AMD Radeon 680M vs NVIDIA Tesla M2090 Comparison

Head-to-Head Benchmarks

The recorded data contains a single directly comparable benchmark between the AMD Radeon 680M and the NVIDIA Tesla M2090: the Geekbench OpenCL test. In this measurement, the AMD Radeon 680M scores 23,468 points against the NVIDIA Tesla M2090's 13,075 points. That represents a 79.5% advantage for the AMD part, a decisive margin by any standard. The delta is large enough that the Tesla M2090 would need a substantial architectural revision to close the gap, but the data shows no such improvement is present.

The AMD Radeon 680M also carries additional benchmark results that the Tesla M2090 lacks entirely. In 3DMark Steel Nomad DX12, the 680M records 378 points, while its Geekbench Vulkan score is 21,965. These figures cannot be compared directly to the Tesla M2090, as the database contains no corresponding Vulkan or Steel Nomad results for the Fermi-era card. However, the OpenCL comparison alone establishes a clear hierarchy: the 680M is the faster compute device in this pairing.

Looking at the average benchmark scores, the AMD Radeon 680M sits at 15,270 across all recorded tests, while the NVIDIA Tesla M2090 averages 13,075. The 680M's percentile rank among all GPUs is 57, versus 53 for the Tesla M2090. Both cards sit in the mid-range of the database, but the 680M occupies a higher position, consistent with its OpenCL victory.

The nearest rivals in the database further contextualize these results. The AMD Radeon 680M is bracketed by the NVIDIA GeForce GTX 580 (average score 15,283, delta -0.1%), the NVIDIA GeForce RTX 2060 (15,290, delta -0.1%), the NVIDIA GeForce RTX 3050 OEM (15,199, delta 0.5%), and the AMD Radeon RX 7600 (15,171, delta 0.7%). This clustering means the 680M trades blows with a previous-generation flagship (GTX 580), a mainstream ray-tracing card (RTX 2060), and a modern budget GPU (RX 7600). The differences are within 0.7% in either direction, indicating that the 680M delivers performance essentially at parity with those discrete cards in aggregate.

For the NVIDIA Tesla M2090, the nearest rivals are the NVIDIA GeForce GTX 1660 SUPER (average score 12,986, delta 0.7%), the NVIDIA GeForce GTX 950 (13,189, delta -0.9%), the NVIDIA GeForce RTX 3050 Ti Mobile (12,940, delta 1%), and the AMD Radeon RX 580 (12,928, delta 1.1%). The Tesla M2090 is within 1.1% of all four rivals, meaning it performs on par with a GTX 1660 SUPER or an RX 580 in the database's aggregate metric. That is a respectable result for a compute card released over a decade prior, but it is still 15.9% below the 680M's average score in absolute terms.

Where Each One Wins

The AMD Radeon 680M wins the only head-to-head benchmark, the Geekbench OpenCL test, by 79.5%. This is a compute-oriented workload that exercises the GPU's general-purpose processing capabilities, and the 680M's RDNA 2.0 architecture clearly handles it more efficiently than the Fermi 2.0 design in the Tesla M2090. The 680M also has a Vulkan result of 21,965 and a Steel Nomad DX12 score of 378, which demonstrate its capability in modern graphics APIs. The Tesla M2090 has no recorded results for those tests, so the 680M wins by default in any API-specific comparison involving Vulkan or DirectX 12.

The NVIDIA Tesla M2090 does not have any benchmark wins in this comparison. Its only recorded result is the OpenCL score of 13,075, which is lower than every benchmark score listed for the 680M. However, the Tesla M2090 does have a structural advantage in memory configuration: it features 6 GB of dedicated GDDR5 memory on a 384-bit bus, delivering 177.4 GB/s of bandwidth. The 680M uses system-shared memory, with bandwidth described as "System Dependent." In workloads where dedicated memory bandwidth and capacity are critical, such as large matrix operations or data-intensive compute tasks, the Tesla M2090's fixed memory pool could be preferable, even if its raw compute throughput is lower.

The Tesla M2090 also has a higher texture fill rate per ROP count: it packs 64 TMUs and 48 ROPs, against 48 TMUs and 32 ROPs for the 680M. Yet the 680M's pixel rate of 70.40 GPixel/s and texture rate of 105.6 GTexel/s dwarf the Tesla M2090's 20.83 GPixel/s and 41.66 GTexel/s. The 680M's higher clock speeds, 2000 MHz base and 2200 MHz boost versus the Tesla M2090's unspecified core clocks, drive those fill-rate advantages. The data suggests that the 680M wins in raw throughput, while the Tesla M2090 retains an edge in memory architecture that could matter for specific workloads.

The Verdict

The data points to the AMD Radeon 680M as the faster GPU in this pairing. Its OpenCL score of 23,468 is 79.5% higher than the Tesla M2090's 13,075, and its average benchmark score of 15,270 exceeds the Tesla M2090's 13,075 by roughly 16%. The 680M also holds a higher percentile rank (57 vs 53) and has results in modern APIs that the Tesla M2090 cannot match. For any user prioritizing compute performance, modern API support, or efficiency, the 680M is the clear choice.

