AMD Radeon Pro 560X vs NVIDIA Tesla M2090 Comparison

FAQ

Q: Which GPU has a higher average benchmark score?

A: The AMD Radeon Pro 560X has an average benchmark score of 15082, while the NVIDIA Tesla M2090 scores 13075. The AMD part sits at the 57th percentile of all GPUs, versus the 53rd percentile for the NVIDIA part.

Q: How do the two compare in the single shared benchmark test?

A: In Geekbench OpenCL, the NVIDIA Tesla M2090 scores 13075 against 9663 for the AMD Radeon Pro 560X, a 26.1% advantage for the NVIDIA card.

Q: What memory configurations do these cards use?

A: The AMD Radeon Pro 560X has 4 GB of GDDR5 on a 128-bit bus with 94.08 GB/s of bandwidth. The NVIDIA Tesla M2090 has 6 GB of GDDR5 on a 384-bit bus with 177.4 GB/s of bandwidth.

Q: What are the power requirements for each card?

A: The AMD Radeon Pro 560X has a TDP of 75 W and requires no power connectors. The NVIDIA Tesla M2090 has a TDP of 250 W, requires one 6-pin and one 8-pin connector, and recommends a 600 W power supply.

Q: Which card is more transistor-dense despite using a larger process node?

A: The AMD Radeon Pro 560X, built on a 14 nm process at GlobalFoundries, packs 3,000 million transistors into a 123 mm² die, giving 24.4M transistors per mm². The NVIDIA Tesla M2090 uses TSMC's 40 nm process with the same 3,000 million transistors on a 520 mm² die, for only 5.8M per mm².

Q: What API support does each card offer?

A: The AMD Radeon Pro 560X supports DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.3. The NVIDIA Tesla M2090 supports DirectX 12 (11_0) and OpenGL 4.6, with no Vulkan support listed.

Architecture Differences

The AMD Radeon Pro 560X is built on the Polaris 21 chip using GCN 4.0 architecture, manufactured on a 14 nm process at GlobalFoundries. The NVIDIA Tesla M2090 uses the GF110 chip with Fermi 2.0 architecture, manufactured on TSMC's 40 nm process. This process gap is significant: both chips pack the same 3,000 million transistors, but the AMD die measures only 123 mm² while the NVIDIA die spans 520 mm². The transistor density difference is stark, 24.4M per mm² versus 5.8M per mm², reflecting the seven-year gap between their release dates.

The two GPUs take fundamentally different approaches to shader organization. The AMD Radeon Pro 560X deploys 1024 shading units, 64 texture mapping units, and 16 ROPs. The NVIDIA Tesla M2090 uses 512 shading units, 64 TMUs, and 48 ROPs. Despite having half the shading units, the NVIDIA part achieves a higher pixel rate of 20.83 GPixel/s versus 16.06 GPixel/s for the AMD card, thanks to its much larger ROP count. The texture rate tells the opposite story: the AMD card reaches 64.26 GTexel/s against 41.66 GTexel/s for the NVIDIA part.

Memory architecture differs substantially. The AMD card uses a 128-bit bus with 4 GB of GDDR5 running at 1470 MHz (5.9 Gbps effective), yielding 94.08 GB/s of bandwidth. The NVIDIA card runs 6 GB of GDDR5 on a 384-bit bus at 924 MHz (3.7 Gbps effective), producing 177.4 GB/s. The Tesla M2090's wider bus more than compensates for its slower memory clock.

Compute capabilities also diverge. The AMD Radeon Pro 560X delivers 2.056 TFLOPS of FP32 and matches that with 2.056 TFLOPS of FP16 at a 1:1 ratio. The NVIDIA Tesla M2090 provides 1,332.2 GFLOPS of FP32 and lists no FP16 support. The AMD card is clearly designed for modern workloads that can exploit half-precision throughput, while the Tesla's focus lies elsewhere.

The interface and physical design reflect their different intended environments. The AMD Radeon Pro 560X is an integrated GPU package (IGP) with a PCIe 3.0 x8 interface and portable-device-dependent display outputs. The NVIDIA Tesla M2090 is a dual-slot card measuring 248 mm (9.8 inches), using PCIe 2.0 x16, and providing no display outputs at all. It is a compute-only accelerator.

Head-to-Head Benchmarks

The database contains one directly comparable benchmark between these two cards: Geekbench OpenCL. The NVIDIA Tesla M2090 wins this test with a score of 13075 against the AMD Radeon Pro 560X's 9663. The delta is 26.1% in favor of the NVIDIA card. This is a substantial margin in a compute-oriented workload.

Contextualizing the AMD card's score through its nearest rivals helps clarify its position. The Radeon Pro 560X sits within a tight cluster: it is 0.1% ahead of the NVIDIA GeForce GTX 660 Ti (15063), 0.6% behind the AMD Radeon RX 7600 (15171), 0.8% behind the NVIDIA GeForce RTX 3050 OEM (15199), and 1.2% behind the AMD Radeon 680M (15270). These deltas are all within a couple of percentage points, meaning the Radeon Pro 560X's average score of 15082 places it in a dense competitive band. Its overall average is lifted by strong showings in Geekbench Metal (18763) and Geekbench Vulkan (16819), tests the Tesla M2090 cannot run.

The NVIDIA Tesla M2090's nearest rivals tell a similar story of tight competition. It is 0.7% ahead of the NVIDIA GeForce GTX 1660 SUPER (12986), 0.9% behind the NVIDIA GeForce GTX 950 (13189), 1% behind the NVIDIA GeForce RTX 3050 Ti Mobile (12940), and 1.1% behind the AMD Radeon RX 580 (12928). The Tesla M2090's single recorded benchmark score of 13075 therefore slots it just above the middle of this group.

