AMD Radeon Instinct MI60 vs NVIDIA GeForce RTX 5080 Comparison

Head-to-Head Benchmarks

The benchmark database pits the AMD Radeon Instinct MI60 against the NVIDIA GeForce RTX 5080 in two recorded tests, and the results are decisive. In Geekbench OpenCL, the RTX 5080 scores 235,901 points, which is 60.8% higher than the MI60’s 92,488 points. The margin is even more pronounced in Geekbench Vulkan, where the RTX 5080 reaches 255,450 points, a 63.8% lead over the MI60’s 92,444 points. These deltas are the largest recorded between the two cards in the database, and they clearly favor the newer NVIDIA part.

The MI60, despite being an end-of-life product, still holds its own in the broader context of the database. Its average benchmark score sits at 92,466, placing it in the 93rd percentile among all GPUs. That is a strong position for a card released in 2018. Its nearest rivals include the NVIDIA RTX A4500 (average score 91,671, 0.9% behind), the RTX A4500 Mobile (91,134, 1.5% behind), the AMD Radeon Pro VII (97,131, 4.8% ahead), and the AMD Radeon RX 7900M (97,487, 5.2% ahead). So the MI60 is competitive with a wide range of modern workstation and mobile parts, even if it trails the RTX 5080 by a massive margin.

The RTX 5080’s average benchmark score is 56,083, but that figure is skewed by a diverse set of tests. Its percentile rank is 87th, which is lower than the MI60’s 93rd, reflecting the fact that the RTX 5080’s scores are spread across many different workloads, including DirectX 9, 10, 11, and 12, as well as 2D and compute tests. Its nearest rivals are the AMD Radeon 8060S (55,757, 0.6% behind), the AMD Radeon RX 6750 GRE 12 GB (55,698, 0.7% behind), the AMD Radeon Pro W5700X (54,828, 2.3% behind), and the AMD Radeon RX 9070 GRE (57,367, 2.2% ahead). The RTX 5080’s average is only slightly above its closest competitors, which suggests that its performance is highly workload-dependent.

When comparing the two directly, the RTX 5080 wins both head-to-head tests. The MI60 has zero wins, while the RTX 5080 has two. The largest single margin is the Vulkan test, where the RTX 5080’s 255,450 score is 63.8% higher than the MI60’s 92,444. This is a clear indication that the RTX 5080’s architecture is far more efficient in modern graphics APIs, while the MI60’s GCN 5.1 design, which predates the widespread adoption of Vulkan, struggles to keep pace.

Where Each One Wins

The data points to a clear split: the RTX 5080 wins in every recorded benchmark category, but the nature of those wins matters. In Geekbench OpenCL, the RTX 5080’s 235,901 score versus the MI60’s 92,488 shows a 60.8% advantage. This is a compute-heavy test, and the RTX 5080’s 56.28 TFLOPS FP32 performance versus the MI60’s 14.75 TFLOPS explains the gap. For workloads that rely on raw FP32 throughput, such as scientific simulation or certain rendering tasks, the RTX 5080 is the clear choice.

In Vulkan, the RTX 5080’s 255,450 score versus 92,444 is a 63.8% lead. Vulkan is a low-level API that benefits from modern hardware features like dedicated ray tracing cores and tensor cores, both of which the RTX 5080 has (84 RT cores and 336 tensor cores), while the MI60 has none. This makes the RTX 5080 far more suitable for real-time graphics, gaming, and any workload that leverages Vulkan’s advanced features.

The MI60, however, still has a role. Its 32 GB of HBM2 memory with a 4096-bit bus and 1.02 TB/s bandwidth is a significant asset. The RTX 5080, with 16 GB of GDDR7 on a 256-bit bus, offers 960.0 GB/s of bandwidth. The MI60’s memory capacity is double that of the RTX 5080, which can be critical for large datasets that must reside entirely in GPU memory, such as certain machine learning models or large-scale data processing. The RTX 5080’s higher compute throughput might not compensate if the workload cannot fit in its smaller memory pool.

