Intel Arc Pro B60 Dual vs NVIDIA GB10 Comparison

Head-to-Head Benchmarks

The recorded data for the Intel Arc Pro B60 Dual shows no benchmark scores in the database, while the NVIDIA GB10 has two recorded Geekbench scores. The GB10 achieves 120,137 in Geekbench OpenCL and 114,648 in Geekbench Vulkan, producing an average benchmark score of 117,393. This places the GB10 in the 95th percentile among all GPUs in the database. The Arc Pro B60 Dual holds a 50th percentile ranking, though its average score is listed as zero due to missing benchmark entries.

The GB10's average score of 117,393 puts it close to several professional workstation competitors. The NVIDIA RTX 4000 SFF Ada Generation scores 117,088, which is 0.3% lower than the GB10. The AMD Radeon PRO W7700 scores 118,976, which is 1.3% higher than the GB10. The NVIDIA Tesla V100 SXM2 16 GB scores 114,395, sitting 2.6% behind, and the NVIDIA RTX A5500 Mobile scores 113,944, trailing by 3%. These deltas show the GB10 is competitive with modern workstation GPUs, edging out the RTX 4000 SFF Ada Generation by a narrow margin while remaining slightly behind the Radeon PRO W7700.

Since the Arc Pro B60 Dual has no benchmark entries, direct head-to-head comparisons using equal test conditions cannot be performed from the available data. However, the GB10's compute capabilities can be inferred from its architecture and specifications. The GB10 delivers 29.71 TFLOPS of FP32 performance and 29.71 TFLOPS of FP16 performance at a 1:1 ratio. The Arc Pro B60 Dual delivers 12.29 TFLOPS of FP32 and 24.58 TFLOPS of FP16 at a 2:1 ratio. The FP32 figure for the GB10 is more than double that of the Arc Pro B60 Dual. The texture rate for the GB10 is 928.5 GTexel/s versus 384.0 GTexel/s for the Intel card. The GB10 also has 6,144 shading units compared to 2,560 on the Arc Pro B60 Dual.

In pixel throughput, the Arc Pro B60 Dual counters with 192.0 GPixel/s, which is higher than the GB10's 116.1 GPixel/s. This indicates the Intel card has a stronger raw pixel fill capability despite its lower compute throughput. Memory bandwidth also favors the Arc Pro B60 Dual: it reaches 456.0 GB/s over a 192-bit bus with GDDR6 memory, while the GB10 manages 273.2 GB/s over a 256-bit bus with LPDDR5X. The GB10 compensates with a much larger 128 GB memory pool versus 24 GB on the Intel card.

Where Each One Wins

The NVIDIA GB10 wins decisively in compute-heavy workloads. Its FP32 throughput of 29.71 TFLOPS and FP16 throughput of 29.71 TFLOPS make it suitable for general compute, AI inference, and scientific simulation tasks. The 384 tensor cores on the GB10 provide hardware acceleration for matrix operations, which the Arc Pro B60 Dual lacks entirely. The GB10's 48 RT cores also outnumber the Arc Pro B60 Dual's 20 RT cores, giving it an advantage in ray-traced rendering workloads. The GB10's 128 GB of LPDDR5X memory is a massive capacity advantage, suited for large datasets and models that exceed the 24 GB capacity of the Intel card.

The Intel Arc Pro B60 Dual wins in display and rendering output scenarios. Its 192.0 GPixel/s pixel rate exceeds the GB10's 116.1 GPixel/s, meaning faster rasterization for viewport rendering and high-resolution display output. The Arc Pro B60 Dual provides four mini-DisplayPort 2.1 outputs, while the GB10 provides only a single HDMI output. This makes the Intel card more practical for multi-monitor professional setups. The Arc Pro B60 Dual also supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, whereas the GB10 reports no API support for DirectX, OpenGL, or Vulkan. The GB10 is an integrated graphics processor (IGP) with no dedicated power connector and a 140 W TDP, while the Arc Pro B60 Dual is a dual-slot card with a 1x 16-pin connector and a 400 W TDP.

For memory bandwidth, the Arc Pro B60 Dual's 456.0 GB/s is substantially higher than the GB10's 273.2 GB/s. This helps the Intel card in bandwidth-sensitive workloads such as high-resolution texture streaming and certain compute kernels. However, the GB10's 128 GB capacity versus 24 GB changes the practical usage model: the GB10 can hold far larger working sets in memory, while the Intel card moves data faster once it is resident. The GB10's PCIe 5.0 x16 interface provides double the lane count of the Arc Pro B60 Dual's PCIe 5.0 x8, which can reduce transfer bottlenecks for external data movement.

