GPU Comparison

Intel Iris Pro Graphics P580 and Intel UHD Graphics P750 are both end-of-life integrated graphics solutions from Intel, but they belong to different generations and target different performance profiles. The P580, built on the Skylake GT4e chip with Generation 9.0 architecture, carries a 39th percentile ranking among all GPUs, while the P750, based on Rocket Lake with Generation 12.1 architecture, sits at the 38th percentile. Despite the newer architecture in the P750, benchmark data shows the older P580 delivers significantly higher raw performance in the one available head-to-head test, making this a case where execution width and memory subsystem advantages outweigh generational improvements.

Head-to-Head Benchmarks

The sole direct comparison in the data is the Geekbench OpenCL test, and it is not close. The Intel Iris Pro Graphics P580 scores 9082 points, while the Intel UHD Graphics P750 manages 6554 points. That translates to a 38.6% advantage for the P580, a massive gap for two integrated parts. The P580 wins the only head-to-head benchmark, giving it a 1–0 record in direct comparisons.

Context from the nearest rivals underscores how meaningful this lead is. The P580’s average benchmark score of 7170 places it within 0.2% of the NVIDIA GeForce GTX 970 (7157) and within 0.5% of the AMD Radeon Vega 8 Mobile (7203). It also sits 0.7% above the NVIDIA GeForce GTX 750 (7222). The P750, by contrast, averages 6554, which is 0.4% behind the AMD Radeon HD 7730M (6581) and 0.9% behind the AMD Radeon R7 M460 (6612). The P750 does edge out the NVIDIA GeForce GTX 670M (6513) by 0.6% and the NVIDIA GeForce GT 555M (6493) by 0.9%, but those are far older mobile parts. In absolute terms, the P580’s 9082 OpenCL score is roughly 38.6% higher than the P750’s 6554, a difference that would be visible in any compute-heavy workload.

The P580’s advantage comes from its execution resources. It packs 576 shading units, 72 texture mapping units, and 9 raster output pipelines. The P750, despite being newer, has only 256 shading units, 64 TMUs, and 32 ROPs. That means the P580 has more than twice the shading units and a higher texture unit count, though the P750’s ROP count is over three times higher. The P580 also posts a higher FP32 throughput of 1,152.0 GFLOPS versus 665.6 GFLOPS for the P750, and FP16 performance of 2.304 TFLOPS (2:1) versus 1,331.2 GFLOPS (2:1). These are not marginal differences; they represent a fundamental gap in compute capacity that the P750’s newer architecture cannot overcome in this benchmark.

Where Each One Wins

The P580 wins in raw compute performance, which is the only area with direct head-to-head data. Its 38.6% lead in Geekbench OpenCL suggests it is better suited for tasks that leverage general-purpose GPU compute, such as OpenCL-accelerated filters, encoding helpers, or scientific workloads that offload math to the iGPU. The higher shading unit count and FP32 throughput make it the stronger choice for any application that scales with shader parallelism.

The P750, however, is not without strengths, though they are not reflected in the single benchmark. Its pixel rate of 41.60 GPixel/s is dramatically higher than the P580’s 9.000 GPixel/s, a 4.6x difference. Its texture rate of 83.20 GTexel/s also exceeds the P580’s 72.00 GTexel/s. These figures indicate that the P750 is built for rasterization efficiency, higher ROP count and faster pixel throughput suggest better fill-rate-bound performance in traditional 3D rendering, even if its raw shader compute is lower. The P750 also has a higher boost clock of 1300 MHz versus 1000 MHz for the P580, which helps in clock-bound scenarios.

For light gaming or desktop composition, the P750’s higher pixel and texture rates could translate to smoother output at lower resolutions, but the data does not include a gaming benchmark to confirm this. The P580’s compute lead makes it the pick for OpenCL-heavy tasks, while the P750’s fill-rate advantage points to a different optimization focus. The P750 also supports Vulkan 1.4, while the P580 is limited to Vulkan 1.3, which matters for newer applications that leverage the latest Vulkan extensions.

FAQ

Q: Which GPU has a higher Geekbench OpenCL score?

