Intel HD Graphics 630 vs NVIDIA GeForce GTX 970M Comparison

The GeForce GTX 970M and Intel HD Graphics 630 represent two entirely different approaches to mobile graphics: a dedicated, high-power MXM module from NVIDIA versus an integrated processor graphics solution from Intel. The benchmark database shows a clear performance hierarchy, but the separation is not uniform across all workloads, and the architectural gulf between the two is substantial. The recorded data indicates the GTX 970M is the dominant performer, yet the HD 630 occupies a distinct niche where its integration and efficiency matter more than raw throughput.

Head-to-Head Benchmarks

The database contains two direct head-to-head benchmark comparisons between the GeForce GTX 970M and Intel HD Graphics 630. In both tests, the NVIDIA part wins decisively, with margins that are nothing short of overwhelming.

In the Geekbench OpenCL test, the GTX 970M scores 18,946 points, while the HD 630 manages only 3,587 points. This represents a delta of 428.2% in favor of the NVIDIA GPU. The scale of this difference is not incremental; it is a multiple of roughly five times the performance. For compute-oriented workloads that leverage OpenCL, the GTX 970M is in a completely different performance class.

The Geekbench Vulkan test shows a similar pattern. The GTX 970M posts a score of 18,292, compared to 3,540 for the HD 630. The delta here is 416.7%. Vulkan is a low-level API that can expose the capabilities of a dedicated GPU more fully, and the data reflects that. The NVIDIA architecture's 1280 shading units and dedicated memory subsystem simply dwarf the integrated solution.

The head-to-head record is 2 wins for the GTX 970M and 0 for the HD 630. There is no benchmark in the database where the Intel part takes a victory. The average benchmark score across all recorded tests reinforces this: the GTX 970M sits at 4,628, while the HD 630 rests at 4,075. That 553-point gap, about a 13.6% difference in averages, is notable, though it is far smaller than the head-to-head deltas because the averages include different mixes of tests. The direct comparisons, however, leave no ambiguity about which GPU is faster.

Where Each One Wins

The GTX 970M wins in every measurable compute scenario recorded in the database. Its PassMark G3D score of 5,704 is a strong indicator of general 3D rendering capability, placing it at the 27th percentile among all GPUs. The HD 630, with no PassMark scores listed, cannot be directly compared there, but its 24th percentile overall standing suggests a closer overall position than the head-to-head tests imply. The difference is that the GTX 970M's percentile is driven by a broad suite of tests, while the HD 630's is based on a narrower set.

For gaming and 3D rendering, the GTX 970M is the clear choice. Its 49.82 GPixel/s pixel rate and 83.04 GTexel/s texture rate provide the raw fill rates needed for modern game engines. The HD 630's 3.000 GPixel/s and 24.00 GTexel/s are orders of magnitude lower. The NVIDIA GPU also has a 2.657 TFLOPS FP32 throughput, versus 384.0 GFLOPS for the Intel part. That is a 6.9x advantage in single-precision compute, which directly translates to faster shader execution and physics calculations.

The HD 630's wins are not in performance but in integration. Its 15 W TDP is a fraction of what the GTX 970M requires, though the database does not list the NVIDIA part's TDP. The Intel GPU is an IGP on a Ring Bus, meaning it shares system memory and requires no power connectors. The GTX 970M is an MXM Module with no power connectors listed, but its slot width and dedicated 6 GB GDDR5 frame buffer indicate a much larger power envelope. The HD 630 is "System Shared" for memory, size, type, and bus width, making it dependent on the host system's RAM, whereas the GTX 970M has 120.3 GB/s of dedicated bandwidth. For users with a laptop that lacks a discrete GPU slot, the HD 630 is the only option; for those with an MXM bay, the GTX 970M is vastly superior.

Architecture Differences

The architectural divide between these two GPUs is profound. The GeForce GTX 970M is built on the GM204 chip using NVIDIA's Maxwell 2.0 architecture, fabricated at TSMC on a 28 nm process. The chip contains 5,200 million transistors on a 398 mm² die, yielding a transistor density of 13.1M per mm². Intel's HD Graphics 630 is the Kaby Lake GT2, using Intel's Generation 9.5 architecture, built on a 14 nm++ process at Intel's own fabs. The database lists no transistor count or die size for the Intel part, as it is integrated into the CPU package.

The compute resources are starkly different. The GTX 970M has 1280 shading units, 80 texture mapping units, and 48 ROPs. The HD 630 has 192 shading units, 24 TMUs, and only 3 ROPs. That ROP count is critically low, which explains the massive pixel rate difference: 49.82 GPixel/s versus 3.000 GPixel/s. The NVIDIA GPU's memory architecture is equally distinct: 6 GB of GDDR5 on a 192-bit bus delivering 120.3 GB/s of bandwidth, with a memory clock of 1253 MHz (5 Gbps effective). The HD 630 uses system shared memory with system dependent bandwidth, clocked at 350 MHz base and 1000 MHz boost.

