Releases·Hardware Release
Framework lists 192GB motherboard option on its website
A Reddit user spotted a 192GB Framework motherboard SKU now listed on the company's website. Based on Framework's existing memory-tier pricing (32GB, 64GB, 128GB), the poster estimates the board at roughly $4.5k. The listing suggests Framework is extending its high-memory desktop board lineup toward workstation/local-LLM territory.
1150 engagement1 sourceSun, Aug 30, 2026, 05:39 AM
AI Overview
- Poster estimates ~$4.5k for the motherboard by extrapolating Framework's current 32/64/128GB memory SKU price tiers
- PCIe slot reportedly open at the back of the board, per what the poster has heard
- Poster speculates the board might deliver 75W and that smaller-SKU board revisions could follow
Related
Social·Community
User seeks critiques on local LLM server upgrade plan
A user running a local LLM server on a NAS with an Intel i9-12900K and multiple GPUs is planning an upgrade to an EPYC-based system with 8-channel memory and more PCIe slots, asking for community advice. The plan targets value and future scalability.
A user running a local LLM server on a NAS with an Intel i9-12900K and multiple GPUs is planning an upgrade to an EPYC-based system with 8-channel memory and more PCIe slots, asking for community advice. The plan targets value and future scalability.
18 engagement1 sourceredditThu, Jul 16, 2026, 09:07 PM
Social·Community
12GB VRAM users weigh upgrade vs. MoE fine-tunes as dense models dominate
A Reddit user with 12GB VRAM asks the community for a plan, noting that current focus is on dense models like Muse Glimmer 30B and Qwen 3.8 27B, which are too large for their setup. They consider upgrading to 24GB VRAM as the only option, while acknowledging fine-tuned MoEs as a temporary workaround.
A Reddit user with 12GB VRAM asks the community for a plan, noting that current focus is on dense models like Muse Glimmer 30B and Qwen 3.8 27B, which are too large for their setup. They consider upgrading to 24GB VRAM as the only option, while acknowledging fine-tuned MoEs as a temporary workaround.
75 engagement1 sourceredditTue, Aug 11, 2026, 07:31 AM
Social·Community
Users weigh Lucebox vs Framework Desktop for DeepSeek-V4-Flash-0731
A Reddit user asks whether to buy Lucebox's hardware for running DeepSeek-V4-Flash-0731 or wait for the upcoming Framework Desktop with a Ryzen AI Max+ PRO 495 and 192 GB RAM, possibly paired with a Radeon AI PRO R9700 via a PCIe adapter. The thread reflects community deliberation on optimal local inference hardware.
A Reddit user asks whether to buy Lucebox's hardware for running DeepSeek-V4-Flash-0731 or wait for the upcoming Framework Desktop with a Ryzen AI Max+ PRO 495 and 192 GB RAM, possibly paired with a Radeon AI PRO R9700 via a PCIe adapter. The thread reflects community deliberation on optimal local inference hardware.
45 engagement1 sourceredditWed, Aug 19, 2026, 05:55 AM
Releases·Hardware Release
Leaked 50-series Super specs show 5070 Ti and 5080 jumping to 24GB VRAM
Leaked specs for Nvidia's rumored 50-series Super refresh indicate the 5070 Ti and 5080 will double from 16GB to 24GB VRAM, and the 5070 to 18GB, using new 3GB GDDR7 modules. The 24GB Ti-class card is seen as a boon for local AI model inference, potentially reducing the need for used 3090s or 90-class cards.
Leaked specs for Nvidia's rumored 50-series Super refresh indicate the 5070 Ti and 5080 will double from 16GB to 24GB VRAM, and the 5070 to 18GB, using new 3GB GDDR7 modules. The 24GB Ti-class card is seen as a boon for local AI model inference, potentially reducing the need for used 3090s or 90-class cards.
157 engagement1 sourceredditTue, Aug 11, 2026, 07:15 AM
Social·Community
Imaging pro seeks AM4 board for dual RTX 3090 ComfyUI setup
An imaging professional asks for AM4 motherboard recommendations to run dual RTX 3090s for ComfyUI batch image edits, video editing, and local agent work with Qwen models. The user seeks a board with at least two PCIe slots, ideally a third for display output, on a budget that avoids AM5.
An imaging professional asks for AM4 motherboard recommendations to run dual RTX 3090s for ComfyUI batch image edits, video editing, and local agent work with Qwen models. The user seeks a board with at least two PCIe slots, ideally a third for display output, on a budget that avoids AM5.
40 engagement1 sourceredditSat, Aug 22, 2026, 09:11 PM