Can Jetson Xavier run with Quadro GP100?

I’ve installed a Quadro GP100 into Jetson as an experiment for doing some testing.

I can see the Quadro on my PCIe bus

| ±01.0-[04]–
| ±02.0-[05]–±00.0 NVIDIA Corporation GP100GL
| | -00.1 NVIDIA Corporation Device 0fb1
| ±03.0-[06]–
| ±04.0-[07]–
| ±05.0-[08]–
| ±06.0-[09]–
| ±07.0-[0a]–
| -08.0-[0b]–
±[0001:00]—00.0-[01-ff]----00.0 Marvell Technology Group Ltd. Device 9171

But I CUDA 10.0 does not see it
CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 1 CUDA Capable device(s)

Device 0: “Xavier”
CUDA Driver Version / Runtime Version 10.0 / 10.0
CUDA Capability Major/Minor version number: 7.2
Total amount of global memory: 15819 MBytes (16587653120 bytes)
( 8) Multiprocessors, ( 64) CUDA Cores/MP: 512 CUDA Cores
GPU Max Clock rate: 1500 MHz (1.50 GHz)
Memory Clock rate: 1500 Mhz
Memory Bus Width: 256-bit
L2 Cache Size: 524288 bytes
Maximum Texture Dimension Size (x,y,z) 1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)
Maximum Layered 1D Texture Size, (num) layers 1D=(32768), 2048 layers
Maximum Layered 2D Texture Size, (num) layers 2D=(32768, 32768), 2048 layers
Total amount of constant memory: 65536 bytes
Total amount of shared memory per block: 49152 bytes
Total number of registers available per block: 65536
Warp size: 32
Maximum number of threads per multiprocessor: 2048
Maximum number of threads per block: 1024
Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
Max dimension size of a grid size (x,y,z): (2147483647, 65535, 65535)
Maximum memory pitch: 2147483647 bytes
Texture alignment: 512 bytes
Concurrent copy and kernel execution: Yes with 1 copy engine(s)
Run time limit on kernels: No
Integrated GPU sharing Host Memory: Yes
Support host page-locked memory mapping: Yes
Alignment requirement for Surfaces: Yes
Device has ECC support: Disabled
Device supports Unified Addressing (UVA): Yes
Device supports Compute Preemption: Yes
Supports Cooperative Kernel Launch: Yes
Supports MultiDevice Co-op Kernel Launch: Yes
Device PCI Domain ID / Bus ID / location ID: 0 / 0 / 0
Compute Mode:
< Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 10.0, CUDA Runtime Version = 10.0, NumDevs = 1
Result = PASS

Do I need to do anything else to instruct CUDA to go out and look for the additional GPU? CUDA was installed on the Jetson before installation of the GP100.



May I know how do you link GP100 into Xavier.
Could you share more detail about it?


We have a Gen4 PCIe Switch which provides high density Fanout. We connect the Xavier to the upstream port of our switch (Gen4 x8) and then we connect the GP100 to our evaluation board slot (x16 Gen3)

The topology appears as follows

Xavier —8x16G – PCIe Switch – 16x8G – GP100

We can see the GP100 in our topology but CUDA does not appear to detect it. I am also aware that GP100 does not have 64-bit linux support in the drivers so this was one possible reason.

When I run the deviceQuery utility within CUDA 10.0 I only see the Xavier

Hi wesley.yung, the drivers for discrete GPUs are not currently supported on Xavier, please refer to this FAQ:

Ok thanks Dusty_nv

Do you know if another Jetson is supported in device mode? THat is out other option.


Hi Wes, only the host Jetson’s onboard integrated GPU is supported. In device mode you could connect to another Jetson over PCIe and perform DMA, but the GPU will be accessible from the Jetson that it resides in.

Keep in mind that the GPU in the Jetson is integrated directly to the memory controller, but the PCIe versions require a different driver for PCIe access. The version for the iGPU can’t work with the PCIe GPU…and the drivers published for PCIe GPUs are for x86_64 architecture, not aarc64/arm64. You could use the PCIe version on a Jetson if you had the ARM version driver, but those don’t exist.

Any PCIe video card will not work. The integrated GPU is the only one supported. The problem is that the drivers for desktop video cards don’t exist for this architecture, and the existing drivers are for GPUs directly integrated to the memory controller (PCIe cards are not integrated).

Hi GeForceX, as mentioned the support will be added in a future release of JetPack, the timeline is still TBD as it requires pretty extensive changes to the driver infrastructure on Jetson since it is based in userspace today for integrated GPU. FYI at this time the proposed support plan will be starting with RTX discrete GPU for compute.

We’re sorry to hear that. Please contact the distributor that you purchased the unit through about if they are able to make you a refund.