Intel UHD Graphics 730 vs NVIDIA GeForce GTX 670M Comparison

Where Each One Wins

The recorded benchmark data splits cleanly between these two parts. The NVIDIA GeForce GTX 670M holds a single head-to-head victory in Geekbench OpenCL, scoring 6513 against the Intel UHD Graphics 730’s 5988. That is an 8.8% advantage, a meaningful margin for compute workloads that scale with raw shader throughput.

The GTX 670M’s OpenCL result also places it at the 38th percentile among all GPUs in the database, while the Intel part sits at the 33rd percentile. That five-percentile gap reflects a consistent edge in general-purpose compute, not just a single favorable test. The NVIDIA part’s 336 shading units, 56 texture mapping units, and 24 ROPs give it substantially more parallel hardware to throw at OpenCL kernels.

The Intel UHD Graphics 730, however, counters with a feature the NVIDIA part lacks entirely: a Vulkan score. The database records 5870 in Geekbench Vulkan for the Intel part, while the GTX 670M has no Vulkan entry at all. For any workload or application that relies on Vulkan, the Intel part is the only one of these two with recorded data to stand on.

The Intel part also wins on efficiency in a narrow sense: its TDP is 15 W versus 75 W for the NVIDIA part. That is a fivefold difference in thermal envelope, which matters for systems where cooling and battery life take priority over raw compute. The GTX 670M is an MXM module with no power connectors listed, designed for portable devices, but it still draws five times the power of the Intel integrated solution.

Architecture Differences

The architectural divide here is generational and fundamental. The NVIDIA GeForce GTX 670M uses the GF114 chip built on Fermi 2.0 architecture, manufactured by TSMC on a 40 nm process. It packs 1,950 million transistors into a 332 mm² die, yielding a transistor density of 5.9 million transistors per square millimeter. The Intel UHD Graphics 730 uses a Rocket Lake chip on Intel’s Generation 12.1 architecture, built on a 14 nm+++ process. Intel does not report transistor count or die size for this part, so direct density comparisons are impossible from the data.

The memory systems could hardly be more different. The GTX 670M has 1536 MB of dedicated GDDR5 on a 192-bit bus, with memory clocked at 750 MHz (3 Gbps effective) and bandwidth of 72.00 GB/s. The Intel part uses System Shared memory, with System Shared type, System Shared bus width, and bandwidth listed as System Dependent. That means the Intel part borrows from main system RAM, and its memory performance varies with the host platform, whereas the NVIDIA part has fixed, dedicated memory resources.

Compute resources strongly favor NVIDIA. The GTX 670M has 336 shading units, 56 TMUs, and 24 ROPs. The Intel UHD Graphics 730 has 192 shading units, 12 TMUs, and 8 ROPs. That is roughly 43% fewer shading units, 79% fewer TMUs, and 67% fewer ROPs for the Intel part. Pixel rate tells the same story in reverse: the Intel part achieves 10.40 GPixel/s versus 8.372 GPixel/s for NVIDIA, a 24% advantage for Intel in rasterization throughput. Texture rate, however, goes decisively to NVIDIA at 33.49 GTexel/s versus 15.60 GTexel/s, a 115% lead.

FP32 compute is also lopsided: the GTX 670M delivers 803.7 GFLOPS against 499.2 GFLOPS for the Intel part. The Intel part does offer FP16 at 998.4 GFLOPS (2:1), a feature the NVIDIA part does not list. Clock behavior differs as well: the Intel part has a base clock of 300 MHz and a boost clock of 1300 MHz, while the NVIDIA part lists no base or boost clock, only its memory clock.

API support splits along similar lines. Both support DirectX 12 and OpenGL 4.6, but the Intel part supports DirectX 12 (12_1) and Vulkan 1.4, while the NVIDIA part is limited to DirectX 12 (11_0) and has no Vulkan entry. The GTX 670M’s bus interface is MXM-B (3.0) with slot width MXM Module; the Intel part is an IGP on a Ring Bus. Display outputs are Portable Device Dependent for NVIDIA and Motherboard Dependent for Intel, reflecting their different integration paths.

The Verdict

The data points to a straightforward split based on workload and platform constraints. For OpenCL compute and raw shader throughput, the NVIDIA GeForce GTX 670M is the stronger part. Its 6513 OpenCL score beats the Intel part’s 5988 by 8.8%, its FP32 output of 803.7 GFLOPS nearly doubles the Intel part’s 499.2 GFLOPS, and its texture rate of 33.49 GTexel/s is more than double the Intel part’s 15.60 GTexel/s. The GTX 670M also sits five percentile points higher in the overall GPU distribution.

For systems that need Vulkan support, the Intel UHD Graphics 730 is the only choice with recorded data. Its Vulkan score of 5870 is the sole Vulkan result between the two parts. The Intel part also delivers higher pixel rate at 10.40 GPixel/s, and it does so within a 15 W TDP, one-fifth the NVIDIA part’s 75 W envelope. For compact, low-power, integrated builds where compute peaks are less important than sustained modest output and broad API coverage, the Intel part has clear appeal.

