Highpoint 7101A - PCIe 3.0 SSD performance for the cMP

[Adam Chuan]

First, do the following commands in Terminal.app (Terminal.app is found in /Applications/Utilities):

svn checkout https://github.com/pciutils/pciutils/trunk pciutils
cd pciutils
make
sudo su
make install
grep /usr/local/sbin /etc/paths || echo /usr/local/sbin >> /etc/paths
update-pciids
exit

The commands do the following:

1. The first command (svn) downloads the pciutils source code to a folder called pciutils in the current folder which is usually the user folder (/Users/your_user_name or just ~ for short).
2. The second command (cd) changes the current directory to the new pciutils directory that was just created.
3. The third command (make) compiles the pciutils source code (you might need Xcode and Xcode command line utilities)
4. The fourth command (sudo su) switches the current to the super user (root)
5. The fifth command (make install) copies the pciutils files to the various folder (/usr/local/sbin)
6. The sixth command (grep) updates the /etc/paths files to add the /usr/local/sbin directory to the list of paths (if it doesn't exist in the list of paths yet). This is necessary to run the pciutils commands (lspci, setpci, and update-pciids) without entering their full path.
7. The seventh command (update-pciids) updates the pci ids database (which originates from The PCI ID Repository)
8. The eighth command (exit) exits from super user mode

After the above has completed successfully, close the Terminal window and open a new Terminal window. Run the ./pcitree.sh command and copy the text results in a reply to this thread.


You should post a screen shot of what happens if you encounter a problem.


They are basically the same cards with slightly different designs. You need to check the PCIe link width and link speed.

it's so amazing! it work!

The screenshots are my PCItree test.

[automerge]1571418629[/automerge]

@Adam Chuan Are you sure you are selecting the Raid0 3-SSD volume in the benchmarking app? Also, can you remove 1 of the SSDs (so the total number of SSDs attached to the card is 3 ... 1 for system and 2 for raid0) and benchmark a Raid0 2-SSD volume?

I will try your step after I check the PCI link Spped and link width.
thanks!

 

[joevt]

it's so amazing! it work!

The screenshots are my PCItree test

Instead of screenshots, post as text because text is easier to read and can be searched.

The output from pcitree.sh shows the link speed and link width of each PCIe device.
g1 = PCIe 1.0 = 2.5 GT/s
g2 = PCIe 2.0 = 5 GT/s
g3 = PCIe 3.0 = 8 GT/s

The problem appears to be that the slot (00:03.0) is running with lower link speed than expected. "g2x16 > g1x16" means that the slot is capable of PCIe 2.0 x16 but is currently running at only PCIe 1.0 x16.

This explains the slow speed (PCIe 1.0 x16 = PCIe 2.0 x8 ≈ PCIe 3.0 x4). The new firmware installed by Mojave is supposed to make PCIe 3.0 cards boot properly at PCIe 2.0 speed instead of PCIe 1.0. Double check the firmware version. Your screen shot shows 144.0.0.0.0 which I believe is correct.

Try increasing the speed from g1 to g2 using the following command (download or create the fast.sh script):

sudo ./fast.sh 2 00:03.0

Then try pcitree.sh again and your benchmark test.

 

[Adam Chuan]

Instead of screenshots, post as text because text is easier to read and can be searched.

The output from pcitree.sh shows the link speed and link width of each PCIe device.
g1 = PCIe 1.0 = 2.5 GT/s
g2 = PCIe 2.0 = 5 GT/s
g3 = PCIe 3.0 = 8 GT/s

The problem appears to be that the slot (00:03.0) is running with lower link speed than expected. "g2x16 > g1x16" means that the slot is capable of PCIe 2.0 x16 but is currently running at only PCIe 1.0 x16.

This explains the slow speed (PCIe 1.0 x16 = PCIe 2.0 x8 ≈ PCIe 3.0 x4). The new firmware installed by Mojave is supposed to make PCIe 3.0 cards boot properly at PCIe 2.0 speed instead of PCIe 1.0. Double check the firmware version. Your screen shot shows 144.0.0.0.0 which I believe is correct.

Try increasing the speed from g1 to g2 using the following command (download or create the fast.sh script):

sudo ./fast.sh 2 00:03.0

Then try pcitree.sh again and your benchmark test.

So it means my all slots are running slow now?

Because my 7101 is on the slot 1 and the GPU Titan is on the slot 2.

where can I got the ./fast.sh ?

If I use the ./fast.sh the PCIe slots will run normally speed?

and the slot 2 will read the 7101?is it?

 

[Adam Chuan]

Instead of screenshots, post as text because text is easier to read and can be searched.

The output from pcitree.sh shows the link speed and link width of each PCIe device.
g1 = PCIe 1.0 = 2.5 GT/s
g2 = PCIe 2.0 = 5 GT/s
g3 = PCIe 3.0 = 8 GT/s

The problem appears to be that the slot (00:03.0) is running with lower link speed than expected. "g2x16 > g1x16" means that the slot is capable of PCIe 2.0 x16 but is currently running at only PCIe 1.0 x16.

