Too wordy to put into Twitter.

Basically I just installed a 160GB Intel SSD 320 into my 2008 Mac Pro. Performance…  I really couldn’t tell how it was after enabling TRIM, but I felt like something was off after doing so.

By the way, for those of you with TRIM-supporting SSDs, you can bypass Apple’s arbitrary restriction on OS TRIM support by following the instructions HERE.

Using this, and later finding this page regarding benchmarks based on both the TRIM setting and the volume mounting option ‘noatime’, I decided to try the benchmarks myself since the settings were of interest to me.

Using the same settings in Postmark, namely:

set buffering false
 set size 500 100000
 set read 4096
 set write 4096
 set number 10000
 set transactions 20000

I ran the benchmark with the stock configuration, with TRIM but not noatime, and with noatime but not TRIM, and compiled everything.

The results are as follows:

SSD +TRIM -noatime
Time:
    17 seconds total
    12 seconds of transactions (1666 per second)
Data:
    557.87 megabytes read (32.82 megabytes per second)
    1165.62 megabytes written (68.57 megabytes per second)

SSD -TRIM -noatime
Time:
    16 seconds total
    11 seconds of transactions (1818 per second)
Data:
    557.87 megabytes read (34.87 megabytes per second)
    1165.62 megabytes written (72.85 megabytes per second)

SSD -TRIM +noatime
Time:
    16 seconds total
    10 seconds of transactions (2000 per second)
Data:
    557.87 megabytes read (34.87 megabytes per second)
    1165.62 megabytes written (72.85 megabytes per second)

And finally, in graph form:

In conclusion, it seems that (with Intel SSDs, at least) for best performance you’ll probably want to keep both TRIM and noatime turned off, at least until Apple gets its shit together and makes TRIM work better.

I saw TRIM lower throughput by around 6%, and transactions by around 9%. Noatime on its own was responsible for an additional gain of almost 10% in transactions per second!

I’m not sure if this affects stock Apple SSDs, but it might be worth trying to knock out TRIM and see how that affects performance, and I’m joining the growing crowd that recommends that noatime be enabled for any SSD – not just for longevity concerns (which may or may not be baseless), but because it hurts nothing, but improves the speed of operations.

Sorry, the comment form is closed at this time.

Nick’s Monster Budget Tracker (taxes/shipping/labour included)

1. 1999 Subaru Impreza L Sedan Feb 12, 2012 $4500
2. Valeo Ultimate wiper blades Feb 16, 2012 $50
3. Misc parts: Armrest extension, cargo net, intermittent wiper stalk, JDM tail lights, JDM clear corners, trunk lamp Feb 24, 2012 $517
4. JDM Impreza WRX STi RA Version 6 DCCD gauge cluster Apr 18, 2012 $280
5. Subaru Impreza JDM front grille Apr 28, 2012 $159
6. Perrin Turbo Inlet Hose May 02, 2012 $255
7. DCCDPro DCCD controller w/ OEM controls May 05, 2012 627$
8. Defi Racer (white) Boost Gauge May 29, 2012 $205
9. Tomei Timing Belt Guide May 29, 2012 $55
10. OEM 16" Wheels (sold) Jun 20, 2012 -$300
11. Hayward and Scott 2.5" Axle-back Muffler May 30, 2012 $345
12. Subaru EJ22 Engine (sold) June 12, 2012 -$500
13. JDM Impreza WRX STi Type-RA Aluminum hood June 17, 2012 $400
14. Prodrive GC-07C Wheels - 17x7.5" (used) June 18, 2012 $1250
15. Centering rings + lug nuts for wheels June 20, 2012 $120
16. 1999 Impreza L rear axles (sold) June 20, 2012 -$50
17. Deposit for full JDM Impreza WRX STi Version 8 drivetrain May 24, 2012 $9500
Total $174518

(via Kelley Blue Book as of May 30, 2012) 2004 Subaru WRX STi          -         $18,326 - 21,126
© 2010 N-Bomb.com Suffusion theme by Sayontan Sinha