The GTX 670M is end-of-life, as is the Intel UHD Graphics 730. Neither part is current. The NVIDIA part was released in 2012, the Intel part in 2021. The GTX 670M’s nearest rivals in the database are the NVIDIA GeForce GT 555M (0.3% behind), NVIDIA Quadro M5000M (0.5% behind), AMD Radeon Vega 10 Mobile (0.6% behind), and Intel UHD Graphics P750 (0.6% ahead). The Intel UHD Graphics 730’s nearest rivals are AMD Radeon HD 8730M (0.4% ahead), NVIDIA Quadro K620M (0.5% ahead), AMD Radeon HD 8750M (0.7% ahead), and NVIDIA Quadro K4000 (0.9% ahead). Both parts cluster tightly among their peers, but the GTX 670M’s cluster sits at a higher absolute score level.

FAQ

Q: Which GPU has the higher OpenCL benchmark score?

A: The NVIDIA GeForce GTX 670M scores 6513 in Geekbench OpenCL, while the Intel UHD Graphics 730 scores 5988, giving NVIDIA an 8.8% lead.

Q: Does the Intel UHD Graphics 730 support Vulkan?

A: Yes. The database records a Geekbench Vulkan score of 5870 for the Intel UHD Graphics 730. The NVIDIA GeForce GTX 670M has no Vulkan entry in the database.

Q: How do their memory configurations differ?

A: The GTX 670M has 1536 MB of dedicated GDDR5 on a 192-bit bus with 72.00 GB/s bandwidth. The Intel UHD Graphics 730 uses System Shared memory with System Dependent bandwidth.

Q: Which part has higher FP32 compute?

A: The NVIDIA GeForce GTX 670M delivers 803.7 GFLOPS of FP32, compared to 499.2 GFLOPS for the Intel UHD Graphics 730.

Q: What are the TDP ratings for these two GPUs?

A: The NVIDIA GeForce GTX 670M has a TDP of 75 W, while the Intel UHD Graphics 730 has a TDP of 15 W.

Q: Which part has more shading units?

A: The NVIDIA GeForce GTX 670M has 336 shading units, while the Intel UHD Graphics 730 has 192 shading units.

Head-to-Head Benchmarks

The database contains one direct head-to-head benchmark between these two parts: Geekbench OpenCL. The NVIDIA GeForce GTX 670M wins with 6513 points against 5988 for the Intel UHD Graphics 730, a delta of 8.8%. That is the only recorded test where both parts have scores, so it carries the full weight of direct comparison.

Breaking down the components behind that OpenCL result, the GTX 670M’s advantage comes from its larger compute array. With 336 shading units, 56 TMUs, and 24 ROPs, it has 144 more shading units, 44 more TMUs, and 16 more ROPs than the Intel part. FP32 throughput of 803.7 GFLOPS versus 499.2 GFLOPS gives NVIDIA a 61% edge in raw single-precision math. Texture rate of 33.49 GTexel/s versus 15.60 GTexel/s is a 115% advantage. Those are the hardware figures that translate directly into OpenCL kernel execution speed.

The Intel part’s counterpoints are pixel rate and API coverage. At 10.40 GPixel/s, the Intel UHD Graphics 730 beats the GTX 670M’s 8.372 GPixel/s by 24%. That suggests the Intel part can fill pixels faster when the workload is rasterization-bound rather than compute-bound. Additionally, the Intel part’s DirectX 12 (12_1) support and Vulkan 1.4 support exceed the NVIDIA part’s DirectX 12 (11_0) and no Vulkan. The Intel part’s FP16 throughput of 998.4 GFLOPS also gives it a path for mixed-precision workloads that the GTX 670M cannot match, since the NVIDIA part lists no FP16 capability.

Nearest rival data confirms the positioning. The GTX 670M’s closest competitor, the NVIDIA GeForce GT 555M, averages 6493, just 0.3% behind. The Intel UHD Graphics 730’s closest rival, the AMD Radeon HD 8730M, averages 5955, 0.4% ahead. The NVIDIA part’s rivals all sit within a 0.6% band around its score, while the Intel part’s rivals sit within a 0.9% band. Both parts are firmly embedded in their respective performance tiers, with the GTX 670M’s tier starting roughly 500 points higher.

The overall distribution matches. The GTX 670M is at the 38th percentile of all GPUs, the Intel UHD Graphics 730 at the 33rd. The 8.8% OpenCL gap between them is larger than the gap between either part and its nearest rivals, which means the difference between these two is not noise; it is a real, measurable performance separation in compute workloads. For anyone choosing between a dedicated Fermi-era mobile GPU and a modern Rocket Lake integrated GPU, the OpenCL data says the older discrete part still has the compute edge, while the integrated part holds the efficiency and API advantages.