Highpoint 7102

[AidenShaw]

SSD7110 is a great if you need the SAS/SATA ....

*WARNING* SSD 7110 defaults to an unsafe configuration *WARNING*

http://highpoint-tech.com/USA_new/series-ssd7110-resource.htm

Page 13 of the user guide:

"Write-back" cache policy without persistent backup is a disaster waiting to happen. All of the RAID controllers that I buy have a lithium ion battery or supercap, and flash memory big enough to hold all of cache and any necessary controller metadata.

In the event of a power failure, the battery or supercap powers the card long enough to copy the cache to the flash memory. On reboot, the dirty data in the cache is written to the drives - avoiding potentially catastrophic corruption. (Some older or cheaper controllers use the battery to maintain the cache memory - which means that you have to get the system back up within the small number of days before the battery drains.)

An interesting question for HighPoint would be "is the cache memory is preserved across a reboot", so that catastrophic data loss won't happen because of a crash or reboot. (A reboot will send an "initialize" command to the controller - and the controller would need to understand that there is dirty data in the cache and not initialize the cache.)

Note that a UPS will protect you from a brief mains failure, but won't protect you from an accidental cord disconnect (or the cat walking across a power strip and stepping on "off").

 

[wjesse]

Did you type it instead of copy and paste it?
Give me the output of "diskutil list > diskutillist.txt".

I made a mistake in the ioreg command. It should be "ioreg -lfiw0 >ioregall.txt".

I copied and pasted it, not worked.

I upload diskutillist.txt and ioregall.txt as attachments.

 

[joevt]

I copied and pasted it, not worked.

I upload diskutillist.txt and ioregall.txt as attachments.

I don't see a problem in your diskutillist.txt file. What is the output of this command?

diskutil list | sed -nE '/.*(disk[0-9]+s[0-9]+)$/s//printf "\1\\n"; diskutil info \/dev\/\1 | grep UUID;/p'

If it doesn't fail, can you execute each line of that output manually?

Was the RS7101Controller in your ioreg output attached to IOResources? If so, then IOResources would be a parent node that I asked to be included (just so we get the full hierarchy). RS7101Controller is a RSNVMeController which is a IOService and it is using IOResourceMatch so it must be a child of IOResources as expected. RSNVMeBlockDevice is a IOBlockStorageDevice representing the entire RAID. The IOMedia (nodes for entire disk and individual partitions), and IOGUIDPartitionScheme are all there. The IOMedia for disk10s2 has the GUID of boot volume as specified in your "nvram -p" output.

One thing to learn from the output is that the IOUserClientClass "RSNVMeControllerClient" attached to the RS7101Controller shows that "raidman-httpsd" is the name of a daemon (the name ends with a "d") that talks to the controller. http probably means it's web based. raidman is probably short for RAID Manager. You already know that there's a Web based RAID Manager but now we know that it talks to the RS7101Controller using a UserClient interface from a process named "raidman-httpsd".

Anyway, none of this gives a further clue about the EFI side of booting this RAID. You would need rEFInd for that.

To install rEFInd:
You can use Disk Utility.app to create a small FAT partition (200 MB), then install rEFInd to it and the EFI Shell.

Download the latest version of rEFInd from:
https://sourceforge.net/projects/refind/files/latest/download

Start at step 2. The latest version is 0.11.4 instead of 0.11.3 so don't forget to change that. Also change BOOTMGR to the name of the partition you created for rEFInd.
https://apple.stackexchange.com/questions/338574/bless-error-when-installing-refind

You need to down load an EFI Shell and place it in the BOOTMGR/EFI/refind/tools_x64 folder as ShellX64.efi
I'm not sure which EFI Shell you should get.
5.1 Obtaining UEFI Shell

Make sure the config file for rEFInd has "shell" listed.
Adjusting the Global Configuration

When you boot into rEFInd, you should be able to select the EFI Shell and start it.

