NVIDIA Tesla M60

The NVIDIA Tesla M60 is a dual-slot, end-of-life accelerator built on the Maxwell 2.0 architecture, designed for compute and virtualization workloads rather than direct display output. Its average benchmark score of 29,969 places it at the 73rd percentile among all GPUs, indicating a solid mid-to-upper tier standing despite its 2015 release date. The data shows a device that holds its own in raw compute tasks, though its architectural age is reflected in specific workload characteristics.

Benchmark Performance

The Tesla M60’s benchmark results are remarkably consistent across the two recorded tests. In Geekbench OpenCL, it scores 29,377, while its Vulkan score is slightly higher at 30,560. This narrow gap of roughly 4% suggests the card’s performance is not heavily dependent on the API layer, which is typical for a compute-oriented accelerator with no display outputs. The average score of 29,969 serves as the reference point for all comparative analysis.

Against its nearest rivals, the M60 sits in a tightly contested performance band. The closest competitor is the NVIDIA GeForce RTX 5070 Mobile, which averages 29,928 — a mere 0.1% lower than the Tesla M60. This effectively places the two cards at parity in synthetic benchmarks. The AMD Radeon RX 7700 XT posts a slightly higher average of 30,097, translating to a 0.4% lead over the M60. The NVIDIA GeForce RTX 5060 follows at 30,109, a 0.5% advantage. Finally, the AMD Radeon RX 570 averages 30,158, which is 0.6% ahead. These deltas are all within a single percentage point, indicating that the M60 delivers performance statistically indistinguishable from these modern rivals in compute-oriented workloads.

The narrow spread of scores — from 29,377 to 30,560 across all tests and rivals — suggests that the M60’s raw compute throughput remains competitive. However, the 73rd percentile ranking clarifies that a majority of all GPUs still outperform it. The M60 achieves 4.825 TFLOPS of FP32 compute, a figure that explains its ability to keep pace with newer cards in OpenCL and Vulkan tests, which often scale with raw shader throughput. The 2,048 shading units operating at a boost clock of 1178 MHz provide the foundation for this performance. Pixel rate stands at 75.39 GPixel/s, and texture rate at 150.8 GTexel/s, both figures that align with its mid-pack standing.

Power and Cooling

The Tesla M60 carries a thermal design power (TDP) of 300 W, a substantial draw that reflects its dual-GPU board design (the GM204 chip appears twice, though the specification lists a single chip). This power requirement dictates a suggested power supply unit rating of 700 W, ensuring sufficient headroom for the card’s peak demands and the rest of the system. The card’s power delivery relies on a single 8-pin power connector, a straightforward requirement given its enterprise positioning.

Cooling is handled by a dual-slot cooler, which is standard for a 300 W component in a server or workstation chassis. The physical dimensions are 267 mm in length, or 10.5 inches, which is compatible with most full-size server racks and workstation cases. The card has no display outputs, meaning it is entirely dependent on host system management for configuration and monitoring. The 28 nm process node from TSMC, using 5,200 million transistors on a 398 mm² die, contributes to the power envelope; the transistor density of 13.1M per mm² is modest by modern standards, explaining why the 300 W TDP is required to sustain the 1178 MHz boost clock.

Memory Subsystem

The Tesla M60 is equipped with 8 GB of GDDR5 memory, configured across a 256-bit bus. This yields a total memory bandwidth of 160.4 GB/s, with the memory clocked at 1253 MHz, translating to 5 Gbps effective. The 256-bit bus width is a critical specification here, as it directly determines the bandwidth ceiling. For high-resolution workloads, this bandwidth figure is a limiting factor compared to modern cards with wider buses or faster memory technologies.

At 4K resolution, the 160.4 GB/s bandwidth means the M60 can handle textures and frame buffers, but it will likely bottleneck in scenarios requiring frequent large data transfers, such as high-detail 3D rendering or large compute datasets. The 8 GB capacity is adequate for many workloads, but the effective bandwidth per GB is lower than newer parts. The pixel rate of 75.39 GPixel/s and texture rate of 150.8 GTexel/s provide context: the memory system can feed the render pipeline at moderate resolutions, but the gap between the M60’s bandwidth and its compute throughput (4.825 TFLOPS) suggests that memory-bound tasks will see diminishing returns. For compute tasks that fit within the 8 GB frame buffer, the memory subsystem is sufficient; for streaming large datasets, it becomes a constraint.

