the workshop's gaming laptop: an RTX 5090 Laptop GPU for the diffusion rematch, deliberately unused for the CPU-only bench
It holds — RTX 5090 Laptop GPU (24 GB).
Language — decode, prefill and first token
Sorted by model, then by power cap ascending — alphabetical and numeric, never by value. Every condition the page states is printed; a condition it does not state reads not stated, which is a third thing and not an uncapped card. The columns a narrow screen cannot hold are printed under the row rather than scrolled to.
| value | what was measured | model | power cap | runtime | measurement window |
|---|---|---|---|---|---|
| 5.0 tok/s | decode, median of five scored runs | gemma4:12b | not stated | Ollama · 0.32.15 | 2026-08-25 (UTC) |
| gemma4:12bnot statedOllama · 0.32.152026-08-25 (UTC) | |||||
| Q4_K_Ma 4,096-token contextOllama's own default of 4 threadsnot statedtwo-hours-on-battery · the-one-table-to-remember · carried by hand, proved against that page’s own row · identity from the table’s own column head | |||||
| 22.2 s | time to first token, median of five cold-prefix runs | gemma4:12b | not stated | Ollama · 0.32.15 | 2026-08-25 (UTC) |
| gemma4:12bnot statedOllama · 0.32.152026-08-25 (UTC) | |||||
| Q4_K_Ma 4,096-token context536 prompt tokens, Ollama's own default of 4 threadsnot statedtwo-hours-on-battery · the-wait-prefill-is-the-real-cpu-tax · carried by hand, proved against that page’s own row · identity from the table’s own column head | |||||
| 13.3 tok/s | decode, median of five scored runs | gemma4:26b | not stated | Ollama · 0.32.15 | 2026-08-25 (UTC) |
| gemma4:26bnot statedOllama · 0.32.152026-08-25 (UTC) | |||||
| Q4_K_Ma 4,096-token contextOllama's own default of 4 threadsnot statedtwo-hours-on-battery · the-one-table-to-remember · carried by hand, proved against that page’s own row · identity from the table’s own column head | |||||
| 8.5 s | time to first token, median of five cold-prefix runs | gemma4:26b (MoE) | not stated | Ollama · 0.32.15 | 2026-08-25 (UTC) |
| gemma4:26b (MoE)not statedOllama · 0.32.152026-08-25 (UTC) | |||||
| Q4_K_Ma 4,096-token context536 prompt tokens, Ollama's own default of 4 threadsnot statedtwo-hours-on-battery · the-wait-prefill-is-the-real-cpu-tax · carried by hand, proved against that page’s own row · identity from the table’s own column head | |||||
| 2.7 tok/s | decode, median of five scored runs | mistral-small3.2:24b | not stated | Ollama · 0.32.15 | 2026-08-25 (UTC) |
| mistral-small3.2:24bnot statedOllama · 0.32.152026-08-25 (UTC) | |||||
| Q4_K_Ma 4,096-token contextOllama's own default of 4 threadsnot statedtwo-hours-on-battery · the-one-table-to-remember · carried by hand, proved against that page’s own row · identity from the table’s own column head | |||||
| 82.4 s | time to first token, median of five cold-prefix runs | mistral-small3.2:24b | not stated | Ollama · 0.32.15 | 2026-08-25 (UTC) |
| mistral-small3.2:24bnot statedOllama · 0.32.152026-08-25 (UTC) | |||||
| Q4_K_Ma 4,096-token context1,023 prompt tokens, Ollama's own default of 4 threadsnot statedtwo-hours-on-battery · the-wait-prefill-is-the-real-cpu-tax · carried by hand, proved against that page’s own row · identity from the table’s own column head | |||||
| 15.4 tok/s | decode, median of five scored runs | nemotron-3.5-lightning:30b-a3b | not stated | Ollama · 0.32.15 | 2026-08-25 (UTC) |
| nemotron-3.5-lightning:30b-a3bnot statedOllama · 0.32.152026-08-25 (UTC) | |||||
| Q4_K_Ma 4,096-token contextOllama's own default of 4 threadsnot statedtwo-hours-on-battery · the-one-table-to-remember · carried by hand, proved against that page’s own row · identity from the table’s own column head | |||||
| 9.9 s | time to first token, median of five cold-prefix runs | nemotron-3.5-lightning:30b-a3b (MoE) | not stated | Ollama · 0.32.15 | 2026-08-25 (UTC) |
| nemotron-3.5-lightning:30b-a3b (MoE)not statedOllama · 0.32.152026-08-25 (UTC) | |||||
| Q4_K_Ma 4,096-token context532 prompt tokens, Ollama's own default of 4 threadsnot statedtwo-hours-on-battery · the-wait-prefill-is-the-real-cpu-tax · carried by hand, proved against that page’s own row · identity from the table’s own column head | |||||