Type the following command to get a list of text modes

mode

Choose the one with the most vertical lines and change the current mode to that one

mode 100 60

Then you need to find the rEFInd file system, by typing commands like:

FS0:
ls

Increment the number each time (FS1:, FS2:, ..., ) until you find the correct one. It may help to place a uniquely named folder on the volume containing rEFInd to make this easier.

Then you can make a directory:

mkdir shelloutput
cd shelloutput

Then you can output the results of commands to the shelloutput folder:

help > help.txt
help * > helpall.txt
ver > ver.txt
pci > pci.txt
devices > devices.txt
set > set.txt
mode > mode.txt
memmap > memmap.txt
map -v > map_v.txt
dmpstore -all > dmpstore_all.txt
dh > dh.txt
dh -d > dh_d.txt
devtree > devtree.txt
devtree -d > devtree_d.txt
drivers > drivers.txt
bcfg boot dump -v > bcfg_boot_dump_v.txt
bcfg driver dump -v > bcfg_driver_dump_v.txt
alias > alias.txt
vol > vol.txt

Get a list of pci devices:

pci

For the devices we care about, dump its register information. For example, the upstream port of the PEX8747 in slot 2 is at bus 10 (0x0A), device 8, function 0 so use this command:

pci 0A 08 00 -i > pci_A_8_0.txt

Do that for all the slot 2 devices using the first 3 decimal numbers of the pcidebug property in the ioregslot2.txt file (convert the numbers to hex).

Type the following to get back to the rEFInd UI:

exit

 

[startergo]

I don't see a problem in your diskutillist.txt file. What is the output of this command?

diskutil list | sed -nE '/.*(disk[0-9]+s[0-9]+)$/s//printf "\1\\n"; diskutil info \/dev\/\1 | grep UUID;/p'

If it doesn't fail, can you execute each line of that output manually?

Was the RS7101Controller in your ioreg output attached to IOResources? If so, then IOResources would be a parent node that I asked to be included (just so we get the full hierarchy). RS7101Controller is a RSNVMeController which is a IOService and it is using IOResourceMatch so it must be a child of IOResources as expected. RSNVMeBlockDevice is a IOBlockStorageDevice representing the entire RAID. The IOMedia (nodes for entire disk and individual partitions), and IOGUIDPartitionScheme are all there. The IOMedia for disk10s2 has the GUID of boot volume as specified in your "nvram -p" output.

One thing to learn from the output is that the IOUserClientClass "RSNVMeControllerClient" attached to the RS7101Controller shows that "raidman-httpsd" is the name of a daemon (the name ends with a "d") that talks to the controller. http probably means it's web based. raidman is probably short for RAID Manager. You already know that there's a Web based RAID Manager but now we know that it talks to the RS7101Controller using a UserClient interface from a process named "raidman-httpsd".

Anyway, none of this gives a further clue about the EFI side of booting this RAID. You would need rEFInd for that.

To install rEFInd:
You can use Disk Utility.app to create a small FAT partition (200 MB), then install rEFInd to it and the EFI Shell.

Download the latest version of rEFInd from:
https://sourceforge.net/projects/refind/files/latest/download

Start at step 2. The latest version is 0.11.4 instead of 0.11.3 so don't forget to change that. Also change BOOTMGR to the name of the partition you created for rEFInd.
https://apple.stackexchange.com/questions/338574/bless-error-when-installing-refind

You need to down load an EFI Shell and place it in the BOOTMGR/EFI/refind/tools_x64 folder as ShellX64.efi
I'm not sure which EFI Shell you should get.
5.1 Obtaining UEFI Shell

Make sure the config file for rEFInd has "shell" listed.
Adjusting the Global Configuration

When you boot into rEFInd, you should be able to select the EFI Shell and start it.

Type the following command to get a list of text modes

mode

Choose the one with the most vertical lines and change the current mode to that one

mode 100 60

Then you need to find the rEFInd file system, by typing commands like:

FS0:
ls

Increment the number each time (FS1:, FS2:, ..., ) until you find the correct one. It may help to place a uniquely named folder on the volume containing rEFInd to make this easier.

