AMD Radeon Instinct MI308X vs NVIDIA GeForce RTX 5080 SUPER Comparison

Head-to-Head Benchmarks

The recorded data contains no direct head-to-head benchmark comparisons between the AMD Radeon Instinct MI308X and the NVIDIA GeForce RTX 5080 SUPER. The database shows zero wins for either product in direct matchup tests, and the head-to-head benchmark list is empty. This absence of comparative testing data means the analysis must rely on the individual specifications and the one available benchmark score for the RTX 5080 SUPER.

The RTX 5080 SUPER has a single recorded benchmark result: a 3DMark Steel Nomad DX12 score of 3075 points. This places the card at the 19th percentile among all GPUs in the database. Its nearest recorded rivals provide context: the NVIDIA Quadro P1000 scores 3163 points, which is 2.8% higher than the RTX 5080 SUPER. The Intel Arc Pro B60 scores 3182 points, sitting 3.4% above. On the other side, the NVIDIA GeForce 820A scores 2983 points, which is 3.1% lower, and the NVIDIA GeForce GTX 860M scores 2967 points, 3.6% lower. These deltas indicate that the RTX 5080 SUPER sits in a tight cluster where the measured performance spread among these five products is only about 7 percentage points from top to bottom.

The AMD Radeon Instinct MI308X has no benchmark scores recorded in the database. Its percentile ranking is 50, which is the median position, but this ranking is not backed by any actual test results. The average benchmark score for the MI308X is listed as zero, and the nearest rivals field is empty. This means the database has no measured performance data for the AMD accelerator, making direct numerical comparison impossible.

The compute throughput figures, however, offer a clear separation. The MI308X delivers 81.72 TFLOPS of FP32 performance, while the RTX 5080 SUPER delivers 56.28 TFLOPS. This puts the AMD part approximately 45% higher in raw FP32 throughput. For FP16, the gap widens dramatically: the MI308X achieves 653.7 TFLOPS (8:1 ratio), while the RTX 5080 SUPER achieves 56.28 TFLOPS (1:1 ratio). The AMD accelerator offers over 11 times the FP16 throughput, though this comes with the caveat of the 8:1 ratio versus the 1:1 ratio on the NVIDIA part.

The pixel and texture rates also show divergence. The RTX 5080 SUPER has a pixel rate of 293.1 GPixel/s, while the MI308X is recorded at 0 MPixel/s, effectively no pixel output capability. The texture rate tells a different story: the MI308X achieves 2,553.6 GTexel/s versus 879.3 GTexel/s for the RTX 5080 SUPER, a nearly 3x advantage for the AMD part.

Where Each One Wins

The AMD Radeon Instinct MI308X wins in raw compute workloads. The 81.72 TFLOPS FP32 figure and 653.7 TFLOPS FP16 figure indicate a design optimized for heavy parallel computation, such as AI training and scientific simulation. The texture rate of 2,553.6 GTexel/s further supports this, as texture-heavy operations in compute shaders benefit from massive texture unit counts (1,216 TMUs). The memory subsystem reinforces this positioning: 192 GB of HBM3 across an 8192-bit bus delivers 10.3 TB/s of bandwidth, a figure that dwarfs typical consumer memory capacity and throughput. This configuration suits large model inference, high-performance computing, and data center workloads where memory capacity and bandwidth are primary constraints.

The NVIDIA GeForce RTX 5080 SUPER wins in graphics rendering and consumer-facing workloads. The 293.1 GPixel/s pixel rate, 84 ray tracing cores, and 336 tensor cores indicate dedicated hardware for rasterization, ray tracing, and AI-accelerated features. The card supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, making it compatible with modern game engines and graphics APIs. The 112 ROPs and 24 GB of GDDR7 memory provide sufficient resources for high-resolution gaming and content creation. The display outputs, 1x HDMI 2.1b and 3x DisplayPort 2.1b, confirm this is a product intended to drive monitors, whereas the MI308X has no display outputs at all.

The RTX 5080 SUPER also wins in efficiency per watt, based on the recorded specifications. The NVIDIA part consumes 415 W TDP, while the MI308X consumes 750 W. For the same 56.28 TFLOPS FP32 on the RTX 5080 SUPER, the power draw is significantly lower than the 81.72 TFLOPS on the MI308X, though the absolute performance per watt favors the AMD part in FP32 (0.109 TFLOPS/W versus 0.136 TFLOPS/W for AMD). The RTX 5080 SUPER, however, delivers its FP16 performance at the same rate as FP32, meaning no power penalty for half-precision workloads, whereas the MI308X likely uses a different execution path for its 8:1 FP16 ratio.

Architecture Differences

The AMD Radeon Instinct MI308X uses the CDNA 3.0 architecture, specifically designed for data center compute. The chip, codenamed Aqua Vanjaram, is built on a 5 nm TSMC process and contains 153,000 million transistors on a 1017 mm² die. This yields a transistor density of 150.4 million transistors per square millimeter. The architecture is a compute-first design, with no RT cores, no tensor cores, and no ROPs listed. The shading units number 19,456, with 1,216 TMUs. The memory interface is 8192 bits wide, using HBM3 memory totaling 192 GB. The MI308X is a member of the Radeon Instinct generation, released on December 5, 2023, and it succeeds the FirePro Data Center line. It uses a PCIe 5.0 x16 bus interface and draws power via an OAM Module slot, with no traditional power connectors. The suggested power supply is 1150 W.

The NVIDIA GeForce RTX 5080 SUPER uses the Blackwell 2.0 architecture, built on the GB203 chip. It is also fabricated on a 5 nm TSMC process, but with 45,600 million transistors on a 378 mm² die, resulting in a transistor density of 120.6 million per square millimeter. The Blackwell 2.0 architecture includes 84 ray tracing cores and 336 tensor cores, along with 10,752 shading units, 336 TMUs, and 112 ROPs. The memory is 24 GB of GDDR7 on a 256-bit bus, delivering 1.02 TB/s bandwidth. The card is part of the GeForce 50 series, released on December 31, 2025, and is currently marked as Active in production. It uses a PCIe 5.0 x16 interface, a dual-slot form factor, and a single 16-pin power connector. The dimensions are 304 mm length, 137 mm height, and 40 mm width. The launch MSRP is 999 USD.

The process node is identical at 5 nm and both use TSMC as the foundry, but the transistor counts and die sizes differ substantially. The MI308X has over three times the transistor count of the RTX 5080 SUPER and a die that is nearly 2.7 times larger. The CDNA 3.0 architecture prioritizes memory bandwidth and compute density, while Blackwell 2.0 balances rasterization, ray tracing, and AI acceleration. The MI308X has no display outputs, no DirectX support, no OpenGL support, and no Vulkan support, reflecting a compute-only design. The RTX 5080 SUPER supports all three major graphics APIs.

Specification Differences

The two products differ in nearly every measurable specification. The AMD MI308X has 19,456 shading units versus 10,752 on the RTX 5080 SUPER. The TMU count is 1,216 versus 336. The MI308X has 0 ROPs, while the RTX 5080 SUPER has 112. The RTX 5080 SUPER has 84 RT cores and 336 tensor cores, while the MI308X lists neither. The clock speeds differ: the MI308X has a base clock of 1000 MHz and a boost clock of 2100 MHz, while the RTX 5080 SUPER has a base clock of 2295 MHz and a boost clock of 2617 MHz. The memory clock is 2525 MHz with 10.1 Gbps effective on the MI308X, versus 2000 MHz with 32 Gbps effective on the RTX 5080 SUPER.

Memory capacity and type are major differentiators: 192 GB of HBM3 versus 24 GB of GDDR7. The bus width is 8192 bits versus 256 bits. Bandwidth is 10.3 TB/s versus 1.02 TB/s, a tenfold difference. The TDP is 750 W for the MI308X and 415 W for the RTX 5080 SUPER. The MI308X uses an OAM Module slot width with no power connectors, while the RTX 5080 SUPER uses a dual-slot design with a single 16-pin connector. The suggested PSU is 1150 W for the AMD part; the NVIDIA part has no suggested PSU listed. Display outputs: none for the MI308X, versus 1x HDMI 2.1b and 3x DisplayPort 2.1b for the RTX 5080 SUPER. The MI308X has no listed dimensions, while the RTX 5080 SUPER measures 304 mm by 137 mm by 40 mm. Release dates differ by roughly two years: December 5, 2023, for the MI308X and December 31, 2025, for the RTX 5080 SUPER.

The FP32 throughput is 81.72 TFLOPS for the MI308X versus 56.28 TFLOPS for the RTX 5080 SUPER. FP16 is 653.7 TFLOPS (8:1) versus 56.28 TFLOPS (1:1). Pixel rate is 0 MPixel/s versus 293.1 GPixel/s. Texture rate is 2,553.6 GTexel/s versus 879.3 GTexel/s. The MI308X has no launch MSRP recorded; the RTX 5080 SUPER has a launch MSRP of 999 USD.

FAQ

Q: Which product has higher FP32 compute performance?

A: The AMD Radeon Instinct MI308X delivers 81.72 TFLOPS of FP32 performance, compared to 56.28 TFLOPS for the NVIDIA GeForce RTX 5080 SUPER.

Q: What is the memory capacity difference?

A: The MI308X has 192 GB of HBM3 memory, while the RTX 5080 SUPER has 24 GB of GDDR7 memory. The MI308X also has a much wider 8192-bit bus, yielding 10.3 TB/s bandwidth versus 1.02 TB/s.

Q: Does the RTX 5080 SUPER support ray tracing?

A: Yes, the RTX 5080 SUPER has 84 ray tracing cores and 336 tensor cores. The MI308X does not list any RT cores or tensor cores, indicating no dedicated ray tracing hardware.

Q: What graphics APIs does each product support?

A: The RTX 5080 SUPER supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The MI308X has no API support listed for DirectX, OpenGL, or Vulkan.

Q: What is the power consumption of each product?

A: The MI308X has a TDP of 750 W and a suggested PSU of 1150 W. The RTX 5080 SUPER has a TDP of 415 W and no suggested PSU listed.

Q: What is the process node for both chips?

A: Both the MI308X and the RTX 5080 SUPER are fabricated on a 5 nm process at TSMC. The MI308X has 153,000 million transistors on a 1017 mm² die, while the RTX 5080 SUPER has 45,600 million transistors on a 378 mm² die.