NVIDIA GeForce GTX 980 Ti vs NVIDIA Tesla M60 Comparison

Both the NVIDIA Tesla M60 and the NVIDIA GeForce GTX 980 Ti are Maxwell 2.0-generation products aimed at entirely different workloads, yet their benchmark data reveals a surprising overlap in raw compute potential. The Tesla M60 is a dual-GPU datacenter card designed for virtualization and cloud graphics, while the GTX 980 Ti is a single-GPU consumer flagship. The core question is whether the M60’s enterprise features justify its existence when the 980 Ti delivers significantly higher raw scores in the shared benchmarks.

Where Each One Wins

The GTX 980 Ti wins every single head-to-head benchmark comparison available in the data. In the Geekbench OpenCL test, the 980 Ti scores 43,513, which is 32.2% higher than the Tesla M60’s 29,506. The gap widens slightly in the Geekbench Vulkan test, where the 980 Ti achieves 47,724 versus the M60’s 31,473, a 34.1% advantage. This is a clean sweep for the consumer card in raw compute performance.

However, the Tesla M60’s strengths lie outside these synthetic compute tests. The M60 features 8 GB of GDDR5 memory, which is 2 GB more than the 980 Ti’s 6 GB. This additional capacity is critical for virtual desktop infrastructure (VDI) scenarios where multiple users share the GPU’s memory pool. The M60 also has no display outputs, indicating it is designed for server racks where rendering happens off-screen and frames are streamed to clients. The 980 Ti, by contrast, offers 1x DVI, 1x HDMI 2.0, and 3x DisplayPort 1.2 outputs, making it a direct-to-display solution.

The M60’s average benchmark score of 30,490 puts it in the 75th percentile of all GPUs, while the 980 Ti’s average of 28,020 places it in the 73rd percentile. This is counterintuitive given the 980 Ti’s dominance in the two shared tests. The M60’s higher percentile likely reflects its dual-GPU nature, which may boost its performance in other unlisted benchmarks. The M60 also has a lower transistor density of 13.1M per mm² versus the 980 Ti’s 13.3M per mm², despite having fewer total transistors (5,200 million vs 8,000 million).

Architecture Differences

Both GPUs are built on the Maxwell 2.0 architecture and use TSMC’s 28 nm process node. The similarities end there. The Tesla M60 uses the GM204 chip, which is smaller at 398 mm², while the GTX 980 Ti uses the larger GM200 chip at 601 mm². The M60’s die houses 5,200 million transistors, while the 980 Ti packs 8,000 million. This explains the raw compute disparity: the 980 Ti has 2,816 shading units, 176 texture mapping units, and 96 ROPs, compared to the M60’s 2,048 shaders, 128 TMUs, and 64 ROPs.

Clock speeds tell another story. The M60 runs at a 557 MHz base clock and 1178 MHz boost, while the 980 Ti starts at 1000 MHz base and boosts to 1076 MHz. Despite the M60’s higher boost clock relative to its base, the 980 Ti’s larger chip and higher base clock give it a significant throughput advantage. The M60’s memory runs at 1253 MHz (5 Gbps effective) on a 256-bit bus, yielding 160.4 GB/s bandwidth. The 980 Ti’s memory runs at 1753 MHz (7 Gbps effective) on a 384-bit bus, delivering 336.6 GB/s — more than double the bandwidth.

Power and cooling also diverge. The M60 is rated at 300 W TDP with a single 8-pin power connector and a suggested 700 W PSU. The 980 Ti consumes 250 W but requires both a 6-pin and an 8-pin connector, with a 600 W suggested PSU. Both are dual-slot cards and share the same 267 mm length. The M60 has no display outputs, reinforcing its server role, while the 980 Ti provides a full complement of modern display connections.

FAQ

Q: Why does the GTX 980 Ti score so much higher than the Tesla M60 in OpenCL and Vulkan?

A: The 980 Ti has 2,816 shading units versus 2,048 on the M60, along with a 384-bit memory bus versus 256-bit. This translates to 6.060 TFLOPS FP32 versus 4.825 TFLOPS, and 336.6 GB/s bandwidth versus 160.4 GB/s.

Q: Does the Tesla M60 have any performance advantage at all?

A: The M60’s average benchmark score of 30,490 is higher than the 980 Ti’s 28,020, and it sits in the 75th percentile versus the 73rd. This suggests the M60 may perform better in workloads not covered by the shared OpenCL and Vulkan tests.

Q: Can the Tesla M60 be used for gaming?

A: The M60 has no display outputs, making direct gaming impossible. It is designed for server-side rendering and virtualized environments where output is streamed remotely.

