Intel Arc Pro B65 vs NVIDIA GB10 Comparison

Head-to-Head Benchmarks

The data for the Intel Arc Pro B65 contains no recorded benchmark scores, while the NVIDIA GB10 has two entries in the database. The GB10's Geekbench OpenCL score is 120,137, and its Geekbench Vulkan score is 114,648, giving it an average benchmark score of 117,393. With no benchmark results for the Intel Arc Pro B65, the GB10 holds all recorded wins in the head-to-head comparison.

The GB10 sits at the 95th percentile among all GPUs in the database, placing it near the top of the field. Its nearest rival, the NVIDIA RTX 4000 SFF Ada Generation, scores 117,088, which puts the GB10 just 0.3 percent ahead. The AMD Radeon PRO W7700 scores 118,976, meaning the GB10 trails it by 1.3 percent. Against the NVIDIA Tesla V100 SXM2 16 GB, the GB10 leads by 2.6 percent, and it is 3 percent ahead of the NVIDIA RTX A5500 Mobile. These deltas are small, indicating the GB10 lands in a tightly contested performance band rather than a dominant position.

The Intel Arc Pro B65 has a percentile rank of 50, exactly the midpoint of the database. Without benchmark scores, its position is defined by its average score of zero, which cannot be compared directly to the GB10's measured results. The data shows that the GB10's recorded performance is substantial, while the Arc Pro B65's lack of recorded scores leaves its standing undefined in practical terms.

Architecture Differences

The Intel Arc Pro B65 uses the BMG-G21 chip built on the Xe2-HPG architecture, part of the Battlemage Pro Series. It is fabricated on a 5 nm process at TSMC with 19,600 million transistors on a 272 mm² die, yielding a transistor density of 72.1 million per square millimeter. The NVIDIA GB10 uses the GB20B chip based on the Blackwell 2.0 architecture, in the Server Blackwell Bxx generation. It is also on a 5 nm TSMC process, but its transistor count is unknown, and its die size is larger at 382 mm².

Clock behavior differs significantly. The Arc Pro B65 runs at a fixed 2400 MHz for both base and boost, with memory at 2375 MHz for 19 Gbps effective. The GB10 has a base clock of 1665 MHz and a boost clock of 2418 MHz, with memory at 1067 MHz for 8.5 Gbps effective. The GB10 boosts higher, but the Arc Pro B65 holds a higher base clock.

Memory configurations diverge sharply. The Arc Pro B65 has 32 GB of GDDR6 on a 256 bit bus, delivering 608.0 GB/s of bandwidth. The GB10 has 128 GB of LPDDR5X on a 256 bit bus, but its bandwidth is 273.2 GB/s, less than half of the Arc Pro B65's throughput. The GB10 trades bandwidth for capacity.

Compute resources are distributed differently. The Arc Pro B65 has 2,560 shading units, 160 texture mapping units, and 80 raster operation units, with 20 ray tracing cores and no tensor cores listed. The GB10 has 6,144 shading units, 384 TMUs, and only 48 ROPs, with 48 ray tracing cores and 384 tensor cores. The GB10 has more shading units, TMUs, and ray tracing cores, but fewer ROPs. The pixel rate reflects this: the Arc Pro B65 achieves 192.0 GPixel/s, while the GB10 manages 116.1 GPixel/s. Texture rate favors the GB10 at 928.5 GTexel/s versus the Arc Pro B65's 384.0 GTexel/s.

Peak floating point performance also diverges. The Arc Pro B65 delivers 12.29 TFLOPS for FP32 and 24.58 TFLOPS for FP16 at a 2:1 ratio. The GB10 delivers 29.71 TFLOPS for both FP32 and FP16 at a 1:1 ratio, meaning it does not double FP16 throughput. The GB10's FP32 output is more than double the Arc Pro B65's.

Power and physical characteristics differ as well. The Arc Pro B65 has a 200 W TDP, a dual-slot profile, and a 1x 8-pin power connector, with a suggested PSU of 550 W. The GB10 has a 140 W TDP, is an IGP (integrated graphics processor) with no power connectors, and a suggested PSU of 300 W. The Arc Pro B65 offers four DisplayPort 2.1 outputs, while the GB10 offers a single HDMI output. API support is another split: the Arc Pro B65 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, while the GB10 lists N/A for DirectX, OpenGL, and Vulkan. The GB10 instead relies on compute-oriented paths, which aligns with its server positioning.

FAQ

Q: Which GPU has higher FP32 compute performance?

A: The NVIDIA GB10 delivers 29.71 TFLOPS for FP32, more than double the Intel Arc Pro B65's 12.29 TFLOPS.

Q: How does memory bandwidth compare between the two?

A: The Intel Arc Pro B65 has 608.0 GB/s of bandwidth from 32 GB of GDDR6 on a 256 bit bus. The NVIDIA GB10 has 273.2 GB/s from 128 GB of LPDDR5X on a 256 bit bus. The Arc Pro B65 provides more than twice the bandwidth.

Q: What is the GB10's position relative to its nearest rivals?

A: The GB10 is 0.3 percent ahead of the NVIDIA RTX 4000 SFF Ada Generation, 1.3 percent behind the AMD Radeon PRO W7700, 2.6 percent ahead of the NVIDIA Tesla V100 SXM2 16 GB, and 3 percent ahead of the NVIDIA RTX A5500 Mobile. It sits at the 95th percentile overall.

Q: Does the Arc Pro B65 support modern graphics APIs?

A: Yes, it supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The GB10 lists N/A for these APIs, indicating a different focus.

Q: What are the power requirements for each card?

