Intel Iris Pro Graphics P580 vs NVIDIA GRID K2 Comparison

The NVIDIA GRID K2 and the Intel Iris Pro Graphics P580 occupy unusual positions in this database: one is a datacenter virtualization card with no display outputs at all, the other is an integrated graphics solution soldered into the CPU package. Comparing them directly is still informative, because the recorded benchmark data shows two parts that land surprisingly close together in overall performance despite radically different design philosophies. The database places the GRID K2 at an average benchmark score of 8080 and the 42nd percentile of all GPUs, while the P580 averages 7170 and sits at the 39th percentile. That is a narrow gap, roughly 12.7% in the K2's favor, but the two wins it rests on come from very different test APIs, which shapes the whole analysis.

Head-to-Head Benchmarks

Only one test appears in both parts' recorded results: Geekbench OpenCL. There, the GRID K2 scores 10602 against the P580's 9082, a 16.7% advantage for the Kepler card. That single shared data point is the entire head-to-head record, and the database credits the K2 with one win and the P580 with zero.

The context around that result matters. The K2's OpenCL score of 10602 is its stronger showing; its Geekbench Metal score of 5557 drags its average down to 8080. The P580's average of 7170 is built from its OpenCL 9082 result and a Geekbench Vulkan score of 5258. So while the K2 wins the only direct comparison, the P580's Vulkan result of 5258 is nearly identical to the K2's Metal result of 5557, which suggests that when each part is pushed onto a modern low-overhead API, they land in the same neighborhood. The head-to-head sample is simply too thin to declare a blowout either way.

Rival placement reinforces the closeness. The GRID K2's average sits within a fraction of a percent of the GeForce GTX 650 Ti Boost (8067, 0.2% behind the K2), the GeForce 945M (8099, 0.2% ahead), the GTX 650 Ti (8053, 0.3% behind), and the GTX 880M (8040, 0.5% behind). The P580 clusters around the GTX 560 SE (7171, dead even), the GTX 970 (7157, 0.2% behind the P580's average), the Radeon Vega 8 Mobile (7203, 0.5% ahead), and the GTX 750 (7222, 0.7% ahead). Both parts are effectively locked into tightly packed performance tiers separated by under one percent from their nearest competitors, so the real difference between them is the 11% average gap that separates those two clusters.

The Verdict

The data favors the GRID K2 on raw compute. It wins the only shared benchmark by 16.7%, holds a higher average score (8080 versus 7170), and ranks three percentile points higher across the database (42 versus 39). Its 2.289 TFLOPS of FP32 throughput is nearly double the P580's 1152.0 GFLOPS, and its dedicated 4 GB of GDDR5 with 160.0 GB/s of bandwidth removes it from dependence on system memory entirely.

The P580's case rests on efficiency and API currency. It does all of this within a 15 W TDP, compared to the K2's 225 W, a fifteenfold difference in power envelope, and it requires no slot, no power connectors, and no suggested PSU rating at all since it is an integrated part. It also supports DirectX 12 feature level 12_1 versus the K2's 11_0, and Vulkan 1.3 versus 1.2.175. For any workload that values modern API support or tight power constraints over absolute OpenCL throughput, the recorded data points toward the P580. For pure GPU compute per the shared benchmark, the K2 is the clear answer.

FAQ

Q: Which GPU wins the only direct benchmark comparison?

A: The NVIDIA GRID K2. In Geekbench OpenCL it scores 10602 versus 9082 for the P580, a 16.7% advantage.

Q: How close are the two in overall database standing?

A: Very close. The GRID K2 averages 8080 and sits at the 42nd percentile; the P580 averages 7170 and sits at the 39th percentile, a three-point percentile gap.

Q: How much power does each consume?

A: The GRID K2 has a 225 W TDP and needs one 6-pin plus one 8-pin power connector, with a suggested 550 W PSU. The P580 is an integrated solution with a 15 W TDP and no power connectors.

Q: Do they support the same graphics APIs?

A: Both support DirectX 12 and OpenGL 4.6, but at different feature levels and Vulkan versions. The P580 supports DirectX 12 feature level 12_1 and Vulkan 1.3, while the K2 supports feature level 11_0 and Vulkan 1.2.175.