Q: Which card has more memory bandwidth?

A: The GTX 980 Ti has 336.6 GB/s, which is more than double the M60’s 160.4 GB/s. This is due to the 980 Ti’s wider 384-bit bus versus the M60’s 256-bit bus.

Q: Are these cards from the same generation?

A: Both use Maxwell 2.0 architecture and were released in 2015. The M60 launched on 2015-08-29, while the 980 Ti launched earlier on 2015-06-01.

Q: What is the transistor density difference?

A: The M60 has 13.1 million transistors per mm², while the 980 Ti has 13.3 million per mm². The 980 Ti’s larger die (601 mm²) allows for 8,000 million transistors versus the M60’s 5,200 million.

Specification Differences

The two cards differ in nearly every measurable specification except architecture, process node, and PCIe interface. The Tesla M60 uses the GM204 chip, while the GTX 980 Ti uses GM200. The M60 has 8 GB of GDDR5 on a 256-bit bus, while the 980 Ti has 6 GB on a 384-bit bus. Memory bandwidth is 160.4 GB/s versus 336.6 GB/s. The M60 has 2,048 shading units, 128 TMUs, and 64 ROPs; the 980 Ti has 2,816 shading units, 176 TMUs, and 96 ROPs.

Clock speeds differ substantially: the M60’s base is 557 MHz with an 1178 MHz boost, while the 980 Ti runs at 1000 MHz base and 1076 MHz boost. FP32 performance is 4.825 TFLOPS for the M60 versus 6.060 TFLOPS for the 980 Ti. Pixel rate is 75.39 GPixel/s versus 103.3 GPixel/s, and texture rate is 150.8 GTexel/s versus 189.4 GTexel/s. The M60 has a 300 W TDP with one 8-pin connector, while the 980 Ti has a 250 W TDP with a 6-pin and 8-pin connector. The M60 has no display outputs; the 980 Ti has 1x DVI, 1x HDMI 2.0, and 3x DisplayPort 1.2. The M60’s suggested PSU is 700 W versus 600 W for the 980 Ti.

Head-to-Head Benchmarks

The two shared benchmarks paint a consistent picture. In Geekbench OpenCL, the GTX 980 Ti scores 43,513 against the Tesla M60’s 29,506, a 32.2% advantage. This is expected given the 980 Ti’s 25% more shading units and 110% more memory bandwidth. The Vulkan test shows an even larger relative gap: 47,724 versus 31,473, a 34.1% difference. The M60’s lower clock speeds and narrower memory bus clearly handicap it in these compute-heavy workloads.

What is more interesting is the M60’s overall percentile ranking. Despite losing both head-to-head tests decisively, the M60 ranks in the 75th percentile of all GPUs, two points higher than the 980 Ti’s 73rd percentile. The M60’s average benchmark score of 30,490 also exceeds the 980 Ti’s 28,020. This suggests that the M60’s dual-GPU configuration (implied by its 8 GB memory and server positioning) may excel in parallelized workloads that are not captured by the OpenCL and Vulkan tests listed. The nearest rivals for the M60 include the NVIDIA CMP 70HX (0% delta) and AMD Radeon RX 6700 (0.2% delta), indicating it competes with modern mid-range cards. The 980 Ti’s nearest rivals are the AMD Radeon Pro W5500X (0.2% delta) and AMD FirePro S7150 (-0.3% delta), placing it in professional workstation territory.

The Verdict

The data is unambiguous for raw compute: the GTX 980 Ti outperforms the Tesla M60 by over 30% in both OpenCL and Vulkan. Anyone needing maximum FP32 throughput, higher pixel and texture rates, or broader memory bandwidth should choose the 980 Ti. Its 6.060 TFLOPS, 103.3 GPixel/s, and 336.6 GB/s are decisive advantages. The 980 Ti also offers display outputs, making it a flexible choice for direct rendering tasks.

The Tesla M60, however, is not without merit. Its 8 GB memory capacity is 33% larger than the 980 Ti’s, and its higher average benchmark score and percentile ranking suggest it handles certain server-side workloads better. The M60’s lack of display outputs and higher TDP (300 W vs 250 W) indicate it is optimized for datacenter deployment, where multiple users share resources and memory capacity matters more than raw speed. Its single 8-pin power connector is simpler than the 980 Ti’s 6-pin + 8-pin requirement.

For a consumer or workstation user, the GTX 980 Ti is the clear winner. For an enterprise deploying virtualized GPU instances, the Tesla M60’s larger memory pool and server-oriented design may be the better fit, despite its lower benchmark scores. The choice ultimately hinges on whether you need the 980 Ti’s brute force or the M60’s capacity and virtualization features.