I love you Mint more than I can bear, but... we had our time together. And I have to let you go. Goodbye Xorg, hello Wayland.

#Linux #Wayland #Xorg #LinuxMint #FedoraKDE #Performance #Gaming #HDR #MultiBoot #MultiGPU #MultiDisplay

Akshay (@akshay_pachaar)

딥러닝 모델은 기본 설정으로는 여러 GPU가 있어도 보통 단일 GPU만 사용한다는 지적. 이상적인 학습은 학습 부하를 여러 GPU에 분산하는 것이라며, 다중 GPU 훈련을 위한 네 가지 전략을 그래픽으로 소개한다는 내용(멀티-GPU 분산 학습 기법 소개).

https://x.com/akshay_pachaar/status/2026649685243654194

#multigpu #distributedtraining #gpu #deeplearning

Akshay 🚀 (@akshay_pachaar) on X

By default, deep learning models only utilize a single GPU for training, even if multiple GPUs are available. An ideal way to train models is to distribute the training workload across multiple GPUs. The graphic depicts four strategies for multi-GPU training👇

X (formerly Twitter)

Avi Chawla (@_avichawla)

Multi-GPU 트레이닝을 위한 4가지 전략을 시각 자료로 설명한 게시물입니다. 대규모 모델 학습에서의 병렬화·데이터/모델 분할·메모리 최적화 등 다양한 멀티-GPU 접근법을 한눈에 비교해 이해를 돕는 내용으로 보입니다.

https://x.com/_avichawla/status/2018935482382684460

#multigpu #distributedtraining #deeplearning #gpu

Avi Chawla (@_avichawla) on X

4 strategies for Multi-GPU training, explained visually:

X (formerly Twitter)

Chủ đề: Dùng dàn máy cũ có 1 PCIe3 × 16 và 7 PCIe2 × 16 để chạy 8 GPU, mỗi GPU nạp một mô hình AI riêng. Lo ngại tốc độ PCIe và RAM khiến thời gian load chậm nhưng có thể giữ mô hình luôn trong bộ nhớ. Tensor parallel không hiệu quả vì cần giao tiếp liên tục giữa các GPU. Có thể tách nhiệm vụ, mỗi GPU xử lý rồi gộp kết quả cuối. #AI #MultiGPU #PCIe #GPU #MachineLearning #Tech #Vietnamese

https://www.reddit.com/r/LocalLLaMA/comments/1ougl1d/individual_models_or_data_sets_for_multigpu/

Giải pháp tạm thời cho vấn đề VRAM bị unloading sau thời gian idle trên setup đa GPU sử dụng Vulkan runtime. Mã code giữ 1MB trên VRAM và giữ GPU "Awake" bằng cách ping every 1 giây. #VRAM #Vulkan #MultiGPU #GPU #Laptop #Computer #CôngNghệ #VramUnloading #ĐaGPU

https://www.reddit.com/r/LocalLLaMA/comments/1oo3pdz/workaround_for_vram_unloading_after_idle_period/

If you have 2 MI50s and a 7900XT, can you use the 7900XT for prompt processing in MOE models while leveraging the faster VRAM of the MI50s? Or combine them for better performance? Thrilling tech experimentation! #TechTalk #VRAM #MOEmodels #GamingPC #Multigpu #C FamilienTechTalk #VRAMbộnhiền #môihínhthùng #GamePC #bộgpuđa

https://www.reddit.com/r/LocalLLaMA/comments/1oksprm/is_it_possible_to_use_vram_like_ram_is_multigpu/

OpenSUSE got new Switcherooctl tool, integrated into GNOME and KDE Plasma, to switch between GPUs on multi-GPU machines:
https://ostechnix.com/effortless-gpu-switching-on-opensuse-with-switcherooctl/
(It's now also preinstalled on Debian, I can right click on an icon on dash, and get the option to switch GPU.)

Fedora plans new COSMIC spin in version 42:
https://ostechnix.com/fedora-cosmic-spin/

Manjaro 24.2 released with Linux kernel 6.12 GNOME 47, KDE Plasma 6.2, Xfce 4.18:
https://ostechnix.com/manjaro-24-2-yonada-released/

Linux kernel 6.13 RC2 released with driver updates, improvements for ARM64, LoongArch and x86 architectures, networking and BPF fixes, filesystem and memory management fixes, watchdog and power management improvements:
https://ostechnix.com/linux-kernel-6-13-rc2-released/

KDE Gear 24.12 released, Dolphin file manager is ported to mobile, got revamped main view to work with screen readers, better file sorting, improved keyboard navigation, Checksum and Permissions tab in the Properties dialog; Tokodon Mastodon client got redesigned registration process and welcome page; Kate text editor got Git status visualization; Alligator RSS reader got option to bookmark posts; KDE Connect got Bluetooth support; Kdenlive video editor got option to resize multiple items on the timeline at once etc.:
https://9to5linux.com/kde-gear-24-12-officially-released-finally-ports-dolphin-to-mobile-devices

KWin 6.3 released with revamped fractional scaling, Zoom effect now offers a sharp representation at high zoom levels, double-clicking on window edges to maximize can now be disabled etc.:
https://ostechnix.com/kwin-6-3-brings-crystal-clear-fractional-scaling-to-kde-plasma/

KDE Frameworks 6.9 released with new symbolic icon for Wine, conversion to and from the CFP franc currency in KRunner, improved Breeze icons, various bug fixes:
https://9to5linux.com/kde-frameworks-6-9-released-with-new-symbolic-icon-for-wine-various-fixes

Proton 9.0-4 released with more supported games (including Total War: SHOGUN 2, Warhammer 40,000: Space Marine 2 and more), bug fixes and improved support for various games (including DOOM Eternal and Diablo IV) etc.:
https://9to5linux.com/proton-9-0-4-adds-total-war-shogun-2-warhammer-40k-space-marine-2-support

2/3
(see reply)

#WeeklyNews #FOSS #OpenSource #Linux #LinuxNews #OpenSUSE #Fedora #LinuxKernel #Manjaro #KDE #Proton #MultiGPU #DistroRelease #LinuxGaming #KWin #FosseryTech

The result is something that can be trivially parallelized with #OpenMP.
As an alternative, it's possible to use the #multiGPU support in #GPUSPH to run code in parallel. This is obviously less efficient, although it may be a good idea to use it in #NUMA setups (one thread per NUMA node, OpenMP for the cores in the same node). This is not implemented yet.

Finally there is some attention on multi-monitor woes on #GNOME!

Hopefully this makes using another monitor more performant and stable. CC to #NVIDIA as well.

https://thisweek.gnome.org/posts/2024/10/twig-170/

#Linux #FOSS #OSS #OpenSource #MultiMonitor #Display #Displays #Fedora #Ubuntu #Laptops #Laptop #MultiGPU #GPU #Graphics #Computers #Computer #PC #Desktop #DesktopPC #Computing

#170 Portal Updates · This Week in GNOME