| 4.4 tok/s | decode, median of five scored runs | phi4:14b | not stated | Ollama · 0.32.15 | 2026-08-25 (UTC) |
| phi4:14bnot statedOllama · 0.32.152026-08-25 (UTC) | |||||
| Q4_K_Ma 4,096-token contextOllama's own default of 4 threadsnot statedtwo-hours-on-battery · the-one-table-to-remember · carried by hand, proved against that page’s own row · identity from the table’s own column head | |||||
| 26.4 s | time to first token, median of five cold-prefix runs | phi4:14b | not stated | Ollama · 0.32.15 | 2026-08-25 (UTC) |
| phi4:14bnot statedOllama · 0.32.152026-08-25 (UTC) | |||||
| Q4_K_Ma 4,096-token context505 prompt tokens, Ollama's own default of 4 threadsnot statedtwo-hours-on-battery · the-wait-prefill-is-the-real-cpu-tax · carried by hand, proved against that page’s own row · identity from the table’s own column head | |||||
| 7.5 tok/s | decode, median of five scored runs | qwen3.5:9b | not stated | Ollama · 0.32.15 | 2026-08-25 (UTC) |
| qwen3.5:9bnot statedOllama · 0.32.152026-08-25 (UTC) | |||||
| Q4_K_Ma 4,096-token contextOllama's own default of 4 threadsnot statedtwo-hours-on-battery · the-one-table-to-remember · carried by hand, proved against that page’s own row · identity from the table’s own column head | |||||
| 14.1 s | time to first token, median of five cold-prefix runs | qwen3.5:9b | not stated | Ollama · 0.32.15 | 2026-08-25 (UTC) |
| qwen3.5:9bnot statedOllama · 0.32.152026-08-25 (UTC) | |||||
| Q4_K_Ma 4,096-token context537 prompt tokens, Ollama's own default of 4 threadsnot statedtwo-hours-on-battery · the-wait-prefill-is-the-real-cpu-tax · carried by hand, proved against that page’s own row · identity from the table’s own column head | |||||
Pictures — seconds per render
1 row — same rule as above.
| value | what was measured | model | power cap | runtime | measurement window |
|---|---|---|---|---|---|
| 5,191,824 bytes (5.2 MB), 1.12 s | the whole board as one sprite, 144 cells | not stated | not stated | not stated | 2026-09-09 15:45:32Z to 15:46:47Z |
| not statednot statednot stated2026-09-09 15:45:32Z to 15:46:47Z | |||||
| not statednot stated144 cells, 128 pixels a cellnot statedhow-the-print-lab-works · numbers-so-it-isnt-just-talk · carried by hand, proved against that page’s own row · identity from the same section | |||||
Power, temperature, and what it costs
1 row — same rule as above.
| value | what was measured | model | power cap | runtime | measurement window |
|---|---|---|---|---|---|
| 59.3 GB/s | memory bandwidth, best triad | not stated | not stated | a STREAM triad, gcc -O3 -march=native -fopenmp · version not stated | 2026-08-25 (UTC), 17:57–19:46Z |
| not statednot stateda STREAM triad, gcc -O3 -march=native -fopenmp · version not stated2026-08-25 (UTC), 17:57–19:46Z | |||||
| not statednot stateda 2,289 MiB working set, best of 9 timed repsnot statedtwo-hours-on-battery · read-this-before-the-table-cpu-speed-is-a-memory-bandwidth · carried by hand, proved against that page’s own row · identity from the row itself | |||||
Where it appears
Every released exhibit that names this identity, in the shelf’s own order. A count beside an exhibit is how many of its figures are carried above — and it links them, because a number with no way back to what it counts is the thing this shelf exists against.
- A Rig Your Friend Already Owns — named, and no figure carried from it — yet
- How the print lab works — 1 figure on this page
- Two Hours at 12 tok/s — 13 figures on this page
How this one is named here
The canonical name above is the full product name. Everything below is a spelling a published page or kit has actually used for it, kept so a reader who searches an older page’s words lands on this row. The set is closed and historical: no page cut after 2026-09-17 writes a memory class where it can write a card.
- Rig B — the laptop: 24G-VRAM class, GPU deliberately unused here.
- Rig B laptop
- Rig B †
- a gaming laptop
- cpu-laptop
- test Rig B, a laptop: 24 logical CPUs - 8 performance cores with no hyperthreading plus 16 efficiency cores; hybrid Intel part; a 24G-VRAM-class GPU present and parked
- the gaming laptop
- the workshop's laptop