Cold Prefill TPS · Peak AI SSD speedup
×1.08
gpt-oss:20b · CTX 128K · Input 1K
Performance Lab / Verified Data
Multi-platform, multi-model inference comparison between AI SSD and standard SSD. Charts use measured results from the same platform.
AI SSD significantly improves hot Prefill and Decode over standard SSD. Each model shows results at its maximum measured context: cyan is AI SSD, gray dashed is standard SSD, and OOM means insufficient memory.
Cold Prefill TPS · Peak AI SSD speedup
×1.08
gpt-oss:20b · CTX 128K · Input 1K
Hot Decode TPS · Peak AI SSD speedup
×1.39
Qwen3.5_122B_A10B · CTX 256K · Input 8K
Cold TTFT · Peak AI SSD speedup
×1.07
gpt-oss:20b · CTX 128K · Input 1K
热 TTFT · Peak AI SSD speedup
×349.2
Gemma4:26B-A4B-IT · CTX 128K · Input 64K
| Input | Mode | Cold TTFT (ms) | Cold Prefill TPS (tokens/s) | Hot Decode TPS (tokens/s) | 热 TTFT (ms) | Hot Prefill speedup |
|---|---|---|---|---|---|---|
| 1K | AI SSD | 969.0 | 1302 | 70.57 | 154.0 | ×9.42 |
| 1K | Standard SSD | 1037 | 1209 | 66.91 | 1034 | ×1.00 |
| 2K | AI SSD | 1652 | 1392 | 65.95 | 234.0 | ×9.37 |
| 2K | Standard SSD | 1621 | 1412 | 64.88 | 1646 | ×0.98 |
| 4K | AI SSD | 3144 | 1389 | 62.93 | 266.0 | ×16.2 |
| 4K | Standard SSD | 3117 | 1403 | 59.28 | 3161 | ×0.99 |
| 8K | AI SSD | 6463 | 1316 | 62.81 | 368.0 | ×24.7 |
| 8K | Standard SSD | 6304 | 1349 | 55.44 | 6352 | ×0.99 |
| 16K | AI SSD | 14417 | 1161 | 59.31 | 558.0 | ×36.2 |
| 16K | Standard SSD | 14251 | 1176 | 52.10 | 14202 | ×1.00 |
| 32K | AI SSD | 37052 | 895.2 | 56.06 | 903.0 | ×57.0 |
| 32K | Standard SSD | 36234 | 916.2 | 48.61 | 36468 | ×0.99 |
| 64K | AI SSD | 99152 | 665.8 | 42.06 | 1649 | ×83.8 |
| 64K | Standard SSD | 97228 | 678.9 | 40.61 | 97632 | ×1.00 |
| 128K | AI SSD | - | - | - | - | - |
| 128K | Standard SSD | - | - | - | - | - |
| Input | Mode | Cold TTFT (ms) | Cold Prefill TPS (tokens/s) | Hot Decode TPS (tokens/s) | 热 TTFT (ms) | Hot Prefill speedup |
|---|---|---|---|---|---|---|
| 1K | AI SSD | 2088 | 590.5 | 51.11 | 198.0 | ×16.4 |
| 1K | Standard SSD | 2217 | 549.7 | 51.56 | 2197 | ×1.01 |
| 2K | AI SSD | 3458 | 655.0 | 50.43 | 235.0 | ×22.5 |
| 2K | Standard SSD | 3604 | 625.6 | 49.18 | 3625 | ×1.00 |
| 4K | AI SSD | 6362 | 680.1 | 49.55 | 290.0 | ×32.5 |
| 4K | Standard SSD | 6436 | 671.8 | 43.57 | 6512 | ×0.99 |
| 8K | AI SSD | 12937 | 652.3 | 48.10 | 404.0 | ×46.1 |
| 8K | Standard SSD | 12931 | 651.9 | 41.98 | 12948 | ×1.00 |