The head-to-head result is notable because it inverts the overall average benchmark hierarchy. The AMD card's average score is 15082, well above the Tesla's 13075, yet in the one test both share, the Tesla wins decisively. This suggests the AMD part's advantage in the database stems from its Metal and Vulkan results, which are not available for the NVIDIA card. The OpenCL result alone favors the older Fermi architecture.

Specification Differences

The two cards differ across nearly every major specification category. The AMD Radeon Pro 560X uses a 14 nm process; the NVIDIA Tesla M2090 uses 40 nm. Both have 3,000 million transistors, but the die sizes diverge at 123 mm² versus 520 mm², yielding transistor densities of 24.4M per mm² and 5.8M per mm² respectively.

Memory capacities differ: 4 GB on the AMD card versus 6 GB on the NVIDIA card. Both use GDDR5, but the bus widths are 128-bit and 384-bit respectively. Clock speeds and resulting bandwidths differ as well, with the AMD memory running at 1470 MHz with 94.08 GB/s and the NVIDIA memory at 924 MHz with 177.4 GB/s.

Shader resources differ in both count and balance. The AMD card has 1024 shading units versus 512, with both sharing 64 TMUs. ROPs are heavily weighted toward the NVIDIA card at 48 versus 16. Pixel rates favor NVIDIA at 20.83 GPixel/s over 16.06 GPixel/s; texture rates favor AMD at 64.26 GTexel/s over 41.66 GTexel/s.

Compute throughput favors AMD in raw numbers: 2.056 TFLOPS FP32 versus 1,332.2 GFLOPS, with FP16 support only on the AMD side. Power profiles are dramatically different, with TDPs of 75 W and 250 W. The AMD card is an IGP with no power connectors, while the NVIDIA card is a dual-slot design with one 6-pin and one 8-pin connector and a 600 W suggested PSU.

Interface and output capabilities differ as well. The AMD card uses PCIe 3.0 x8 with portable-device-dependent outputs; the NVIDIA card uses PCIe 2.0 x16 with no display outputs. API support overlaps on OpenGL 4.6 but diverges elsewhere: DirectX 12 (12_0) for AMD versus DirectX 12 (11_0) for NVIDIA, and Vulkan 1.3 for AMD with no Vulkan support for NVIDIA.

The Verdict

The data presents two very different profiles. The AMD Radeon Pro 560X is the newer, more efficient part with a 14 nm process, a 75 W TDP, and support for modern APIs including Vulkan 1.3 and DirectX 12 (12_0). Its average benchmark score of 15082 places it at the 57th percentile of all GPUs, comfortably ahead of the Tesla M2090's 13075 average and 53rd percentile.

However, in the single directly comparable workload, Geekbench OpenCL, the NVIDIA Tesla M2090 wins by 26.1%. The Tesla also offers more memory (6 GB versus 4 GB) and substantially higher memory bandwidth (177.4 GB/s versus 94.08 GB/s), along with nearly three times the ROP count. Its pixel rate of 20.83 GPixel/s exceeds the AMD card's 16.06 GPixel/s.

The AMD card is the better all-rounder in the database's broader benchmark suite, with strong Metal and Vulkan results that the Tesla cannot match. The NVIDIA card is the more specialized compute accelerator, winning the shared OpenCL test despite its older architecture and far higher power draw. Buyers seeking modern API support, efficiency, and overall benchmark versatility should favor the AMD Radeon Pro 560X. Those prioritizing raw OpenCL compute throughput in a legacy compute context should consider the Tesla M2090, provided the 250 W power requirement and dual-slot footprint are acceptable.

Where Each One Wins

The AMD Radeon Products 560X wins in modern API-heavy workloads: it is the only one of the two with Vulkan support, and its DirectX 12 (12_0; the NVIDIA part only reaches 11_0, while its OpenGL 4.6 support is shared by both. Its FP16 throughput of 2.056 TFLOPS opens up half-precision compute that the Tesla M2090 cannot access at all, and its 2.056 TFLOPS FP32 is a larger number than the Tesla M2090 GFLOPS 1,332. The AMD card's 16.06 GPixel/s is half-bus 94.08 GB/s bandwidth, 4 GB of memory, and a 75 W TDP, which makes it the only one without power connectors.

The NVIDIA Tesla M2090 wins in compute-bound OpenCL scenarios: its 13075 Geekbench OpenCL score beats the 9663 AMD score, and the 26.1% delta is the largest margin in any shared test. Its 6 GB frame buffer, 384-bit bus, and 177.4 GB/s bandwidth give it a memory bandwidth advantage that the AMD card cannot match. The Tesla M2090 also leads in pixel throughput at 20.83 GPixel/s thanks to 48 ROPs, and it offers a 1.1% advantage over the AMD card in its average score relative to its nearest rivals, meaning it holds its own in its competitive tier. The Tesla M2090's 248 mm dual-slot design with 1x 6-pin and 1x 8-pin connectors, paired with a 600 W suggested PSU, positions it as a dedicated compute accelerator, while the AMD card functions as an integrated solution with portable-device-dependent outputs.

For use-case selection, the AMD Radeon Pro 560X suits environments where modern API coverage, low power draw, and overall benchmark diversity matter, such as portable or integrated systems. The NVIDIA Tesla M2090 suits dedicated compute nodes where OpenCL performance and large memory bandwidth are the priority, and where the lack of display outputs and high power consumption are acceptable trade-offs.