PB G4 SSD using 1.8" SATA to IDE

[grockk]

XBench disk test. Scores are the average of 5 runs.

 

[Dannyshing27]

I have recently installed an 32gb mSATA crucial SSD to PATA drive in my A1001 tibook following this thread. I will post some benchmarks when i get home from work!

Results 120.64
System Info
Xbench Version 1.3
System Version 10.5.8 (9L31a)
Physical RAM 1024 MB
Model PowerBook3,4
Processor PowerPC G4 @ 667 MHz
Version 7455 (Apollo) v2.1
L1 Cache 32K (instruction), 32K (data)
L2 Cache 256K @ 667 MHz
L3 Cache 1024K @ 167 MHz
Bus Frequency 134 MHz
Video Card ATY,RageM7
Drive Type M4-CT032M4SSD3 M4-CT032M4SSD3
Disk Test 120.64
Sequential 81.73
Uncached Write 84.29 51.76 MB/sec [4K blocks]
Uncached Write 74.92 42.39 MB/sec [256K blocks]
Uncached Read 66.58 19.48 MB/sec [4K blocks]
Uncached Read 114.82 57.71 MB/sec [256K blocks]
Random 230.28
Uncached Write 159.91 16.93 MB/sec [4K blocks]
Uncached Write 139.03 44.51 MB/sec [256K blocks]
Uncached Read 1796.56 12.73 MB/sec [4K blocks]
Uncached Read 296.98 55.11 MB/sec [256K blocks]

 

[Cox Orange]

@ Glockk:

I can understand, that you have ignored my post, since it was so long winded.
Well, now it is too late anyway. I was interested to see a comparision of a file copied on the IDE disc and a file copied on the SSD (internally), stopping the time and calculate the MB/s. (I know it's about access time, what all care about, but shouldn't a SSD saturate the ATA-66 BUS and show a small time advantage over IDE at transfer speeds, too?)

@ all:
To see how "reliable" XBench is, please compare the pictures in post 7 http://cubeuser.de/showthread.php?t=1456
(this thread is not by me, btw.)

 

[grockk]

@ Glockk:

I can understand, that you have ignored my post, since it was so long winded.
Well, now it is too late anyway.

Sorry for not replying but it is not too late. I plan to do the SSD portion still and give you a comparison. Problem is it is very time consuming to sit with a stop watch. Xbench is easy.

I was interested to see a comparision of a file copied on the IDE disc and a file copied on the SSD (internally), stopping the time and calculate the MB/s. (I know it's about access time, what all care about, but shouldn't a SSD saturate the ATA-66 BUS and show a small time advantage over IDE at transfer speeds, too?)

SSD on an SATA drive should saturate, but clearly there is some overhead in the adapter. I think my speeds are an incredible advance for IDE, but if this were a macbook with SATA, I'd have been disappointed.

@ all:
To see how "reliable" XBench is, please compare the pictures in post 7 http://cubeuser.de/showthread.php?t=1456
(this thread is not by me, btw.)

This is why I used averaging. I don't know of any other free PPC disk benchmarks. My standard deviations were only between 2-5 on any of the tests so I considered it at least good enough for showing the improved change.

I'll find time this weekend to do the SSD side.

 

[Cox Orange]

Sorry for not replying but it is not too late. I plan to do the SSD portion still and give you a comparison. Problem is it is very time consuming to sit with a stop watch. Xbench is easy.

I must have misunderstood you, I thought the XBench results were of your SSD and not of the IDE HDD.
Well sitting with a stopwatch does not take that long. If the throughput is 40MB/s a 14GB file should take somewhere around 6-7Minutes. Well, of cause one has to wait. OTHER WAY: If the system displays it right, you can copy the file and afterword highlight it in finder and look, when the file was created and when it was last edited (the last edited is the end time of the copying process).
If, you do not want to sit and watch, you can choose to copy the file and look what OS X says, when it is ready, be sure to be back 2 Minutes before the end. (Mac OS X is very good at predicting how long it takes.)

