#nvidia

Posts on X2Q tagged nvidia. Updated as new posts are published.

• Jul 13, 2026

  First look at the NVIDIA DGX Spark — what the box actually reports (2026)

  The NVIDIA DGX Spark reports itself as a 20-core Arm machine (10× Cortex-X925 + 10× Cortex-A725), 121 GiB of unified LPDDR5x memory, a single GB10 Blackwell GPU on CUDA 13, and a 4 TB Samsung NVMe — all in a gold box the size of a thick paperback. This first-look post reads the whole machine back from the command line (lscpu, free -h, nvidia-smi, lspci, lsblk) and shows it doing the one thing no consumer GPU can: holding a 67 GB llama4:scout and a 65 GB gpt-oss:120b model in memory at the same time.

• Jun 22, 2026

  Apple M1 Pro vs RTX 4070 Ti for local AI — GPU matmul and LLM benchmarks (2026)

  RTX 4070 Ti hits 75 TFLOPS at fp16 via tensor cores; M1 Pro MPS reaches 4.2 TFLOPS — an 18× compute gap. But on real LLM inference (minicpm-v4.6, qwen3.5, gemma4:12b via ollama), the RTX is only 3–4× faster. LLM inference at batch=1 is memory-bandwidth-bound, not compute-bound, and the M1's unified-memory architecture closes most of the gap. Apple's AMX extensions also give its CPU 2.7× better matmul throughput than an Intel i5 with the same thread count.

• Jun 7, 2026

  Extending video clips with AI on a 12 GB GPU — six models compared

  You have a short machine/action clip and want it longer. 'Extending' a clip is really image-to-video: take the last frame and generate a plausible continuation. I fed the same shot to six models — Wan2.2-TI2V-5B, Wan2.2-I2V-A14B (GGUF), Stable Video Diffusion, LTX-Video and CogVideoX-5B — all on a single 12 GB RTX 4070 Ti. The interesting part isn't which looks best; it's the memory-offload tricks that make a 14B video model (and 720p) fit in 12 GB at all.

• Jun 2, 2026

  Five ways to get b-roll from one ride — all on a local GPU

  B-roll is the cutaway footage that makes an edit breathe — and onboard POV footage from a single camera has none. I tried five ways to manufacture it from one ATV ride: CV auto-mining of the real footage, generative stills (SDXL-Turbo → RealVisXL img2img), AI Ken Burns via Depth-Anything-V2 parallax, true generated motion via Stable Video Diffusion, and optical-flow slow-motion. All local on a 12 GB card. Here's what each is good for, and the gotchas.