03Why are GPUs so much faster than CPUs for deep learning? Be specific about what the silicon is doing differently.▼easyNewNVIDIAGoogle4 repliesunlockedNot 'more cores.' The real answers are a 10x memory system, a 16x matrix datapath, and a design that spends transistors on lanes and registers instead of on making one thread wait less. With the numbers for a two-socket server against one H100.Open full answer →
08How much faster is an H200 than an H100, really? Which workloads see the gain and which do not?▼mediumNewNVIDIACoreWeaveLambda4 repliesunlockedThe H200 has the same compute die as the H100 and costs more per hour. The datasheet gives two ratios, 1.43x bandwidth and 1.76x memory, and those two numbers decide exactly which workloads pay back the premium and which ones lose money on it.Open full answer →
12Explain what HBM is and why memory bandwidth, not compute, is the wall for LLM inference.▼mediumNewNVIDIAAMDMicron4 replies○ sign inHBM stacks DRAM dies on top of each other and wires them to the GPU through a silicon interposer with a 1,024-bit bus per stack. That design sets how much bandwidth and capacity a card can have, why the two scale together, and why they have grown more slowly than FLOPS across three generations.Open full answer →
29If you could change one thing about GPU architecture for LLMs, what would it be, and what argues against it?▼expertNewNVIDIAAnthropicOpenAI4 replies◆ premiumThe workload asks for bytes and the chip delivers FLOPs: bytes per FLOP halved from A100 to H100 and held level on B200. A defensible thesis is more bandwidth and capacity per FLOP, the shoreline and power arithmetic for why it has not happened, and the counterargument that software already routes around the wall.Open full answer →
26nvidia-smi shows 30 percent utilization on our serving fleet. Is that a problem, and what would you look at instead?▼mediumNewBasetenModal4 replies◆ premiumThe number nvidia-smi calls utilization measures whether any kernel was running, not whether the chip was busy. A decode step that saturates HBM can read as thirty percent. The right dashboard has four other numbers on it.Open full answer →