guys, that will never work !
see specs pcie. see specs usb
this isn't handlebar in software ...
GPIOs are to slow to communicate as software pcie x1 emulator
Topic: Running a GPU with BBB (Read 995 times)
Klondike, you are correct with #1, it is backwards
& I thought the USB had 9 connections but could only see 4 in the cable.
I was thinking we could modify the USB cable to eliminate the power going to the 16x, BUT if that line is used to send signals, a new type of USB driver can not remove the hard wired voltage to the USB port on the BBB. --Thanks. I understand
your idea with the GPIO pins seems the way to go.
& I thought the USB had 9 connections but could only see 4 in the cable.
I was thinking we could modify the USB cable to eliminate the power going to the 16x, BUT if that line is used to send signals, a new type of USB driver can not remove the hard wired voltage to the USB port on the BBB. --Thanks. I understand
your idea with the GPIO pins seems the way to go.
te other 5 are tucked deeper into the connector.
Klondike, you are correct with #1, it is backwards
& I thought the USB had 9 connections but could only see 4 in the cable.
I was thinking we could modify the USB cable to eliminate the power going to the 16x, BUT if that line is used to send signals, a new type of USB driver can not remove the hard wired voltage to the USB port on the BBB. --Thanks. I understand
your idea with the GPIO pins seems the way to go.
& I thought the USB had 9 connections but could only see 4 in the cable.
I was thinking we could modify the USB cable to eliminate the power going to the 16x, BUT if that line is used to send signals, a new type of USB driver can not remove the hard wired voltage to the USB port on the BBB. --Thanks. I understand
your idea with the GPIO pins seems the way to go.
okay - theres a lot wrong here - so much it hurts.
1) your picture of the riser is backwards. it does touch the enire first section of the riser. (ever wonder why its the perfect width to plug in backwards of the way you have it?
2) the USB cable does not carry 12V, or 5V, for the riser board. thats the entire reason they are used, because you can have the GPU on a seperate PSU without frying the mobo over the common 12V/GND lines (because this eliminates those)
3) The USB cable is a USB 3.0, and AFAIK thats because it uses more than just 4 wires to push the PCIe 1x data. either way:
4) plugging that usb in will push a 5V/GND signal to the 16x board, over lines that were not meant to support either. You'll probably damage or lock-up any cards in the 16x slot doing so
TLDR: its not meant for a USB connection.
what you really need to do to achieve this is emulate a PCIe 1x connection from the GPIO pins, and wire that to a custom USB3.0 port for the riser. AFAIK it hasnt been achieved yet
1) your picture of the riser is backwards. it does touch the enire first section of the riser. (ever wonder why its the perfect width to plug in backwards of the way you have it?
2) the USB cable does not carry 12V, or 5V, for the riser board. thats the entire reason they are used, because you can have the GPU on a seperate PSU without frying the mobo over the common 12V/GND lines (because this eliminates those)
3) The USB cable is a USB 3.0, and AFAIK thats because it uses more than just 4 wires to push the PCIe 1x data. either way:
4) plugging that usb in will push a 5V/GND signal to the 16x board, over lines that were not meant to support either. You'll probably damage or lock-up any cards in the 16x slot doing so
TLDR: its not meant for a USB connection.
what you really need to do to achieve this is emulate a PCIe 1x connection from the GPIO pins, and wire that to a custom USB3.0 port for the riser. AFAIK it hasnt been achieved yet
So, if there are only 4 wires in the USB cable - I am not sure which 4 in the red box are being used.
My guess it is 4 before the mechanical key.
IF these 4 wires go to pins 5, 6, 7 & 8, then we could cut the 3.3V line, make sure the ground goes to ground
And then figure out a way to make the USB communicate with the SMBus Clock and Data (5 & 6)
I am pretty sure that the SATA power connector would provide the 12 & 3.3 V to the PCIe connector.
I have no idea what 4 lines the usb is connected to other than 5 thru 22. So the above is totally guessing.
It does seem POSSIBLE, I just dont know if it is feasible.
My guess it is 4 before the mechanical key.
IF these 4 wires go to pins 5, 6, 7 & 8, then we could cut the 3.3V line, make sure the ground goes to ground
And then figure out a way to make the USB communicate with the SMBus Clock and Data (5 & 6)
I am pretty sure that the SATA power connector would provide the 12 & 3.3 V to the PCIe connector.
I have no idea what 4 lines the usb is connected to other than 5 thru 22. So the above is totally guessing.