Then you can make a directory:

mkdir shelloutput
cd shelloutput

Then you can output the results of commands to the shelloutput folder:

help > help.txt
help * > helpall.txt
ver > ver.txt
pci > pci.txt
devices > devices.txt
set > set.txt
mode > mode.txt
memmap > memmap.txt
map -v > map_v.txt
dmpstore -all > dmpstore_all.txt
dh > dh.txt
dh -d > dh_d.txt
devtree > devtree.txt
devtree -d > devtree_d.txt
drivers > drivers.txt
bcfg boot dump -v > bcfg_boot_dump_v.txt
bcfg driver dump -v > bcfg_driver_dump_v.txt
alias > alias.txt
vol > vol.txt

Get a list of pci devices:

pci

For the devices we care about, dump its register information. For example, the upstream port of the PEX8747 in slot 2 is at bus 10 (0x0A), device 8, function 0 so use this command:

pci 0A 08 00 -i > pci_A_8_0.txt

Do that for all the slot 2 devices using the first 3 decimal numbers of the pcidebug property in the ioregslot2.txt file (convert the numbers to hex).

Type the following to get back to the rEFInd UI:

exit

I like this:
Add a label for this new startup option. Without a label the Startup Manager will show EFI Bootwhich may be confusing if you have other partitions with an /EFI/BOOT/bootx64.efi file.

cd /Volumes/BOOTMGR/EFI
bless --folder BOOT --label rEFInd

I was looking for this for a long time

 

[wjesse]

What is the output of this command?
If it doesn't fail, can you execute each line of that output manually?

Was the RS7101Controller in your ioreg output attached to IOResources?

Anyway, none of this gives a further clue about the EFI side of booting this RAID. You would need rEFInd for that.

To install rEFInd:
You can use Disk Utility.app to create a small FAT partition (200 MB), then install rEFInd to it and the EFI Shell.

I think the command failed, the out put is:
sed: illegal option -- /
usage: sed script [-Ealn] [-i extension] [file ...]
sed [-Ealn] [-i extension] [-e script] ... [-f script_file] ... [file ...]

and yes, I finally find out IOResources is a parent node. there is too much information in ioregall.txt, it is difficult for me to understand.

I will try to install rEFInd.(seems difficult too)

 

[joevt]

I think the command failed, the out put is:
sed: illegal option -- /
usage: sed script [-Ealn] [-i extension] [file ...]
sed [-Ealn] [-i extension] [-e script] ... [-f script_file] ... [file ...]

I need a screenshot of what you typed. Is there a space between -nE and ' ?

Try this one:

diskutil list | sed -nE '/.*(disk[0-9]+s[0-9]+)$/s//echo \1/p'

 

[startergo]

I need a screenshot of what you typed. Is there a space between -nE and ' ?

Try this one:

diskutil list | sed -nE '/.*(disk[0-9]+s[0-9]+)$/s//echo \1/p'

On my Maverics Vmware machine

Maverickss-Mac:~ mavericks$ diskutil list | sed -nE '/.*(disk[0-9]+s[0-9]+)$/s//echo \1/p'
echo disk0s1
echo disk0s2
Maverickss-Mac:~ mavericks$ diskutil list | sed -nE '/.*(disk[0-9]+s[0-9]+)$/s// \1/p'
 disk0s1
 disk0s2

[doublepost=1558807547][/doublepost]

Maverickss-Mac:~ mavericks$ diskutil list | sed -nE '/.*(disk[0-9]+s[0-9]+)$/s//printf "\1\\n"; diskutil info \/dev\/\1 | grep UUID;/p'
printf "disk0s1\n"; diskutil info /dev/disk0s1 | grep UUID;
printf "disk0s2\n"; diskutil info /dev/disk0s2 | grep UUID;

 

[joevt]

On my Maverics Vmware machine

startergo, you can try the eval command from #91. For some reason it's not working for wjesse - something in the sed command.

 

[wjesse]

I need a screenshot of what you typed. Is there a space between -nE and ' ?
Try this one:

diskutil list | sed -nE '/.*(disk[0-9]+s[0-9]+)$/s//echo \1/p'

It is strange that I copy and paste the previous command to run again(no any difference from before), it runs. Anyway, I get the UUIDs output.

It seems the Volume UUID:0E239BC6-F960-3107-89CF-1C97F78BB46B in disk0s1 to disk5s1 and disk10s1 are the same.
What that means? can that points out the reason why my cMP can't find Windows boot drive to boot WIN10, when SSD7102 is plugged into PCIe slot2?
I usually boot MacOS from disk2(in first or second 3.5-inch sata drive bay), my Windows10 bootcamp drive is disk3(in first optical drive bay ), but they become non bootable when SSD7102 is plugged into PCIe slot3 or slot4.

 

[joevt]

It is strange that I copy and paste the previous command to run again(no any difference from before), it runs. Anyway, I get the UUIDs output.

Maybe it was a problem with Japanese language differences in the OS? I don't know.

It seems the Volume UUID:0E239BC6-F960-3107-89CF-1C97F78BB46B in disk0s1 to disk5s1 and disk10s1 are the same.
What that means?

I don't think it means anything. I have six EFI partitions and only one of them have that UUID. The rest do not have a UUID. I executed the following commands to look at the FAT file system boot sector of all my EFI partitions:

diskutil list | grep " EFI " | sed -nE '/.*(disk[0-9]+s[0-9]+)$/s//dd if=\/dev\/\1 count=1 > \/tmp\/\1.bin 2> \/dev\/null; file \/tmp\/\1.bin;/p' > /tmp/script.sh
chmod +x /tmp/script.sh
sudo /tmp/script.sh

The output looks like this:

/tmp/disk0s1.bin: DOS/MBR boot sector, code offset 0x58+2, OEM-ID "BSD  4.4", sectors/track 32, heads 16, hidden sectors 40, sectors 409600 (volumes > 32 MB), FAT (32 bit), sectors/FAT 3151, serial number 0x0, label: "EFI        "
/tmp/disk1s1.bin: DOS/MBR boot sector, code offset 0x58+2, OEM-ID "BSD  4.4", sectors/track 32, heads 16, hidden sectors 40, sectors 409600 (volumes > 32 MB), FAT (32 bit), sectors/FAT 3151, serial number 0x0, label: "EFI        "
/tmp/disk2s1.bin: DOS/MBR boot sector, code offset 0x58+2, OEM-ID "BSD  4.4", sectors/track 32, heads 16,                    sectors 409600 (volumes > 32 MB), FAT (32 bit), sectors/FAT 3151, serial number 0x67e317ed, label: "EFI        "
/tmp/disk3s1.bin: DOS/MBR boot sector, code offset 0x58+2, OEM-ID "BSD  4.4", sectors/track 32, heads 16, hidden sectors 40, sectors 409600 (volumes > 32 MB), FAT (32 bit), sectors/FAT 3151, serial number 0x0, label: "EFI        "
/tmp/disk4s1.bin: DOS/MBR boot sector, code offset 0x58+2, OEM-ID "BSD  4.4", sectors/track 32, heads 16, hidden sectors 40, sectors 409600 (volumes > 32 MB), FAT (32 bit), sectors/FAT 3151, serial number 0x0, label: "EFI        "
/tmp/disk6s1.bin: DOS/MBR boot sector, code offset 0x58+2, OEM-ID "BSD  4.4", sectors/track 32, heads 16, hidden sectors 40, sectors 409600 (volumes > 32 MB), FAT (32 bit), sectors/FAT 3151, serial number 0x0, label: "EFI        "

Only disk2s1 has a serial number (Volume ID) of 0x67e317ed which is stored at offset 0x43 as ed17e367.

xxd < /tmp/disk2s1.bin | grep 040:

Result:

00000040: 0000 29ed 17e3 6745 4649 2020 2020 2020  ..)...gEFI

If your EFI partitions have the same Volume ID, then I suspect the 16 byte UUID is a hash of the 4 byte Volume ID (because FAT volumes don't have a 16 byte identifier).

can that points out the reason why my cMP can't find Windows boot drive to boot WIN10, when SSD7102 is plugged into PCIe slot2?

