" AMUG has posted a review of the NORCO-4629 4-Port PCI Express SATA II host adapter. This 8x PCIe card can support 4 direct connect SATA external hard drives or twenty drives when using up to four SATA PM enclosures. The NORCO-4629 utilizes a serial to parallel PCI bridge interface that allows the Silicon Image SiI-3124 PCI/PCI-X chip to function with a PCIe bus. The card supports deep sleep and hot swap but is not bootable. Using twenty hard drives the NORCO-4629 provided read performance of 770MB/sec. The article contains performance results for 5, 10, 15, 16, 20 drive single card configurations and 16, 20 and 24 drive setups using dual 4629 cards. The Norco-4629 retails for $189 which is a great value for the performance it provides.
Have a great weekend,
Michael Bean
TEAM AMUG "

" I am using a freshly bought Highpoint RocketRaid 1742 card in my FW800 (single) 1GHz MDD. I have two internal Samsung SP2004C drives connected to it. During the initial testing with AJA Kona System Test, I got pretty nice throughput (around 100MB/sec. read AND write rate).

But: the transfer rate will drop randomly to around 6MB/sec. At the sime time the harddisks will start sounding like boiling water, as if they are copying four files within one disks at the same time.
(If you leave Activity Monitor open do you see any increases in CPU usage, etc. when the drive performance nosedives? Any related info in any of the logs? Almost reminds me of Thermal Recalibration but I didn't think that was an issue for modern drives.-Mike)

I stopped Spotlight from indexing the Array, force-jumpered the harddisks to 1.5GB/s or 3GB/s - nothing helps :(

I have not found any correlation to the sudden and random drops of throughput.

If you get any news, I'd be VERY happy not to be forced to return the card.
Regards, Klaus"

Last week there was another G4 tower/Rocketraid 1742 card owner that mentioned very poor performance with his ext. Seagate 750GB drives in the May 8th, 2007 news page. Are you running OS X 10.4.9? (just curious, as I've had several recent reports on SATA card/drive problems from 10.4.9 users lately - different mac towers, different SATA cards and drives but just curious as to the OS version you're using.) He later wrote:

" I am writing to you only because of that post from May 8th 2007. In the meantime Highpoint support has answered my several mails and promised to report the issue to engineering, but also said that it was successfully tested on older Towers and PCI-X G5s. And they suggested contacting Samsung support :/

I just managed to make a full Test without any issues, but the next run was "successful": There are no changes in CPU usage, and system.log and console.log show no changes while the test is running :(

On the german forum I used (www.apfeltalk.de) someone else reported issues with the RR1742 and internal Seagate drives in a MDD.
If you get any good news I'd be really happy to hear them.
Best Regards, Klaus"

If anyone has any suggestions, let me know.
A later reply today:

"While I don't have a RR1742 card, I do have a RR2224 and RR2322 card, and when I first configured my array I changed the default cache policy from "write-back" to "write-through", preferring the 'safer' mode since I'm a video editor.
This caused the write performance to plummet down to less throughput than a single FW400 drive performs, and it took a while of 'cold sweating' before I tried changing the write cache back to it's default setting, and that did the trick!

Because this behavior happened on BOTH of my Highpoint RR cards on different Mac's, I offer it as a possible fix to the individuals who've written to you re: very poor performance. Thanks for a great website!
Best Regards,
-Kevin "

I wrote to Klaus to ask about this but from his "around 100MB/sec" read/write Kona benchmark comment I suspect that wasn't his problem. Here's his reply:

"Mike, I have to say, your site and what you are doing with it is incredible. Really, keep up the good work! But: I saw the box for "Cache Policy" and no cache policy was specified at all. After consulting even the manual for Windows/BIOS based management, I found that I can only set a cache policy when initializing a RAID 5 Array. I have initialized a RAID 0 Array, since i have only 2 Disks (and only two internal connectors).