A: The Arc Pro B65 has a 200 W TDP with a 1x 8-pin power connector and a suggested PSU of 550 W. The GB10 has a 140 W TDP, no power connectors, and a suggested PSU of 300 W.

Q: How many display outputs does each provide?

A: The Arc Pro B65 provides four DisplayPort 2.1 outputs. The GB10 provides one HDMI output.

The Verdict

The recorded data favors the NVIDIA GB10 by a wide margin in raw compute benchmarks, since it has measurable scores while the Intel Arc Pro B65 has none. The GB10's 95th percentile ranking and an average score of 117,393 place it among high performers, with close competition from the RTX 4000 SFF Ada Generation and the Radeon PRO W7700. The Arc Pro B65's 50th percentile ranking and zero average score mean it cannot be compared numerically, but its specifications suggest a different design intent.

The GB10 is the clear choice for workloads that need massive FP32 throughput, tensor cores for AI tasks, and 128 GB of memory capacity. Its 29.71 TFLOPS FP32 output and 384 tensor cores indicate a compute-oriented part. The Arc Pro B65 is the better option for tasks that rely on memory bandwidth, rasterization throughput, and display connectivity. Its 608.0 GB/s bandwidth, 192.0 GPixel/s pixel rate, and four DisplayPort 2.1 outputs point toward graphics rendering and multi-display setups.

The GB10's launch MSRP is 3,999 USD. Its predecessor is Server Hopper, and its successor is Server Rubin, confirming its place in a server-focused lineup. The Arc Pro B65 has no predecessor or successor listed, and its active production status with a 2026 release date indicates a newer entry. Buyers should select based on workload: the GB10 for compute-heavy server tasks, the Arc Pro B65 for professional graphics with high bandwidth and display flexibility.

Specification Differences

| Field | Intel Arc Pro B65 | NVIDIA GB10 |

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

| Architecture | Xe2-HPG | Blackwell 2.0 |

| Generation | Battlemage (Pro Series) | Server Blackwell (Bxx) |

| Die Size | 272 mm² | 382 mm² |

| Transistors | 19,600 million | unknown |

| Base Clock | 2400 MHz | 1665 MHz |

| Boost Clock | 2400 MHz | 2418 MHz |

| Memory Size | 32 GB | 128 GB |

| Memory Type | GDDR6 | LPDDR5X |

| Memory Clock | 2375 MHz | 1067 MHz |

| Bandwidth | 608.0 GB/s | 273.2 GB/s |

| Shading Units | 2560 | 6144 |

| TMUs | 160 | 384 |

| ROPs | 80 | 48 |

| RT Cores | 20 | 48 |

| Tensor Cores | null | 384 |

| Pixel Rate | 192.0 GPixel/s | 116.1 GPixel/s |

| Texture Rate | 384.0 GTexel/s | 928.5 GTexel/s |

| FP32 | 12.29 TFLOPS | 29.71 TFLOPS |

| FP16 | 24.58 TFLOPS (2:1) | 29.71 TFLOPS (1:1) |

| TDP | 200 W | 140 W |

| Slot Width | Dual-slot | IGP |

| Power Connectors | 1x 8-pin | None |

| Suggested PSU | 550 W | 300 W |

| Display Outputs | 4x DisplayPort 2.1 | 1x HDMI |

| API Support | DX12U, OpenGL 4.6, Vulkan 1.4 | N/A for DX, GL, Vulkan |

| Release Date | 2026-03-31 | 2025-10-14 |

Where Each One Wins

The NVIDIA GB10 wins in every measured benchmark category, as it is the only one with recorded scores. Its FP32 compute is 29.71 TFLOPS, which is 2.4 times the Arc Pro B65's 12.29 TFLOPS. Its texture rate of 928.5 GTexel/s is 2.4 times higher. Its shading unit count of 6,144 is 2.4 times higher, and its 384 tensor cores provide dedicated AI processing hardware that the Arc Pro B65 lacks entirely. The GB10 also offers 128 GB of memory, four times the Arc Pro B65's 32 GB, which matters for large datasets.

The Intel Arc Pro B65 wins in several specification categories that affect graphics output and memory throughput. Its memory bandwidth of 608.0 GB/s is 2.2 times the GB10's 273.2 GB/s. Its pixel rate of 192.0 GPixel/s is 1.65 times higher, driven by 80 ROPs versus 48. Its base clock of 2400 MHz is 44 percent higher than the GB10's 1665 MHz, and its memory clock of 2375 MHz is more than double the GB10's 1067 MHz. The Arc Pro B65 also provides four DisplayPort 2.1 outputs against a single HDMI, and it supports DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4, while the GB10 lists none of these APIs.

The power profile favors the GB10 in efficiency, with a 140 W TDP versus 200 W, and no power connectors required, while the Arc Pro B65 needs a 1x 8-pin connector. The GB10 also has a smaller physical footprint as an IGP, with dimensions of 150 mm by 51 mm by 150 mm, while the Arc Pro B65's dimensions are not listed. The Arc Pro B65's higher TDP and larger power draw suggest it dedicates that budget to memory bandwidth and raster throughput, while the GB10 uses its lower power budget for compute density and tensor operations.

For rendering tasks that rely on fill rate and memory bandwidth, the Arc Pro B65 is the stronger choice based on specifications. For compute tasks that rely on FP32 throughput, tensor cores, and memory capacity, the GB10 is clearly ahead. The GB10's 95th percentile ranking and benchmark scores confirm its compute strength, while the Arc Pro B65's 50th percentile and absent benchmarks leave its real-world performance unmeasured, making its specification advantages the only basis for selection.