how he can truly know for sure whether or just how clean/virgin that SSD is.

Not only that, he explicitly filled the drive with zeroes using Disk Utility. He probably should have run HDDerase (or a similar utility from Samsung, if available) from a bootable USB stick to clear it instead.

As an aside, the SSD firmwares -could- detect pages filled with ones or zeros and use that knowledge to transparently do a "TRIM light", at the cost of doubling the size of the "block in use" bitmap (to track 'free', 'in use', 'all zeroes', 'all ones') and added code complexity.

For the SSD makers it probably isn't worth the effort and debugging, since TRIM support is well on its way to ubiquity. Should be an interesting experiment though, if nobody has done it yet.

There is a reason why there was no performance degradation: They ran the tests on "dirty" disk both times.

To clear an SSD, specials methods must be used: They are vendor specific but the SECURE ERASE command (ATA spec) will mostly do.

That is correct. They just wrote data over the whole damn disk. ATA SECURE ERASE as I believe directs the controller/firmware to bring the cells to a factory state, which takes a little time because the cells only need to be marked clean, not rewritten or anything like that.

The method I use on my SSD when reinstalling:

    hdparm --user-master u --security-set-pass 12345 /dev/sda
    hdparm --user-master u --security-erase 12345 /dev/sda

Actually each cell (each bit) is reset to an initial state.

It is still blazingly fast, as the erase is:

  1) Done in large chunks

  2) Done in parallel (most SSDs have multiple dies)

  3) Does not require data transfers out of the drive.

The ERASE also resets the internal data structures (dirty/etc lists).

I have a MacBook Pro from last year with the Intel X-25 SDD and I can attest, on real life (not benchmark programs) the hdd grows slower (more noticeable on Spotlight and Safari). I even have a ram disk for most scratch operations (compiling code for example) but even so, after a few months of use, the sdd needs a good wipe (hdparm does the trick).

I just bit the bullet and bought a 700 dollar SSD for my MacBook (which is over half the price I paid for the MacBook Pro itself!)

It's really hardest on the wallet for us laptop users. On the desktop, you can get by with a 40GB+ SSD and a secondary traditional disk (in my case, 2xVelociraptors in RAID0). But on a laptop, you've got to shell out the extra bucks for the largest SSDs on the market, because it's ridiculous to walk around with anything less than 200GBs in your notebook these days.

I just did the same, probably a similar disk. If you don't use the optical drive much you can always add an SSD and keep your current disk with 320 or 500GB.

I think 240GB will be enough. If not I might end up replacing the optical drive.

It's a lot, yeah, but the way I look at it is I just spent top dollar on a shiny new notebook and using a rotational disk in it is cheating myself out of a good investment. I couldn't justify not getting one any longer, even though I know that winter will bring higher capacities and lower prices.

Exactly the same reasoning. I don't use my optical drive at all (maybe twice in 2 years??) but I just don't like the idea of castrating my MBP (funny, because I had no such qualms about any of my old devices).

I got the C300 RealSSD: http://www.amazon.com/Crucial-Technology-RealSSD-Solid-CTFDD...

I had been waiting for the F240 from Corsair to come out, and as I was waiting, this one popped up on the market and it blew all the competition away.

Nothing new here. It's coming, just a matter of when and how broad.

All the reviews I've read regarding the Sandforce 1200 indicate that slowdowns are not a big problem for that controller. They don't mention it in this article, neither do they mention it in their test from February.

Just do your homework before buying a disk. It only takes an hour to check out a few reviews.

Ugh. As pointed out in the Slashdot comments, this article is not even wrong.

tl;dr SSDs that run Windows need TRIM for a vast performance boost, but macs either already have TRIM and are not advertising it or do not need TRIM because of intrinsic qualities of the OS.

... but only for the relatively ancient and poorly performing SSD that Apple ships with the MacBook Air. Whether the results generalize to other SSDs, or the reason for why there's no major performance degradation, are both open questions.

There's a big confounding factor in their Windows benchmarks: To test performance without TRIM, they used older firmwares. It's unclear how much of the performance increase is from firmware improvements unrelated to TRIM support.

Someone pointed me to the OWC Mercury Extreme Pros, which are extremely fast and maintain performance without TRIM:

http://macperformanceguide.com/SSD-RealWorld-Seasoning.html

http://www.barefeats.com/mbpp19.html

I took the plunge and am quite happy I did.

Reply on news.ycombinator.com