Intel Iris Xe MAX Graphics vs NVIDIA Tesla M2090 Comparison

FAQ

Q: How do the Intel Iris Xe MAX Graphics and NVIDIA Tesla M2090 compare in raw compute performance?

A: In the Geekbench OpenCL benchmark, the Intel Iris Xe MAX Graphics scores 14315, while the NVIDIA Tesla M2090 scores 13075. This gives Intel a 9.5% lead in this particular test.

Q: What is the memory capacity difference between the two cards?

A: The Intel Iris Xe MAX Graphics comes with 4 GB of LPDDR4X memory on a 128-bit bus, while the NVIDIA Tesla M2090 offers 6 GB of GDDR5 memory on a much wider 384-bit bus.

Q: Which card has higher memory bandwidth?

A: The NVIDIA Tesla M2090 has significantly higher memory bandwidth at 177.4 GB/s, compared to the Intel Iris Xe MAX Graphics' 68.26 GB/s. This is despite the Intel card running at a higher effective memory speed of 4.3 Gbps versus 3.7 Gbps for the Tesla.

Q: How do the two cards compare in terms of power draw?

A: The Intel Iris Xe MAX Graphics has a TDP of 25 W with a suggested PSU of 200 W, while the NVIDIA Tesla M2090 has a TDP of 250 W and requires a suggested PSU of 600 W. The Tesla also needs 1x 6-pin and 1x 8-pin power connectors.

Q: Which card supports newer graphics APIs?

A: The Intel Iris Xe MAX Graphics supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The NVIDIA Tesla M2090 supports DirectX 12 (11_0) and OpenGL 4.6, but has no Vulkan support listed.

Q: What are the physical form factor differences?

A: The Intel Iris Xe MAX Graphics is an IGP (integrated graphics processor) with no display outputs and no specified length. The NVIDIA Tesla M2090 is a dual-slot card measuring 248 mm (9.8 inches) in length, also with no display outputs.

The Verdict

The benchmark data presents a clear but nuanced picture. The Intel Iris Xe MAX Graphics wins the only head-to-head benchmark recorded, beating the NVIDIA Tesla M2090 by 9.5% in Geekbench OpenCL. It also sits at the 56th percentile of all GPUs, compared to the Tesla M2090's 53rd percentile. For compute workloads measured by OpenCL, the Intel card is the better choice.

However, the NVIDIA Tesla M2090 has its own strengths that matter in different contexts. It offers 6 GB of memory versus 4 GB, and its memory bandwidth is 2.6 times higher at 177.4 GB/s compared to 68.26 GB/s. This makes it better suited for memory-intensive workloads, despite its lower raw compute score. The Tesla also has more texture mapping units (64 versus 48) and more render output units (48 versus 24), which could benefit certain graphics pipelines.

The choice depends on workload characteristics. For general compute and OpenCL tasks, the Intel Iris Xe MAX Graphics is the data-backed winner. For memory-heavy operations where bandwidth matters more than raw FP32 throughput, the NVIDIA Tesla M2090 holds an advantage that the benchmark scores do not fully capture. The Intel part is also far more power-efficient, drawing 25 W versus 250 W, which makes it viable in systems with modest power supplies.

Head-to-Head Benchmarks

The only recorded head-to-head benchmark is Geekbench OpenCL, and the Intel Iris Xe MAX Graphics takes it decisively. Intel scores 14315 against NVIDIA's 13075, a 9.5% advantage. This places the Intel part within striking distance of several strong rivals: it trails the AMD Radeon Vega 11 by just 0.3% (14315 versus 14352), sits 0.3% ahead of the NVIDIA GeForce GTX 1070 Ti (14277), and is 0.4% behind the NVIDIA GeForce GTX TITAN (14373). The AMD Radeon RX Vega 11 leads Intel by 0.5% with a score of 14385.

The NVIDIA Tesla M2090, meanwhile, lands in a different competitive tier. Its score of 13075 puts it 0.7% ahead of the NVIDIA GeForce GTX 1660 SUPER (12986), 0.9% behind the NVIDIA GeForce GTX 950 (13189), 1% ahead of the NVIDIA GeForce RTX 3050 Ti Mobile (12940), and 1.1% ahead of the AMD Radeon RX 580 (12928). The gap between the two cards in question, 9.5%, is larger than the spread within either card's immediate rival group.

In practical terms, the Intel card's 9.5% lead means it is roughly one full competitor-tier above the Tesla M2090 in OpenCL compute. The Intel part's nearest rivals include the GTX 1070 Ti and GTX TITAN, while the Tesla M2090 competes with the GTX 1660 SUPER and RX 580. That places the Intel Iris Xe MAX roughly one performance class higher, despite being an integrated part with far lower power consumption.

Specification Differences

The two cards differ substantially across nearly every specification category. The Intel Iris Xe MAX Graphics is built on a 10 nm process by Intel, while the NVIDIA Tesla M2090 uses a 40 nm process from TSMC. The Intel chip, designated DG1, has a die size of 95 mm², whereas the NVIDIA GF110 chip measures 520 mm² and packs 3,000 million transistors at a density of 5.8M per mm².

