Intel UHD Graphics P750 vs NVIDIA GeForce GTX 670M Comparison

Intel UHD Graphics P750 and NVIDIA GeForce GTX 670M are both end-of-life graphics solutions, but they represent fundamentally different approaches to mobile computing. The Intel part is an integrated processor graphics unit (IGP) built into the Rocket Lake chip, while the NVIDIA part is a discrete mobile module based on the Fermi 2.0 architecture. Their benchmark scores land them in the same performance tier, with the Intel P750 edging ahead by a razor-thin margin in the available OpenCL test, but the underlying hardware tells a story of different strengths and limitations that matter depending on your workload.

Head-to-Head Benchmarks

The only direct benchmark result available is the Geekbench OpenCL test, and it is extremely close. The Intel UHD Graphics P750 scores 6554, while the NVIDIA GeForce GTX 670M scores 6513. This gives the Intel part a 0.6% deltaPct advantage, meaning it is just 0.6% faster than the GTX 670M in this specific workload. In practical terms, this difference is negligible — a single percentage point or less is well within run-to-run variance for most synthetic benchmarks, and neither card will feel meaningfully faster than the other in compute tasks that scale linearly.

Looking at the nearest rivals for each card puts this performance in context. The Intel P750 sits between the AMD Radeon HD 7730M (6581, 0.4% higher) and the NVIDIA GeForce GT 555M (6493, 0.9% lower). The GTX 670M, meanwhile, is bracketed by the GeForce GT 555M (6493, 0.3% lower) and the AMD Radeon Vega 10 Mobile (6476, 0.6% lower). Both cards occupy the 38th percentile of all GPUs, which places them in the lower-midrange of the overall performance distribution. The data shows a cluster of GPUs within roughly 1% of each other here, so the head-to-head result is best interpreted as a statistical tie rather than a decisive victory.

The wins tally confirms this: Intel takes 1 win, NVIDIA takes 0. But with a delta of only 0.6%, the practical takeaway is that these two solutions are interchangeable in raw compute performance for OpenCL-based applications. If you are choosing between them solely on benchmark numbers, you will need to look at other factors like power draw, memory bandwidth, and driver support to make a decision.

Architecture Differences

The architectural gap between these two is substantial. The Intel UHD Graphics P750 is built on Intel's Generation 12.1 architecture, fabricated on a 14 nm+++ process at Intel's own foundry. It is part of the HD Graphics-W (Rocket Lake) generation, which means it shares the processor die and uses system memory for both its frame buffer and its working set. The GTX 670M, by contrast, uses NVIDIA's Fermi 2.0 architecture, built on a 40 nm process at TSMC. The chip, codenamed GF114, is a discrete design with its own dedicated memory.

The transistor counts tell a story of design philosophy. The GTX 670M packs 1,950 million transistors into a 332 mm² die, giving it a transistor density of 5.9M per mm². The Intel P750 has no listed transistor count or die size in the data, but its integrated nature means it shares the Rocket Lake package rather than being a standalone chip. The Fermi 2.0 architecture is older and less efficient per transistor, which is evident in the power figures: the GTX 670M has a 75 W TDP, while the Intel P750 sips just 15 W.

Memory configuration differs dramatically. The Intel P750 uses "System Shared" memory for everything — size, type, bus width, and bandwidth are all dependent on the host system's RAM. The GTX 670M has a dedicated 1536 MB of GDDR5 on a 192-bit bus, delivering 72.00 GB/s of bandwidth. Its memory clock is 750 MHz, or 3 Gbps effective. This dedicated memory means the GTX 670M does not compete with the CPU for bandwidth, which can be a significant advantage in texture-heavy workloads.

Compute resources also differ in their allocation. The Intel P750 has 256 shading units, 64 texture mapping units (TMUs), and 32 raster operation pipelines (ROPs). The GTX 670M has 336 shading units, 56 TMUs, and 24 ROPs. Despite having fewer shading units, the Intel part achieves higher pixel and texture rates: 41.60 GPixel/s and 83.20 GTexel/s, versus 8.372 GPixel/s and 33.49 GTexel/s for the NVIDIA card. This is because the Intel P750's clock speeds are much higher — a 350 MHz base and 1300 MHz boost — while the GTX 670M has no base or boost clock listed, only its memory clock. The FP32 throughput tells a different story: the GTX 670M delivers 803.7 GFLOPS, while the Intel P750 delivers 665.6 GFLOPS. The Intel part also lists FP16 performance at 1,331.2 GFLOPS with a 2:1 ratio, while the GTX 670M has no FP16 figure.

