AMD Radeon 890M vs NVIDIA B300 SXM6 AC Comparison

Head-to-Head Benchmarks

The database contains a single directly comparable benchmark between the AMD Radeon 890M and the NVIDIA B300 SXM6 AC: Geekbench OpenCL. In this test, the NVIDIA B300 SXM6 AC records a score of 369,831, while the AMD Radeon 890M achieves 37,254. The delta percentage of -89.9% indicates the AMD part trails by nearly 90 percent in this compute-oriented workload. This is not a close contest by any measure; the NVIDIA accelerator delivers roughly 9.9 times the raw OpenCL performance of the integrated AMD graphics solution.

The NVIDIA B300 SXM6 AC sits at the 100th percentile among all GPUs in the database, meaning no other recorded device scores higher in average benchmark results. Its average benchmark score of 369,831 places it in a class occupied exclusively by data center accelerators. The AMD Radeon 890M, by contrast, holds a 45th percentile ranking, with an average score of 9,210. The gap between the two is approximately 40-fold in average terms, though the only shared test metric shows a narrower but still decisive margin.

When the AMD Radeon 890M is compared to its nearest rivals, the data places it among mainstream and older discrete GPUs. The Radeon 890M trails the AMD Radeon Vega 8 by -0.1%, the NVIDIA GeForce GTX 960 by -0.7%, the NVIDIA GeForce GTX 465 by -0.9%, and the NVIDIA GeForce GTX 850M by -1%. These deltas are all within a single percentage point, indicating the 890M performs essentially at parity with those established parts. The NVIDIA B300 SXM6 AC, in its nearest rival comparisons, leads the NVIDIA B200 by 7%, the NVIDIA H200 NVL by 10.4%, the AMD Instinct MI300X by 16.3%, and the NVIDIA L40S by 25%. These are substantial margins even against other top-tier accelerators.

The benchmark results confirm that the B300 SXM6 AC is not merely faster; it is in a different performance tier altogether. The 890M delivers 5.939 TFLOPS of FP32 compute, whereas the B300 SXM6 AC reaches 76.99 TFLOPS, a 13-fold difference in theoretical peak throughput. Texture rate tells a similar story: 185.6 GTexel/s for the AMD part versus 1,202.9 GTexel/s for NVIDIA. Pixel rate is the one metric where the AMD part shows an advantage, at 92.80 GPixel/s compared to 48.77 GPixel/s, though this reflects the B300's server-oriented design with fewer ROPs relative to its massive compute array.

Where Each One Wins

The AMD Radeon 890M wins in the pixel throughput category, posting 92.80 GPixel/s against the B300 SXM6 AC's 48.77 GPixel/s. This advantage stems from the 890M's 32 ROPs operating at a boost clock of 2900 MHz, versus the B300's 24 ROPs at a lower 2032 MHz boost. For rasterization-bound scenarios that depend heavily on fill rate, the integrated AMD part holds a measurable edge.

The NVIDIA B300 SXM6 AC dominates every compute-oriented metric recorded. Its FP32 throughput of 76.99 TFLOPS dwarfs the 890M's 5.939 TFLOPS. Texture rate favors NVIDIA at 1,202.9 GTexel/s versus 185.6 GTexel/s. The B300 also brings 592 tensor cores, a feature the 890M lacks entirely, positioning it for AI and machine learning workloads where tensor operations are essential. The 890M counters with 16 ray tracing cores, though the B300's RT core count is not recorded in the database.

Memory capacity and bandwidth are decisive differentiators. The B300 SXM6 AC ships with 288 GB of HBM3e memory across an 8192-bit bus, delivering 8.19 TB/s of bandwidth. The 890M uses system-shared memory with bandwidth described as "System Dependent," which ties performance to the host platform's memory subsystem. In practice, the B300's dedicated memory pool provides a stable, predictable bandwidth profile that shared-memory designs cannot match.

The 890M does win on power efficiency in absolute terms, with a 15 W TDP versus the B300's 1100 W. However, the performance per watt picture is inverted when normalized: the B300 delivers 76.99 TFLOPS at 1100 W (approximately 70 GFLOPS per watt), while the 890M delivers 5.939 TFLOPS at 15 W (approximately 396 GFLOPS per watt). The AMD part is roughly 5.7 times more efficient in FP32 per watt, a notable advantage for battery-constrained mobile platforms.

Architecture Differences

The AMD Radeon 890M uses the Strix Point chip built on RDNA 3.5 architecture, fabricated on TSMC's 4 nm process. The NVIDIA B300 SXM6 AC uses the GB110 chip on Blackwell Ultra architecture, also fabricated by TSMC but on a 5 nm node. The transistor counts differ dramatically: 34,000 million for the AMD chip across a 233 mm² die, versus 208,000 million for NVIDIA across a 1,628 mm² die. Transistor density favors AMD at 145.9 million per mm², compared to 127.8 million per mm² for NVIDIA, reflecting the denser 4 nm process.

The core configurations are fundamentally different in scale. The 890M has 1,024 shading units, 64 TMUs, and 32 ROPs, with 16 dedicated ray tracing cores. The B300 SXM6 AC has 18,944 shading units, 592 TMUs, and only 24 ROPs, with 592 tensor cores and no recorded RT core count. The NVIDIA part's 18.5-times shading unit advantage explains its massive FP32 lead, while the 890M's higher ROP count relative to its shading units explains its pixel rate advantage.

