AMD Steam Deck OLED GPU vs Intel Arc Pro B390 Comparison

The Verdict

The AMD Steam Deck OLED GPU and Intel Arc Pro B390 serve fundamentally different roles, and the recorded data makes that split clear. The AMD part is a 15 W console-class chip built for a fixed handheld device, while the Intel part is an 80 W integrated graphics processor aimed at mobile workstations and portable devices. For anyone needing raw throughput, the Intel Arc Pro B390 is the obvious choice: it delivers 7.680 TFLOPS FP32 against the AMD's 1.638 TFLOPS, a 4.7x gap in raw compute. For anyone constrained by power or thermal limits, the AMD Steam Deck OLED GPU offers a far more efficient package at 15 W, and it includes a dedicated 16 GB LPDDR5 memory pool, which the Intel part lacks entirely.

The data shows two different design philosophies. AMD built a self-contained graphics solution with its own memory and a conservative 1000 MHz base clock, boosting to 1600 MHz. Intel built an integrated GPU that borrows system memory, runs at a 300 MHz base clock, and boosts aggressively to 2500 MHz. Neither part has benchmark scores in the database, so the verdict rests on specifications and architectural comparison. The Intel Arc Pro B390 wins decisively on compute, fill rates, and API support. The AMD Steam Deck OLED GPU wins on power efficiency, dedicated memory, and a smaller physical footprint.

Architecture Differences

The AMD Steam Deck OLED GPU uses the RDNA 2.0 architecture on a 6 nm TSMC process. The chip is named Sephiroth and belongs to the Console GPU (Valve) generation. It packs 2,400 million transistors into a 131 mm² die, yielding a transistor density of 18.3 million per square millimeter. The Intel Arc Pro B390 uses the Xe3-LPG architecture on a 3 nm Intel process. Its chip is named Panther Lake and belongs to the Arc Graphics-WM (Panther Lake) generation. Transistor count and die size are not recorded in the database for the Intel part, so no density comparison is possible.

The AMD GPU has 512 shading units, 32 texture mapping units, 16 raster output units, and 8 ray tracing cores. The Intel GPU has 1536 shading units, 48 TMUs, 24 ROPs, and 12 ray tracing cores. That means the Intel part has three times the shading units, 1.5 times the TMUs, 1.5 times the ROPs, and 1.5 times the ray tracing cores. The Intel part also supports Vulkan 1.4, while the AMD part supports Vulkan 1.3. Both support DirectX 12 Ultimate (12_2) and OpenGL 4.6.

Memory architecture differs sharply. The AMD Steam Deck OLED GPU uses 16 GB of LPDDR5 on a 128 bit bus with 176.0 GB/s of bandwidth and a memory clock of 1375 MHz (11 Gbps effective). The Intel Arc Pro B390 uses system shared memory, with system dependent bandwidth and no dedicated memory clock. This is a critical architectural difference: the AMD part has guaranteed memory bandwidth, while the Intel part depends entirely on the host system's memory configuration.

The AMD GPU has a base clock of 1000 MHz and a boost clock of 1600 MHz. The Intel GPU has a base clock of 300 MHz and a boost clock of 2500 MHz. The Intel part relies on a much wider clock range, which suggests aggressive boost behavior under load. The AMD part runs at a more modest but stable clock envelope, appropriate for its 15 W power budget. The Intel part's 80 W TDP is over five times higher, which explains the substantial performance advantage.

Head-to-Head Benchmarks

The database contains no direct head-to-head benchmark runs between these two GPUs, and neither part has an average benchmark score recorded. The comparison must therefore come from specification-derived metrics. The Intel Arc Pro B390 delivers 7.680 TFLOPS FP32 against the AMD's 1.638 TFLOPS, which is 4.69 times higher. In FP16, the Intel part delivers 15.36 TFLOPS against the AMD's 3.277 TFLOPS, again 4.69 times higher. Both use a 2:1 FP16 to FP32 ratio.

Pixel throughput follows the same pattern. The Intel part achieves 60.00 GPixel/s, while the AMD part achieves 25.60 GPixel/s. That is a 2.34x advantage for Intel. Texture throughput shows a wider gap: 120.0 GTexel/s versus 51.20 GTexel/s, a 2.34x advantage. These fill rate numbers reflect the Intel part's higher ROP and TMU counts combined with its 2500 MHz boost clock.

Clock speed alone does not tell the whole story. The AMD part's 1600 MHz boost is lower than the Intel part's 2500 MHz boost, but the AMD part has a much higher base clock at 1000 MHz versus 300 MHz. The Intel part's low base clock means it relies heavily on boost behavior, which in an integrated GPU depends on the host device's thermal and power headroom. The AMD part's narrower clock range suggests more predictable performance in a fixed handheld design.

The only area where the AMD part shows a clear structural advantage is memory. With 16 GB of dedicated LPDDR5 and 176.0 GB/s of bandwidth, the AMD GPU does not compete with the CPU for memory access. The Intel part uses system shared memory, so its effective bandwidth is system dependent and could be substantially lower or higher depending on the host platform. The database records no specific bandwidth figure for the Intel part.

FAQ

Q: Which GPU has higher raw compute performance?