FAQ

Q: Which GPU has a higher benchmark score?

A: Only the NVIDIA GB10 has recorded benchmark scores. Its Geekbench OpenCL score is 120,137, and its Geekbench Vulkan score is 114,648, giving an average of 117,393. The Intel Arc Pro B60 Dual has no benchmark entries in the database.

Q: How does the GB10 compare to its nearest rivals?

A: The GB10's average score of 117,393 is 0.3% higher than the NVIDIA RTX 4000 SFF Ada Generation (117,088), 1.3% lower than the AMD Radeon PRO W7700 (118,976), 2.6% higher than the NVIDIA Tesla V100 SXM2 16 GB (114,395), and 3% higher than the NVIDIA RTX A5500 Mobile (113,944).

Q: Which GPU has more memory?

A: The NVIDIA GB10 has 128 GB of LPDDR5X memory on a 256-bit bus with 273.2 GB/s bandwidth. The Intel Arc Pro B60 Dual has 24 GB of GDDR6 memory on a 192-bit bus with 456.0 GB/s bandwidth.

Q: What is the compute performance difference?

A: The GB10 delivers 29.71 TFLOPS of FP32 and 29.71 TFLOPS of FP16 at a 1:1 ratio. The Arc Pro B60 Dual delivers 12.29 TFLOPS of FP32 and 24.58 TFLOPS of FP16 at a 2:1 ratio. The GB10's FP32 throughput is roughly 2.4 times higher.

Q: Which card supports more display outputs?

A: The Intel Arc Pro B60 Dual has four mini-DisplayPort 2.1 outputs. The NVIDIA GB10 has one HDMI output.

Q: What are the power requirements?

A: The Intel Arc Pro B60 Dual has a 400 W TDP, requires a 1x 16-pin power connector, and suggests an 800 W power supply. The NVIDIA GB10 has a 140 W TDP, uses no power connector, and suggests a 300 W power supply.

Specification Differences

The two GPUs differ substantially in nearly every measured specification. The Intel Arc Pro B60 Dual uses the BMG-G21 chip with 19,600 million transistors on a 272 mm² die, giving a transistor density of 72.1M per mm². The NVIDIA GB10 uses the GB20B chip with a 382 mm² die and unknown transistor count. Both are fabricated on a 5 nm process at TSMC.

Memory configurations diverge sharply. The Arc Pro B60 Dual uses 24 GB of GDDR6 with a 192-bit bus and 456.0 GB/s bandwidth. The GB10 uses 128 GB of LPDDR5X with a 256-bit bus and 273.2 GB/s bandwidth. Memory clocks also differ: the Intel card runs at 2375 MHz with 19 Gbps effective speed, while the NVIDIA card runs at 1067 MHz with 8.5 Gbps effective speed.

Compute resources favor the GB10. It has 6,144 shading units, 384 TMUs, 48 ROPs, 48 RT cores, and 384 tensor cores. The Arc Pro B60 Dual has 2,560 shading units, 160 TMUs, 80 ROPs, 20 RT cores, and no tensor cores. The ROP count is higher on the Intel card (80 versus 48), which aligns with its higher pixel rate. Clock speeds are similar at the boost level: 2400 MHz for Intel and 2418 MHz for NVIDIA. Base clocks differ: 2000 MHz for Intel versus 1665 MHz for NVIDIA.

Power and physical characteristics differ greatly. The Arc Pro B60 Dual is a dual-slot card measuring 300 mm long, 110 mm tall, and 40 mm wide. It uses a 1x 16-pin power connector and has a 400 W TDP with an 800 W suggested power supply. The GB10 is an IGP (integrated graphics processor) measuring 150 mm long, 51 mm tall, and 150 mm wide. It uses no power connector and has a 140 W TDP with a 300 W suggested power supply. The bus interface differs as well: PCIe 5.0 x8 for Intel versus PCIe 5.0 x16 for NVIDIA.

