Intel UHD Graphics P750 vs NVIDIA GRID K2 Comparison

Where Each One Wins

The recorded benchmark data delivers a clear split: the NVIDIA GRID K2 wins the only shared head-to-head test, while the Intel UHD Graphics P750 has no recorded benchmark victory over it. In the Geekbench OpenCL test, the GRID K2 scores 10,602 versus 6,554 for the Intel part, a 61.8% advantage. That single result defines the competitive landscape: the GRID K2 is the compute-oriented part, the Intel UHD Graphics P750 is the efficiency-oriented integrated solution.

The GRID K2’s winning profile stems from its dedicated memory subsystem and higher raw throughput. Its 4 GB of GDDR5 memory on a 256-bit bus delivers 160.0 GB/s of bandwidth, while the Intel part relies on system shared memory with bandwidth described as “System Dependent.” For workloads that stress memory bandwidth, the GRID K2’s advantage is structural, not marginal. The Intel UHD Graphics P750, by contrast, has no dedicated VRAM, so its performance scales with the host system’s memory configuration, a variable the database cannot control.

Where the Intel part does win is in efficiency and integration. Its 15 W TDP is a fraction of the GRID K2’s 225 W, and it is an IGP, meaning it requires no expansion slot, no power connectors, and no separate cooling solution. The GRID K2 is a dual-slot card, 267 mm long, requiring a 550 W suggested power supply and both 6-pin and 8-pin power connectors. For a system builder prioritizing low power draw and minimal physical footprint, the Intel part is the only choice between the two. The GRID K2 has no display outputs; the Intel part’s outputs are motherboard dependent, so it can drive displays in configurations where the NVIDIA card cannot.

In terms of benchmark percentiles, the GRID K2 sits at the 42nd percentile of all GPUs in the database, while the Intel part sits at the 38th. The GRID K2’s average benchmark score is 8,080, compared to 6,554 for the Intel part. That 1,526-point gap is substantial and consistent with the head-to-head result. The GRID K2’s nearest rivals include the GeForce GTX 650 Ti Boost (delta 0.2%), the GeForce 945M (delta -0.2%), the GeForce GTX 650 Ti (delta 0.3%), and the GeForce GTX 880M (delta 0.5%). The Intel part’s nearest rivals include the Radeon HD 7730M (delta -0.4%), the GeForce GTX 670M (delta 0.6%), the Radeon R7 M460 (delta -0.9%), and the GeForce GT 555M (delta 0.9%). These deltas place both parts in the same performance tier, but the GRID K2 consistently edges ahead in compute tests.

Architecture Differences

The two parts come from fundamentally different design philosophies. The NVIDIA GRID K2 uses the GK104 chip on the Kepler architecture, built on a 28 nm process at TSMC. It packs 3,540 million transistors into a 294 mm² die, yielding a transistor density of 12.0 million per square millimeter. The Intel UHD Graphics P750 uses the Rocket Lake chip on the Generation 12.1 architecture, built on Intel’s 14 nm+++ process. The database lists no transistor count, die size, or density for the Intel part, reflecting its integrated nature where the GPU shares the die with CPU cores.

The GRID K2’s compute resources are substantial: 1,536 shading units, 128 texture mapping units, and 32 raster operations pipelines. The Intel part has 256 shading units, 64 TMUs, and 32 ROPs. That 6:1 ratio in shading units explains the massive FP32 gap: the GRID K2 delivers 2.289 TFLOPS, while the Intel part delivers 665.6 GFLOPS. However, the Intel part supports FP16 at 1,331.2 GFLOPS (2:1 ratio), a feature the GRID K2 lacks entirely, as its FP16 field is null. This makes the Intel part more flexible for mixed-precision workloads, even if its absolute throughput is lower.

Memory architecture diverges completely. The GRID K2 uses dedicated GDDR5 with a 256-bit bus and fixed 160.0 GB/s bandwidth. The Intel part uses system shared memory, with bandwidth dependent on the host platform. The GRID K2’s memory clock is 1250 MHz (5 Gbps effective), while the Intel part’s base clock is 350 MHz with a boost of 1300 MHz. The GRID K2 has no base or boost clock listed, indicating its clocks are fixed for its workload.

API support differs in direct ways. The GRID K2 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The Intel part supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The Intel part’s higher DirectX feature level (12_1 vs 11_0) and newer Vulkan version suggest better support for modern graphics features, even though its raw performance is lower. Both are end-of-life products, with the GRID K2 released in May 2013, while the Intel part has no release date listed.

Head-to-Head Benchmarks

The only shared benchmark in the database is Geekbench OpenCL. The NVIDIA GRID K2 scores 10,602, while the Intel UHD Graphics P750 scores 6,554. That is a delta of 61.8% in favor of the NVIDIA part. This is not a marginal win; it is a decisive margin that reflects the GRID K2’s dedicated memory and 6x shading unit advantage.

To contextualize the GRID K2’s score: its average benchmark score across all recorded tests is 8,080, which places it 0.2% above the GeForce GTX 650 Ti Boost (8,067) and 0.3% above the GeForce GTX 650 Ti (8,053). It trails the GeForce 945M by 0.2% (8,099) and beats the GeForce GTX 880M by 0.5% (8,040). The Intel part’s 6,554 score puts it 0.4% below the Radeon HD 7730M (6,581), 0.6% above the GeForce GTX 670M (6,513), 0.9% below the Radeon R7 M460 (6,612), and 0.9% above the GeForce GT 555M (6,493). So while both parts sit near mid-range mobile and older desktop GPUs, the GRID K2’s absolute score is roughly 61.8% higher in the shared test.

