You're right; I certainly didn't read your opening post fully at first.
I don't have a third ATA or SATA HD attached to this machine.
Still, I had thought the Xbench stats for both drives might be illuminating -- the lack of consistency in tests that you report for the a drive with Xbench, however, casts considerable doubt on such results.
Perhaps what follows may be more helpful, if not definitively what you request:
(the HD is on the native ATA bus, the SSD is on the PCI SATA card)
making a copy of an 8.62 .mov file on the Seagate HD -- 6:20
making a copy of the same file on the Kingston SSD -- 3:10
AJA System Test (used by Macworld and Storagereview.com as part of their drive testing routine -- and perhaps a bit more realistic):
I can also note that file transfers to the SSD from another machine over gigabit ethernet result in times about 30 seconds less than to the HD (2:50 vs. 3:21). Copies from the SSD to the HD on the G4, and vice versa, were within six seconds of each other (3:34 vs. 3:40).
I am running a pATA133 RAID-0, hardware striped RAID, with 2x500GB drives. I think they are getting close to the limit of the system bus throughput.
Maybe a SATA-PCI single drive would perform about the same in the G4 MDD, but with less complexity of setting up a RAID.
Surprised the SATA-PCI SSD is not a little bit faster. There is usually some variation with these tests when you repeat them.
Thank you Myaka!
So for the ATA-HDD you get 22,68MB/s and for the SSD about 45MB/s. That is interesting, since you get results near to my results on the SATA-Bus with normal HDDs (remember I had between 33-39MB/s) and not the 80-100MB/s as elsewhere stated for SATA-Drives etc.
@ Glen Dogget:
Glen, could you copy a file for me and count the time?
You said "There is usually some variation with these tests when you repeat them." Yes, I know, but in my case it was the small values that changed from 1-2MB to 18MB/s and the big values varied from 20-33MB/s. Try Xbench several time and save the results and than at the end compare them. I also always did a restart and waited until I did not hear "crunching" from the Boot-Drives anymore and then started the tests. Maybe sometimes I did not wait long enough or the boot-drive did take different times to complete boot-up everytime.
Hey, Japamac, if your are here again, I would like you to answer in this thread! You made benchmarks with your G4 and several connection types. As my first post explains, counting the time without a benchmark tool it seems to be much slower.
Another thing I recently discovered is, that formatting a Drive to FAT-32 takes long. Transfering big files 5-7GB results in transfer rates of 1-2MB/s via SATA-PCI.
Could you please transfer a 5GB file or bigger, manually via SATA-PCI from one drive to another (both HFS+) and count the time? (Mac OS X is very near to exact when prediciting the time it will take, btw.!). Then 2nd task, could you write a files from a HFS+ HDD to a FAT-32 Drive via SATA-PCI, please?
I only have two SATA-PCI cards and currently only one PowerMac G4 assembled, so I can not do comparative tests on my own.
Again, please do not use a benchmark tool, stop the time and devide the file-size by the time elapsed.
I would be so very thankful!
Off the top of my head as it's been a few weeks, but I added a Firmtek 2 port external E-SATA card to a Quicksilver DP 1GHz (2002) and seem to recall getting 1-2 GB per minute transferring between 7500 RPM SATA drives. One was an IOMEGA 4-way enclosure while the other was a Sans Digital RAID 1 striped. I was transferring large files up to almost 2 TB.
Another experience was trying to do a Migration Assistant transfer from the internal PATA drive to one of the E-SATA drives and I got a estimated time count of almost 24 hours, this was about a 170 GB file as the largest part was my iTunes.
I had a backup of this internal drive on another E-SATA drive and think I did the HD to HD migration which is an option, and it only took a few hours.
I notice Firmtek claims these E-SATA cards offload transfers without using the computer's processing, perhaps why they go so fast. I also use these external drives for boot drives (Leopard) and have found Safari is much faster and opens more tabs as it tends to use a lot of VM spooling on the boot drive, a good reason to also leave 20% empty headroom on the drive. Otherwise I had to frequently relaunch Safari as the VM grows to 1.5GB or so.
1.5 GB or RAM helps here too as compared to the 1 GB I had before. Overall performance has been comparable to a MBP 17" (early 2008) 2GB/5000 RPM when running Leopard, except too many tabs will often require relaunching Safari in the G4 as it doesn't seem to release the VM which accumulates.
thanks for participating and sharing your experiences. So, you had 1-2GB/min., that is about 16-32MB/s?
Do not know, whether the external SATA ports are really faster, but I do not believe, that they use another firmware, than they use for the Firmtek/Sonnet cards with internal ports. (I have a Sonnet 2xinternal, and Sonnet Cards are known to have the Firmtek firmware and register as such in the system profiler. Also I tested it under both 10.4 and 10.5. There are different firmwares for 10.4 and 10.5, though both work under both OSes.)
I assume using Migration Assistant takes that long, because it is indexing and searching and defragmenting files and trying to keep the overal structure. The same is the case with Super Duper, it is very slow, when making a system backup, because it defragments all files.