I don't think so. Windows boots using BIOS mode on the classic Mac Pro. Apple's BIOS compatibility layer is dumb and can't handle some devices. Sometimes it's the combination of devices (using a different graphics card might work, for example). Maybe changing the slots might work. I've experienced these problems before on my MacPro3,1. I thought the MacPro5,1 BIOS mode would suck less. Now I only have disks connected to my four internal drive bays, or USB, or optical bays.

I usually boot MacOS from disk2(in first or second 3.5-inch sata drive bay), my Windows10 bootcamp drive is disk3(in first optical drive bay ), but they become non bootable when SSD7102 is plugged into PCIe slot3 or slot4.

What connector is your bootcamp drive connected to? It should probably be one of the four internal drive bay SATA connectors.

 

[wjesse]

Sometimes it's the combination of devices (using a different graphics card might work, for example). Maybe changing the slots might work.

What connector is your bootcamp drive connected to? It should probably be one of the four internal drive bay SATA connectors.

Now my bootcamp drive is connected to internal optical bay SATA connector, I used to connect to one of the four internal drive bay SATA connectors but have the same issue.
I will try to use a different graphics card, thank you for your explanation and suggestion.

 

[Mac_User 0101]

Now my bootcamp drive is connected to internal optical bay SATA connector, I used to connect to one of the four internal drive bay SATA connectors but have the same issue.
I will try to use a different graphics card, thank you for your explanation and suggestion.

Hey guys, I remember hearing about this issue a few years ago. It was happening to all the 3,1 users installing a RAID controller in PCIe to upgrade internal SATA bus. The drives installed into the optical bays became unbootable. People started figuring out that even with RAID card installed in a PCIe slot but not connected to the molex on the backplane, the issue was still present. It sounds to me like the same issue but I wasn't aware of it affecting 4,1 & 5,1 too. For all I know, this issue could be limited to certain backplane serials or certain PCIe slot(s). Try booting to the optical bay with the Highpoint card pulled and see what happens. If the drive in optical bay boots with the Highpoint card pulled, I would try it in a different slot just to see if that makes a difference. Did this setup work before you made the hardware RAID with the 7102? Perhaps the issue wasn't there when the 7102 was in JBOD mode.

 

[wjesse]

If the drive in optical bay boots with the Highpoint card pulled, I would try it in a different slot just to see if that makes a difference. Did this setup work before you made the hardware RAID with the 7102? Perhaps the issue wasn't there when the 7102 was in JBOD mode.

The Win10 drive in optical bay boots with the SSD7102 pulled, the issue was already there before I install SSD7101-A driver and made the hardware RAID.

I'm not sure are the SSD7101-A users experience the same issue, and what about the new SSD7103? You may also test it when you get the SSD7103.

 

[Mac_User 0101]

The Win10 drive in optical bay boots with the SSD7102 pulled, the issue was already there before I install SSD7101-A driver and made the hardware RAID.

Does it happen no matter what slot the SSD7102 is in? Is the problem only with Windows in optical bay? joevt recommended you try another graphics card, have you done that?

I'm not sure are the SSD7101-A users experience the same issue, and what about the new SSD7103? You may also test it when you get the SSD7103.

I could test macOS from optical bay when SSD7103 gets here. They start shipping mid June. If the problem is only with Win10 (or Windows in general) I'm afraid I can't test that. I don't own Windows or Boot Camp.

 

[joevt]

The Win10 drive in optical bay boots with the SSD7102 pulled, the issue was already there before I install SSD7101-A driver and made the hardware RAID.

The macOS driver is not going to affect the BIOS drivers. The OS has no affect at this stage of booting.
[doublepost=1558914748][/doublepost]

Does it happen no matter what slot the SSD7102 is in? Is the problem only with Windows in optical bay?

