Intel Iris Pro Graphics P580 vs NVIDIA GeForce 945M Comparison

The NVIDIA GeForce 945M and the Intel Iris Pro Graphics P580 are two end-of-life mobile graphics solutions from 2015 that represent fundamentally different philosophies: a discrete MXM module on an older but larger process, versus an integrated GPU on a modern node. The recorded benchmark data shows a clear winner in the one head-to-head test, and the gap is wider than the raw specifications might suggest. The Iris Pro P580's average benchmark score of 7170 sits at the 39th percentile against all GPUs in the database, while the GeForce 945M's 8099 average places it at the 42nd percentile. But the single direct comparison between them tells a more surprising story.

Head-to-Head Benchmarks

The only head-to-head test recorded in the database is Geekbench OpenCL, and it goes decisively to the Iris Pro Graphics P580: 9082 points against 8099 for the GeForce 945M, a margin of 10.8 percent. That is a substantial gap for two parts whose theoretical specifications sit much closer together on paper.

Consider the compute numbers. The 945M delivers 1,305.6 GFLOPS of FP32 throughput against 1,152.0 GFLOPS for the P580, which suggests the NVIDIA part should lead by a comfortable margin. It does not. The benchmark results indicate the P580 extracts more real-world OpenCL performance from its hardware than the 945M does, despite the deficit in shading resources: 576 shading units against 640, and a lower boost clock of 1000 MHz versus 1020 MHz.

Context from each part's rival cluster reinforces how unusual this result is. The GeForce 945M's average score of 8099 lands it within a fraction of a percent of the NVIDIA GRID K2 (8080, a 0.2 percent gap), the AMD Radeon R9 M360 (8129, 0.4 percent behind), the NVIDIA GeForce GTX 650 Ti Boost (8067, 0.4 percent ahead), and the GeForce GTX 950M (8135, 0.4 percent behind). In other words, the 945M performs exactly like a mid-tier Maxwell-class mobile discrete GPU should.

The P580's average of 7170 puts it dead even with the GeForce GTX 560 SE (7171, a 0 percent gap), marginally ahead of the GeForce GTX 970 average cluster figure of 7157 (0.2 percent), just behind the Radeon Vega 8 Mobile (7203, 0.5 percent) and the GeForce GTX 750 (7222, 0.7 percent). That is strong company for an integrated GPU, and in the OpenCL discipline specifically, it beats the 945M outright.

The P580 also holds a Vulkan result of 5258 in Geekbench Vulkan, a test for which no comparable 945M score exists in the database, so no cross-comparison can be drawn there. The overall head-to-head record stands at one win for the P580 and zero for the 945M.

The Verdict

Based strictly on the recorded data, the Iris Pro Graphics P580 is the stronger performer where it counts: the one shared benchmark favors it by 10.8 percent. For a buyer choosing between laptops carrying these two parts, and caring about GPU compute throughput, the P580 is the data-backed pick.

The 945M's case rests entirely on its specification sheet rather than its results. It has a dedicated 2 GB of DDR3 memory on a 128-bit bus with 28.80 GB/s of bandwidth, where the P580 shares system memory with bandwidth that is entirely system dependent. If consistent memory bandwidth independent of the host system's RAM configuration matters, the discrete design has a structural advantage the benchmark data cannot fully capture. It also posts higher raw pixel output at 16.32 GPixel/s against 9.000 GPixel/s, and higher FP32 compute at 1,305.6 GFLOPS against 1,152.0 GFLOPS.

Power is the other axis, and it is lopsided. The P580 is specified at 15 W as an integrated GPU, while the 945M carries a 75 W TDP as an MXM module requiring its own slot. A five-fold difference in TDP is the price of the discrete approach here.

Net verdict: the P580 wins on measured performance and power efficiency; the 945M wins on dedicated memory and theoretical rasterization throughput, but could not convert either into a benchmark win in the recorded data.

Architecture Differences

These two parts could hardly be more different architecturally. The GeForce 945M is built on the GM107 chip, Maxwell architecture, GeForce 900M generation, manufactured by TSMC on a 28 nm process. It packs 1,870 million transistors into a 148 mm² die, for a density of 12.6M transistors per square millimeter. The Iris Pro P580 uses the Skylake GT4e chip, Intel's Generation 9.0 graphics architecture, produced in-house by Intel on a 14 nm+ process. Transistor count, die size, and density are not recorded for the P580, but the node gap is nearly a full decade of process scaling in Intel's favor.