A: The Intel Arc Pro B390 has 7.680 TFLOPS FP32 and 15.36 TFLOPS FP16, compared to the AMD Steam Deck OLED GPU's 1.638 TFLOPS FP32 and 3.277 TFLOPS FP16. The Intel part is roughly 4.7 times faster in both metrics.

Q: Which GPU uses less power?

A: The AMD Steam Deck OLED GPU has a 15 W TDP, while the Intel Arc Pro B390 has an 80 W TDP. The AMD part consumes one fifth of the power of the Intel part.

Q: Do both GPUs support ray tracing?

A: Yes. The AMD Steam Deck OLED GPU has 8 ray tracing cores, and the Intel Arc Pro B390 has 12 ray tracing cores.

Q: Which GPU has dedicated video memory?

A: The AMD Steam Deck OLED GPU has 16 GB of LPDDR5 on a 128 bit bus with 176.0 GB/s bandwidth. The Intel Arc Pro B390 uses system shared memory with system dependent bandwidth.

Q: Which GPU supports a newer Vulkan version?

A: The Intel Arc Pro B390 supports Vulkan 1.4, while the AMD Steam Deck OLED GPU supports Vulkan 1.3. Both support DirectX 12 Ultimate (12_2) and OpenGL 4.6.

Q: What process nodes do the two GPUs use?

A: The AMD Steam Deck OLED GPU uses a 6 nm process from TSMC. The Intel Arc Pro B390 uses a 3 nm process from Intel.

Where Each One Wins

The AMD Steam Deck OLED GPU wins in power efficiency. Its 15 W TDP makes it suitable for a handheld console with a fixed thermal envelope. The 1000 MHz base clock provides a solid floor, and the 1600 MHz boost is modest enough to sustain in a compact chassis. The dedicated 16 GB LPDDR5 pool with 176.0 GB/s bandwidth means the GPU does not depend on system memory configuration, which is an advantage in a purpose-built device. The 131 mm² die on a 6 nm process from TSMC with 2,400 million transistors shows a design focused on density and efficiency. The 1x USB Type-C display output fits a portable device profile.

The Intel Arc Pro B390 wins in raw performance across every compute and fill rate metric. Its 1536 shading units, 48 TMUs, and 24 ROPs provide a strong foundation for graphics workloads. The 7.680 TFLOPS FP32 rating is in a different performance class entirely. The 2500 MHz boost clock, while dependent on thermal headroom, offers a high ceiling for burst workloads. The 60.00 GPixel/s pixel rate and 120.0 GTexel/s texture rate give it a clear advantage in resolution-heavy and texture-heavy scenarios. The 12 ray tracing cores exceed the AMD part's 8. Vulkan 1.4 support adds a compatibility edge for newer applications. The 3 nm Intel process node indicates a more advanced manufacturing technology, and the IGP form factor with no power connectors and no dedicated slot width requirement means it can fit into systems without discrete GPU space.

The use-case split is straightforward. The AMD Steam Deck OLED GPU suits fixed handheld gaming devices where power draw and dedicated memory matter more than peak throughput. The Intel Arc Pro B390 suits mobile workstations or portable devices where compute density, fill rates, and API support take priority, and where the host system can supply adequate memory bandwidth and 80 W of power. There is no overlap in the data suggesting either part could substitute for the other.

Specification Differences

The two GPUs differ in nearly every recorded specification field. The AMD part uses the Sephiroth chip on RDNA 2.0 architecture, while the Intel part uses the Panther Lake chip on Xe3-LPG architecture. The AMD part is built on a 6 nm TSMC process; the Intel part uses a 3 nm Intel process. The AMD part has 2,400 million transistors and a 131 mm² die; the Intel part has unknown transistor count and die size.

Clock behavior differs substantially. The AMD part has a 1000 MHz base and 1600 MHz boost. The Intel part has a 300 MHz base and 2500 MHz boost. Memory differs entirely: the AMD part uses 16 GB LPDDR5 on a 128 bit bus with 176.0 GB/s bandwidth, while the Intel part uses system shared memory with system dependent bandwidth.

Compute resources differ by multiples. The AMD part has 512 shading units, 32 TMUs, 16 ROPs, and 8 ray tracing cores. The Intel part has 1536 shading units, 48 TMUs, 24 ROPs, and 12 ray tracing cores. Pixel rate is 25.60 GPixel/s for AMD and 60.00 GPixel/s for Intel. Texture rate is 51.20 GTexel/s for AMD and 120.0 GTexel/s for Intel. FP32 is 1.638 TFLOPS for AMD and 7.680 TFLOPS for Intel. FP16 is 3.277 TFLOPS for AMD and 15.36 TFLOPS for Intel, both at a 2:1 ratio.

Power draw differs by more than five times: 15 W for the AMD part versus 80 W for the Intel part. The AMD part has no slot width, power connector, bus interface, or suggested PSU recorded. The Intel part is an IGP with no power connectors and a portable device dependent display output. The AMD part has a 1x USB Type-C display output and physical dimensions of 298 mm length, 117 mm height, and 49 mm width. The Intel part has no recorded dimensions. The AMD part released on 2023-11-08, while the Intel part released on 2026-01-26. The Intel part's predecessor is HD Graphics-WM; the AMD part has no recorded predecessor or successor. Neither part has a recorded launch MSRP. Both parts are currently active in production and both sit at the 50th percentile in the database's GPU rankings.