I could test macOS from optical bay when SSD7103 gets here. They start shipping mid June. If the problem is only with Win10 (or Windows in general) I'm afraid I can't test that. I don't own Windows or Boot Camp.

macOS uses EFI to boot so there will be no problem. It's the Apple BIOS compatibility layer used for Windows booting that is the problem. You could try booting Windows using EFI but I don't think that's recommended for classic Mac Pro.

Linux booting using BIOS would have the same problems as Windows booting using BIOS.

 

[Mac_User 0101]

I just received the Highpoint SSD7103 today. I only unboxed it so far. It's beautiful and has a great weight to it. I have 3 new Addlink S70 2TB to test this with. I'll take photos and test benchmarks and report back. If joevt wants ioreg and diskutillist output, I'd be happy to provide. Also, I'm posting here since it's a direct replacement for SSD7102. But if you guys think it's better try start a SSD7103 thread, I could do that.

 

[joevt]

I just received the Highpoint SSD7103 today. I only unboxed it so far. It's beautiful and has a great weight to it. I have 3 new Addlink S70 2TB to test this with. I'll take photos and test benchmarks and report back. If joevt wants ioreg and diskutillist output, I'd be happy to provide. Also, I'm posting here since it's a direct replacement for SSD7102. But if you guys think it's better try start a SSD7103 thread, I could do that.

I would like an ioreg and maybe an lscpi

 

[Mac_User 0101]

I would like an ioreg and maybe an lscpi

I can do that, but do I need Homebrew for the lscpi? You'll have to walk me through that one.

 

[startergo]

I can do that, but do I need Homebrew for the lscpi? You'll have to walk me through that one.

The easiest way is to run darwindumper with LSPCI option only:
https://bitbucket.org/blackosx/darwindumper/downloads/

 

[joevt]

The easiest way is to run darwindumper with LSPCI option only:
https://bitbucket.org/blackosx/darwindumper/downloads/

Does that version show PCIe extended capabilities and registers > 0x100?

sudo lspci -nnvvvxxxx

I would use the latest version instead from https://github.com/pciutils/pciutils . See #223

 

[startergo]

Does that version show PCIe extended capabilities and registers > 0x100?

sudo lspci -nnvvvxxxx

I would use the latest version instead from https://github.com/gittup/pciutils. See #223

it does automatically all these modes. Maybe only pciid needs to be updated:
[doublepost=1561550394][/doublepost]Or here is the brew formula:
https://github.com/gmerlino/homebrew-pciutils

 

[joevt]

it does automatically all these modes. Maybe only pciid needs to be updated:
[doublepost=1561550394][/doublepost]Or here is the brew formula:
https://github.com/gmerlino/homebrew-pciutils

That screenshot is missing the PCIe Capabilities. The -xxxx option is supposed to dump 4096 registers but I see only 256 in the screenshot. The screenshot says version 3.2.0 but the latest is 3.6.2 from https://github.com/pciutils/pciutils (I posted the wrong link in my previous post. updated.)

 

[Mac_User 0101]

My Thoughts On The Highpoint SSD7103
tested in Mac Pro 5,1​

Initial impression is that the card is very well built. The board itself is high quality and I was pleasantly surprised to see the silicon on the pair of PEX8747. I wish I had nothing but good things to say, unfortunately that's not entirely true.

I installed x3 Addlink S70 2TB and formatted a single blade to run a quick speed test with AJA:

Next, I was able to install the Mac driver for SSD7101A along with the Highpoint webGUI and it works. The webGUI provides info for the blades and ability to create and monitor RAID arrays. When creating an array, I was presented with options for RAID 0, 1, 5. If I had a fourth blade installed, I'd also be able to select RAID 10 (1+0).

I created a RAID 0 array with 3 blades using the webGUI then formatted it with Disk Utility. Disk Utility does not see the physical devices attached to the logical device. I ran a quick speed test with AJA to verify RAID 0 performance:

The speed tests seem to be pretty much on par with SSD7101A & SSD7102 performance. I then used Carbon Copy Cloner to clone my boot drive to the RAID 0 array and successfully booted High Sierra 10.13.6 formatted HFS+. The real test would be booting Mojave formatted APFS. I probably won't get around to trying that, (keep reading)...