Clock speeds also differ. The GTX 970M runs at 924 MHz base and 1038 MHz boost. The HD 630 boosts to 1000 MHz, which is close, but with 6.7x fewer shading units, the aggregate throughput is vastly different. The FP32 output is 2.657 TFLOPS for the NVIDIA part versus 384.0 GFLOPS for Intel. Interestingly, the HD 630 has a listed FP16 performance of 768.0 GFLOPS (2:1 ratio), which is double its FP32, whereas the GTX 970M has no FP16 data listed. This suggests the Intel part has some fast-path FP16 capability, but it is a minor footnote given the overall performance deficit.

API support is similar on the surface: both support DirectX 12 (12_1), OpenGL 4.6, and Vulkan (1.4 for NVIDIA, 1.3 for Intel). The GTX 970M also has a broader benchmark footprint, including PassMark tests for DirectX 9, 10, 11, and 12, plus G2D and GPU compute. The HD 630 only has Geekbench Metal, OpenCL, and Vulkan results. The NVIDIA part also has a Geekbench Metal score listed in its general benchmarks? No, it does not; the HD 630 has a Metal score of 5,099, which is not part of the head-to-head but is notable for macOS compatibility. The GTX 970M's production status is end-of-life, as is the HD 630's, and the GTX 970M released on 2014-10-06, while the HD 630 came later on 2016-08-29. The GTX 970M has a predecessor in the GeForce 800M and a successor in the GeForce 10 Mobile, while the HD 630 lists no predecessor or successor.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The GeForce GTX 970M has an average benchmark score of 4,628, while the Intel HD Graphics 630 averages 4,075. The NVIDIA part is ahead by 553 points, or roughly 13.6%.

Q: How much faster is the GTX 970M in OpenCL compute?

A: In the Geekbench OpenCL test, the GTX 970M scores 18,946 versus 3,587 for the HD 630, a delta of 428.2%. This means the NVIDIA GPU delivers more than five times the OpenCL performance.

Q: Does the Intel HD Graphics 630 win any benchmark in the database?

A: No. The head-to-head record is 2 wins for the GTX 970M and 0 for the HD 630. Intel does not win any recorded test, though it has a Geekbench Metal score of 5,099 that is not compared directly against the NVIDIA part.

Q: What are the pixel fill rates of these two GPUs?

A: The GTX 970M has a pixel rate of 49.82 GPixel/s, while the HD 630 is limited to 3.000 GPixel/s. This is a direct consequence of the 48 ROPs on the NVIDIA chip versus only 3 ROPs on the Intel integrated GPU.

Q: What memory configurations do they use?

A: The GTX 970M uses 6 GB of dedicated GDDR5 memory on a 192-bit bus, providing 120.3 GB/s of bandwidth. The HD 630 uses system shared memory with a system dependent bus width and bandwidth, meaning it relies on the host system's RAM.

Q: Which GPU supports Vulkan, and at what version?

A: Both support Vulkan. The GTX 970M supports Vulkan 1.4, while the HD 630 supports Vulkan 1.3. In the Geekbench Vulkan test, the GTX 970M scores 18,292 versus 3,540 for the HD 630.

The Verdict

The data is unambiguous for performance-driven users: the GeForce GTX 970M is the superior GPU in every recorded benchmark. Its 2.657 TFLOPS of FP32 compute, 120.3 GB/s of dedicated memory bandwidth, and 1280 shading units make it a capable part for gaming and compute workloads. The 428.2% OpenCL lead and 416.7% Vulkan lead over the HD 630 are not small margins; they represent entirely different tiers of graphics hardware. The GTX 970M's 27th percentile ranking among all GPUs, while modest, is still ahead of the HD 630's 24th percentile, and the average score gap of 13.6% confirms a consistent advantage.

The Intel HD Graphics 630 is not a gaming or compute part; it is an integrated solution designed for basic display output and low-power operation. Its 15 W TDP, system shared memory, and 3 ROPs make it suitable for office work, video playback, and light 2D tasks. The GTX 970M, by contrast, is an MXM module that requires a compatible slot and likely a larger thermal solution, though the database does not list its power draw. The HD 630's sole advantage is its integration into the CPU, eliminating the need for a separate GPU purchase.

For a laptop user who needs to play modern 3D games or run GPU-accelerated compute tasks, the GTX 970M is the only viable choice. For a user who needs basic graphics with minimal power consumption and no discrete GPU slot, the HD 630 suffices. The benchmark record shows no scenario where the HD 630 matches the GTX 970M, and the architectural differences explain why: 80 TMUs versus 24, 48 ROPs versus 3, and 6 GB GDDR5 versus shared system memory. The verdict is a function of use case. Power-constrained, light-duty systems will choose the HD 630 out of necessity; any other requirement points to the GTX 970M.