80% of the time I get performance between 90MB/s and 100MB/s, so the poor performance Kevin is writing about is not my problem, rather the random drops to 6MB/s.
Best Regards, Klaus"

Thanks, I doubted that was your problem but wanted to confirm it. From your comments it almost sounds like the older drive's thermal recalibration, but I don't think that applies to modern drives anymore. And if it was some sort of flushing of a write cache even that should not cause the kind of thing you're seeing.
Update: On June 11th, Klaus sent a follow-up mail with notes that using a different hard drive (not the Samsung) seems to have solved the problem:

"Re: RocketRaid 1742 in MDD Tower
Hey Mike, I don't know if the topic is still interesting to you and your readers, but I thought I'd send you a follow-up:
I took a deep breath, got out my credit card and bought two of those new Seagate 7200.10 320GB 16MB cache SATA drives. I have run AJA Kona System test on them for many many times now and everything appears to be fine! I get around 97 MB/s Write and 84 MB/s Read, which is a little bit slower than the 98 MB/s Write AND Read with the Samsung drives I used before, but the random drops of throughput are completely gone!

So it seems that the RR1742 is incompatible with at least the Samsung SP2004C drives I had before. Or the drives were faulty, I have no other card or system at hand to confirm this.

My budget is pretty much maxed out (I'm a student) with the new HDs and I don't intend to do further testing with the Samsung drives (I'm a student, the time all this has cost me...)
But I have a happy Mac now again ;)
Regards, Klaus "

Hope the drive swap solves it. What firmware version is your 7200.10 drives? (last I heard there was still lower than expected random performance with them, but maybe they've revised the firmware again)

DVD+R/RW + DVD-R/RW Drives:
• IDE LG (HL-DT-ST) GSA-H42N in G4/AGP (OS X 10.4.9)
  (Illustrated Optical drive install guide here covers G4 towers up to the Digital Audio Model, Beige G3 MT and B&W G3 Macs)
• IDE NEC Optiarc AD-7170A in Mac Pro (OS X 10.4.x)
  (noted using patchburn but typically for OS X 10.4.x that's not actually needed.)
• IDE Pioneer DVR-200 (retail boxed DVR-111C) in Mac Pro (OS X 10.4.9)
  (reported as "Apple Shipped/Supported")
  
  Hard Drives:
• IDE (SerialATA onboard) Seagate Momentus 80GB (ST980825AS) in Intel-CPU Mini (OS X 10.4.9)
• IDE SerialATA PCI Card (OWC) connected Western Digital Raptor 150GB in (CPU upgraded Digital Audio) G4/AGP (OS X 10.4.9)
  (noted audio stops playing for 1-2 seconds during disk activity)
• IDE SerialATA PCI card (RocketRaid 1742) connected Samsung SP2004C 200GB (x2) in (MDD) G4 single CPU DDR (OS X 10.4.9)
  (noted periodic drops in drive performance. See copy of his comments in the news above. Last week there was a G4 Tower/Rocketraid 1742 card owner that mentioned very poor performance with his ext. Seagate 750GB drives in the May 8th, 2007 news page)

You can find full owner reports (latest shown first) by searching the database by drive/brand/interface/mac models (the latest reports are shown first in searches). The database includes reports on DVD +/-R, CDRWs, Combo DVD-CDRW, DVD-ROM, Hard Drives and Removables (tape drive, ORB, ZIP, MO drives, NAS, CF/Smartmedia readers, etc.) in all interface types (IDE, IDE RAID cards, SATA, Firewire, SCSI, USB, adapters).
If you've added a IDE, SATA, Firewire, USB, or SCSI, hard drive, Optical drive/burner, NAS (including Airport Extreme base connected USB drive), tape drive, etc., send me the details (drive info/mac model/OS used, etc. - *Please take a minute to test for booting from CD with Optical drives*. Thanks.)

• MacBook, MacBook Pro: Noises from the optical drive (Apple)
• Xsan: Client computer stops responding after moving item(s) to Trash (Apple)
• iChat: Some Buddy List icons don't look like they should (Apple)
• Audio in Rich Media - Telling Stories with Sound (Apple)
• Music, video phones may ride iPhone wave
• Intensity 1.4 (adds support for Blackmagic Intensity Pro card with HDMI)
• TurretWars 1.2 (game)
• FCC approves the iPhone
• Apple, Amazon may hold future of DRM-free music
• Top 10 HDTV Myths: Fact vs. Fiction, Hi-Def Style
• 5 FireFox Tricks You Probably Didn't Know About
• Dog beats Man at Wii Boxing
• PhysX game shows promise (CellFactor: Revolution)
• Modified Xbox 360 Consoles Now Banned From Xbox Live
• Game turns moviegoers into human joysticks
• AMD's new Puma stalks Intel's Centrino
• Eidos resurrecting Deus Ex?