Intel UHD Graphics P630

The Intel UHD Graphics P630 is an integrated GPU built on Intel's Generation 9.5 architecture, using the Comet Lake GT2 chip on a 14 nm+++ process. It was released on 2020-05-12 and is now end-of-life. The GPU carries 192 shading units, 24 texture mapping units, and 3 ROPs, with a base clock of 350 MHz and a boost clock of 1200 MHz. In benchmark data, it ranks at the 33rd percentile among all GPUs and posts an average benchmark score of 5760, placing it in a near dead heat with a group of older discrete and professional GPUs.

How It Compares

Against the NVIDIA GeForce GTX 670MX, the UHD Graphics P630 is effectively tied. The P630's average score of 5760 is 0.3% above the rival's 5742. That is a meaningless margin for real-world workloads.

Against the NVIDIA GeForce GTX 550 Ti, the P630 is 0.5% ahead, with an average of 5760 versus 5731. The data does not establish any meaningful separation between the two.

Against the NVIDIA Quadro K4000, the P630 leads by 0.7%, with averages of 5760 and 5723. A professional-oriented card lands at essentially the same performance point as this integrated processor.

Against the Intel Iris Pro Graphics P6300, the closest internal-family comparison, the P630 is 0.8% ahead: 5760 versus 5712. These two Intel integrated parts are effectively interchangeable in aggregate benchmark performance.

Ray Tracing and Feature Set

The P630 specification set includes no dedicated RT cores and no tensor cores. Ray tracing acceleration is therefore not part of this hardware feature set. The design relies on conventional shader and fixed-function throughput: 192 shading units, 24 TMUs, and 3 ROPs. The FP32 rate is 460.8 GFLOPS, while FP16 reaches 921.6 GFLOPS at a 2:1 ratio.

API support is broad for an integrated part: DirectX 12 at feature level 12_1, OpenGL 4.6, and Vulkan 1.3. This means the hardware can expose modern graphics APIs. However, without dedicated ray tracing or tensor hardware, specialized workloads in those areas do not have dedicated silicon acceleration. The architecture is Generation 9.5, built as Comet Lake GT2 on Intel's 14 nm+++ process.

Benchmark Performance

The Geekbench OpenCL result is 5664, and the Geekbench Vulkan result is 5856. The Vulkan score is the higher of the two, indicating that the newer API path extracts a bit more performance from the same hardware. The average benchmark score of 5760 is the aggregate used for ranking and comparison.

Relative to its nearest rivals, the P630 is 0.3% above the GTX 670MX, 0.5% above the GTX 550 Ti, 0.7% above the Quadro K4000, and 0.8% above the Iris Pro P6300. All of those deltas are small enough that run-to-run variation and system configuration could flip the ordering. The 33rd-percentile ranking places it in the lower portion of the database, not in a position to challenge mainstream discrete GPUs.

These scores also indicate a better fit for API-accelerated 2D and light 3D work than for sustained compute. With 460.8 GFLOPS of FP32 throughput, the theoretical compute ceiling is modest. The nearest rivals confirm this: each sits within a 0.8% band, so no nearby competitor delivers a meaningful performance escape.

Power and Cooling

The P630 has a TDP of 15 W. Its slot width is IGP, meaning there is no add-in card. No power connectors are listed, so the GPU does not require supplemental graphics power. No suggested PSU is listed; the integrated nature means the host system's existing power delivery covers this GPU.

Display outputs are motherboard dependent, so the physical video connections are determined by the board rather than by the GPU itself. The bus interface is Ring Bus. Thermal demands are low at 15 W, but the actual cooling solution is the one provided for the processor package. The combination of a 15 W TDP, no power connectors, and no discrete slot makes integration straightforward in systems that support the processor.

FAQ

Q: What GPU architecture does the Intel UHD Graphics P630 use?

A: It uses Intel's Generation 9.5 architecture with the Comet Lake GT2 chip, manufactured on a 14 nm+++ process. It belongs to the HD Graphics-W (Comet Lake) generation.

Q: How much dedicated VRAM does it have?

A: It has no dedicated VRAM. Memory size, memory type, and bus width are all System Shared, and bandwidth is System Dependent.

Q: What clock speeds does it run at?

A: The base clock is 350 MHz, with a boost clock of 1200 MHz.

Q: Which graphics APIs are supported?

A: It supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.3.

Q: Does it have ray tracing cores?

A: No dedicated RT cores are listed in its specifications, and it also has no tensor cores. Ray tracing acceleration is not present in the hardware feature set.

Q: When was it released and is it still produced?

A: Its release date is 2020-05-12, and its production status is end-of-life.

Who Should Consider It

Users with workloads that fit within an average benchmark score of 5760 and a 33rd-percentile ranking. This is not a GPU for high-end gaming or hardware ray tracing. The absence of RT and tensor cores closes those paths. The 3.600 GPixel/s pixel rate and 28.80 GTexel/s texture rate put a strict cap on fill-rate-heavy scenes.

For systems that already use this integrated GPU, it can serve as a competent output device for basic desktop use, light 2D applications, and undemanding 3D workloads at lower settings. At higher resolutions, the limited ROP count of 3 and the System Shared memory arrangement become constraints. Buyers should note that the part is end-of-life, so any evaluation should treat it as a legacy integrated solution rather than a forward-looking graphics platform.

Memory Subsystem

Memory size is System Shared, memory type is System Shared, bus width is System Shared, and bandwidth is System Dependent. These fields tell the core story: the P630 has no dedicated frame buffer and no fixed memory pipeline. The host system's RAM does double duty as CPU memory and GPU memory.

This arrangement is acceptable for low-overhead tasks, but it means high-resolution rendering is limited by platform memory bandwidth as well as GPU clocks. The fixed graphics rates do not change with system memory: the pixel rate is 3.600 GPixel/s and the texture rate is 28.80 GTexel/s. With 3 ROPs, final pixel output is a serious bottleneck for high-resolution fill work. The 192 shading units can handle shader workloads, but that data eventually passes through the low ROP count and the shared memory interface. For high resolutions, the memory subsystem is the fundamental limiter.