The pixel and texture rates reinforce this gap. The GRID K2 has a pixel rate of 23.84 GPixel/s and a texture rate of 95.36 GTexel/s. The Intel part has a pixel rate of 41.60 GPixel/s and a texture rate of 83.20 GTexel/s. Interestingly, the Intel part wins in pixel rate, despite losing in texture rate. This suggests the Intel architecture is more efficient at raster operations per clock, but the GRID K2’s raw shader count dominates compute workloads. The FP32 numbers tell a similar story: 2.289 TFLOPS versus 665.6 GFLOPS, a 3.44x difference in favor of NVIDIA.

The GRID K2’s Geekbench Metal score is 5,557, which the database records but the Intel part has no equivalent test. That Metal score is lower than its OpenCL score, indicating the Kepler architecture is more optimized for OpenCL compute than Metal. The Intel part has no Metal benchmark, so cross-API comparisons are unavailable.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The NVIDIA GRID K2 has an average benchmark score of 8,080, compared to 6,554 for the Intel UHD Graphics P750.

Q: What is the exact performance difference in the shared Geekbench OpenCL test?

A: The GRID K2 scores 10,602, the Intel part scores 6,554, giving the GRID K2 a 61.8% lead.

Q: Does the Intel UHD Graphics P750 support any features the GRID K2 lacks?

A: Yes. The Intel part supports DirectX 12 (12_1) and Vulkan 1.4, while the GRID K2 supports DirectX 12 (11_0) and Vulkan 1.2.175. The Intel part also has FP16 support at 1,331.2 GFLOPS, which the GRID K2 does not list.

Q: How do these parts compare in power consumption?

A: The GRID K2 has a TDP of 225 W, while the Intel UHD Graphics P750 has a TDP of 15 W. The GRID K2 also requires a 550 W suggested power supply and dual-slot cooling, while the Intel part is an IGP with no separate power connectors.

Q: Which GPU is better for display output?

A: The GRID K2 has no display outputs, while the Intel UHD Graphics P750’s outputs are motherboard dependent. For any display task, the Intel part is the only option.

Q: How do their nearest rivals compare?

A: The GRID K2 is 0.2% above the GeForce GTX 650 Ti Boost and 0.5% above the GeForce GTX 880M. The Intel part is 0.6% above the GeForce GTX 670M and 0.9% below the Radeon R7 M460.

The Verdict

The data points to one conclusion: the NVIDIA GRID K2 is the stronger compute performer, while the Intel UHD Graphics P750 is the only viable option for integrated, low-power systems. If the workload is OpenCL compute, the GRID K2 wins by 61.8% in the head-to-head test, and its average benchmark score is 23.3% higher overall. No recorded benchmark shows the Intel part winning any test.

However, the GRID K2 cannot function in a typical desktop without significant infrastructure. It requires a PCIe 3.0 x16 slot, a 550 W power supply, and dual-slot clearance. It has no display outputs, so it is strictly a compute or virtualization accelerator. The Intel part, by contrast, is a system-on-chip GPU that works with any motherboard supporting Rocket Lake, draws 15 W, and can drive displays. Its DirectX 12_1 and Vulkan 1.4 support are newer than the GRID K2’s, making it more future-proof for graphics API compatibility.

For a user prioritizing raw compute in a server or workstation with existing power and cooling headroom, the GRID K2 is the clear pick. For a user needing a low-power integrated GPU with modern API support and display output, the Intel UHD Graphics P750 is the only choice. The GRID K2’s 4 GB GDDR5 and 160 GB/s bandwidth are decisive for memory-bound workloads, but the Intel part’s system shared memory is flexible, though slower. The verdict is not about which is better in absolute terms, but which fits the intended system and workload. The database shows one decisive win for NVIDIA, and the architecture differences explain why.

Specification Differences

| Field | NVIDIA GRID K2 | Intel UHD Graphics P750 |

|-------|---------------|------------------------|

| Chip | GK104 | Rocket Lake |

| Architecture | Kepler | Generation 12.1 |

| Process Node | 28 nm | 14 nm+++ |

| Foundry | TSMC | Intel |

| Transistors | 3,540 million | null |

| Die Size | 294 mm² | null |

| Transistor Density | 12.0M / mm² | null |

| Base Clock | null | 350 MHz |

| Boost Clock | null | 1300 MHz |

| Memory Clock | 1250 MHz (5 Gbps effective) | System Shared |

| Memory Size | 4 GB | System Shared |

| Memory Type | GDDR5 | System Shared |

| Memory Bus Width | 256 bit | System Shared |

| Memory Bandwidth | 160.0 GB/s | System Dependent |

| Shading Units | 1536 | 256 |

| TMUs | 128 | 64 |

| ROPs | 32 | 32 |

| Pixel Rate | 23.84 GPixel/s | 41.60 GPixel/s |

| Texture Rate | 95.36 GTexel/s | 83.20 GTexel/s |

| FP32 | 2.289 TFLOPS | 665.6 GFLOPS |

| FP16 | null | 1,331.2 GFLOPS (2:1) |

| TDP | 225 W | 15 W |

| Slot Width | Dual-slot | IGP |

| Power Connectors | 1x 6-pin + 1x 8-pin | null |

| Suggested PSU | 550 W | null |

| Bus Interface | PCIe 3.0 x16 | Ring Bus |

| Display Outputs | No outputs | Motherboard Dependent |

| DirectX | 12 (11_0) | 12 (12_1) |

| OpenGL | 4.6 | 4.6 |

| Vulkan | 1.2.175 | 1.4 |

| Launch MSRP | 5,199 USD | null |

| Release Date | 2013-05-10 | null |

| Production Status | End-of-life | End-of-life |