| 16K | AI SSD | 28650 | 581.3 | 41.00 | 630.0 | ×62.6 |
| 16K | Standard SSD | 28250 | 589.6 | 39.61 | 28415 | ×0.99 |
| 32K | AI SSD | 69630 | 474.9 | 40.28 | 1048 | ×89.5 |
| 32K | Standard SSD | 68523 | 482.5 | 35.86 | 68629 | ×1.00 |
| 64K | AI SSD | 175855 | 374.7 | 32.68 | 1915 | ×123.1 |
| 64K | Standard SSD | 173387 | 380.0 | 30.11 | 173736 | ×1.00 |
| 128K | AI SSD | - | - | - | - | - |
| 128K | Standard SSD | - | - | - | - | - |
| Input | Mode | Cold TTFT (ms) | Cold Prefill TPS (tokens/s) | Hot Decode TPS (tokens/s) | 热 TTFT (ms) | Hot Prefill speedup |
|---|---|---|---|---|---|---|
| 1K | AI SSD | 4630 | 261.8 | 29.27 | 745.0 | ×6.44 |
| 1K | Standard SSD | 4005 | 299.7 | 22.21 | 4024 | ×1.00 |
| 2K | AI SSD | 7794 | 287.8 | 28.94 | 768.0 | ×10.5 |
| 2K | Standard SSD | 7101 | 314.1 | 21.57 | 7069 | ×1.00 |
| 4K | AI SSD | 14725 | 291.8 | 27.93 | 862.0 | ×17.8 |
| 4K | Standard SSD | 13444 | 318.7 | 20.80 | 13503 | ×1.00 |
| 8K | AI SSD | 27905 | 301.4 | 28.37 | 953.0 | ×30.7 |
| 8K | Standard SSD | 26866 | 312.7 | 20.44 | 26861 | ×1.00 |
| 16K | AI SSD | 57648 | 288.3 | 25.64 | 1191 | ×51.3 |
| 16K | Standard SSD | 55674 | 298.4 | 20.19 | 55803 | ×1.00 |
| 32K | AI SSD | 127878 | 258.4 | 24.85 | 1545 | ×87.2 |
| 32K | Standard SSD | 124014 | 266.4 | 19.14 | 124298 | ×1.00 |
| 64K | AI SSD | 304732 | 216.3 | 23.30 | 1731 | ×194.3 |
| 64K | Standard SSD | 298399 | 220.9 | 17.87 | 298668 | ×1.00 |
| 128K | AI SSD | 815325 | 161.4 | 19.71 | 2877 | ×322.7 |
| 128K | Standard SSD | 794568 | 165.6 | 16.00 | 794622 | ×1.00 |
| Input | Mode | Cold TTFT (ms) | Cold Prefill TPS (tokens/s) | Hot Decode TPS (tokens/s) | 热 TTFT (ms) | Hot Prefill speedup |
|---|---|---|---|---|---|---|
| 1K | AI SSD | 1742 | 700.6 | 68.60 | 304.0 | ×6.42 |
| 1K | Standard SSD | 1212 | 999.8 | 57.06 | 1217 | ×1.00 |
| 2K | AI SSD | 2371 | 954.0 | 68.26 | 284.0 | ×9.07 |
| 2K | Standard SSD | 2098 | 1073 | 56.75 | 2106 | ×1.00 |
| 4K | AI SSD | 4174 | 1035 | 67.10 | 331.0 | ×13.9 |
| 4K | Standard SSD | 3861 | 1117 | 55.75 | 3862 | ×1.00 |
| 8K | AI SSD | 8347 | 1011 | 64.90 | 402.0 | ×23.5 |
| 8K | Standard SSD | 7802 | 1080 | 54.16 | 7828 | ×1.00 |
| 16K | AI SSD | 17789 | 936.4 | 61.18 | 503.0 | ×40.4 |
| 16K | Standard SSD | 16986 | 979.9 | 52.08 | 16993 | ×1.00 |
| 32K | AI SSD | 41348 | 800.5 | 56.23 | 707.0 | ×68.0 |
| 32K | Standard SSD | 39499 | 837.7 | 48.00 | 39620 | ×1.00 |