This explains the slow speed (PCIe 1.0 x16 = PCIe 2.0 x8 ≈ PCIe 3.0 x4). The new firmware installed by Mojave is supposed to make PCIe 3.0 cards boot properly at PCIe 2.0 speed instead of PCIe 1.0. Double check the firmware version. Your screen shot shows 144.0.0.0.0 which I believe is correct.

Try increasing the speed from g1 to g2 using the following command (download or create the fast.sh script):

sudo ./fast.sh 2 00:03.0

Then try pcitree.sh again and your benchmark test.

I have try to use the sudo ./fast.sh 2 00:03.0 in Terminal, but it shows me command not found.

I have download the fast.sh in ＃82 .

 

[joevt]

First, change the directory to the directory that contains the fast.sh file.

cd ~/Downloads
./fast.sh 00:03.0

 

[Adam Chuan]

First, change the directory to the directory that contains the fast.sh file.

cd ~/Downloads
./fast.sh 00:03.0

it didn't run anything.

which step is wrong?

 

[joevt]

which step is wrong?

I am not sure. Try this:
1) download pcitree_fast.zip (it should unzip and create a pcitree_fast folder in the ~/Downloads folder)
2) run the following commands:

cd ~/Downloads/pcitree_fast
chmod 744 fast.sh
chmod 744 pcitree.sh
xattr -d com.apple.quarantine fast.sh
xattr -d com.apple.quarantine pcitree.sh
sudo ./fast.sh 2 00:03.0
sudo ./pcitree.sh

 

[Adam Chuan]

I am not sure. Try this:
1) download pcitree_fast.zip (it should unzip and create a pcitree_fast folder in the ~/Downloads folder)
2) run the following commands:

cd ~/Downloads/pcitree_fast
chmod 744 fast.sh
chmod 744 pcitree.sh
xattr -d com.apple.quarantine fast.sh
xattr -d com.apple.quarantine pcitree.sh
sudo ./fast.sh 2 00:03.0
sudo ./pcitree.sh

The code can work!

but the PCIe speed still on the 1.0 .

Have anything didn't repair done?

 

[joevt]

The code can work!
but the PCIe speed still on the 1.0 .
Have anything didn't repair done?

It seems that there's a problem with the card because it fails in other slots and other computers. I think you may need to replace the card.

Post the result of "sudo ./pcitree.sh" for each slot to see if they all have the same link speed problem. Try the "sudo ./fast.sh 2 xx:xx.x" command for each slot (replace xx:xx.x with the parent device (root port) of the first PEX 8747 device (upstream port of the card's PCIe bridge); for example - the PEX 8747 upstream port in your last pcitree output is 05:00.0 and the parent root port is 00:03.0).

 

[Slash-2CPU]

View attachment 870851

The code can work!

but the PCIe speed still on the 1.0 .

Have anything didn't repair done?

When you RMA the card, do not say the words "apple" or "mac." Tell them it's not working in any Windows PC you try it in. They will make it difficult if you mention it's a Mac.

 

[Adam Chuan]

It seems that there's a problem with the card because it fails in other slots and other computers. I think you may need to replace the card.

Post the result of "sudo ./pcitree.sh" for each slot to see if they all have the same link speed problem. Try the "sudo ./fast.sh 2 xx:xx.x" command for each slot (replace xx:xx.x with the parent device (root port) of the first PEX 8747 device (upstream port of the card's PCIe bridge); for example - the PEX 8747 upstream port in your last pcitree output is 05:00.0 and the parent root port is 00:03.0).

So it’s only my 7101 has problem?

or my other PCIe slot have something problem?

because the GPU Titan is running 2.5 only, can I use your code to change the speed too?
[automerge]1571546192[/automerge]

When you RMA the card, do not say the words "apple" or "mac." Tell them it's not working in any Windows PC you try it in. They will make it difficult if you mention it's a Mac.

Ok! I will sent the card back to replace it.
thank you !

 

[tsialex]

because the GPU Titan is running 2.5 only, can I use your code to change the speed too?

NVIDIA GPUs change the PCIe speed on the fly at load, it's a power saving resource and it's normal to report PCIe 1.0 speeds (2.5GT/s) with SystemReport.

 

[Adam Chuan]

NVIDIA GPUs change the PCIe speed on the fly at load, it's a power saving resource and it's normal to report PCIe 1.0 speeds (2.5GT/s) with SystemReport.

oh! I got it!

so the speed on the Nvidia GPU is normal.

if I download the CUDA driver and run it , it will run the full speed to 5.0.is it?

 

[tsialex]

oh! I got it!

so the speed on the Nvidia GPU is normal.

if I download the CUDA driver and run it , it will run the full speed to 5.0.is it?

NVIDIA GPUs change the PCIe speed connection alone with MP5,1, when in load, you don't have to do anything.

 

[Adam Chuan]

NVIDIA GPUs change the PCIe speed connection alone with MP5,1, when in load, you don't have to do anything.

ok!
thank you.
so it is my 7101 has problem for sure.

