Lenovo P1 Gen 6 with Ada generation AD104 12G 192bit for ML/LLM training

[obriensystems]

The Lenovo P1 is an excellent machine to run NVidia GPUs as a mobile option to building your own around desktop 4090's or A4500s
The machine complements the Macbook Pro 16 inch M1+ model with it's metal GPU under tensorflow

https://github.com/wilicc/gpu-burn

Lenovo P1 Gen 6 - bios is limited to 64g not 96g

The lenovo P1 gen 6 as of Nov 2023 has a persistent bios issue when running 48g dimms in 48g or 96g configurations. every ~24h the machine will enter an unrecoverable ram check on reboot/sleep that is only fixed by replacing the ram with 32g or lower dimms. The fix is to limit the ram to 64g.

Make sure to get the 64G model - not the 96G one
https://forums.lenovo.com/t5/ThinkPad-P-and-W-Series-Mobile-Workstations/P1-Gen6-Bricked-after-BSOD-second-laptop-with-the-same-problem/m-p/5254145?page=2#6148028

Disable the intel GPU and reenable - to move to GPU 1

[obriensystems]

update windows

Install Google Chrome

Git + Git Bash

configure git-bash

http://wiki.obrienlabs.cloud/display/DEV/Developer+Guide#DeveloperGuide-Windows10

Java 21

Springsource Tool Suite 4.4

Gcloud SDK

https://cloud.google.com/sdk/auth_success

C:\soft\gcloiud>gcloud config get project
bigquery-ol

C:\soft\gcloiud>gcloud version
Google Cloud SDK 453.0.0
bq 2.0.98
core 2023.10.27
gcloud-crc32c 1.0.0
gsutil 5.27

Maven 3.9.5

• https://maven.apache.org/download.cgi

[obriensystems]

runtime - run gpuburn on the GPU

micha@p1gen6 MINGW64 /c/wse_github/gpu-burn (master)
$ docker build -t gpu_burn .

micha@p1gen6 MINGW64 /c/wse_github/gpu-burn (master)
$ docker run --gpus all gpu_burn

==========
== CUDA ==
==========

CUDA Version 11.8.0

Container image Copyright (c) 2016-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.

This container image and its contents are governed by the NVIDIA Deep Learning Container License.
By pulling and using the container, you accept the terms and conditions of this license:
https://developer.nvidia.com/ngc/nvidia-deep-learning-container-license

A copy of this license is made available in this container at /NGC-DL-CONTAINER-LICENSE for your convenience.

GPU 0: NVIDIA RTX 3500 Ada Generation Laptop GPU (UUID: GPU-25326b0f-ad93-c319-7027-b0029d4aee8e)
Using compare file: compare.ptx
Burning for 120 seconds.
10.8%  proc'd: 148 (13665 Gflop/s)   errors: 0   temps: 73 C
        Summary at:   Wed Nov  1 18:38:50 UTC 2023

23.3%  proc'd: 333 (13337 Gflop/s)   errors: 0   temps: 79 C
        Summary at:   Wed Nov  1 18:39:05 UTC 2023

35.8%  proc'd: 518 (13367 Gflop/s)   errors: 0   temps: 80 C
        Summary at:   Wed Nov  1 18:39:20 UTC 2023

45.8%  proc'd: 629 (13367 Gflop/s)   errors: 0   temps: 80 C
        Summary at:   Wed Nov  1 18:39:32 UTC 2023

56.7%  proc'd: 814 (13207 Gflop/s)   errors: 0   temps: 81 C
        Summary at:   Wed Nov  1 18:39:45 UTC 2023

69.2%  proc'd: 999 (13303 Gflop/s)   errors: 0   temps: 83 C
        Summary at:   Wed Nov  1 18:40:00 UTC 2023

79.2%  proc'd: 1110 (12973 Gflop/s)   errors: 0   temps: 82 C
        Summary at:   Wed Nov  1 18:40:12 UTC 2023

90.0%  proc'd: 1295 (13059 Gflop/s)   errors: 0   temps: 83 C
        Summary at:   Wed Nov  1 18:40:25 UTC 2023

100.0%  proc'd: 1480 (13182 Gflop/s)   errors: 0   temps: 84 C
Killing processes with SIGTERM (soft kill)

[obriensystems]

NVidia tensor cores only

micha@p1gen6 MINGW64 /c/wse_github/gpu-burn (master)
$ cat Dockerfile
ARG CUDA_VERSION=11.8.0
ARG IMAGE_DISTRO=ubi8

FROM nvidia/cuda:${CUDA_VERSION}-devel-${IMAGE_DISTRO} AS builder

WORKDIR /build

COPY . /build/

RUN make

FROM nvidia/cuda:${CUDA_VERSION}-runtime-${IMAGE_DISTRO}

COPY --from=builder /build/gpu_burn /app/
COPY --from=builder /build/compare.ptx /app/

WORKDIR /app

CMD ["./gpu_burn", "-tc", "120"]
#CMD ["./gpu_burn", "120"]

   10  docker build -t gpu_burn .
   11  docker run --gpus all gpu_burn



micha@p1gen6 MINGW64 /c/wse_github/gpu-burn (master)
$ docker run --gpus all gpu_burn

==========
== CUDA ==
==========

CUDA Version 11.8.0

Container image Copyright (c) 2016-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.

