Intel HD Graphics 630

The Intel HD Graphics 630 is an integrated graphics processor built on Intel's Generation 9.5 architecture using a 14 nm++ process. It was released on August 29, 2016, and has since been marked end-of-life. The GPU features 192 shading units, 24 texture mapping units, and 3 ROPs, with base and boost clocks of 350 MHz and 1000 MHz respectively. Its FP32 throughput is 384.0 GFLOPS, and FP16 reaches 768.0 GFLOPS (2:1). With a TDP of 15 W, it is designed for low-power systems, and its Ring Bus interface connects it to the host processor. As an integrated part, it relies entirely on system memory, which shapes its performance profile.

Benchmark Performance

The HD 630's average benchmark score across Geekbench compute tests is 4977. This places it in the 28th percentile of all GPUs in the database, meaning it outperforms only 28% of the GPUs tracked. That is a low-end ranking, consistent with an integrated GPU from 2016. The three individual scores show meaningful variation: Metal scores 5099, Vulkan scores 5267, and OpenCL scores 4566. The OpenCL result is significantly lower than the other two, suggesting that the driver stack is less optimized for OpenCL compute on this hardware. The Vulkan score is the highest, indicating that Vulkan's lower-level API allows the GPU to achieve better throughput. The average of 4977 sits between the Metal and OpenCL scores, pulled down by the OpenCL outlier.

When compared to its nearest rivals, the HD 630 is effectively tied with all of them. The AMD Radeon R7 M360 averages 4978, a delta of 0% — a statistical dead heat. The AMD Radeon R5 M430 scores 4981, a delta of -0.1%, so the HD 630 is marginally slower. The NVIDIA Quadro 4000 posts 5000, a delta of -0.5%, and the AMD Radeon HD 8670M scores 5012, a delta of -0.7%. None of these deltas exceed 1%, which means that in compute benchmarks, the HD 630 is indistinguishable from these discrete mobile GPUs. The 28th percentile, however, is a stark reminder that this is not a gaming or rendering part; it is intended for basic 2D acceleration, video decode, and light productivity. The narrow spread among the rivals suggests that all of these GPUs are clustered at the very bottom of the performance scale.

Memory Subsystem

The HD 630 uses system shared memory for all graphics operations. There is no dedicated VRAM; the memory size, type, and bus width are all listed as "System Shared". The bandwidth is "System Dependent", meaning it relies on the host system's memory controller and RAM speed. This architecture has profound implications for performance. Because the GPU and CPU share the same memory bus, available bandwidth is a finite resource that must be divided. At higher resolutions, the frame buffer and texture storage demands increase, and the system shared memory can become a bottleneck. The lack of dedicated VRAM also limits the amount of texture data that can be quickly accessed, which can cause stuttering in memory-intensive applications. The base clock of 350 MHz and boost clock of 1000 MHz apply to the GPU core, but memory clocks are not specified because they are tied to the system's RAM configuration. For users considering this GPU for modern games, the memory subsystem would likely be the primary limiter, especially with large texture packs or high-resolution displays. The system dependent bandwidth also means that performance can vary significantly between different platforms, depending on whether the system uses dual-channel memory or faster DDR4 modules.

How It Compares

AMD Radeon R7 M360: The R7 M360 averages 4978, a 0% delta relative to the HD 630. This is a dead heat; the 1-point difference in scores is negligible. Both GPUs are entry-level mobile parts, and their compute performance is essentially identical. The R7 M360 is a discrete GPU, but its performance is so low that it offers no advantage over the integrated Intel solution.

AMD Radeon R5 M430: The R5 M430 scores 4981, giving a delta of -0.1%. The HD 630 is slightly slower, but the margin is within measurement noise. Users would not perceive any difference in day-to-day tasks. The R5 M430 is also a discrete GPU, yet its performance is only 0.1% higher than the integrated HD 630, highlighting how far integrated graphics have come in closing the gap with entry-level discrete parts.

NVIDIA Quadro 4000: The Quadro 4000 averages 5000, a delta of -0.5%. Although the Quadro is a workstation-oriented GPU, its compute score is only 0.5% higher than the HD 630. This suggests that the integrated Intel part can match a legacy workstation GPU in synthetic benchmarks. The Quadro's advantage is marginal, and in real-world workstation tasks, other factors such as driver certification and memory size would matter more.

AMD Radeon HD 8670M: The HD 8670M scores 5012, a delta of -0.7%. This is the largest gap among the four rivals, but still under 1%. The HD 630 trails by a small margin, likely due to differences in memory bandwidth or driver efficiency. The HD 8670M is a older discrete GPU, and its performance is only slightly better than the integrated Intel part, underscoring the HD 630's position at the very bottom of the performance curve.

FAQ

Q: What is the TDP of the Intel HD Graphics 630?

A: The TDP is 15 W, making it suitable for ultra-portable laptops and low-power desktop systems.

Q: Does the HD 630 support DirectX 12?

A: Yes, it supports DirectX 12 (12_1), as well as OpenGL 4.6 and Vulkan 1.3.

Q: How much VRAM does the HD 630 have?

A: It has no dedicated VRAM; it uses system shared memory, with bandwidth that is system dependent.

Q: What are the pixel and texture fill rates?

A: The pixel rate is 3.000 GPixel/s, and the texture rate is 24.00 GTexel/s.

Q: What is the release date and production status?

A: It was released on August 29, 2016, and is now end-of-life.

Q: What is the FP32 performance?

A: The FP32 throughput is 384.0 GFLOPS, with FP16 at 768.0 GFLOPS (2:1).

Ray Tracing and Feature Set

The HD 630 does not include dedicated ray tracing cores or tensor cores; both fields are null in the specification. This means ray tracing workloads must rely on compute shaders, which is inefficient compared to hardware-accelerated RT. The GPU does support DirectX 12_1, which includes features like conservative rasterization and rasterizer ordered views, but not hardware ray tracing. Vulkan 1.3 support allows for compute-heavy workloads, but again without RT acceleration. The GPU's compute capabilities are modest: 192 shading units, 24 TMUs, and 3 ROPs. The low ROP count (3) severely limits pixel output, which is why the pixel rate is only 3.000 GPixel/s. The texture rate of 24.00 GTexel/s is also modest. The bus interface is Ring Bus, typical for integrated graphics, and display outputs are motherboard dependent, so connectivity varies by platform. The 14 nm++ process node and Generation 9.5 architecture are mature, but the GPU's feature set is basic. For users needing modern ray tracing or AI acceleration, the HD 630 is not suitable; it is a legacy integrated GPU aimed at basic display output and light compute tasks. Its API support, however, is surprisingly modern, with Vulkan 1.3 and DirectX 12_1, which ensures compatibility with current operating systems and some modern applications, albeit at low performance levels.