It's saddens me to say that this card will not work out for me in the long run. As soon as I installed it and powered up my Mac Pro (5,1), I was shocked at the level of fan noise I was hearing. This is not the annoying frequency emanated by the card's fan, rather my PCI and PSU fans. Although system temps were completely fine across the board, it seems that the SMC handles the SSD7103 like this :

The card itself has a pretty loud fan but that can be rectified like user W1SS has done with the SSD7101A. I attempted troubleshooting this the best I could in a desperate attempt to get my system fans spinning closer to their default values. First, I disconnected all peripherals and removed all my PCIe cards, then used the SMC reset button on the backplane. Next, I only installed my graphics card in slot 1 and SSD7103 in slot 2 but that didn't change anything. So I tried the card with only one blade just to see if less NVMe installed would make any difference. When that didn't work either, I tried the SSD7103 in slot 3 (x4), pretty much the same SMC behavior there too. The last thing I did was remove all the blades and disconnected the SSD7103 fan, then booted while being mindful to not overheat the two PEX8747. I noticed right away that the only difference was I no longer heard the card's noisy fan. Unfortunately SMC behavior was the same, so I immediately shut down the computer. I then concluded that it's likey that the pair of PEX8747 and their combined power draw could be responsible for this. Highpoint is using custom firmware on them as well and it's probably not optimized for our machines. I'm assuming this is not an issue on a modern PC. That is all speculation so if any of you guys have a better idea of why the SSD7103 is causing this SMC behavior in cMP, please share it with me. Also, if anyone has observed similar behavior with SSD7101A/SSD7102, I'd like to know.

Unfortunately, I'm more than likely going to contact Highpoint and request a return authorization because of this. I'm an audio engineer and this is a workstation primarily for that work. My love for the classic Mac Pro is why I developed the hobby of upgrading it and a general interest in computers and hardware. Some of you may ask why I don't use Macs Fan Control app to regulate the fan speed. Simply because I would like close to stock fan noise without relying on a 3rd party app to achieve that. It's really a shame because it's a great card with great performance. If the machine wasn't in my control room, I'd keep it without a second thought.

Attached below are a few terminal outputs and LSPCI outputs right after the RAID 0 array was created using the Highpoint webGUI. If you guys would like to see, or have me test anything else before I return the card, please let me know.

 

[joevt]

My Thoughts On The Highpoint SSD7103
tested in Mac Pro 5,1​

Interesting. Looks very similar to the 7102 where I said I couldn't find an expansion rom for the pci1b4b,7102 raid controller (#80). For the 7103, the lspci output seems to indicate that the raid controller pci1103,7103 has a 64K expansion rom. The info for the 7102 did not include lspci and the presence of an expansion rom is not always indicated in ioreg output.

The 7102 used a Marvel controller with Highpoint subsystem vendor id (1103). The 7103 uses a Highpoint raid controller with same subsystem vendor id. Maybe it's a Marvel controller with modified vendor ID? I think we need a picture of the 7103 raid controller.

The raid 0 speed seems strange since you got less than twice the speed of a single blade when using three blades. I wonder if Disk Utility or Soft RAID would give better results (but they wouldn't allow booting using the HighPoint method).

The version of lspci that you're using is old and is missing stuff from later versions. Also, the method it uses to read PCIe registers doesn't work for capabilities and registers ≥ 0x100. Your lspci output is missing Capabilities > ff.

Extracting the option rom might be interesting. I haven't tried doing this before. Maybe use Linux (but I read this might not work?):
http://etherboot.org/wiki/romdumping

 

[handheldgames]

My Thoughts On The Highpoint SSD7103
tested in Mac Pro 5,1​

View attachment 845378 View attachment 845379
Initial impression is that the card is very well built. The board itself is high quality and I was pleasantly surprised to see the silicon on the pair of PEX8747. I wish I had nothing but good things to say, unfortunately that's not entirely true.