A: The Intel Iris Pro Graphics P580 scores 9082, which is 38.6% higher than the Intel UHD Graphics P750’s 6554.

Q: How do these GPUs compare to discrete graphics cards in their nearest rival lists?

A: The P580’s average score of 7170 is nearly identical to the NVIDIA GeForce GTX 970 (7157, 0.2% difference) and AMD Radeon Vega 8 Mobile (7203, 0.5% difference). The P750’s average of 6554 is 0.4% behind the AMD Radeon HD 7730M and 0.6% ahead of the NVIDIA GeForce GTX 670M.

Q: What is the shading unit count for each GPU?

A: The P580 has 576 shading units, while the P750 has 256 shading units. This is a 2.25x difference in favor of the P580.

Q: Does the newer P750 have any performance advantages?

A: Yes. The P750 has a pixel rate of 41.60 GPixel/s versus 9.000 GPixel/s for the P580, and a texture rate of 83.20 GTexel/s versus 72.00 GTexel/s. Its boost clock is also higher at 1300 MHz versus 1000 MHz.

Q: What are the FP32 and FP16 performance figures?

A: The P580 delivers 1,152.0 GFLOPS FP32 and 2.304 TFLOPS FP16 (2:1). The P750 delivers 665.6 GFLOPS FP32 and 1,331.2 GFLOPS FP16 (2:1).

Q: Do both GPUs support the same API levels?

A: Both support DirectX 12 (12_1) and OpenGL 4.6. However, the P750 supports Vulkan 1.4, while the P580 supports Vulkan 1.3.

Specification Differences

The two GPUs differ in several key specifications. The P580 is based on the Skylake GT4e chip with a 14 nm+ process node, while the P750 uses the Rocket Lake chip on a 14 nm+++ node. Shading units are 576 for the P580 versus 256 for the P750. Texture mapping units are 72 versus 64, and raster output pipelines are 9 versus 32. The boost clock is 1000 MHz for the P580 and 1300 MHz for the P750; the base clock is identical at 350 MHz. Pixel rate is 9.000 GPixel/s for the P580 and 41.60 GPixel/s for the P750. Texture rate is 72.00 GTexel/s versus 83.20 GTexel/s. FP32 performance is 1,152.0 GFLOPS versus 665.6 GFLOPS, and FP16 performance is 2.304 TFLOPS (2:1) versus 1,331.2 GFLOPS (2:1). Vulkan support differs, with the P580 at 1.3 and the P750 at 1.4. The P580 has a release date of 2015-08-31, while the P750 has no release date listed. Both have a TDP of 15 W, use System Shared memory, and feature a Ring Bus interface.

Architecture Differences

The architectural divide is clear. The P580 uses Intel’s Generation 9.0 architecture, specifically the Skylake GT4e chip, fabricated on a 14 nm+ process. The P750 moves to Generation 12.1, based on the Rocket Lake chip, on a more refined 14 nm+++ node. The P580’s GT4e designation indicates a high-end integrated part with a large execution unit array, which explains its 576 shading units and 72 TMUs. The P750, despite being newer, is a more streamlined design with 256 shading units and 64 TMUs, but it compensates with a much higher ROP count of 32 and a substantially higher pixel rate.

The P580’s higher FP32 and FP16 throughput stems from its wider shader array, making it more compute-oriented. The P750’s higher pixel and texture rates, combined with a 30% higher boost clock, suggest a design tuned for fill-rate and clock efficiency rather than raw parallel math. Both use system-shared memory with bandwidth described as system dependent, so memory performance hinges on the host platform. The P580’s older architecture supports Vulkan 1.3, while the P750’s Generation 12.1 adds Vulkan 1.4 support, a meaningful update for API compatibility. Both are end-of-life products, and the P580’s release date of 2015-08-31 predates the P750’s unspecified launch by several years. The P750’s higher pixel rate (41.60 GPixel/s) and texture rate (83.20 GTexel/s) indicate a different optimization focus, likely targeting higher-resolution display output and rasterization-heavy tasks, while the P580’s compute lead makes it the better choice for OpenCL workloads.