AMD Radeon 890M vs Intel Arc B370 Comparison

FAQ

Q: Which GPU wins the only shared benchmark, 3DMark Steel Nomad DX12?

A: The Intel Arc B370 wins decisively. It scores 1184 points, while the AMD Radeon 890M scores 590 points, giving Intel a 50.2% lead.

Q: How do the two GPUs compare in raw compute throughput?

A: The Intel Arc B370 delivers 6.144 TFLOPS of FP32 compute, slightly ahead of the AMD Radeon 890M's 5.939 TFLOPS. In FP16, the difference is larger: Intel reaches 12.29 TFLOPS (2:1 ratio) versus AMD's 5.939 TFLOPS (1:1 ratio).

Q: What are the process node differences between the two chips?

A: The AMD Radeon 890M is built on TSMC's 4 nm process node, while the Intel Arc B370 uses Intel's 3 nm process node. Intel's chip is manufactured on a smaller node.

Q: Which GPU has more shading units and texture mapping units?

A: The Intel Arc B370 has 1280 shading units, which is 256 more than the AMD Radeon 890M's 1024. However, AMD has more TMUs (64 versus 40) and more ROPs (32 versus 20).

Q: What is the thermal design power difference between the two?

A: The Intel Arc B370 has a TDP of 25 W, which is 10 W higher than the AMD Radeon 890M's 15 W TDP.

Q: How do the two GPUs rank in the overall GPU percentile standings?

A: The AMD Radeon 890M sits at the 45th percentile of all GPUs in the database, while the Intel Arc B370 ranks at the 5th percentile. This reflects the sparse benchmark data available for the Intel part, which only has one recorded test.

Architecture Differences

The AMD Radeon 890M and Intel Arc B370 represent fundamentally different integrated graphics approaches. AMD's chip is based on the RDNA 3.5 architecture, using the Strix Point chip, and belongs to the Navi III IGP generation. Intel's part uses the Xe3-LPG architecture, built on the Panther Lake chip, and belongs to the Arc Graphics-M (Panther Lake) generation.

The manufacturing processes differ significantly. AMD fabricates the 890M on a 4 nm TSMC node, while Intel produces the B370 on a 3 nm Intel node. AMD's die size is 233 mm² with 34,000 million transistors, yielding a transistor density of 145.9M per mm². Intel has not disclosed transistor counts or die size for the B370.

Clock behavior also diverges. The AMD part has a 400 MHz base clock and a 2900 MHz boost clock. The Intel part starts at 300 MHz base and boosts to 2400 MHz. Both use system-shared memory, meaning memory capacity, bus width, and bandwidth are dependent on the host system rather than fixed specifications.

The shading and texture configurations differ sharply. Intel fields 1280 shading units, 40 TMUs, and 20 ROPs, while AMD uses 1024 shading units, 64 TMUs, and 32 ROPs. Ray tracing cores are present on both: 16 on AMD versus 10 on Intel. Neither GPU includes dedicated tensor cores.

Pixel and texture throughput reflect these configurations. AMD achieves 92.80 GPixel/s pixel rate and 185.6 GTexel/s texture rate. Intel delivers 48.00 GPixel/s and 96.00 GTexel/s, roughly half of AMD's rates in both categories.

Compute throughput tells a different story. AMD's FP32 performance is 5.939 TFLOPS, while Intel reaches 6.144 TFLOPS. In FP16, AMD maintains a 1:1 ratio at 5.939 TFLOPS, whereas Intel uses a 2:1 ratio to reach 12.29 TFLOPS. This suggests Intel's architecture is optimized for workloads that can exploit half-precision math.

Both GPUs support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Both are integrated graphics (IGP) with no power connectors and portable-device-dependent display outputs. The AMD part uses a PCIe 4.0 x8 bus interface, while Intel's bus interface is simply listed as IGP. Power consumption differs, with AMD rated at 15 W TDP and Intel at 25 W TDP.

The Verdict

The recorded data points to a clear winner in the single head-to-head comparison. Intel Arc B370 dominates the 3DMark Steel Nomad DX12 test with a score of 1184, which is 50.2% higher than the AMD Radeon 890M's 590. The Intel part also edges ahead in raw FP32 compute (6.144 TFLOPS versus 5.939 TFLOPS) and offers substantially higher FP16 throughput.

However, the verdict is not one-sided across all metrics. The AMD Radeon 890M leads in pixel fill rate (92.80 GPixel/s versus 48.00 GPixel/s), texture fill rate (185.6 GTexel/s versus 96.00 GTexel/s), and has more TMUs and ROPs. Its 15 W TDP also makes it more power-efficient on paper than Intel's 25 W part.

The Intel Arc B370's 5th percentile ranking in the overall GPU database is misleading. That percentile is calculated from a single benchmark score of 1184, while the AMD part has ten recorded benchmarks with an average score of 9210 and a 45th percentile standing. The AMD part's nearest rivals include the AMD Radeon Vega 8 (average 9221, delta -0.1%), NVIDIA GeForce GTX 960 (average 9273, delta -0.7%), and NVIDIA GeForce GTX 465 (average 9294, delta -0.9%). The Intel part's nearest rivals are far older hardware: ATI Mobility Radeon HD 5570 (average 1186, delta -0.2%), ATI Radeon HD 5770 (average 1190, delta -0.5%), and AMD Radeon HD 7650M (average 1192, delta -0.7%).

