Price!!!
en.gamegpu.com/news/zhelezo/…
Sources at major retail chains report that NVIDIA has updated its internal backend schedules and plans to discontinue the flagship GeForce RTX 5090 graphics card as early as the beginning of 2027. Retailers use these dates for long-term financial planning.
At the same time, all product listings for the unreleased GeForce RTX 5080 24 GB have been completely removed from retailer databases. Although engineering samples have been in partner labs since early 2025, the model is now effectively canceled.
This early discontinuation of the current flagship strongly suggests NVIDIA is clearing the top tier for the GeForce RTX 6090, which is expected to launch in late 2027 or early 2028 at $2500 to $3000.
#NVIDIA #RTX5090 #RTX5080 #RTX6090 #GeForce #GPU #Hardware
5
Can't understand why many mini PC makers are still STUPID, not realizing RAM issue because of @SKhynix & @SamsungUS !!!
9
Still no Android version, yet!
Meet the new Claude mobile app!
So much craft and polish across the app, including a new tab bar. Huge shoutout to @daniel__designs, @khalilcader, @pranathiperii, Andrew, and Graham for all their work across iOS and Android.
Rolling out to all users, starting today!
12
I love the idea 💡 Local LLM with Omarchy is the way!
🆕New AI Hardware! Phison wants to put an AI processor inside the SSD!
And give the SSD its own 64GB of high-speed memory.
Earlier, I posted about Phison's aiDAPTIV+ technology, which uses SSDs to help run huge AI models on PCs with limited RAM?
Well, Phison just revealed something very different at Supercomputing 2026.
Instead of using the SSD only as an extended memory tier, the SSD actually runs AI computations itself? 👀
Phison has unveiled a new computational-storage architecture.
Here are the proposed specs ...
🧠 Integrated NPU
💾 64GB onboard memory
🚀 130 GB/s internal memory bandwidth
⚙️ 36,800 DMIPS reported compute metric
And Phison says it's designed to support familiar AI frameworks such as
🐍 PyTorch
🚀 vLLM
🦙 llama.cpp
🔧 OpenVINO
Phison claims the architecture could reduce data transfers between storage and the host by 10–100× for applicable workloads.
Here's the idea
TODAY
💾 SSD stores model weights
🧠 RAM caches model data
🎮 GPU performs the computations
PHISON'S NEW APPROACH
💾 SSD stores data
🧠 SSD has its own 64GB memory
⚡ SSD has an integrated AI accelerator
🔗 Host receives processed results
Instead of always transferring MoE expert weights from storage into GPU memory, perhaps some experts could be processed directly on the drive, returning intermediate results.
⚠️ This is still an enterprise SSD design under development. No LLM generation benchmarks or shipping product have been announced.
But consider where this could go.
Links in ALT.

ALT https://www.phison.com/sc26-phison-unveils-three-next-generation-storage-innovations-for-the-new-compute-era/ https://www.tomshardware.com/pc-components/ssds/phison-details-ssd-with-a-built-in-npu-and-64gb-of-memory-that-supports-pytorch-and-llama-cpp-running-on-the-drive-second-design-merges-fast-pslc-cache-and-qlc-capacity-behind-one-controller
18
AGI is NOT achieved, because of memory loss, no predictions of what's needs to be done next and also I have always to ask for audit and QA! @OpenAI, @GeminiApp and @AnthropicAI
9
Speed it UP! Local LLM will reach a higher level 🚄
La DDR6 se précise et ça promet du lourd : jusqu'à 12 800 MT/s et des barrettes beaucoup plus compactes ! 🚀
👉 f.mtr.cool/r6sh6d0whf
12
I have a Big heart ❤️ for many Linux distros and here are my favorite:
1- @omarchy as best Agentic AI distro
2- @Linux_Mint as best simplicity Distro
3- @fedora still my daily driver
4-@pop_os_official best Gaming and multi GPUs distro with Cosmic DE for personalization.
5- Debian for Severs
2
25
So, now I'm tired of playing ping pong 🏓 switching between @grok bot and @AnthropicAI Claude code Dispatch which plays well as an alternative to Duet 😔 So, Duet needs to be available soon and Grok Bot needs to deliver Grok 4.8!
4
1
121
This is going to be fantastic! But, we need a new @elonmusk RAM factory 🏭!
Future of Local AI is looking bright !!
JDEC previews DDR6 and i really hope it makes it's way to desktops soon.
The target: 12.8 GT/s with 14.4 GT/s and beyond on the roadmap.
How it translates.
I'm running 2x32 GB of DDR5-6000 right now. If i replace that with DDR6 12800 MT/s two sticks, that'd be 3.2x the theortetical memory bandwidth at the same capacity.
So instead of 96 GB/s i could be running around 307 GB/s !!
That's small graphics card territory !
That could be a big deal for running MoE models locally, especially when some of the experts have to sit in system RAM.
videocardz.com/newz/jedec-an…
2
3
125
Use TrueNAS the Linux Os*
‼️ BREAKING: Synology warns of an unpatched critical flaw that lets remote attackers read or write any file on them, no login needed.
The bug, CVE-2026-120382, sits in the OAuth sign-in feature of its DSM operating system and scores 9.8 out of 10.
Patches for DSM 7.3 and 7.4 are due by 20 November. Older DSM versions and BeeStation have no fix date yet.
Until then, Synology says to switch off the OAuth Service package.
13
I have to admit I Love Grok Bot, very reliable, BUT in tokens usage it's the worst! Claude Duet from @AnthropicAI and Dots from @OpenAI still have many things to learn how it works out in BOTS World!
42
Linux too, Thanks ! Dont push us to buy @intel CPUs 😂
More possibilities for local AI, across connected @Windows PCs.
AMD is collaborating with @Microsoft on the RDMA over USB4 developer preview for Windows, helping developers explore larger models across connected PCs.
"At AMD, we believe clustering is a critical part in a developer's experience to unlocking new possibilities for local AI and high-performance compute. We're excited to be working alongside Microsoft to support their new RDMA over USB4 driver to enhance the experience." - Andrej Zdravkovic, Senior Vice President and Chief Software Officer, AMD
Learn more: techcommunity.microsoft.com/…
8
Cool 😎
Unsloth Desktop Now Supports Using The @AMD @AMDRyzen AI NPU - Including On Linux
phoronix.com/news/Unsloth-De…
9
Yeah why not!
Introducing the Firecrawl plugin for Grok Bot.
Give your agents 3,300+ data tools, specialized indexes & the live web, all from Firecrawl Alexandria, the knowledge library for agents.
Lift answer quality 21% over @bot's built-in web tools.
x.ai/bot/plugin/789
4






















