Hi,
I’m trying to profile my application running on the Pegasus (Xavier A) platform.
I understand the dGPU is Turing architecture, should I expect to have the same performance (specific kernel run time) as my RTX 2080 on a desktop computer (which also has Turing architecture)?
Moving this topic to the DRIVE forum.
Dear @yotam.nachmias,
The performance of any application depends on the computation power of GPU. You can quickly compare two GPU configuration by running CUDA deviceQuery sample.
Hi SivaRamaKrishnaNV,
The deviceQuery is not part of the cuda-10.2 that was installed in the Pegasus (Xavier) platform using the Nvidia SDKManager.
So how do you recommend to run this test (or similar one) on the Pegasus?
Dear @yotam.nachmias,
You can follow the below steps to run any CUDA sample on board
Cross compile the CUDA sample for aarch64( check make arch=aarch64. For more details check cuda sample cross compilation section at
CUDA Samples :: CUDA Toolkit Documentation)
Copy the binary to board
Setup ld_library_path to include cuda libs on board and run the sample
Hi SivaRamaKrishnaNV,
I was able to run the device query on both target and host and the main difference I see is at “memory clock rate” (1440 MHz on target vs 7000MHz on host)
could that explain difference (857[ms] VS 123[ms]) I see in my Kernel execution time?
attached are the deviceQuery plots:
Xavier target:
Device 0: “Graphics Device”
CUDA Driver Version / Runtime Version 10.2 / 10.2
CUDA Capability Major/Minor version number: 7.5
Total amount of global memory: 7680 MBytes (8052998144 bytes)
(44) Multiprocessors, ( 64) CUDA Cores/MP: 2816 CUDA Cores
GPU Max Clock rate: 1500 MHz (1.50 GHz)
Memory Clock rate: 1440 Mhz
Memory Bus Width: 256-bit
L2 Cache Size: 4194304 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: 1024
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 3 copy engine(s)
Run time limit on kernels: No
Integrated GPU sharing Host Memory: No
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: 1 / 1 / 0
host:
Device 0: “GeForce RTX 2080”
CUDA Driver Version / Runtime Version 11.0 / 10.2
CUDA Capability Major/Minor version number: 7.5
Total amount of global memory: 7979 MBytes (8366784512 bytes)
(46) Multiprocessors, ( 64) CUDA Cores/MP: 2944 CUDA Cores
GPU Max Clock rate: 1710 MHz (1.71 GHz)
Memory Clock rate: 7000 Mhz
Memory Bus Width: 256-bit
L2 Cache Size: 4194304 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: 1024
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 3 copy engine(s)
Run time limit on kernels: Yes
Integrated GPU sharing Host Memory: No
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 / 2 / 0
Compute Mode:
< Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >
Dear @yotam.nachmias,
I could see number of Multiprocessors, GPU clock speed, Memory clock speed are different. All these will effect the kernel execution time.
In general, You can profile your application using nvprof(or nisght) and see the bottlenecks in application and suggestions for improving the overall kerenel execution time.