For users prioritizing 3D rendering performance in the tested workload, the Intel Arc B370 is the stronger choice. For users who value pixel and texture throughput, higher clock speeds, and lower power draw, the AMD Radeon 890M holds advantages. The data does not support a universal recommendation, as the benchmark suite is too limited for the Intel part.

Specification Differences

| Specification | AMD Radeon 890M | Intel Arc B370 |

|---|---|---|

| Architecture | RDNA 3.5 | Xe3-LPG |

| Chip | Strix Point | Panther Lake |

| Generation | Navi III IGP (Strix Point Mobile) | Arc Graphics-M (Panther Lake) |

| Process Node | 4 nm (TSMC) | 3 nm (Intel) |

| Die Size | 233 mm² | unknown |

| Transistors | 34,000 million | unknown |

| Transistor Density | 145.9M / mm² | null |

| Base Clock | 400 MHz | 300 MHz |

| Boost Clock | 2900 MHz | 2400 MHz |

| Shading Units | 1024 | 1280 |

| TMUs | 64 | 40 |

| ROPs | 32 | 20 |

| RT Cores | 16 | 10 |

| Pixel Rate | 92.80 GPixel/s | 48.00 GPixel/s |

| Texture Rate | 185.6 GTexel/s | 96.00 GTexel/s |

| FP32 | 5.939 TFLOPS | 6.144 TFLOPS |

| FP16 | 5.939 TFLOPS (1:1) | 12.29 TFLOPS (2:1) |

| TDP | 15 W | 25 W |

| Bus Interface | PCIe 4.0 x8 | IGP |

| Release Date | 2024-07-14 | 2026-01-26 |

| Predecessor | Navi II IGP | None |

Head-to-Head Benchmarks

The only direct comparison available in the database is the 3DMark Steel Nomad DX12 test. The Intel Arc B370 scores 1184 points, while the AMD Radeon 890M scores 590 points. The delta percentage is -50.2% relative to the AMD part, meaning Intel's score is roughly double AMD's score. This is a substantial margin that indicates Intel's discrete-class architecture delivers significantly better performance in this modern DirectX 12 workload.

Beyond the single shared test, the database provides separate benchmark suites for each GPU. The AMD Radeon 890M has ten recorded benchmarks. Its Geekbench OpenCL score is 37254, and its Geekbench Vulkan score is 40808. In Passmark tests, it scores 8076 in G3D, 4157 in GPU Compute, 979 in G2D, 97 in DirectX 9, 73 in DirectX 11, 37 in DirectX 12, and 33 in DirectX 10. Its average benchmark score across all tests is 9210.

The Intel Arc B370 has only one recorded benchmark besides the head-to-head: the 3DMark Steel Nomad DX12 score of 1184, which also serves as its average benchmark score. The absence of additional benchmarks for the Intel part limits direct comparisons in compute, legacy DirectX, and 2D workloads.

The nearest rival data reinforces the performance gap. The AMD part's closest competitor is the AMD Radeon Vega 8 with an average score of 9221, a delta of -0.1%. The Intel part's closest competitor is the AMD FirePro M2000 with an average score of 1168, a delta of +1.4% in Intel's favor. Intel's nearest rivals are all older GPUs with substantially lower average scores than the AMD part's rivals.

Where Each One Wins

Intel Arc B370 wins in modern 3D rendering. The 3DMark Steel Nomad DX12 result is the clearest signal: a 50.2% advantage over the AMD part. The Intel GPU also leads in FP32 compute (6.144 TFLOPS versus 5.939 TFLOPS) and dominates FP16 throughput at 12.29 TFLOPS versus 5.939 TFLOPS. Applications that leverage half-precision math, such as certain AI inference workloads or compute shaders, would benefit from Intel's 2:1 FP16 ratio.

AMD Radeon 890M wins in fill-rate-bound workloads. With 92.80 GPixel/s and 185.6 GTexel/s, the AMD part outperforms Intel's 48.00 GPixel/s and 96.00 GTexel/s by 93% in both categories. Games or applications that are limited by pixel shading or texture fetching would favor the AMD part. The higher count of TMUs (64 versus 40) and ROPs (32 versus 20) supports this advantage.

AMD Radeon 890M wins on clock speed. The 2900 MHz boost clock on AMD's part is 500 MHz higher than Intel's 2400 MHz boost. Higher clocks typically benefit latency-sensitive workloads and can offset architectural differences in some scenarios.

AMD Radeon 890M wins on power efficiency. At 15 W TDP versus 25 W TDP, the AMD part consumes 40% less power while still providing competitive compute throughput. For thin-and-light laptops where thermal headroom is limited, the AMD part is the more practical choice.

AMD Radeon 890M wins on ray tracing core count. With 16 RT cores versus 10 on the Intel part, AMD has 60% more ray tracing hardware. This could translate to better ray-traced performance in supported titles, though no direct benchmark in the database confirms this.

Intel Arc B370 wins on shading unit count. With 1280 shading units versus 1024, Intel has 25% more execution units for general shader work. This likely contributes to its strong performance in the Steel Nomad test.

Intel Arc B370 wins on manufacturing process. The 3 nm node represents a smaller geometry than AMD's 4 nm node, which could offer density or efficiency advantages in the long term, though the higher TDP suggests Intel uses the process for performance rather than efficiency.

AMD Radeon 890M wins on benchmark coverage and maturity. The database contains ten benchmarks for the AMD part spanning DirectX 9 through 12, OpenCL, Vulkan, and compute workloads. The Intel part's single benchmark leaves its broader performance profile unverified. Users considering the Intel part should note that only one data point exists, making its 5th percentile ranking unreliable for general comparisons.