What we hold, and what we do not
Live completeness for every tracked field across 4,042 processors and 3,222 graphics cards. Fields below 25 percent are marked weak. We publish this because a database that only advertises its strengths gives you no way to judge the value of a blank field.
CPU field coverage
| Field | Coverage | Records |
|---|---|---|
| Socket | 92.1% | 3,722 / 4,042 |
| TDP | 96.2% | 3,889 / 4,042 |
| Architecture | 74.9% | 3,026 / 4,042 |
| Launch date | 96.7% | 3,908 / 4,042 |
| Physical cores | 99.7% | 4,028 / 4,042 |
| Logical threads | 99.2% | 4,011 / 4,042 |
| Base frequency | 98.4% | 3,976 / 4,042 |
| Max turbo frequency | 73.2% | 2,959 / 4,042 |
| L3 cache | 83.4% | 3,370 / 4,042 |
| Max memory | 57.4% | 2,321 / 4,042 |
| PCIe revision | 67.6% | 2,732 / 4,042 |
| Instruction feature matrix | 77.0% | 3,111 / 4,042 |
GPU field coverage
| Field | Coverage | Records |
|---|---|---|
| PassMark G3D score | 82.5% | 2,658 / 3,222 |
| Launch date | 48.3% | 1,555 / 3,222 |
| Codename | 46.4% | 1,494 / 3,222 |
| Board power (TDP) | 27.0% | 870 / 3,222 |
| Video memory | 33.4% | 1,076 / 3,222 |
| Architecture | 31.4% | 1,011 / 3,222 |
| Shading units | 26.7% | 861 / 3,222 |
| Memory type | 43.4% | 1,399 / 3,222 |
| Memory bus width | 26.8% | 862 / 3,222 |
| Memory bandwidth | 29.7% | 957 / 3,222 |
| Boost clock (weak) | 16.1% | 518 / 3,222 |
| Die size (weak) | 17.1% | 551 / 3,222 |
| PassMark workload tests (DirectX 9/10/11/12, Compute) | 52.4% | 1,687 / 3,222 |
| PassMark listing facts (bus, max memory, clocks) | 82.5% | 2,658 / 3,222 |
| Graphics API support (DirectX, OpenGL, Vulkan, OpenCL, CUDA) | 39.1% | 1,261 / 3,222 |
| Feature matrix (supported and explicitly unsupported) (weak) | 20.5% | 660 / 3,222 |
| Media engine (codec encode and decode) (weak) | 13.5% | 434 / 3,222 |
| Display outputs (weak) | 7.5% | 242 / 3,222 |
| Theoretical throughput (TFLOPS, TOPS, fill rates) (weak) | 11.2% | 362 / 3,222 |
| Board power, connectors and dimensions (weak) | 7.0% | 226 / 3,222 |
Why GPU coverage is thinner than CPU coverage
CPU data benefits from Intel ARK and AMD.com, which publish complete, structured specifications for essentially every part they have shipped. There is no equivalent vendor source for the full history of graphics hardware, and the most complete public GPU database is one we deliberately do not collect from. The result is that modern graphics cards are well covered while older, OEM and workstation parts often carry benchmark scores and little else. Those pages will show a low coverage meter, which is the honest reading.
How coverage is calculated
Coverage counts entities where the field holds a non-null value that passed validation. It is computed live from the database when you load this page, not cached from a report. The per-entity meter on each page uses the same ten key specifications for GPUs, so a page reading "7/10" refers to this same list.