Q: What rivals does each part compete with?

A: The K2's average is within 0.5% of the GTX 650 Ti Boost, GeForce 945M, GTX 650 Ti, and GTX 880M. The P580 matches the GTX 560 SE and sits within 0.7% of the GTX 970, Radeon Vega 8 Mobile, and GTX 750.

Q: Are both products still in production?

A: No. The database lists both as end-of-life. The K2 launched with a 5,199 USD launch MSRP; no launch MSRP is recorded for the P580, which shipped as an integrated part.

Specification Differences

The two parts differ on nearly every measurable specification. The GRID K2 is a discrete dual-slot card on PCIe 3.0 x16 with 4 GB of GDDR5 on a 256-bit bus delivering 160.0 GB/s, with memory running at 1250 MHz (5 Gbps effective). The P580 is an integrated solution using system-shared memory of system-dependent bandwidth, connected via a ring bus rather than PCIe.

Rendering resources also diverge sharply. The K2 fieldss 1536 shading units, 128 TMUs, and 32 ROPs, producing 23.84 GPixel/s of pixel rate and 95.36 GTexel/s of texture rate. The P580 has 576 shading units, 72 TMUs, and just 9 ROPs, yielding 9.000 GPixel/s and 72.00 GTexel/s. The K2 more than doubles pixel throughput and beats texture throughput by roughly a third.

FP32 compute favors the K2, 2.289 TFLOPS versus 1152.0 GFLOPS, but FP16 tells a different story: the P580 records 2.304 TFLOPS at a 2:1 rate, which actually edges past the K2's FP32 figure, while no FP16 figure is recorded for the K2 at all. The K2 measures 267 mm (10.5 inches) long; the P580, as an IGP, has no card dimensions. The K2 has no display outputs because it is a virtualization card; the P580's outputs depend on the motherboard.

Architecture Differences

These are architecturally distant products. The GRID K2 uses NVIDIA's Kepler architecture on the GK104 chip, fabricated on TSMC's 28 nm process, with 3,540 million transistors packed into a 294 mm² die for a density of 12.0M per mm². The P580 is Intel's Generation 9.0 graphics, built on the Skylake GT4e chip at Intel's own 14 nm+ node. No transistor count, die size, or density figures are recorded for it.

Clock behavior differs in kind. The K2's database entry lists no base, boost, or game clock, only memory speed. The P580 records a 350 MHz base and a 1000 MHz boost clock. Feature support diverges on modern APIs, with the P580 carrying the newer DirectX feature level and Vulkan version, while both share OpenGL 4.6. Neither part has RT cores or tensor cores. The release dates are also separated: the K2 arrived on 2013-05-10, the P580 on 2015-08-31, meaning the P580 benefits from a newer process node and API stack while the K2 brings a much larger dedicated silicon budget.

Where Each One Wins

The GRID K2 wins on raw compute throughput. Its OpenCL score of 10602 leads by 16.7%, its FP32 output of 2.289 TFLOPS is nearly double the P580's, and its dedicated GDDR5 bandwidth of 160.0 GB/s is unaffected by whatever the host system provides. Workloads that lean on OpenCL compute, dense pixel output, or high texture throughput, 95.36 GTexel/s versus 72.00, are firmly in K2 territory. Its rival bracket, clustered with the GTX 650 Ti Boost and GTX 880M, confirms mainstream-discrete-class throughput.

The P580 wins on efficiency and platform integration. It delivers 9082 in OpenCL, within striking distance of the K2, at a 15 W TDP versus 225 W, with no power connectors, no slot occupancy, and no discrete memory to buy or cool. Its Vulkan result of 5258 shows it holding its own on a modern API the K2 only partially supports, and its DirectX 12_1 feature level exceeds the K2's 11_0. Its rival bracket, spanning the GTX 560 SE, GTX 970, GTX 750, and Vega 8 Mobile, places it alongside entry discrete cards and newer mobile integrated solutions. In short: the K2 is the compute card that happens to be near this tier, and the P580 is the integrated part that climbs into this tier, and the benchmark record shows both statements are true at once.