Clock speeds tell a nuanced story. The 890M has a base clock of 400 MHz and a boost clock of 2900 MHz, a 7.25-times boost range. The B300 SXM6 AC operates at a 1665 MHz base and 2032 MHz boost, a much narrower 1.22-times range. The AMD part's high boost clock compensates partially for its smaller core count, but not enough to close the compute gap.

Memory architecture is entirely different. The 890M uses system-shared memory with no dedicated VRAM, relying on the host's RAM with a bus width described as "System Shared." The B300 SXM6 AC has 288 GB of HBM3e on an 8192-bit bus, with memory clocked at 2000 MHz (8 Gbps effective) to produce 8.19 TB/s of bandwidth. The NVIDIA part also uses PCIe 6.0 x16, while the 890M uses PCIe 4.0 x8, doubling both the generation and lane count for the server part.

API support diverges notably. The 890M supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The B300 SXM6 AC lists all three APIs as "N/A," indicating it is not intended for conventional graphics rendering. Display outputs differ accordingly: the 890M has "Portable Device Dependent" outputs, while the B300 has "No outputs." The B300's form factor is an SXM Module with a 1500 W suggested PSU, whereas the 890M is an IGP with no power connectors.

FAQ

Q: How much faster is the NVIDIA B300 SXM6 AC in Geekbench OpenCL compared to the AMD Radeon 890M?

A: The B300 SXM6 AC scores 369,831 versus the 890M's 37,254, a delta of -89.9% from the AMD part's perspective. This translates to roughly 9.9 times higher raw score.

Q: Does the AMD Radeon 890M have any performance advantage over the B300 SXM6 AC?

A: Yes, in pixel fill rate. The 890M delivers 92.80 GPixel/s compared to 48.77 GPixel/s for the B300, due to its higher ROP count (32 versus 24) and higher boost clock (2900 MHz versus 2032 MHz).

Q: What is the memory configuration difference between these two parts?

A: The 890M uses system-shared memory with bandwidth described as "System Dependent." The B300 SXM6 AC has 288 GB of HBM3e on an 8192-bit bus, with 8.19 TB/s of dedicated bandwidth.

Q: Which part has tensor cores for AI workloads?

A: The NVIDIA B300 SXM6 AC has 592 tensor cores. The AMD Radeon 890M has no tensor cores listed in the database, though it does have 16 ray tracing cores.

Q: How do their average benchmark scores compare?

A: The 890M has an average benchmark score of 9,210 across ten recorded tests, placing it at the 45th percentile. The B300 SXM6 AC has an average score of 369,831 from a single test, placing it at the 100th percentile.

Q: What are their respective process nodes and transistor counts?

A: The 890M uses TSMC's 4 nm process with 34,000 million transistors on a 233 mm² die. The B300 SXM6 AC uses TSMC's 5 nm process with 208,000 million transistors on a 1,628 mm² die.

Specification Differences

| Specification | AMD Radeon 890M | NVIDIA B300 SXM6 AC |

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

| Chip | Strix Point | GB110 |

| Architecture | RDNA 3.5 | Blackwell Ultra |

| Generation | Navi III IGP (Strix Point Mobile) | Server Blackwell (Bxx) |

| Process Node | 4 nm | 5 nm |

| Transistors | 34,000 million | 208,000 million |

| Die Size | 233 mm² | 1,628 mm² |

| Transistor Density | 145.9M / mm² | 127.8M / mm² |

| Base Clock | 400 MHz | 1665 MHz |

| Boost Clock | 2900 MHz | 2032 MHz |

| Memory Size | System Shared | 288 GB |

| Memory Type | System Shared | HBM3e |

| Memory Bus Width | System Shared | 8192 bit |

| Memory Bandwidth | System Dependent | 8.19 TB/s |

| Shading Units | 1,024 | 18,944 |

| TMUs | 64 | 592 |

| ROPs | 32 | 24 |

| RT Cores | 16 | N/A |

| Tensor Cores | N/A | 592 |

| Pixel Rate | 92.80 GPixel/s | 48.77 GPixel/s |

| Texture Rate | 185.6 GTexel/s | 1,202.9 GTexel/s |

| FP32 Performance | 5.939 TFLOPS | 76.99 TFLOPS |

| FP16 Performance | 5.939 TFLOPS (1:1) | 76.99 TFLOPS (1:1) |

| TDP | 15 W | 1100 W |

| Slot Width | IGP | SXM Module |

| Power Connectors | None | N/A |

| Suggested PSU | N/A | 1500 W |

| Bus Interface | PCIe 4.0 x8 | PCIe 6.0 x16 |

| Display Outputs | Portable Device Dependent | No outputs |

| DirectX Support | 12 Ultimate (12_2) | N/A |

| OpenGL Support | 4.6 | N/A |

| Vulkan Support | 1.4 | N/A |

| Release Date | 2024-07-14 | 2025-09-10 |

| Predecessor | Navi II IGP | Server Hopper |

| Successor | N/A | Server Rubin |

| Percentile vs All GPUs | 45 | 100 |

| Average Benchmark Score | 9,210 | 369,831 |