API support is another differentiator. The Intel P750 supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The GTX 670M supports DirectX 12 (11_0), OpenGL 4.6, but has no Vulkan support listed. For modern games or applications that rely on Vulkan, the Intel part has a clear compatibility advantage. The GTX 670M's DirectX 12 support is also limited to the 11_0 feature level, which means it may not support all DirectX 12 features.

The Verdict

From the data, the choice between these two comes down to your platform and workload priorities. The Intel UHD Graphics P750 is the better pick if you are building a system around a Rocket Lake processor and want a capable IGP for light gaming, media playback, or compute tasks that benefit from modern API support. Its 15 W TDP is a massive advantage for battery life and thermal management in thin-and-light laptops. The 0.6% benchmark lead over the GTX 670M, combined with Vulkan 1.4 support and DirectX 12_1, makes it the more future-proof option on paper.

The NVIDIA GeForce GTX 670M is the better pick if you need dedicated memory and higher raw FP32 throughput. Its 803.7 GFLOPS exceeds the Intel part's 665.6 GFLOPS, and the 72.00 GB/s of dedicated GDDR5 bandwidth will not be shared with the CPU. This can matter in applications that are sensitive to memory latency or bandwidth contention. However, the 75 W TDP is five times higher, and the lack of Vulkan support limits its modern API compatibility.

The data does not support a clear winner for general use. Both cards sit at the 38th percentile of all GPUs, and their nearest rivals are nearly identical in score. If you already have a Rocket Lake system, the P750 is effectively free — you simply use the IGP. If you need a discrete mobile GPU, the GTX 670M offers dedicated memory and higher compute throughput, but at a significant power cost. For most users, the Intel part's lower power draw and modern API support make it the more practical choice, provided your CPU supports it.

FAQ

Q: Which GPU is faster in OpenCL compute?

A: The Intel UHD Graphics P750 scores 6554, which is 0.6% higher than the GTX 670M's 6513. This is a negligible difference in practical terms.

Q: Does the GTX 670M support Vulkan?

A: No. The GTX 670M has no Vulkan support listed, while the Intel P750 supports Vulkan 1.4.

Q: How much memory does the GTX 670M have, and what type is it?

A: It has 1536 MB of GDDR5 on a 192-bit bus, with 72.00 GB/s of bandwidth and a 750 MHz (3 Gbps effective) memory clock.

Q: What is the TDP difference between these two?

A: The Intel P750 has a 15 W TDP, while the GTX 670M has a 75 W TDP. That is a five-fold difference in power draw.

Q: What is the process node for each chip?

A: The Intel P750 is built on 14 nm+++ at Intel, while the GTX 670M uses 40 nm at TSMC.

Q: Which has more shading units?

A: The GTX 670M has 336 shading units, while the Intel P750 has 256. However, the Intel part has higher pixel and texture rates due to its higher clocks.

Where Each One Wins

The Intel UHD Graphics P750 wins in scenarios where power efficiency and modern API support are paramount. Its 15 W TDP is a fraction of the GTX 670M's 75 W, making it ideal for ultraportable laptops where battery life and thermals are critical. The Vulkan 1.4 support and DirectX 12_1 feature level mean it can run newer titles and applications that require these APIs, which the GTX 670M cannot. Its higher pixel rate (41.60 GPixel/s vs 8.372 GPixel/s) and texture rate (83.20 GTexel/s vs 33.49 GTexel/s) also give it an edge in fill-rate-bound scenes, despite fewer shading units. The system-shared memory is a disadvantage, but for light workloads it is often sufficient.

The NVIDIA GeForce GTX 670M wins in scenarios that demand dedicated memory and higher raw compute throughput. Its 1536 MB of GDDR5 with 72.00 GB/s bandwidth does not contend with the CPU for memory access, which can be a major advantage in texture-heavy games or compute workloads that thrash the memory bus. Its FP32 output of 803.7 GFLOPS exceeds the Intel part's 665.6 GFLOPS, so applications that are heavily dependent on shader math will see a performance benefit. The 336 shading units also provide more parallel processing lanes for workloads that scale with shader count. The MXM module form factor means it can be replaced or upgraded in compatible laptops, which is not possible with an integrated GPU. However, the 75 W TDP and lack of Vulkan support are significant drawbacks that limit its appeal in modern contexts.

The data shows a split decision: Intel wins on efficiency, API support, and fill-rate; NVIDIA wins on memory bandwidth, FP32 throughput, and shader count. Neither card is a clear winner for all use cases, so the right choice depends entirely on whether your workload is bottlenecked by memory, compute, or API compatibility.