This container image and its contents are governed by the NVIDIA Deep Learning Container License.
By pulling and using the container, you accept the terms and conditions of this license:
https://developer.nvidia.com/ngc/nvidia-deep-learning-container-license

A copy of this license is made available in this container at /NGC-DL-CONTAINER-LICENSE for your convenience.

GPU 0: NVIDIA RTX 3500 Ada Generation Laptop GPU (UUID: GPU-25326b0f-ad93-c319-7027-b0029d4aee8e)
Using compare file: compare.ptx
Burning for 120 seconds.
10.8%  proc'd: 518 (47881 Gflop/s)   errors: 0   temps: 68 C
        Summary at:   Wed Nov  1 21:32:20 UTC 2023

21.7%  proc'd: 1073 (45284 Gflop/s)   errors: 0   temps: 72 C
        Summary at:   Wed Nov  1 21:32:33 UTC 2023

32.5%  proc'd: 1591 (46962 Gflop/s)   errors: 0   temps: 75 C
        Summary at:   Wed Nov  1 21:32:46 UTC 2023

43.3%  proc'd: 2146 (46766 Gflop/s)   errors: 0   temps: 78 C
        Summary at:   Wed Nov  1 21:32:59 UTC 2023

53.3%  proc'd: 2664 (46698 Gflop/s)   errors: 0   temps: 79 C
        Summary at:   Wed Nov  1 21:33:11 UTC 2023

64.2%  proc'd: 3219 (46535 Gflop/s)   errors: 0   temps: 78 C
        Summary at:   Wed Nov  1 21:33:24 UTC 2023

75.0%  proc'd: 3774 (45680 Gflop/s)   errors: 0   temps: 81 C
        Summary at:   Wed Nov  1 21:33:37 UTC 2023

85.8%  proc'd: 4292 (45082 Gflop/s)   errors: 0   temps: 81 C
        Summary at:   Wed Nov  1 21:33:50 UTC 2023

96.7%  proc'd: 4810 (46265 Gflop/s)   errors: 0   temps: 81 C
        Summary at:   Wed Nov  1 21:34:03 UTC 2023

100.0%  proc'd: 5032 (46172 Gflop/s)   errors: 0   temps: 82 C
Killing processes with SIGTERM (soft kill)
Using compare file: compare.ptx
Burning for 120 seconds.
Initialized device 0 with 12281 MB of memory (11119 MB available, using 10007 MB of it), using FLOATS, using Tensor Cores
Results are 268435456 bytes each, thus performing 37 iterations
Freed memory for dev 0
Uninitted cublas
done

Tested 1 GPUs:
        GPU 0: OK

[obriensystems]

Double on RTX-3500

micha@p1gen6 MINGW64 /c/wse_github/gpu-burn (master)
$ docker run --gpus all gpu_burn

==========
== CUDA ==
==========

CUDA Version 11.8.0

Container image Copyright (c) 2016-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.

This container image and its contents are governed by the NVIDIA Deep Learning Container License.
By pulling and using the container, you accept the terms and conditions of this license:
https://developer.nvidia.com/ngc/nvidia-deep-learning-container-license

A copy of this license is made available in this container at /NGC-DL-CONTAINER-LICENSE for your convenience.

GPU 0: NVIDIA RTX 3500 Ada Generation Laptop GPU (UUID: GPU-25326b0f-ad93-c319-7027-b0029d4aee8e)
Using compare file: compare.ptx
Burning for 120 seconds.
51.7%  proc'd: 17 (298 Gflop/s)   errors: 0   temps: 74 C
        Summary at:   Wed Nov  1 21:39:54 UTC 2023

62.5%  proc'd: 17 (298 Gflop/s)   errors: 0   temps: 72 C
        Summary at:   Wed Nov  1 21:40:07 UTC 2023

75.0%  proc'd: 17 (298 Gflop/s)   errors: 0   temps: 70 C
        Summary at:   Wed Nov  1 21:40:22 UTC 2023

87.5%  proc'd: 17 (298 Gflop/s)   errors: 0   temps: 67 C
        Summary at:   Wed Nov  1 21:40:37 UTC 2023

100.0%  proc'd: 17 (298 Gflop/s)   errors: 0   temps: 65 C
        Summary at:   Wed Nov  1 21:40:52 UTC 2023

100.0%  proc'd: 34 (304 Gflop/s)   errors: 0   temps: 65 C
Killing processes with SIGTERM (soft kill)

CMD ["./gpu_burn", "-d", "120"]