Release dates differ by about one month. The Intel card was released on 2025-09-04, and the NVIDIA GB10 followed on 2025-10-14. The GB10's predecessor is Server Hopper, and its successor is Server Rubin. The Intel card has no listed predecessor or successor. The launch MSRP for the Arc Pro B60 Dual is 1,199 USD, and for the GB10 it is 3,999 USD.

Architecture Differences

The Intel Arc Pro B60 Dual is built on the Xe2-HPG architecture from the Battlemage (Pro Series) generation. The NVIDIA GB10 uses the Blackwell 2.0 architecture from the Server Blackwell (Bxx) generation. These architectural differences drive the divergent feature sets.

The Xe2-HPG architecture in the Intel card supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The Blackwell 2.0 architecture in the GB10 reports no API support for DirectX, OpenGL, or Vulkan, indicating it is designed primarily for compute rather than graphics rendering. This is consistent with the GB10's server-oriented generation label and its IGP form factor.

The GB10 includes 384 tensor cores, which the Arc Pro B60 Dual does not have. Tensor cores accelerate AI and deep learning operations, giving the GB10 a clear architectural advantage for neural network workloads. The GB10 also has 48 RT cores versus 20 on the Intel card, doubling ray tracing hardware. The RT core count difference, combined with the lack of tensor cores on the Intel card, shows the GB10 is architected for compute acceleration while the Arc Pro B60 Dual is architected for graphics rendering.

The memory architecture differs fundamentally. The Intel card uses GDDR6 memory, which is conventional for graphics cards and prioritizes bandwidth. The GB10 uses LPDDR5X, which is lower-power memory typically found in integrated or mobile contexts. The GB10's 128 GB capacity is unusual for a GPU and suggests a compute-oriented design where memory capacity for large models takes priority over raw bandwidth. The Intel card's 456.0 GB/s bandwidth versus the GB10's 273.2 GB/s shows a bandwidth-first design on the Intel side.

The FP16 ratio difference is significant. The Intel card achieves 24.58 TFLOPS of FP16 at a 2:1 ratio relative to FP32, meaning it uses paired FP32 units for FP16 work. The GB10 achieves 29.71 TFLOPS of FP16 at a 1:1 ratio, meaning it has dedicated FP16 throughput equal to its FP32 rate. The 1:1 ratio on the GB10 indicates better FP16 efficiency for AI and scientific workloads that rely on reduced precision.

The transistor density also differs: the Intel card has 72.1M transistors per mm² across 272 mm², while the GB10 spreads its (unknown) transistor count across a larger 382 mm² die. The GB10's larger die and higher shading unit count suggest a more complex compute architecture, while the Intel card's smaller die with higher density reflects a focused graphics design.

The Verdict

The data points to a clear functional split. The NVIDIA GB10 is a compute-first device with 29.71 TFLOPS of FP32, 384 tensor cores, 48 RT cores, and 128 GB of memory. Its 95th percentile ranking among all GPUs, supported by Geekbench scores of 120,137 OpenCL and 114,648 Vulkan, confirms strong compute performance. The GB10 is also efficient: its 140 W TDP with no power connector makes it suitable for systems where power and space are constrained. The 3,999 USD launch MSRP reflects its server-oriented positioning.

The Intel Arc Pro B60 Dual is a graphics-first card with higher pixel rate (192.0 GPixel/s versus 116.1 GPixel/s), higher memory bandwidth (456.0 GB/s versus 273.2 GB/s), and four mini-DisplayPort 2.1 outputs. It supports the full DirectX 12 Ultimate and Vulkan 1.4 APIs, making it a viable option for professional rendering and multi-monitor workflows. Its 400 W TDP and dual-slot design require a substantial power supply, and its 1,199 USD launch MSRP is lower than the GB10's.

Users who need large memory capacity for AI models or data-heavy compute tasks should select the GB10. Its 128 GB memory pool and tensor cores provide capabilities the Intel card cannot match. Users who need graphics API support, high pixel throughput, and multiple display outputs should select the Arc Pro B60 Dual. The Intel card's lack of benchmark scores in the database means its real-world performance cannot be quantified from this data, but its specifications indicate a different use case than the GB10. The GB10's recorded performance against the RTX 4000 SFF Ada Generation (0.3% higher), Radeon PRO W7700 (1.3% lower), Tesla V100 SXM2 (2.6% higher), and RTX A5500 Mobile (3% higher) gives it a measured competitive position. The Arc Pro B60 Dual has no comparable data.