Clock speeds show a mixed picture. The Intel card has a base clock of 300 MHz and a boost clock of 1650 MHz. The Tesla M2090 has no listed base or boost clocks, only a memory clock of 924 MHz (3.7 Gbps effective). Intel's memory runs at 2133 MHz (4.3 Gbps effective).

Memory specifications diverge sharply. Intel offers 4 GB of LPDDR4X on a 128-bit bus with 68.26 GB/s bandwidth. NVIDIA offers 6 GB of GDDR5 on a 384-bit bus with 177.4 GB/s bandwidth. The Tesla's bandwidth advantage is 2.6-fold.

The compute units also differ. Intel has 768 shading units, 48 TMUs, and 24 ROPs. NVIDIA has 512 shading units, 64 TMUs, and 48 ROPs. Intel's pixel rate is 39.60 GPixel/s versus 20.83 GPixel/s for NVIDIA. Intel's texture rate is 79.20 GTexel/s versus 41.66 GTexel/s. Intel's FP32 throughput is 2.534 TFLOPS, while NVIDIA manages 1,332.2 GFLOPS (1.332 TFLOPS). Intel also supports FP16 at 5.069 TFLOPS with a 2:1 ratio; NVIDIA has no FP16 listing.

Power and physical specifications could not be more different. Intel's TDP is 25 W with a 200 W suggested PSU, and it is an IGP with no power connectors. NVIDIA's TDP is 250 W with a 600 W suggested PSU, requiring 1x 6-pin plus 1x 8-pin connectors, and it occupies a dual-slot form factor measuring 248 mm (9.8 inches). Both cards have no display outputs.

The bus interface also differs: Intel uses PCIe 4.0 x8, while NVIDIA uses PCIe 2.0 x16. API support favors Intel with DirectX 12 (12_1) and Vulkan 1.4, while NVIDIA only reaches DirectX 12 (11_0) and lacks Vulkan entirely. Both support OpenGL 4.6.

Architecture Differences

The architectural gap between these two GPUs spans over a decade of hardware evolution. The Intel Iris Xe MAX Graphics is built on Intel's Generation 12.1 architecture, part of the Xe Graphics generation, and uses the DG1 chip. This architecture was designed for modern compute workloads with support for DirectX 12 Ultimate features (12_1) and Vulkan 1.4, enabling current-generation graphics APIs and compute paradigms.

The NVIDIA Tesla M2090 uses the Fermi 2.0 architecture, based on the GF110 chip, from the Tesla Fermi (x20xx) generation. Fermi was designed for compute-heavy scientific and professional workloads, but its API support reflects an older era: DirectX 12 (11_0) without Vulkan support. This limits its compatibility with modern graphics and compute software stacks.

The manufacturing process tells a story of two different eras. Intel uses a 10 nm process at its own foundry, while NVIDIA relies on TSMC's 40 nm process. The die sizes reflect this: Intel's DG1 measures 95 mm², while NVIDIA's GF110 is 520 mm², over five times larger. NVIDIA packs 3,000 million transistors into that space, but at a relatively low density of 5.8M per mm² due to the older process.

Compute architecture differences are significant. Intel deploys 768 shading units, which is 50% more than NVIDIA's 512. Yet NVIDIA counters with 64 TMUs versus Intel's 48, and 48 ROPs versus Intel's 24. This suggests NVIDIA's architecture prioritized texture and pixel throughput, while Intel focused on raw shader compute. The FP32 numbers bear this out: Intel reaches 2.534 TFLOPS versus NVIDIA's 1,332.2 GFLOPS, a 1.9x advantage for Intel.

Memory architecture also reflects different design goals. NVIDIA's 384-bit bus with 177.4 GB/s bandwidth is a high-throughput design, whereas Intel's 128-bit bus with 68.26 GB/s is more modest. The Tesla M2090's 6 GB capacity exceeds Intel's 4 GB, but Intel's LPDDR4X memory operates at a higher effective speed of 4.3 Gbps versus 3.7 Gbps.

The feature sets diverge in ways that matter for modern versus legacy workloads. Intel's support for FP16 at 5.069 TFLOPS (2:1 ratio) enables accelerated half-precision compute, which is increasingly important in AI and machine learning inference. NVIDIA's Fermi architecture lacks any FP16 capability. Intel's Vulkan 1.4 support opens the door to cross-platform modern graphics, while NVIDIA's DirectX 12 (11_0) caps out at an earlier DirectX feature level.

Both cards are end-of-life products, but their release dates show the generation gap: Intel launched on 2020-10-30, while NVIDIA launched on 2011-07-24. Intel's predecessor is listed as "Graphics" and successor as "Alchemist"; NVIDIA's predecessor is "Tesla" and successor is "Tesla Kepler." These product lineage names confirm that Intel is iterating within its Xe family, while NVIDIA is moving through the Tesla compute generations.