Author
BitByteCore Silicon Desk
Stories by BitByteCore Silicon Desk, newest first.
12+ stories

Quantization Explained: What Q4, Q8, and FP16 Actually Do to a Local Model
Quantization shrinks AI model weights from 16-bit floats to 4- or 8-bit. Here is exactly what that trade-off costs you, and how to pick between Q4_K_M, Q8, and FP16 for the model you actually want to run, with an interactive memory calculator.
BitByteCore Silicon Desk · Aug 6, 2026 · 8 min read

CUDA Lock-In Is Real: A Precise Cost Accounting of What Switching GPU Vendors Actually Breaks
Switching away from CUDA isn't one migration — it's five separate porting problems, hidden revalidation costs, and an org chart that fights you the whole way.
BitByteCore Silicon Desk · Aug 6, 2026 · 9 min read

Mixture-of-Experts Models: How They Work and Why They Cut Inference Costs
MoE models activate only a fraction of their parameters per token: DeepSeek-V3 fires 37B of 671B, delivering large-model quality at small-model compute cost. How routers, experts, and load balancing actually work, and where the savings are and are not real. With an interactive dense-vs-MoE view.
BitByteCore Silicon Desk · Aug 6, 2026 · 10 min read

Why only a few foundries make the leading-edge chips
Leading-edge chip manufacturing has collapsed to a handful of companies, in practice one clear leader and two challengers. The reasons are structural: staggering cost, an unforgiving learning curve, and a feedback loop that punishes falling behind. Here is how it works.
BitByteCore Silicon Desk · Jul 31, 2026 · 8 min read

What a Nanometer Process Node Really Means
The number on a process node, 7nm, 3nm, now 2nm, is a marketing label, not a measurement of anything physical on the chip. Here is what it actually tracks, and the metrics that describe a chip honestly.
BitByteCore Silicon Desk · Jul 31, 2026 · 9 min read
More stories
Article · chipsCoWoS, SoIC, and Foveros: How Advanced Chip Packaging Actually WorksJul 30, 2026 · 11 min read
Article · chipsTSMC's Pricing Power: How One Foundry Sets the Cost of Every Advanced ChipJul 30, 2026 · 10 min read
Article · chipsHow EUV Lithography Works, in Plain TermsJul 30, 2026 · 8 min read
Article · chipsAI Inference on the Edge: How Embedded Chips in Cars, Cameras, and Appliances Actually WorkJul 29, 2026 · 12 min read
Article · chipsRISC-V vs ARM vs x86: The Honest ComparisonJul 29, 2026 · 10 min read
Article · chipsHBM Explained: Why High-Bandwidth Memory Is the Real Bottleneck in AI ChipsJul 29, 2026 · 12 min read
Article · hardwareWhy Battery Life Is a Chip and Software StoryJul 28, 2026 · 11 min read