If you choose to not just copy a file, but use iMovie and export a DV file as "Quicktime"-DV file to the disc (so it will only has to render were you did cut something), then iMovie will say "remaining time 14Minutes". Here you can go away and get back 3 Minutes before the end. But still, Finder will most likely tell you, when you started copying and when the file was "changed" the last time.

SSD on an SATA drive should saturate, but clearly there is some overhead in the adapter. I think my speeds are an incredible advance for IDE, but if this were a macbook with SATA, I'd have been disappointed.

Yes, of course, there is overhead, but while transfer speed from a HDD varies a SSD (via adapter) should have more constant transfer rates than the native IDE-HDD.

This is why I used averaging. I don't know of any other free PPC disk benchmarks. My standard deviations were only between 2-5 on any of the tests so I considered it at least good enough for showing the improved change.

I'll find time this weekend to do the SSD side.

Again, sorry, I thought you tested the SDD and not the HDD.

Free:
Aja-Test http://mac.softpedia.com/progDownload/AJA-System-Test-Download-81585.html
Intech Demo (not sure, what the limitations are, I do not use benchmark tests, after a nerve wrecking 12hour session with XBench) http://www.macupdate.com/app/mac/15467/speedtools-utilities-pro


An end note, why I dislike benchmarks: why do I prefer copying files over using Benchmarks. There are numerous people having used benchmarks claiming they get 80-100MB/s with a SATA-PCI Card in PowerMac G4s, even the PowerPC-God Japamac (he refuses to answer questions on that issue). Me and "macuser453787" have seen, that when you copy files (I used 2GB, 5GB, 8GB, 15GB, 20B, 800GB) via a Sonnet SATA-PCI Card, you do not get the 80MB/s that all present in their benchmarks. I have 40MB/s max. This is with WD Velociraptors and 2TB and 3TB drives produced in 2012 and 2013, that should easily reach 80MB/s. (older drives get 33MB/s as worst number).

The Sonnet Card is 32bit. A Macally Card, I possess is 32bit as well and produced the same results. I bought a 64bit "PC/Windows"-Card with SIL3124 Chip and downloaded a Mac driver for it. I got 58-62MB/s with Velociraptor drives. (I know Velociraptor drives are stupid, if you compare the price of a SSD, but I got these very cheap, so I went with these). According to Benchmarks on the net I should easily get 100MB/s with Velociraptors.

This was with file copying. Using iMovie instead resulted in:
PM G4 with Sonnet 30seconds
PM G4 with SIL3124 26seconds
PM G5 2,3GHz 15seconds (stock SATA-ports).
(The G4 uses a Sonnet 1,2GHz CPU. Using a 14-15GB file (always the same file and HDDs).

 

[grockk]

I'm testing the SSD speeds now and I'll post both the HDD and SSD file transfer comparisons soon. The Xbench results are of both the HDD and SSD. The HDD transfers of 15 gb files took forever on the HDD so I figured SSD would take awhile too. We'll see.

 

[grockk]

Ok here is what I did.

On each system I made a large file using

mkfile -n 15g ~/Desktop/LargeFile

Then I did

cp ~/Desktop/LargeFile ~/Desktop/LargeFileCopy

and in a second window started a timer with

time cat

and hit cntl-c when i saw the transfer end.

I did this for HDD and SSD. Here are results.

I only did one run for each test but it shows a somewhat improvement for large file transfers. Much less improvement than the random access speeds but several minutes for these tests was pretty good. And the improvement in launch times and system responsiveness are well worth it.

I expected to have the IDE bus saturated but clearly that doesn't happen.

 

[Cox Orange]

Thank you very much! Much appreciated!

 

[grockk]

Don't underestimate the hot glue. I just had to open it up and add more to hold the drive firmly in place. Every time I carried the powerbook I could hear the drive flop around.

I used a piece of plastic bent in an L Shape under the drive with the edge glued to the frame and the drive. Otherwise you'd have to stick it to the logic board and that seemed like a bad idea.