Here http://forums.macrumors.com/showthread.php?p=17582354#post17582354 is another person that had a max. of 50MB/s. So I more and more doubt, that benchmark tools really say anything in the Mac world.
I would really like Japamac to answer in this thread, because he has done several benchmarks and is well respected in the apple community. If a person like him would redo the tests and stop the time, people would believe me and not just claim they are right because of Xbench or AJA. Funny these people do not even want to retest by copying manually. Of course, it would tell them that their super fast SATA-SSDs (which are promoted with 200-500MB/s) can't even bring more than an old SATA-HDD.
I would find it really honorable, if Japamac would do the test and confess that his virtual benchmark testing was wrong. This would even lift up my respect for him, instead of loosing his face, if that is his concern (if that is still the trouble for modern japanese people anyway).
You can read what SeriTek claims about my card here: http://firmtek.stores.yahoo.net/sata1ese2.html
Supports The Latest I-Type eSATA Connectors
The SeriTek/1eSE2 Host Adapter extends Serial ATA's 1.5Gbits/sec performance to the outside of the computer chassis, enabling users to take advantage of external Serial ATA enclosures such as the SeriTek/1EN2. With the latest eSATA shielded connectors, the SeriTek/1eSE2 provides the high bandwidth necessary to meet the needs of performance-hungry applications while offering hot-swap flexibility similar to that of Universal Serial Bus (USB) and FireWire.
SeriTek/1eSE2 works with all Power Macintosh and PC/Windows computers with an available PCI slot.
- Two external eSATA ports with hot-swap capability
- Compatible with all PCI-based Power PC Macintosh computers and PC/Windows computers
- Boosts overall system performance with per-port data transfer rates of up to 150MBytes/sec or 1.5Gbits/sec
- Supports Mac OS 9.XX, and OS X version 10.1.5 or later, Windows 98/98SE/ME/NT/2000/XP/2003
- Provides performance and protection: Supports native OS X RAID 0/1 (OS 9 RAID with 3rd party software), native Windows 2000/XP/2003 RAID
- User-upgradeable firmware
- 48-bit LBA support for drive sizes larger than 137GB
- Perfect for upgrading and expanding legacy and G5 Macintosh computers
- 32-bit, 33 or 66 MHz PCI bus support
- Compliant with 32-bit PCI bus version 2.2
- Bus master operation enhances multitasking during disk transfers and increases CPU efficiency: The CPU is free to handle other tasks during data transfers between the PCI Bus and system memory
- Reduced settings on storage devices (no master/slave) means easier installation and replacement
- Ideal for a variety of Power Macintosh and PC/Windows applications: Desktop publishing, Audio/Video storage and editing, graphics manipulation/publishing, photo storage and editing, servers, and gaming.
This website does a lot of performance reviews going back to legacy computers & hard drives: http://www.barefeats.com/hard43.html
You can also find more SeriTek reviews here for individual products, or search the Barefeats website, they do excellent performance tests: http://www.firmtek.com/seritek/
Yes, Sonnet also claims their cards reach SATA-3 speeds for the newest drives. Even in a G5 a SATA-3 card only works as SATA-1, so that is just marketing.
Thanks for the links. I know barefeats, but I think they also use benchmark tools, when testing drive speeds.
Can you do me a favour and take a big file (2-5GB is ok), copy it manually and stop the time and report back, please? DO NOT USE Benchmark Tools.
On your RAM point in the message before. I tested it, when copying files (not working with them like in Logic studio or Photoshop), it does not matter how much RAM you have. (espeacially when comparing 1GB ot 1,5GB). Just be sure to have 256MB for OS X only. The rest will take 128-512MB. You can test that, too, if you like. It will not change the copy speed, if you have more than 1GB (I didn't test 768MB, but even that could be enough).
Re that copy, can it be on the same drive? I just don't have a second external hooked up at the moment. Also I noticed a quirk in copying, can't remember if it was in Finder or using Carbon Copy Cloner, but it seems something gets spooled during the copy and uses Virtual Memory. For some reason if I canceled a copy and then re-did it--it would crawl, and I might of had to even reboot to get the speed back up. Not sure what caused this though, I think because when 1000's of items get listed in the initial copy it does tie up memory or processor as it must keep calculating the size/time.
Something else to check when copying huge amounts of files is to make the target drive unlisted in Spotlight or that might become active in the background.
have you tried repairing permissions and resetting PMU? If you have the install disc, start from it and run all the disk utility tests. If not (better anyway): start in single user mode and do the following:
1. type: sbin_fsck ßyf (? or ß is another character on english keyboards, it is the second key after the backshift key)
2. press return
4. type: ctrl. d
doing the copying on the same drive is ok, too. It should be faster than 40MB/s still, if it would reach SATA-HDD speeds. If it is slower, than it shows, that the PCI controller can't go over 40MB/s.