PDA

View Full Version : My ATA/66 7300 Deskstar is half as fast as my SCSI?



Ernest House
12-20-2001, 02:19 PM
I examined the xlr8yourmac tests back when I purchased by drives. I have an 18GB IBM Ultrastar (Ultra 2 SCSI) on an Initio U2W card. I also have an 18GB IBM Deskstar 75GXP as the only IDE drive. System is a G4 500DP.

I was about to bag the SCSI because it adds a minute to every startup and won't wake up from a disk spin down. But I ran a test using Final Cut Pro where I export to a DV .MOV file. This operation coalesces all the tracks into a single DV stream (in this case a 2.5GB file). Exporting to the SCSI took 7 minutes. Exporting to the IDE took 15 minutes. Both drives had 12GB free. I expected the performance of the drives to be similar based on the XLR8YOURMAC data.

Should the drives have closer performance and I'm dealing with fragmentation differences or something?

TIA
Ernest

ricks
12-20-2001, 03:34 PM
Hi Ernest,
First thing I'd do is run a benchtest on your drives yourself. Find out how they are working on your system as opposed to what you expected from the XLR8 review.

Run ATTO benchtest on them. That's availiable from either ATTOtech.com in the Express Pro Tools package or from the Gurus ftp site @
ftp://ftp.macgurus.com/ftp/atto_stuff/ExpressPro-Tools%202.5/

After running a test if you find any anomolies in your results it's time to run Disk Warrior and optomize your drives and or repair the drive. Then you can decide whether your bus/drive setups are fast enough for what your doing. Neither of your drives are anywhere near peak performers today, I am surprised that the UW drive was faster than the ATA66, test and find out what you got.

Rick

Ernest House
12-20-2001, 07:13 PM
OK, so I went back to MacBench and the drives test out the same as before: The ATA/66 IDE drive exceeding the SCSI performance slightly. The results of the ATTO Benchmark are here: http://homepage.mac.com/topdown/images/SawtoothATAvsSCSI.gif

According to ATTO benchmark, the ATA drive is better across the board.

So my conclusion is neither benchmark predicts actual real world results for video applications. In fact they seem to predict the opposite. Not too surprising since real world vided applications don't work like benchmarks (unless all one does is Finder copies :-)). I suppose this is why SCSI drives are still alive and well in video workstations.

I guess I'll see if formatting does anything. Thanks.

ricks
12-20-2001, 08:43 PM
From your email I saw you were running the ATTO benchtest Transfer Size at 2mb, when you run them to 8mb you are likly to see a major drop in the indicated performance from the ATA drive.

Real world apps are likely to perform in ways that no test program can really emulate. Your test results with Final Cut Pro are just as valid an indicator of specific drive performance as any benchmark.

I always wanted a test program that would transfer a sustainable and verifiable large file size like 500mb or so to and from a RAM DISK. At least that would show the actual hardware bus/drive speed. In real life too many things are going to have an effect on it for this simplistic approach to mean much in day to day performance.

Rick

Ernest House
12-21-2001, 12:05 AM
I am an idiot. I neglected to account for the fact that the source files were on the same ATA drive that I was also writing to. I put in another ATA drive and the Export time was 5 minutes. I then ran the experiment with everything on a single but defragmented drive. The Export time was 10 minutes.

This is more consistent with the relative benchmark results for each drive. Thanks.