Exxact powers the world's greatest advancements with HPC and AI infrastructures that accelerate innovation and solve complex problems.
Fremont, CA
Joined March 2011
- Tweets3.4K
- Following566
- Followers8.5K
- Likes5.2K
Building a 4× NVIDIA DGX Spark cluster? Here’s the switch, QSFP cabling, and power plan to scale beyond direct-link setups—plus key config tips. Read more. buff.ly/mZldTiH
#NVIDIA
MD jobs failing or crawling? Here are 3 signs your GPU is the bottleneck—and how to pick the right upgrade for AMBER, GROMACS, or NAMD. buff.ly/vwX0MW5
#MolecularDynamics #GPU
Claude Code down or too pricey? Run it locally. Point ANTHROPIC_BASE_URL to DS4 or Ollama, keep code private, and cut usage costs—plus benchmark results on DGX Spark. Read more → buff.ly/kdSN5ys
#Claude #Ollama
One big model or many agents at once? See how Exxact Valence GB300 compares to a 4x RTX PRO 6000 Blackwell workstation for agentic AI. Read more. buff.ly/tfj76ea
#GB300 #AgenticAI
What can two NVIDIA DGX Station GB300 systems unlock for agentic AI? Explore model sharding, long-context inference, high availability, and when to step up to HGX B300. Read the full breakdown. buff.ly/lmb8htE
@NVIDIA @NVIDIAAI
#DGXStation
AMD EPYC 9006 brings more than higher core counts. Explore Zen 6, SP7 vs SP8, and what's changing from Turin to Venice. Read our guide to plan your next server upgrade. buff.ly/scOrVvx
@AMD #AMDEYPC #9006
Run frontier-scale LLMs privately. NVIDIA DGX Station GB300 brings 748GB unified memory + data center AI to the desk. See which workloads benefit most. buff.ly/JGBsiVm
#NVIDIA #DGXStation
One big model or many agents at once? See how Exxact Valence GB300 compares to a 4x RTX PRO 6000 Blackwell workstation for agentic AI. Read more. buff.ly/tfj76ea
#GB300 #AgenticAI
Exxact Corporation retweeted
My home AI lab, powered by enough solar and batteries to be off grid:
2x Nvidia DGX Stations GB300, each with RTX Pro 6000
3x DGX Sparks, Founder Edition
Threadripper linux server, 512GB
Alienware Area 51 interface
Synopsis NAS and Mikrotik
I run local for everything it can do, frontier where it still wins. I'll make sure and post some numbers soon.
@NVIDIAAI
One founder. One NVIDIA DGX Spark™. An entire company. OptiPrime AI runs 53 production services and 56 local models on a single system, including a legal research tool built to return the same cited answer every time. buff.ly/IyaOlFP
#NVIDIA #DGXSpark #LocalAI
Building agentic workflows for research? Keep agents close to your data and tools. Exxact hardware options help you run retrieval + inference + execution with responsive performance. buff.ly/ZNz7k9Q
#LifeSciences #HPC
Ran a 284B model on one NVIDIA DGX Spark. We benchmarked Ollama vs vLLM vs DS4 for speed, concurrency, and tool-calling—and picked the best setup for single-user + multi-tenant. Read the results. buff.ly/tckjtEO
#NVIDIA #DGXSpark
Building a 4× NVIDIA DGX Spark cluster? Here’s the switch, QSFP cabling, and power plan to scale beyond direct-link setups—plus key config tips. Read more. buff.ly/mZldTiH
#NVIDIA
RF-DETR vs YOLO: how transformers boost object detection accuracy, reduce tuning, and scale for real-world CV. Read the comparison. buff.ly/rIINraB
#Computervision #YOLO #Transformers
Can a local LLM behave like a reliable agent? We benchmarked Ollama + OpenClaw on NVIDIA DGX Spark—tool calling, hop depth, speed, and who holds up in real workflows. Read the results. buff.ly/9nZNN7e
#LLM #Ollama
One big model or many agents at once? See how Exxact Valence GB300 compares to a 4x RTX PRO 6000 Blackwell workstation for agentic AI. Read more. buff.ly/tfj76ea
#GB300 #AgenticAI
We benchmarked DeepSeek V4 Flash on a single Exxact Workstation with GB300 and saw reliability across 10 simultaneous 1M-token sessions. Read how we did it & reproduce it yourself. buff.ly/7ruWX0u
#GB300
MD jobs failing or crawling? Here are 3 signs your GPU is the bottleneck—and how to pick the right upgrade for AMBER, GROMACS, or NAMD. buff.ly/vwX0MW5
#MolecularDynamics #GPU
Ansys Fluent 2026 GPU benchmarks are in: RTX PRO 5000 vs 6000 vs H200 NVL, 1–8 GPU scaling, NVLink surprises, plus precision modes that cut VRAM. See the results + hardware recommendations. buff.ly/a3HKEEz
#Ansys #GPU
Ran a 284B model on one NVIDIA DGX Spark. We benchmarked Ollama vs vLLM vs DS4 for speed, concurrency, and tool-calling—and picked the best setup for single-user + multi-tenant. Read the results. buff.ly/WCnebx2
#AgenticAI
