-
- Private work.
- Upgraded macbook to 14.0/sonoma (from 13.5.2/ventura). Skimmed the changelog, nothing affecting my nominal use.
- Never got the (promised) update from liveaquaria last night. Called back again today, and now not 1 but ~6 items are unavailable (half the order). Had her send a message to the team to “ship whatever available, even <75” so we’ll see what happens. After this order, I’m done with liveaquaria – their stock is too unreliable.
- Ordered fresh. Also mealprepped (vermicelli + lentils + pesto, sausage + chicken breast + bacon).
- Computing.
- GPU microarchitectures (enterprise) from NVIDIA: Maxwell -> Pascal -> Volta -> Ampere -> Hopper -> Blackwell (coming soon). These are the full architectures for datacenters (including switches, hw/sw, the actual GPUs, more).
- CUDA = Compute Unified Device Architecture. Software API (usually C++) to GPUs.
- Some GPUs from NVIDIA: Tesla, Titan.
- Some GPUs from Intel: Movidius Myriad, Xeon Phi.
- Hybrid models, CPUs and GPUs working in concert. CPUs handling basic tasks, GPUs handling highly-parallel tasks. Hundreds of cores. Can process a lot. Very quickly.
- Matrix multiplication with tensor cores.
- Supercontest. Confirmed the scorefetching worked properly after the score day UTC fix.
