AMD Radeon 840M vs NVIDIA GeForce RTX 5090 SE Comparison

Head-to-Head Benchmarks

The recorded database contains no direct head-to-head benchmark entries for the AMD Radeon 840M and the NVIDIA GeForce RTX 5090 SE. Both products return an average benchmark score of 0, and the wins tally shows 0 for each side. This absence of measured results means the comparison must rely entirely on architectural specifications and computed throughput figures rather than empirical frame-rate or workload scores.

The raw compute numbers illustrate the scale of separation. The Radeon 840M delivers 1,484.8 GFLOPS of FP32 throughput, while the RTX 5090 SE delivers 66.94 TFLOPS, which is 66,940 GFLOPS. That places the NVIDIA part roughly 45 times higher in raw shader math. Pixel throughput tells a similar story: the integrated part outputs 23.20 GPixel/s, while the discrete card reaches 380.3 GPixel/s, a difference of approximately 16.4 times. Texture rate shows 46.40 GTexel/s for the AMD IGP versus 1,045.9 GTexel/s for the NVIDIA card, about 22.5 times higher.

Memory bandwidth is another decisive gap. The Radeon 840M uses system shared memory with bandwidth described as system dependent, which means it has no dedicated allocation and no fixed figure. The RTX 5090 SE carries 24 GB of GDDR7 on a 384-bit bus, delivering 1.34 TB/s of bandwidth. Without a fixed number for the integrated part, the comparison is qualitative, but the discrete card's bandwidth is an order of magnitude beyond what any shared-memory IGP can typically sustain.

Clock behavior differs as well. The Radeon 840M has a base clock of 400 MHz and a boost of 2900 MHz, a wide dynamic range typical of power-constrained integrated graphics. The RTX 5090 SE runs a base of 1740 MHz and a boost of 2377 MHz, a much narrower range, indicating it operates near its ceiling under load rather than scaling from a very low idle floor.

The RTX 5090 SE's memory clock is listed as 1750 MHz with 28 Gbps effective, which is a fixed specification. The Radeon 840M's memory clock is simply system shared, again underscoring that its performance depends entirely on the host platform's DRAM configuration.

Architecture Differences

The two products occupy opposite ends of the GPU spectrum, and the architectural details make that explicit. The AMD Radeon 840M is an integrated graphics processor built on the Krackan Point chip, using RDNA 3.5 architecture and belonging to the Navi III IGP generation for Strix Point Mobile. It is fabricated on a 4 nm process at TSMC. Transistor count and die size are listed as unknown, so no density figure can be stated. The NVIDIA GeForce RTX 5090 SE uses the GB202 chip with Blackwell 2.0 architecture in the GeForce 50-series. It is built on a 5 nm process, also at TSMC, with 92,200 million transistors on a 750 mm² die, giving a transistor density of 122.9M per mm².

Shading resources differ dramatically. The Radeon 840M has 256 shading units, 16 texture mapping units, 8 raster operation units, and 4 ray tracing cores. It has no tensor cores listed. The RTX 5090 SE has 14,080 shading units, 440 TMUs, 160 ROPs, 110 ray tracing cores, and 440 tensor cores. The ratio of shading units is 55 to 1 in favor of the discrete card.

The FP16 compute figures match the FP32 numbers for both products, each listed as 1:1 ratios. The Radeon 840M provides 1,484.8 GFLOPS in FP16, identical to its FP32 output. The RTX 5090 SE provides 66.94 TFLOPS in FP16, also identical to its FP32 output. Neither part shows a half-rate or double-rate FP16 mode in the data.

API support is identical across both: DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Both are listed as production status active. The Radeon 840M was released on February 28, 2025, while the RTX 5090 SE carries a release date of December 31, 2025. The AMD part lists its predecessor as Navi II IGP, while the NVIDIA part lists GeForce 40 as predecessor and GeForce 60 as successor.

The bus interface differs: PCIe 4.0 x8 for the Radeon 840M versus PCIe 5.0 x16 for the RTX 5090 SE. This affects how quickly the CPU can feed data to each GPU, though the integrated part's system shared memory means it relies on the same memory subsystem as the processor anyway.

Power and physical specifications reinforce the fundamental split. The Radeon 840M has a TDP of 15 W, is an IGP with no slot width, and uses no power connectors. The RTX 5090 SE has a TDP of 500 W, is dual-slot, uses a single 16-pin connector, and lists a suggested power supply of 900 W. The NVIDIA card measures 267 mm in length, 111 mm in height, and 40 mm in width. The AMD part has no listed dimensions because it is integrated into a mobile processor package.

Display outputs also differ. The Radeon 840M lists portable device dependent, meaning outputs are determined by the laptop or handheld manufacturer. The RTX 5090 SE lists 1x HDMI 2.1b and 3x DisplayPort 2.1b as fixed outputs. The NVIDIA card has a launch MSRP of 1,499 USD; the AMD IGP has no MSRP because it is not sold as a standalone product.

