Western Digital SSDs and Legacy Macs

Western Digital SSDs and Legacy Macs(geeklan.co.uk)

28 points by Sevan777 1 year ago | 10 comments

The Broadco- "ServerWorks" K2 chipsets that power G5 Macs SATA controllers are horrible. I have a half dozen G5 Quads and a Dual G5 Xserve. Getting compatible SSD's for them, or even HDD's back in the day has been an exercise in frustration and tedium

All these machines run SATA I (1.5Gbps). The Sata II and SATA III specifications state that it's meant to be backwards compatible and the drive should negotiate down to SATA I speeds (ala 10/100/1000 Ethernet)

For HDD's they used to have one of the jumper pins you could set to "force 1.5Gbps mode" at least on W.D drives at the time. SSD's of course however are meant to be fast and in theory do this all in software. In practice with the K2 controller though they just...tend to break!

I don't blame the SSD makers of course. Why would they test their SSD's in 20~ year old machines? (or even if they do, knowing to make appropriate concessions for PowerPC macs, which are themselves known to be extremely unreliable anyway!)

For extra fun I added an NVME drive to one of my G5 Quad machines via a PCI-E adapter. The boot process goes OpenFirmware (ROM) --> GRUB2 ---> Linux Kernel + initrams on SSD with root=/dev/nvme0n1p1 set

Even with this set up it takes five real world seconds for the K2 controller to report back the single device (of just TWO it can have plugged in!) to the linux kernel. With this driver disabled or moved to a module the kernel boot process takes approximately 3 seconds instead of 8 or 9

The K2 controllers are awful and I am glad they died quietly

Sevan777 1 year ago | |

At least you have expansion slots (PCI-X/PCIe) to add new controllers. Totally stuck with the iMac G5.

vitus 1 year ago |

If anyone's curious, https://i.imgur.com/08vtB4n.png was my initial attempt to recover the garbled screenshot at the end of the post, although I did end up throwing away the vast majority of the pixels. It's at least vaguely recognizable as an OSX screenshot?

vitus 1 year ago | |

Okay. I've fully reconstituted the image, to the best of my ability: https://i.imgur.com/8XemFj2.png

    from PIL import Image
        
    png = Image.open('screenshot.png')
    pxl = Image.frombytes('RGB', (5120,200), png.tobytes()).load()
    
    new_img = Image.new('RGB', (1280, 800))
    new_pxl = new_img.load()
    
    for block in range(80):
        for x in range(16):
            for y in range(200):
                for offset in range(0,4):
                    new_pxl[block*16 + x, 4*y + offset] = pxl[(block*4 + offset) * 16 + x, y]
    
    new_img.show()

Sevan777 1 year ago | | |

very cool :)

justinclift 1 year ago |

WD has all kinds of weird issues with their SSDs. Some pretty major, and it's really hit or miss whether any particular one will be completely problem free or be a total nightmare. :(

• https://github.com/openzfs/zfs/discussions/14793

Best to avoid them at all costs.

Sevan777 1 year ago | |

That's interesting, I have a WD Black SN850 in my MNT Reform with ZFS on it and luckily haven't run into any issues with it, though I'm not stressing it in any kind of way. WD disks and ZFS go back a long way, when the Green HDD line first showed up in the late 2000s they caused a lot of headaches since ZFS expects to be in control & have a response when issuing requests where as the Green drive's firmware would be trying to conserve power, resulting in bogus errors.

shrubble 1 year ago |

I have definitely had issues even with Intel based systems that were Sata-1 in working with Sata-2 drives; I think that being a first generation chipset it was not well supported in terms of anomalies that showed up with how the specification was implemented by different vendors.

Sevan777 1 year ago | |

I suspect you're right and the firmware changes have likely covered compatibility with SATA II and missed these edge cases with legacy SATA I controllers.

1over137 1 year ago |

https://archive.is/E9sWK

from PIL import Image png = Image.open('screenshot.png') pxl = Image.frombytes('RGB', (5120,200), png.tobytes()).load() new_img = Image.new('RGB', (1280, 800)) new_pxl = new_img.load() for block in range(80): for x in range(16): for y in range(200): for offset in range(0,4): new_pxl[block*16 + x, 4*y + offset] = pxl[(block*4 + offset) * 16 + x, y] new_img.show()