I installed x3 Addlink S70 2TB and formatted a single blade to run a quick speed test with AJA:
View attachment 845385
Next, I was able to install the Mac driver for SSD7101A along with the Highpoint webGUI and it works. The webGUI provides info for the blades and ability to create and monitor RAID arrays. When creating an array, I was presented with options for RAID 0, 1, 5. If I had a fourth blade installed, I'd also be able to select RAID 10 (1+0).
View attachment 845386
View attachment 845387
View attachment 845388
I created a RAID 0 array with 3 blades using the webGUI then formatted it with Disk Utility. Disk Utility does not see the physical devices attached to the logical device. I ran a quick speed test with AJA to verify RAID 0 performance:
View attachment 845389
The speed tests seem to be pretty much on par with SSD7101A & SSD7102 performance. I then used Carbon Copy Cloner to clone my boot drive to the RAID 0 array and successfully booted High Sierra 10.13.6 formatted HFS+. The real test would be booting Mojave formatted APFS. I probably won't get around to trying that, (keep reading)...

It's saddens me to say that this card will not work out for me in the long run. As soon as I installed it and powered up my Mac Pro (5,1), I was shocked at the level of fan noise I was hearing. This is not the annoying frequency emanated by the card's fan, rather my PCI and PSU fans. Although system temps were completely fine across the board, it seems that the SMC handles the SSD7103 like this :
View attachment 845390
The card itself has a pretty loud fan but that can be rectified like user W1SS has done with the SSD7101A. I attempted troubleshooting this the best I could in a desperate attempt to get my system fans spinning closer to their default values. First, I disconnected all peripherals and removed all my PCIe cards, then used the SMC reset button on the backplane. Next, I only installed my graphics card in slot 1 and SSD7103 in slot 2 but that didn't change anything. So I tried the card with only one blade just to see if less NVMe installed would make any difference. When that didn't work either, I tried the SSD7103 in slot 3 (x4), pretty much the same SMC behavior there too. The last thing I did was remove all the blades and disconnected the SSD7103 fan, then booted while being mindful to not overheat the two PEX8747. I noticed right away that the only difference was I no longer heard the card's noisy fan. Unfortunately SMC behavior was the same, so I immediately shut down the computer. I then concluded that it's likey that the pair of PEX8747 and their combined power draw could be responsible for this. Highpoint is using custom firmware on them as well and it's probably not optimized for our machines. I'm assuming this is not an issue on a modern PC. That is all speculation so if any of you guys have a better idea of why the SSD7103 is causing this SMC behavior in cMP, please share it with me. Also, if anyone has observed similar behavior with SSD7101A/SSD7102, I'd like to know.

Unfortunately, I'm more than likely going to contact Highpoint and request a return authorization because of this. I'm an audio engineer and this is a workstation primarily for that work. My love for the classic Mac Pro is why I developed the hobby of upgrading it and a general interest in computers and hardware. Some of you may ask why I don't use Macs Fan Control app to regulate the fan speed. Simply because I would like close to stock fan noise without relying on a 3rd party app to achieve that. It's really a shame because it's a great card with great performance. If the machine wasn't in my control room, I'd keep it without a second thought.

Attached below are a few terminal outputs and LSPCI outputs right after the RAID 0 array was created using the Highpoint webGUI. If you guys would like to see, or have me test anything else before I return the card, please let me know.

The speeds look slow for a 3 disk array. With a MacOS raid 0 across 3 disks, you should be seeing close to 6000 MB/s reads/writes.

FWIW... In my testing of the Highpoint SSD7101a with the Highpoint drivers, across the board bandwidth from 4k to Large files was greatly inferior to the performance provided by High Sierra & Mojave's out of the box drivers. For this reason alone, I have stayed away from Highpoint's drivers that are clearly optimized for a slow thunderbolt connection VS a direct PCIe 2.0 x16 connection.