For legacy API compatibility, the MI60 supports DirectX 12 (12_1) and Vulkan 1.3, while the RTX 5080 supports DirectX 12 Ultimate (12_2) and Vulkan 1.4. The MI60’s OpenGL 4.6 support matches the RTX 5080’s. If a user is running older software that relies on DirectX 12_1 features, the MI60 might still function, but the RTX 5080’s newer API support is more future-proof.

Architecture Differences

The architectural divide is stark. The MI60 uses AMD’s GCN 5.1 architecture, built on Vega 20, manufactured on a 7 nm TSMC process. The RTX 5080 uses NVIDIA’s Blackwell 2.0 architecture, built on GB203, manufactured on a 5 nm TSMC process. The process difference is significant: the RTX 5080 packs 45,600 million transistors into a 378 mm² die, yielding a transistor density of 120.6 million per square millimeter. The MI60 has 13,230 million transistors on a 331 mm² die, with a density of 40.0 million per square millimeter. That is a threefold density advantage for the RTX 5080, which directly translates to its higher clock speeds and compute throughput.

Clock speeds tell a similar story. The MI60 runs at a base of 1200 MHz and boosts to 1800 MHz. The RTX 5080 operates at a base of 2295 MHz and boosts to 2617 MHz. This higher clock rate, combined with the larger shader count (10,752 versus 4,096), explains why the RTX 5080’s FP32 throughput is 56.28 TFLOPS compared to the MI60’s 14.75 TFLOPS. The RTX 5080 also achieves 56.28 TFLOPS FP16, while the MI60 reaches 29.49 TFLOPS FP16 (at a 2:1 ratio). The RTX 5080’s FP16 performance is identical to its FP32 performance, indicating a 1:1 ratio, which is a hallmark of modern NVIDIA architectures that can handle mixed-precision workloads efficiently.

The memory subsystems are fundamentally different. The MI60 uses HBM2 with a 4096-bit bus, which is an extremely wide interface that delivers 1.02 TB/s of bandwidth. The RTX 5080 uses GDDR7 on a 256-bit bus, achieving 960.0 GB/s. While the RTX 5080’s bandwidth is slightly lower, its memory clock is 1875 MHz (30 Gbps effective), compared to the MI60’s 1000 MHz (2 Gbps effective). The MI60’s advantage is capacity: 32 GB versus 16 GB. For workloads that need to hold massive datasets, the MI60’s memory is a distinct advantage, but for speed, the RTX 5080’s GDDR7 technology is more efficient per bit.

The RTX 5080 also includes dedicated hardware that the MI60 lacks entirely: 84 RT cores and 336 tensor cores. These enable hardware-accelerated ray tracing and AI-based tensor operations, which are critical for modern games, real-time ray tracing, and machine learning inference. The MI60 has no such dedicated units, relying entirely on its GCN shader array. This is the single most important feature gap in the database.

FAQ

Q: Which GPU has higher FP32 compute performance?

A: The NVIDIA GeForce RTX 5080, with 56.28 TFLOPS FP32, is significantly ahead of the AMD Radeon Instinct MI60’s 14.75 TFLOPS FP32.

Q: What is the memory capacity difference?

A: The AMD Radeon Instinct MI60 has 32 GB of HBM2 memory, while the NVIDIA GeForce RTX 5080 has 16 GB of GDDR7 memory. The MI60 offers twice the capacity.

Q: Does the RTX 5080 support hardware ray tracing?

A: Yes, the RTX 5080 includes 84 RT cores and 336 tensor cores, which enable hardware-accelerated ray tracing and AI workloads. The MI60 has no such dedicated cores.

Q: Which card has a higher bandwidth?

A: The MI60’s 1.02 TB/s bandwidth is slightly higher than the RTX 5080’s 960.0 GB/s, but the RTX 5080 uses a more modern GDDR7 memory type.

Q: What is the release date difference?

A: The AMD Radeon Instinct MI60 was released on 2018-11-17, while the NVIDIA GeForce RTX 5080 was released on 2025-01-29. The RTX 5080 is over six years newer.

Q: Which GPU has a higher average benchmark score?