| 64K | AI SSD | 106536 | 619.3 | 47.80 | 981.0 | ×131.8 |
| 64K | Standard SSD | 101442 | 650.3 | 41.33 | 101738 | ×1.00 |
| 128K | AI SSD | 308657 | 426.7 | 37.21 | 1787 | ×211.6 |
| 128K | Standard SSD | 293804 | 448.2 | 32.89 | 294137 | ×1.00 |
| Input | Mode | Cold TTFT (ms) | Cold Prefill TPS (tokens/s) | Hot Decode TPS (tokens/s) | 热 TTFT (ms) | Hot Prefill speedup |
|---|---|---|---|---|---|---|
| 1K | AI SSD | 6299 | 182.7 | 11.48 | 414.0 | ×16.2 |
| 1K | Standard SSD | 6304 | 187.0 | 11.74 | 6461 | ×0.98 |
| 2K | AI SSD | 11368 | 193.0 | 11.94 | 808.0 | ×14.6 |
| 2K | Standard SSD | 11896 | 185.3 | 11.76 | 11624 | ×1.02 |
| 4K | AI SSD | 21491 | 197.6 | 11.78 | 852.0 | ×26.3 |
| 4K | Standard SSD | 22098 | 192.7 | 11.76 | 22223 | ×0.99 |
| 8K | AI SSD | 42555 | 196.3 | 11.52 | 1153 | ×39.1 |
| 8K | Standard SSD | 44210 | 189.2 | 11.46 | 44289 | ×1.00 |
| 16K | AI SSD | 91388 | 181.0 | 11.00 | 1793 | ×54.6 |
| 16K | Standard SSD | 95081 | 174.1 | 10.94 | 95294 | ×1.00 |
| 32K | AI SSD | 220898 | 149.1 | 10.03 | 3016 | ×79.3 |
| 32K | Standard SSD | 229802 | 143.4 | 10.05 | 229579 | ×1.00 |
| 64K | AI SSD | 613450 | 107.2 | 8.64 | 2052 | ×371.1 |
| 64K | Standard SSD | 642412 | 102.3 | 8.62 | 643426 | ×1.00 |
| 128K | AI SSD | - | - | - | - | - |
| 128K | Standard SSD | - | - | - | - | - |
| Input | Mode | Cold TTFT (ms) | Cold Prefill TPS (tokens/s) | Hot Decode TPS (tokens/s) | 热 TTFT (ms) | Hot Prefill speedup |
|---|---|---|---|---|---|---|
| 1K | AI SSD | 2541 | 475.7 | 51.61 | 349.0 | ×7.65 |
| 1K | Standard SSD | 1329 | 910.6 | 42.46 | 1315 | ×1.01 |
| 2K | AI SSD | 2757 | 805.8 | 51.59 | 363.0 | ×8.03 |
| 2K | Standard SSD | 2481 | 899.4 | 43.77 | 2504 | ×0.99 |
| 4K | AI SSD | 6741 | 633.8 | 50.14 | 425.0 | ×16.8 |
| 4K | Standard SSD | 5620 | 763.0 | 42.98 | 5736 | ×0.98 |
| 8K | AI SSD | 15707 | 533.3 | 49.03 | 449.0 | ×38.1 |
| 8K | Standard SSD | 15130 | 554.5 | 42.06 | 15191 | ×1.00 |
| 16K | AI SSD | 47133 | 351.4 | 45.64 | 625.0 | ×85.7 |
| 16K | Standard SSD | 46161 | 359.1 | 38.58 | 46331 | ×1.00 |
| 32K | AI SSD | 158371 | 208.0 | 41.35 | 951.0 | ×190.9 |
| 32K | Standard SSD | 157088 | 209.9 | 32.30 | 156892 | ×1.00 |
| 64K | AI SSD | 576065 | 114.1 | 33.02 | 1641 | ×407.6 |
| 64K | Standard SSD | 573362 | 114.7 | 27.94 | 573105 | ×1.00 |
| 128K | AI SSD | - | - | - | - | - |
| 128K | Standard SSD | - | - | - | - | - |