How It Compares

NVIDIA GeForce RTX 5070 Mobile: The M60 matches this laptop-class GPU within 0.1%, a statistically negligible difference. This is notable because the RTX 5070 Mobile is a much newer architecture, yet the M60’s higher TDP and dual-chip design allow it to achieve parity in compute benchmarks. The M60’s desktop-oriented power delivery compensates for its architectural age.

AMD Radeon RX 7700 XT: The RX 7700 XT leads by 0.4%, with an average score of 30,097 versus 29,969. This is a minimal margin that falls within typical run-to-run variance. The M60’s performance is effectively equivalent to this modern desktop card, despite the RX 7700 XT’s newer memory technology and higher bandwidth. The M60’s 4.825 TFLOPS of FP32 throughput is the primary driver of this close result.

NVIDIA GeForce RTX 5060: The RTX 5060 is 0.5% faster, posting 30,109. Again, this is a negligible difference. The M60’s 2048 shading units at 1178 MHz boost clock produce raw compute that rivals the RTX 5060’s architecture-specific optimizations. The M60’s 160.4 GB/s bandwidth, however, is significantly lower than what the RTX 5060 would offer, meaning the M60’s competitive scores may not translate to all real-world workloads.

AMD Radeon RX 570: The RX 570 is the fastest rival in this group at 30,158, a 0.6% lead over the M60. This is the largest delta among the nearest rivals, yet still under one percentage point. The RX 570 is an older card itself, but its 256-bit memory bus and GDDR5 memory configuration resemble the M60’s setup. The M60’s higher compute throughput (4.825 TFLOPS) is offset by the RX 570’s slightly more efficient memory utilization in these specific benchmarks.

Who Should Consider It

The Tesla M60 is a compute accelerator, not a gaming card — it has no display outputs. For users with existing server infrastructure, the data shows it performs on par with mid-range modern GPUs in OpenCL and Vulkan compute tasks. The 73rd percentile ranking and average score of 29,969 indicate that it is well-suited for workloads like GPU virtualization, video transcoding, or general-purpose compute that can leverage its 8 GB frame buffer and 160.4 GB/s bandwidth.

At 1080p and 1440p resolutions, the M60’s memory bandwidth is sufficient for many compute tasks, and its 4.825 TFLOPS of FP32 throughput handles moderate parallel workloads. At 4K, the 160.4 GB/s bandwidth becomes a limiting factor; users should expect performance to scale poorly with resolution increases, as the memory subsystem cannot keep pace with the compute units. The card’s 300 W TDP and 700 W PSU requirement make it suitable for workstations with robust power delivery, but not for compact or power-constrained environments. Given its end-of-life status, the M60 is best considered for legacy systems or specific compute deployments where its dual-slot form factor and single 8-pin connector are already accommodated.

FAQ

Q: What is the Tesla M60’s performance percentile among all GPUs?

A: The Tesla M60 sits at the 73rd percentile, meaning it outperforms roughly 73% of all GPUs in the database based on its average benchmark score of 29,969.

Q: How does the Tesla M60 compare to the NVIDIA GeForce RTX 5070 Mobile?

A: The M60 is 0.1% faster than the RTX 5070 Mobile, with an average score of 29,969 versus 29,928. This is effectively a tie in synthetic compute benchmarks.

Q: What is the memory configuration of the Tesla M60?

A: It has 8 GB of GDDR5 memory on a 256-bit bus, providing 160.4 GB/s of bandwidth. The memory runs at 1253 MHz, or 5 Gbps effective.

Q: What power supply is recommended for the Tesla M60?

A: The suggested PSU rating is 700 W. The card itself has a 300 W TDP and requires a single 8-pin power connector.

Q: Does the Tesla M60 support modern graphics APIs?

A: Yes, it supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4, making it compatible with current software standards despite its age.

Q: What is the FP32 compute performance of the Tesla M60?

A: The card delivers 4.825 TFLOPS of FP32 compute, driven by 2,048 shading units operating at a boost clock of 1178 MHz.