The provided PTX was compiled with an unsupported toolchain

Accelerated Computing CUDA CUDA Setup and Installation
1

Hi,

When I run my program (01-vector-add.cu) I 've got this error:

Error addVectorsInto: the provided PTX was compiled with an unsupported toolchain.
FAIL: vector[0] - 0 does not equal 7

I have read this post and this page (550 is compatible with 12.5?) . May be it’s same problem but I’m not sure.

My configuration :
uname -a
LinuxPrecision-3490 6.8.0-38-generic #38-Ubuntu SMP PREEMPT_DYNAMIC Fri Jun 7 15:25:01 UTC 2024 x86_64 x86_64 x86_64 GNU/Linux

nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2024 NVIDIA Corporation
Built on Thu_Jun__6_02:18:23_PDT_2024
Cuda compilation tools, release 12.5, V12.5.82
Build cuda_12.5.r12.5/compiler.34385749_0

±----------------------------------------------------------------------------------------+
| NVIDIA-SMI 550.90.07 Driver Version: 550.90.07 CUDA Version: 12.4 |
|-----------------------------------------±-----------------------±---------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA RTX 500 Ada Gener… Off | 00000000:01:00.0 Off | N/A |
| N/A 42C P8 2W / 30W | 565MiB / 4094MiB | 0% Default |
| | | N/A |
±----------------------------------------±-----------------------±---------------------+

±----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 2402 G /usr/lib/xorg/Xorg 147MiB |
| 0 N/A N/A 2765 G /usr/bin/gnome-shell 161MiB |
| 0 N/A N/A 3411 G …irefox/4539/usr/lib/firefox/firefox 170MiB |
| 0 N/A N/A 6296 G …erProcess --variations-seed-version 75MiB |
±----------------------------------------------------------------------------------------+

2

You cannot use CUDA 12.5 with a r550 driver. Your nvidia-smi output indicates only up to CUDA 12.4 is supported. Update to the latest available driver (e.g. the one that ships with a CUDA 12.5 installer) or switch to CUDA 12.4

3

Thanks for answer
but in table 3 (Table 3 CUDA Application Compatibility Support Matrix CUDA toolkit is compatible with driver 550
So I do not understand relation between toolkit version and driver version

4

you might want to read that link you provided, in its entirety, carefully.

  1. From here: “Applications that compile device code to PTX will not work on older drivers. If the application requires PTX then admins have to upgrade the installed driver.”
  2. From here: “Forward Compatibility is applicable only for systems with NVIDIA Data Center GPUs or select NGC Server Ready SKUs of RTX cards. It’s mainly intended to support applications built on newer CUDA Toolkits to run on systems installed with an older NVIDIA Linux GPU driver from different major release families. This new forward-compatible upgrade path requires the use of a special package called “CUDA compat package”” Your RTX 500 Ada GPU is not one of the GPUs covered by this forward compatibility option.

This forward compatibility path (everything you linked) is not an option for your case. Therefore the relevant compatibility table is table 3 here, for CUDA 12.5, which indicates that a R555 driver is required for CUDA 12.5

5

Thanks. I think i will have to read doc carefully.

I deleted everything and go back to cuda 12.4 and everything is fine now

[quote=“Robert_Crovella, post:4, topic:300132”]
Your RTX 500 Ada GPU is not one of the GPUs covered by this forward compatibility option.
[/quote] I think I can read this here
but may be it’s good now with this one for cuda toolkit 12.5 -I can find NVIDIA RTX 500 Ada Generation Laptop GPU in supported product list. Is it right?

6

I’m not sure what you are asking. If you are trying to find a suitable driver installer for your GPU, use the wizard

If you want to install a particular CUDA toolkit version, the installers are available here and install guides are available and linked from the installer page (“CUDA Documentation/Release Notes”).

7

I ask about this :

How do you know that RTX 500 Ada is not one of the GPUs covered by this forward compatibility option?
Is it written here(Linux x64 (AMD64/EM64T) Display Driver | 555.42.02 | Linux 64-bit | NVIDIA) or elsewhere?

8

You said:

I said/quoted:

A laptop GPU is neither a Data Center GPU nor is it a NGC Server Ready SKU of an RTX card.

9

Forget everything!
With cuda toolkit 12.5 version of NVIDIA Kernel Mode Driver must be >= 555.42.02+
In product supported of page Linux x64 (AMD64/EM64T) Display Driver | 555.42.02 | Linux 64-bit | NVIDIA there is NVIDIA RTX 500 Ada Generation Laptop GPU so Cuda toolkit 12.5 should be compatible with RTX A500 Ada. Is it right?

10

I believe that is approximately what I said here:

Yes, you can use CUDA 12.5 with a RTX 500 Ada

11

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.