The Verdict

The data supports a clear separation of roles. The Radeon 840M is an integrated solution designed for low-power mobile devices, with 15 W TDP, no dedicated memory, and no standalone power delivery. The RTX 5090 SE is a high-end discrete card with 500 W TDP, 24 GB of dedicated GDDR7 memory, and a 900 W suggested power supply. These are not competing products in the same market segment.

The RTX 5090 SE holds every measurable advantage in raw throughput: 45 times the FP32 compute, 16.4 times the pixel rate, 22.5 times the texture rate, and 110 ray tracing cores versus 4. It also has tensor cores, which the AMD part lacks entirely. The discrete card's memory bandwidth of 1.34 TB/s is a fixed specification, while the integrated part's bandwidth is system dependent.

The Radeon 840M's advantages are limited to power consumption and physical integration. At 15 W versus 500 W, it consumes 33 times less power. It requires no power connector, no additional slot space, and no external PSU recommendation. Its 4 nm process is a smaller node than the 5 nm process used by the RTX 5090 SE, though the NVIDIA part packs far more transistors into a much larger die.

Benchmark results are not available for either product, so the verdict rests on architectural specifications. The RTX 5090 SE is the only choice for high-throughput rendering, ray tracing, or tensor workloads. The Radeon 840M is the only choice for a system where a discrete card cannot physically or electrically fit.

FAQ

Q: Which GPU has more shading units?

A: The NVIDIA GeForce RTX 5090 SE has 14,080 shading units, while the AMD Radeon 840M has 256.

Q: What is the memory configuration of each product?

A: The Radeon 840M uses system shared memory with system dependent bandwidth and no dedicated size, type, or bus width. The RTX 5090 SE has 24 GB of GDDR7 on a 384-bit bus with 1.34 TB/s bandwidth.

Q: How do the power requirements compare?

A: The Radeon 840M has a 15 W TDP, no power connectors, and no suggested PSU. The RTX 5090 SE has a 500 W TDP, one 16-pin connector, and a 900 W suggested power supply.

Q: Do both GPUs support the same APIs?

A: Yes, both list DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: What are the release dates?

A: The Radeon 840M was released on February 28, 2025. The RTX 5090 SE has a release date of December 31, 2025.

Q: Does the Radeon 840M have tensor cores?

A: No tensor cores are listed for the Radeon 840M. The RTX 5090 SE has 440 tensor cores.

Where Each One Wins

The RTX 5090 SE wins in every category of raw graphics throughput. Its FP32 output of 66.94 TFLOPS dwarfs the Radeon 840M's 1,484.8 GFLOPS. Its pixel rate of 380.3 GPixel/s versus 23.20 GPixel/s makes it suitable for high-resolution rendering with heavy overdraw. Its texture rate of 1,045.9 GTexel/s versus 46.40 GTexel/s supports detailed material sampling in complex scenes. The 110 ray tracing cores enable hardware-accelerated ray tracing workloads, while the 4 ray tracing cores on the AMD part limit it to minimal RT effects. The 440 tensor cores provide AI acceleration capabilities that the Radeon 840M cannot offer at all.

Memory capacity and bandwidth also favor the discrete card decisively. The RTX 5090 SE's 24 GB GDDR7 pool with 1.34 TB/s bandwidth supports large datasets, high-resolution textures, and multi-frame rendering. The Radeon 840M's system shared memory is entirely dependent on the host platform's DRAM speed and capacity, with no dedicated allocation.

The Radeon 840M wins in power efficiency and physical footprint. Its 15 W TDP allows operation in fanless or passively cooled designs, and its IGP form factor means it occupies no expansion slot. The RTX 5090 SE requires a dual-slot footprint, a 16-pin power connector, and a 900 W PSU. The AMD part's 4 nm process node is smaller than the NVIDIA part's 5 nm node, which contributes to its low power draw even if it cannot compensate for the massive difference in compute resources.

The RTX 5090 SE also wins on interface bandwidth with PCIe 5.0 x16 versus the Radeon 840M's PCIe 4.0 x8. The NVIDIA card's fixed display outputs of 1x HDMI 2.1b and 3x DisplayPort 2.1b provide predictable multi-monitor connectivity, while the AMD part's outputs are portable device dependent. The NVIDIA card has a launch MSRP of 1,499 USD, while the AMD IGP has no standalone pricing because it ships as part of a mobile processor.

For workloads requiring maximum shader throughput, ray tracing, tensor operations, or large dedicated memory pools, the RTX 5090 SE is the only viable option. For ultra-portable systems where power draw must stay at 15 W and no discrete card can be installed, the Radeon 840M is the only viable option. The database shows no overlap in their capability profiles.