The execution-resource split is interesting. The 945M has more shading units (640 versus 576) and more ROPs (16 versus 9), but the P580 actually has far more texture units: 72 against 40. That is why the P580's texture rate of 72.00 GTexel/s comfortably exceeds the 945M's 40.80 GTexel/s, nearly matching it despite the pixel-rate deficit. The P580 also supports FP16 at 2.304 TFLOPS with a 2:1 ratio, a capability for which no figure is recorded on the 945M.

Memory subsystems diverge completely. The 945M has 2 GB of dedicated DDR3 at 900 MHz (1800 Mbps effective) over a 128-bit bus. The P580 shares system memory wholesale: size, type, bus width, and bandwidth are all system dependent. Clock behavior differs too, with the 945M running a tight 928 MHz base to 1020 MHz boost range, while the P580 swings from a 350 MHz base up to a 1000 MHz boost.

Software support slightly favors the P580 on feature level: DirectX 12 with 12_1 features against the 945M's 12 (11_0), and Vulkan 1.3 against 1.4 in the 945M's favor, with both supporting OpenGL 4.6. Neither part has RT cores or tensor cores. Neither has a recorded launch MSRP. Both are end-of-life. The 945M was released on October 26, 2015, succeeding the GeForce 800M line and preceding the GeForce 10 Mobile series; the P580 shipped earlier, on August 31, 2015.

FAQ

Q: Which GPU is faster in the benchmarks?

A: The Intel Iris Pro Graphics P580. It scored 9082 in Geekbench OpenCL against 8099 for the GeForce 945M, winning by 10.8 percent.

Q: Which one has more raw compute throughput?

A: The GeForce 945M, on paper. It delivers 1,305.6 GFLOPS FP32 against the P580's 1,152.0 GFLOPS, yet it still lost the one shared benchmark.

Q: How do they compare in power consumption?

A: The P580 is specified at 15 W as an integrated GPU; the 945M carries a 75 W TDP as an MXM module with no power connectors.

Q: Does the 945M have its own memory?

A: Yes, 2 GB of DDR3 on a 128-bit bus with 28.80 GB/s of bandwidth. The P580 shares system memory with system-dependent bandwidth.

Q: Where does each GPU rank against all GPUs in the database?

A: The 945M sits at the 42nd percentile with an average score of 8099; the P580 sits at the 39th percentile with an average of 7170.

Q: Which supports the newer DirectX feature level?

A: The P580 supports DirectX 12 (12_1); the 945M supports DirectX 12 (11_0).

Where Each One Wins

The P580 wins on measured GPU compute, full stop. The OpenCL result is the only direct evidence in the database, and it favors the Intel part by 10.8 percent. The P580 also wins on power efficiency by a wide TDP margin, on texture throughput at 72.00 GTexel/s versus 40.80 GTexel/s, on FP16 capability, and on the newer DirectX 12_1 feature level. Anyone prioritizing battery-sensitive general compute workloads would be choosing the P580 on this data.

The 945M wins on dedicated memory capacity and bandwidth, on pixel fill rate at 16.32 GPixel/s versus 9.000 GPixel/s, on ROP count (16 versus 9), and on FP32 theoretical throughput. Its Vulkan 1.4 support also edges the P580's 1.3. For workloads that depend on guaranteed memory bandwidth rather than system-shared RAM, and on rasterization output, the 945M holds the theoretical edge, though the database contains no benchmark result demonstrating it.

Specification Differences

The two parts differ on nearly every axis. Process node: 28 nm TSMC for the 945M versus 14 nm+ Intel for the P580. Form factor: MXM-B (3.0) bus interface and MXM module slot for the 945M, versus an integrated GPU on a ring bus. The 945M's die measures 148 mm² with 1,870 million transistors; no equivalent figures are recorded for the P580.

Resource counts split: 640 shading units versus 576, 40 TMUs versus 72, and 16 ROPs versus 9, all in the 945M's favor except TMUs. The 945M has dedicated 2 GB DDR3 memory with a 128-bit bus and 28.80 GB/s bandwidth; the P580 shares everything with the system. Clocks run 928/1020 MHz base/boost on the 945M and 350/1000 MHz on the P580. TDP is 75 W versus 15 W. API support differs at DirectX 12 (11_0) versus 12 (12_1) and Vulkan 1.4 versus 1.3, with OpenGL 4.6 shared. Release dates are October 26, 2015 for the 945M and August 31, 2015 for the P580, and both are end-of-life products.