It does seem POSSIBLE, I just dont know if it is feasible.
The main power for the GPU is from the main pci power connectors, But the power on the pcie slot is powered by an aux power connection from the riser card
picture shows how a SATA power connector will provide the 12 & 3.3 V
--
--
There are ONLY 4 wires in the USB cable going from the motherboard to the GPU.
The USB Riser Connector:
--
--
The USB Riser Connector to Motherboard PCIe slot:
--
--
You can see that the connector does not connect to the first four contacts, so will not touch 12V.
Here is the PCIe pinout, red square shows the connections for the USB riser.
picture shows how a SATA power connector will provide the 12 & 3.3 V
--
--
There are ONLY 4 wires in the USB cable going from the motherboard to the GPU.
The USB Riser Connector:
--
--
The USB Riser Connector to Motherboard PCIe slot:
--
--
You can see that the connector does not connect to the first four contacts, so will not touch 12V.
Here is the PCIe pinout, red square shows the connections for the USB riser.
How do you intend to power the Card?
Only way it would work is if the BBB had the PCIe lane. Which they don't support so it won't work.
I understand about the voltage. If that was all it would be easy to remove those 12V from going to the USB --> you power the 12V on the pcie with the aux pcie power connection.
Problem is the signals on the usb are not standard USB Protocol. To do this it would definately require some custom software and maybe different usb drivers.
Anything is possible with enough time and money. Remember people said we would never fly, never put a man on the moon (and with less than a TSR80 computer). And Bitcoin will never . . . . .
Just curious if someone has figured this out.
Problem is the signals on the usb are not standard USB Protocol. To do this it would definately require some custom software and maybe different usb drivers.
Anything is possible with enough time and money. Remember people said we would never fly, never put a man on the moon (and with less than a TSR80 computer). And Bitcoin will never . . . . .
Just curious if someone has figured this out.
this is a completly other connection. the only thing that use both are the physicaly connectors.
I understand about the voltage. If that was all it would be easy to remove those 12V from going to the USB --> you power the 12V on the pcie with the aux pcie power connection.
Problem is the signals on the usb are not standard USB Protocol. To do this it would definately require some custom software and maybe different usb drivers.
Anything is possible with enough time and money. Remember people said we would never fly, never put a man on the moon (and with less than a TSR80 computer). And Bitcoin will never . . . . .
Just curious if someone has figured this out.
Problem is the signals on the usb are not standard USB Protocol. To do this it would definately require some custom software and maybe different usb drivers.
Anything is possible with enough time and money. Remember people said we would never fly, never put a man on the moon (and with less than a TSR80 computer). And Bitcoin will never . . . . .
Just curious if someone has figured this out.
So-called "USB risers" don't actually use the USB protocal, so that concept is totally impractical.
They just "merge" signaling so that a USB-type CABLE can be used to transfer data.
They just "merge" signaling so that a USB-type CABLE can be used to transfer data.
I think he knows that, but the signal is there so communication in theory is only up to the software.
hope you didn't powered on ...
USB = 5 V PCIE 12 V.
1. When same powerlines are used for the PCIE USB-Style Riser (It is only same connectors!) possible that BBB usb bridge get roasted
2. When powerlines switched, possible that Card and BBB are get roasted...
Will never work !
So-called "USB risers" don't actually use the USB protocal, so that concept is totally impractical.
They just "merge" signaling so that a USB-type CABLE can be used to transfer data.
They just "merge" signaling so that a USB-type CABLE can be used to transfer data.
I think he knows that, but the signal is there so communication in theory is only up to the software.
So-called "USB risers" don't actually use the USB protocal, so that concept is totally impractical.
They just "merge" signaling so that a USB-type CABLE can be used to transfer data.
They just "merge" signaling so that a USB-type CABLE can be used to transfer data.
Interesting idea. I don't know much about it but I don't think a BBB could drive the card even if you'd manage to communicate with it.
But if it would work it would be awesome.
But if it would work it would be awesome.
Does anyone know how to run a GPU (for mining/other code - NOT video) with a BBB as pictured?
Obviously the BBB does not have the PCI* but the signals are sent over the USB Cable when used with these risers. Maybe some sort of pci* usb emulator? Looks like it should be technically feasible. ??
Look forward to everyones ideas.
Obviously the BBB does not have the PCI* but the signals are sent over the USB Cable when used with these risers. Maybe some sort of pci* usb emulator? Looks like it should be technically feasible. ??
Look forward to everyones ideas.
Jump to: