AMD Radeon 840M vs NVIDIA Rubin GPU Comparison

Head-to-Head Benchmarks

The recorded data contains no direct head-to-head benchmark results for the AMD Radeon 840M against the NVIDIA Rubin GPU. The database shows zero wins for either part in a matched comparison, and the average benchmark score for both is recorded as zero. This absence of measured performance data is itself informative: the two devices occupy entirely different segments, so no common workload suite has been applied to both.

The AMD Radeon 840M is an integrated graphics processor within the Krackan Point chip, built for portable devices. Its compute ceiling is defined by 256 shading units, 16 texture mapping units, and 8 render output units, with a peak FP32 rate of 1,484.8 GFLOPS. The NVIDIA Rubin GPU, by contrast, is a server-class accelerator with 28,672 shading units, 896 texture mapping units, and an FP32 throughput of 130.0 TFLOPS. That places the Rubin part at roughly 87.5 times the raw FP32 throughput of the Radeon 840M, though the exact ratio is not stated in the database.

Texture and pixel throughput tell the same story. The Radeon 840M delivers 46.40 GTexel/s and 23.20 GPixel/s. The Rubin GPU delivers 2,031.2 GTexel/s and 54.41 GPixel/s. The texture rate gap is particularly large, reflecting the 896 TMUs on the Rubin part versus 16 on the Radeon. Pixel rate differences are narrower because the Rubin GPU has only 24 ROPs, which is unusual for a part of this scale but consistent with a compute-oriented server design.

Both parts sit at the 50th percentile against all GPUs in the database, but this percentile is derived from an empty benchmark set, so it should not be read as evidence of comparable real-world performance. The absence of head-to-head data means the analysis must rely on architectural and specification differences rather than measured results.

FAQ

Q: Which GPU has more shading units?

A: The NVIDIA Rubin GPU has 28,672 shading units, while the AMD Radeon 840M has 256.

Q: What is the FP32 performance difference?

A: The NVIDIA Rubin GPU reaches 130.0 TFLOPS, and the AMD Radeon 840M reaches 1,484.8 GFLOPS.

Q: How much memory does each GPU use?

A: The NVIDIA Rubin GPU has 288 GB of HBM4 memory on a 16384-bit bus, while the AMD Radeon 840M uses system shared memory with system dependent bandwidth.

Q: What process nodes are used?

A: The AMD Radeon 840M uses a 4 nm TSMC process, and the NVIDIA Rubin GPU uses a 3 nm TSMC process.

Q: Do both support DirectX?

A: The AMD Radeon 840M supports DirectX 12 Ultimate (12_2) and Vulkan 1.4, while the NVIDIA Rubin GPU reports N/A for DirectX, OpenGL, and Vulkan.

Q: What is the power requirement for each?

A: The AMD Radeon 840M has a 15 W TDP as an IGP, and the NVIDIA Rubin GPU has a 2300 W TDP with a suggested PSU of 2700 W.

Architecture Differences

The two GPUs share a foundry but diverge sharply in every architectural choice. The AMD Radeon 840M uses RDNA 3.5 architecture on a 4 nm TSMC node, while the NVIDIA Rubin GPU uses the Rubin architecture on a 3 nm TSMC node. The Rubin GPU is built on the GR100 chip with 336,000 million transistors on a 1456 mm² die, giving a transistor density of 230.8M per mm². The Radeon 840M's transistor count and die size are recorded as unknown, so no density comparison is possible.

Ray tracing support differs in form. The Radeon 840M includes 4 dedicated ray tracing cores, part of its RDNA 3.5 design. The Rubin GPU's ray tracing core count is not recorded in the database. Tensor hardware shows a similar split: the Radeon 840M has no tensor core field populated, while the Rubin GPU carries 896 tensor cores, indicating a strong focus on matrix workloads.

The Radeon 840M supports a full consumer API stack: DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The Rubin GPU reports N/A for DirectX, OpenGL, and Vulkan, which aligns with its server positioning where graphics APIs are not the primary interface. The Rubin GPU has no display outputs, while the Radeon 840M's display outputs are portable device dependent.

FP16 behavior also separates the architectures. The Radeon 840M runs FP16 at a 1:1 ratio with FP32, both at 1,484.8 GFLOPS. The Rubin GPU runs FP16 at a 2:1 ratio, delivering 260.0 TFLOPS compared to 130.0 TFLOPS FP32. This indicates the Rubin part is engineered for throughput-heavy compute where reduced precision is acceptable.

Specification Differences

The AMD Radeon 840M and NVIDIA Rubin GPU differ across nearly every recorded specification field.

Clock speeds: the Radeon 840M runs at a 400 MHz base and 2900 MHz boost. The Rubin GPU runs at a 700 MHz base and 2267 MHz boost. The Radeon part boosts higher, but the Rubin part starts from a higher base.

Memory: the Radeon 840M uses system shared memory with system dependent bandwidth. The Rubin GPU uses 288 GB of HBM4 with a 16384-bit bus and 22.1 TB/s bandwidth, running at 2695 MHz with 10.8 Gbps effective.

Compute units: the Radeon 840M has 256 shading units, 16 TMUs, 8 ROPs, and 4 ray tracing cores. The Rubin GPU has 28,672 shading units, 896 TMUs, 24 ROPs, and 896 tensor cores.

Rates: the Radeon 840M records 23.20 GPixel/s and 46.40 GTexel/s. The Rubin GPU records 54.41 GPixel/s and 2,031.2 GTexel/s.

Power and form factor: the Radeon 840M is an IGP with a 15 W TDP and no power connectors, using PCIe 4.0 x8. The Rubin GPU is an SXM Module with a 2300 W TDP, a suggested PSU of 2700 W, and PCIe 6.0 x16.

Release timing: the Radeon 840M launched on 2025-02-28, and the Rubin GPU launched on 2025-12-31.

Production status and predecessor: both are marked Active. The Radeon 840M succeeds the Navi II IGP, and the Rubin GPU succeeds Server Blackwell.

The Verdict

The data supports a clear segmentation verdict. The AMD Radeon 840M is an integrated solution for portable devices, shipping as an IGP within the Krackan Point chip. Its 15 W TDP, system shared memory, and portable device dependent display outputs confirm this role. The NVIDIA Rubin GPU is a server accelerator, built as an SXM Module with a 2300 W TDP, 288 GB of HBM4, and no display outputs.

For workloads that fit the Radeon 840M's profile, the relevant strengths are its compact integration, its 2900 MHz boost clock, and its consumer API support including DirectX 12 Ultimate and Vulkan 1.4. For workloads that require massive parallel throughput, the Rubin GPU's 28,672 shading units, 896 tensor cores, and 22.1 TB/s memory bandwidth place it in a different class entirely.

A buyer choosing between these parts would effectively be choosing between a portable integrated GPU and a server module. The Radeon 840M cannot approach the Rubin GPU's raw throughput, and the Rubin GPU cannot operate in the portable segment that the Radeon 840M targets. The absence of head-to-head benchmarks reflects this reality: there is no common use case where both would be measured side by side.

Where Each One Wins

The AMD Radeon 840M wins in scenarios defined by portability and integration. It operates within a 15 W TDP, requires no power connectors, and uses PCIe 4.0 x8. Its system shared memory design means it can draw on the host system's memory rather than carrying dedicated VRAM. The 2900 MHz boost clock is higher than the Rubin GPU's 2267 MHz boost, which matters for latency-sensitive tasks that do not scale across thousands of cores. The Radeon 840M also supports DirectX 12 Ultimate and Vulkan 1.4, making it a functional graphics solution for consumer software, whereas the Rubin GPU reports no graphics API support.

The NVIDIA Rubin GPU wins in scenarios defined by scale and memory capacity. Its 28,672 shading units and 896 tensor cores deliver 130.0 TFLOPS FP32 and 260.0 TFLOPS FP16. The 288 GB HBM4 pool with 22.1 TB/s bandwidth dwarfs any system shared memory arrangement. The 896 TMUs produce 2,031.2 GTexel/s, and the 16384-bit memory bus supports data movement at a scale the Radeon 840M cannot approach. The 2300 W TDP and 2700 W suggested PSU indicate a design intended for sustained, high-power compute rather than battery-constrained operation.

In practical terms, the Radeon 840M is suited to devices where power draw and physical footprint are the primary constraints. The Rubin GPU is suited to server installations where throughput and memory capacity justify extreme power requirements. The database records no workload where both parts compete directly, and the specification gap confirms they serve separate markets.