A: The MI60 has a higher average benchmark score of 92,466, while the RTX 5080 averages 56,083, but this is due to the RTX 5080’s broader test suite, which includes older DirectX versions and 2D tests.

Specification Differences

The two cards differ in nearly every measurable specification. The MI60 uses AMD’s GCN 5.1 architecture on a 7 nm process, while the RTX 5080 uses NVIDIA’s Blackwell 2.0 on a 5 nm process. The transistor count is 13,230 million versus 45,600 million, and die size is 331 mm² versus 378 mm², resulting in a transistor density of 40.0M/mm² versus 120.6M/mm².

Clock speeds: the MI60 has a base of 1200 MHz and a boost of 1800 MHz, while the RTX 5080 has a base of 2295 MHz and a boost of 2617 MHz. The memory clock is 1000 MHz (2 Gbps effective) for the MI60 versus 1875 MHz (30 Gbps effective) for the RTX 5080.

Memory: the MI60 has 32 GB of HBM2 on a 4096-bit bus with 1.02 TB/s bandwidth, while the RTX 5080 has 16 GB of GDDR7 on a 256-bit bus with 960.0 GB/s bandwidth. The MI60 has 4,096 shading units, 256 TMUs, and 64 ROPs, while the RTX 5080 has 10,752 shading units, 336 TMUs, and 112 ROPs. The RTX 5080 also has 84 RT cores and 336 tensor cores, which the MI60 lacks.

Pixel and texture rates: the MI60 achieves 115.2 GPixel/s and 460.8 GTexel/s, while the RTX 5080 reaches 293.1 GPixel/s and 879.3 GTexel/s. The TDP is 300 W for the MI60 and 360 W for the RTX 5080. The power connectors differ: 1x 6-pin + 1x 8-pin for the MI60, versus 1x 16-pin for the RTX 5080. The suggested PSU is 700 W for the MI60 and 750 W for the RTX 5080.

Bus interface: the MI60 uses PCIe 4.0 x16, while the RTX 5080 uses PCIe 5.0 x16. Display outputs: the MI60 has 1x mini-DisplayPort 1.4a, while the RTX 5080 has 1x HDMI 2.1b and 3x DisplayPort 2.1b. Dimensions: the MI60 is 267 mm long and 111 mm tall, while the RTX 5080 is 304 mm long, 137 mm tall, and 40 mm wide. DirectX support: the MI60 supports 12 (12_1), while the RTX 5080 supports 12 Ultimate (12_2). Vulkan support: 1.3 versus 1.4. The production status is end-of-life for the MI60 versus active for the RTX 5080.

The Verdict

The data is unambiguous: the NVIDIA GeForce RTX 5080 is the superior performer in every recorded benchmark, with a 60.8% lead in OpenCL and a 63.8% lead in Vulkan over the AMD Radeon Instinct MI60. Its newer architecture, higher clock speeds, and dedicated RT and tensor cores make it the clear choice for any workload that prioritizes compute throughput, modern API support, or real-time graphics.

The MI60, however, should not be dismissed. Its 32 GB of HBM2 memory with 1.02 TB/s bandwidth is a unique asset that the RTX 5080 cannot match in capacity. For users who need to process datasets larger than 16 GB in GPU memory, the MI60 remains a viable option, even if it is slower per core. Its 93rd percentile ranking among all GPUs shows that it is still a capable card for its age.

The RTX 5080’s lower average benchmark score (56,083) compared to the MI60’s 92,466 is misleading, as it reflects the RTX 5080’s inclusion of legacy DirectX 9, 10, and 11 tests, where modern GPUs often score lower due to driver overhead or architectural changes. In the two tests that directly compare them, the RTX 5080 wins decisively.

In the end, the choice depends on the use case. If the workload demands maximum FP32 or FP16 performance, modern API features, or hardware ray tracing, the RTX 5080 is the only rational pick. If the workload requires more than 16 GB of memory and can tolerate lower compute throughput, the MI60’s larger memory pool may be the deciding factor. The database records no scenario where the MI60 wins on performance, so for most users, the RTX 5080 is the recommended choice.