 

[tsialex]

ok!
thank you.
so it is my 7101 has problem for sure.

About the SSD7101-A, did you tried it with a Mac Pro without a TITAN GPU?

Test it before with an Apple OEM GPU or an AMD one with a clean install before sending it to RMA. Do a 3-times sequential NVRAM reset before doing the clean install. Don't install NVIDIA web drivers.

TITAN GPUs are known to have an interdependency with the NVRAM, saving some bogus data in it that can cause some weird problems system wide.

P.S.: as already warned by @Slash-2CPU, never say Mac, Mac Pro or Apple to the HighPoint RMA. They will dismiss your RMA case instantly.

 

[Adam Chuan]

About the SSD7101-A, did you tried it with a Mac Pro without a TITAN GPU?

Test it before with an Apple OEM GPU or an AMD one with a clean install before sending it to RMA. Do a 3-times sequential NVRAM reset before doing the clean install. Don't install NVIDIA web drivers.

TITAN GPUs are known to have an interdependency with the NVRAM, saving some bogus data in it that can cause some weird problems system wide.

P.S.: as already warned by @Slash-2CPU, never say Mac, Mac Pro or Apple to the HighPoint RMA. They will dismiss your RMA case instantly.

I had try some Mac Pro, the model are 2009 to 2012.

these model are running this GPU of this photo too.

but the problem is all the same.

the PCIe slot 2 can't read the 7101, and switch to the GPU slot 1 that speed still stop at 2400/2800.

 

[tsialex]

I had try some Mac Pro, the model are 2009 to 2012.

these model are running this GPU of this photo too. View attachment 871014

but the problem is all the same.

the PCIe slot 2 can't read the 7101, and switch to the GPU slot 1 that speed still stop at 2400/2800.

If you ever used a TITAN GPU with the Mac Pros you tried, the NVRAM bogus settings in the private part of the NVRAM can happen and then you have weird problems. This seems stupid, but you should check before sending the SSD7101-A to the RMA.

 

[joevt]

if I download the CUDA driver and run it , it will run the full speed to 5.0.is it?

Use pcitree.sh to find out.

 

[Adam Chuan]

If you ever used a TITAN GPU with the Mac Pros you tried, the NVRAM bogus settings in the private part of the NVRAM can happen and then you have weird problems.

these devices didn't use the titan.

 

[CNDCodewarrior]

Hi guys,

Long time lurker, first time posting as I feel I have something to add now that I bought the card and a couple of Samsung 970 EVO's 1TB.

During my setup I was having issues with the driver supplied by Highpoint, the driver was not loading even after allowing it thru the Security and Privacy control panel.

Without the driver working I was unable to get the web management of the card to connect. I thought about doing a software raid but I like to use the full functionality of the card if I could.

While trouble shooting I found that the kext was listed as disabled in the software report due to no user consent.

Doing a few more google searches lead me to this blog post about the same issue with Oracle's virtual box.

https://ilgthegeek.wordpress.com/2018/01/27/macos-install-oracle-virtualbox-on-10-13/

This was the ticket for me I booted into recovery mode (disconnected my 4k monitor, annoying) and added the following code in the terminal.

spctl kext-consent add DX6G69M9N2

After the reboot I was able to access the Highpoint's raid management page properly and set up a RAID 0 with a 64k block size and a 4k Sector size.

There are more settings in there that I didn't tackle yet, like the error reporting and such but I can easily see temps and drive status.

Note the card disappears from the NVMExpress and isn't listed in the Hardware RAID sections of the system report but the drive does show up in Storage.

Once the drive popped up in Disk Utility I formatted with APFS and ran some speed tests.

Not a bad little upgrade. I've moved only my home folder over to it as I plan on getting the new Mac Pro tower once it is released, use the internal base storage for the OS and have my home folder located on the Highpoint.

Hope this helps others that may be having issues with the driver.

 

[Executor]

I just got a HighPoint 7101A for my Mac Pro and it works well. Fast and so far very stable. The fan noise however, is driving me insane. Is a passive heat sink solution possible? I saw lenny_'s heat-sink and I think it would work but I have no idea where he got it or how he mounted it. I am assuming it is a custom solution. If any member could post a link where I could buy something similar it would be appreciated.

 

[Matt_Mac]

I removed the fan and put a big cpu-heatsink on the card... now its quite and the SSDs are still cool ?

 

[handheldgames]

I removed the fan and put a big cpu-heatsink on the card... now its quite and the SSDs are still cool ?

Just make sure the PCIe switch is kept well under 70c at all times.

 

[Executor]

I removed the fan and put a big cpu-heatsink on the card... now its quite and the SSDs are still cool ?

But what particular CPU heatsink did you use? and how did you mount it? Any specific links would be helpful.
[automerge]1573352269[/automerge]

Just make sure the PCIe switch is kept well under 70c at all times.

How do you check the temperature of the PCIe switch? iStat doesn't seem to show it, unless I am missing something.