The NVIDIA Tesla M2090, however, is not without merit. Its 6 GB of dedicated GDDR5 memory with 177.4 GB/s bandwidth provides a fixed, predictable memory subsystem that the 680M cannot offer, since the 680M relies on system-shared memory. For applications that demand large, consistent memory allocation and high bandwidth, the Tesla M2090 may still be functional, especially given its proximity to the GTX 1660 SUPER and RX 580 in aggregate performance. But the Tesla M2090 is end-of-life, uses a 40 nm process node, and consumes 250 W, compared to the 680M's 50 W TDP on a 6 nm node. The efficiency gap is enormous, and the performance gap is decisive.

The verdict is straightforward: the AMD Radeon 680M wins on performance, efficiency, API support, and production status. The Tesla M2090 wins only on dedicated memory capacity and bandwidth, which may matter in niche compute scenarios. Most users should select the 680M.

FAQ

Q: How much faster is the AMD Radeon 680M than the NVIDIA Tesla M2090 in OpenCL?

A: The 680M scores 23,468 against the Tesla M2090's 13,075 in Geekbench OpenCL, a 79.5% advantage for the AMD part.

Q: Does the NVIDIA Tesla M2090 have any benchmark wins over the AMD Radeon 680M?

A: No. The database records one head-to-head benchmark, Geekbench OpenCL, and the 680M wins it. The Tesla M2090 has zero wins in this comparison.

Q: What is the average benchmark score for each GPU?

A: The AMD Radeon 680M averages 15,270 across its three recorded benchmarks, while the NVIDIA Tesla M2090 averages 13,075 from its single OpenCL result.

Q: Which GPU has more memory bandwidth?

A: The NVIDIA Tesla M2090 has 177.4 GB/s of bandwidth from 6 GB of GDDR5 on a 384-bit bus. The AMD Radeon 680M uses system-shared memory, and its bandwidth is system dependent.

Q: How does the AMD Radeon 680M compare to its nearest rivals?

A: The 680M is within 0.7% of the NVIDIA GeForce GTX 580, RTX 2060, RTX 3050 OEM, and AMD Radeon RX 7600 in average score, with deltas of -0.1%, -0.1%, 0.5%, and 0.7% respectively.

Q: What API support does each GPU offer?

A: The AMD Radeon 680M supports DirectX 12 Ultimate (12_2), 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 recorded.

Architecture Differences

The AMD Radeon 680M is built on the RDNA 2.0 architecture, using a chip codenamed Rembrandt+ on a 6 nm process node from TSMC. It contains 13,100 million transistors on a 208 mm² die, yielding a transistor density of 63.0M per mm². The NVIDIA Tesla M2090, by contrast, uses the Fermi 2.0 architecture with a GF110 chip on a 40 nm process, also from TSMC. It packs 3,000 million transistors on a much larger 520 mm² die, resulting in a transistor density of just 5.8M per mm². The 680M achieves more than ten times the transistor density of the Tesla M2090, a direct consequence of the process node advantage.

The 680M features 768 shading units, 48 TMUs, 32 ROPs, and 12 ray-tracing cores. It has no tensor cores recorded. The Tesla M2090 has 512 shading units, 64 TMUs, and 48 ROPs, with no ray-tracing or tensor cores. Despite having fewer TMUs and ROPs, the 680M's higher clocks (2000 MHz base, 2200 MHz boost) produce pixel and texture rates that are far superior: 70.40 GPixel/s and 105.6 GTexel/s for the 680M versus 20.83 GPixel/s and 41.66 GTexel/s for the Tesla M2090. The Tesla M2090's memory clocks are listed as 924 MHz, with 3.7 Gbps effective, but its core clocks are not recorded.

Memory architecture differs fundamentally. The 680M uses system-shared memory with a system-dependent bandwidth, meaning it borrows from the host CPU's RAM. The Tesla M2090 has 6 GB of dedicated GDDR5 on a 384-bit bus, delivering a fixed 177.4 GB/s. This makes the Tesla M2090 more predictable in memory-bound workloads, but the 680M benefits from not needing a separate memory pool. Power consumption reflects the architectural gap: the 680M has a 50 W TDP and requires no power connectors, while the Tesla M2090 has a 250 W TDP and needs one 6-pin and one 8-pin connector, with a suggested 600 W power supply.

The 680M is an integrated graphics processor (IGP) with a PCIe 4.0 x8 interface, while the Tesla M2090 is a dual-slot discrete card with PCIe 2.0 x16. The 680M's display outputs are portable-device dependent, whereas the Tesla M2090 has no display outputs at all, reflecting its compute-focused role. The Tesla M2090 measures 248 mm in length (9.8 inches). The 680M's dimensions are not recorded. The 680M supports DirectX 12 Ultimate (12_2) and Vulkan 1.4, while the Tesla M2090 tops out at DirectX 12 (11_0) with no Vulkan support. Both support OpenGL 4.6. The 680M was released in January 2023 and remains active, while the Tesla M2090 was released in July 2011 and is end-of-life. The 680M's predecessor is the Vega II IGP, and its successor is the Navi III IGP. The Tesla M2090's predecessor is listed simply as Tesla, and its successor is Tesla Kepler.