Intel Arc Pro B370 vs NVIDIA B200 SXM6 Comparison

Head-to-Head Benchmarks

The recorded data shows no direct benchmark scores for either the Intel Arc Pro B370 or the NVIDIA B200 SXM6. Both entries carry an average benchmark score of zero, and the head-to-head benchmark list is empty. The win counts for both parts are also zero. This means the database has not yet captured any synthetic or real-world performance runs for these two accelerators.

What the database does provide is a set of raw compute specifications that allow for a theoretical comparison. The most dramatic gap appears in FP32 throughput. The Intel Arc Pro B370 delivers 6.144 TFLOPS of FP32 compute, while the NVIDIA B200 SXM6 reaches 69.34 TFLOPS. That is roughly 11.3 times the FP32 throughput, a margin that dwarfs any other single specification difference between the two.

FP16 performance tells a different story in terms of efficiency ratio. The Intel part achieves 12.29 TFLOPS at a 2:1 ratio, which indicates it is doubling its FP32 rate by using packed FP16 operations. The NVIDIA B200 SXM6 lists FP16 at 69.34 TFLOPS with a 1:1 ratio, meaning it does not gain additional throughput when switching to FP16. The NVIDIA part still has more than 5.6 times the FP16 throughput, but the architectural approach differs fundamentally.

Pixel throughput is one of the few areas where the Intel part leads. The Arc Pro B370 reaches 48.00 GPixel/s, while the B200 SXM6 sits at 43.92 GPixel/s. This is a modest 9.3% advantage for the Intel integrated graphics solution. Texture rate goes the other way decisively: the B200 SXM6 delivers 1,083.4 GTexel/s versus 96.00 GTexel/s for the Arc Pro B370, a factor of roughly 11.3.

The NVIDIA part uses 592 texture mapping units against 40 for the Intel part, while the Intel part fields 20 ROPs against 24 for the NVIDIA part. The Intel pixel rate advantage comes from a boost clock of 2400 MHz combined with its 20 ROPs, whereas the B200 SXM6 boosts to only 1830 MHz but carries 24 ROPs. Clock speeds themselves show a wide gulf: the Intel base clock is 300 MHz and the NVIDIA base clock is 120 MHz, but the boost clocks are 2400 MHz and 1830 MHz, respectively.

Architecture Differences

The two chips occupy opposite ends of the hardware spectrum. The Intel Arc Pro B370 is built on Intel's 3 nm process node, uses the Panther Lake chip, and employs the Xe3-LPG architecture. It belongs to the Arc Graphics-WM (Panther Lake) generation. The NVIDIA B200 SXM6 uses a 5 nm process at TSMC, is built around the GB100 chip, and runs the Blackwell architecture, part of the Server Blackwell (Bxx) generation.

The Intel part is an integrated graphics processor (IGP) with a slot width of IGP and a bus interface of IGP. It draws 25 W of power and requires no power connectors. The NVIDIA part is an SXM module with a 1000 W TDP, a suggested PSU of 1400 W, and a PCIe 6.0 x16 bus interface. The display outputs differ completely: the Intel part uses portable device dependent outputs, while the B200 SXM6 has no outputs at all.

The transistor counts are starkly different. The B200 SXM6 packs 208,000 million transistors on a 1628 mm² die, with a transistor density of 127.8M per mm². The Intel Arc Pro B370 lists its transistor count and die size as unknown, so no direct density comparison can be made from the database.

Memory subsystems could not be more different. The Intel part uses system shared memory, with a system shared bus width and system dependent bandwidth. The NVIDIA part carries 180 GB of HBM3e memory on an 8192 bit bus, delivering 8.19 TB/s of bandwidth. The B200 SXM6 memory clock is listed at 2000 MHz with 8 Gbps effective transfer.

The compute unit counts reflect the divergent design goals. The Intel part has 1280 shading units, 40 TMUs, 20 ROPs, and 10 RT cores, with no tensor cores listed. The NVIDIA part has 18944 shading units, 592 TMUs, 24 ROPs, and 592 tensor cores, with no RT cores listed. The Intel part uses a 12 Ultimate (12_2) DirectX feature level, OpenGL 4.6, and Vulkan 1.4. The B200 SXM6 reports N/A for DirectX, OpenGL, and Vulkan, which confirms it is not designed for conventional graphics API workloads.

The NVIDIA part has a release date of October 31, 2024, and a launch MSRP of 34,999 USD. The Intel part has a release date of January 26, 2026. The Intel predecessor is HD Graphics-WM, while the NVIDIA predecessor is Server Hopper. The NVIDIA successor is listed as Server Rubin. The Intel part has no successor listed.

Where Each One Wins

The Intel Arc Pro B370 claims the pixel rate crown with 48.00 GPixel/s versus the B200 SXM6's 43.92 GPixel/s. This is the only headline compute metric where the Intel part leads, and it stems from the integrated part's higher boost clock of 2400 MHz and its lower ROP count relative to its clock speed. For workloads that are purely ROP-bound, such as certain rasterization passes at fixed resolution, the Intel part holds a narrow edge.

The NVIDIA B200 SXM6 wins every other major compute category. FP32 throughput of 69.34 TFLOPS dwarfs the Intel part's 6.144 TFLOPS. FP16 throughput of 69.34 TFLOPS at 1:1 ratio beats the Intel part's 12.29 TFLOPS at 2:1 ratio. Texture rate of 1,083.4 GTexel/s versus 96.00 GTexel/s is a 11.3x advantage. The 592 tensor cores provide a dedicated deep learning pathway that the Intel part cannot match, as it lists no tensor cores.

Memory bandwidth is where the B200 SXM6 is most dominant. With 8.19 TB/s from HBM3e, it offers a 45.8x advantage over the Intel part's system dependent bandwidth, which is effectively limited by the host platform's shared memory architecture. The 180 GB capacity versus system shared memory also gives the NVIDIA part a massive working set advantage for large models and datasets.

The Intel part wins on power draw, with a 25 W TDP against 1000 W for the NVIDIA part. That is a 40x difference in thermal design power. The Intel part also wins on physical integration, as an IGP with no power connectors, no slot width beyond the integrated design, and portable device dependent display outputs. The B200 SXM6 requires an SXM module slot and a 1400 W suggested PSU.

The API support landscape is another clear split. The Intel part supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, making it usable for traditional graphics workloads. The B200 SXM6 reports N/A for all three APIs, indicating it is not intended for end-user graphics rendering at all.

FAQ

Q: Which part has higher FP32 compute throughput?

A: The NVIDIA B200 SXM6 delivers 69.34 TFLOPS of FP32, compared to 6.144 TFLOPS for the Intel Arc Pro B370. The NVIDIA part is approximately 11.3 times faster in this metric.

Q: Does the Intel Arc Pro B370 win any compute benchmark?

A: The Intel part leads in pixel rate at 48.00 GPixel/s versus the B200 SXM6's 43.92 GPixel/s. It also consumes far less power at 25 W versus 1000 W.

Q: What memory does each part use?

A: The Intel Arc Pro B370 uses system shared memory with system dependent bandwidth. The NVIDIA B200 SXM6 uses 180 GB of HBM3e on an 8192 bit bus with 8.19 TB/s bandwidth.

Q: Are these parts comparable for gaming?

A: The Intel Arc Pro B370 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, making it suitable for graphics APIs. The NVIDIA B200 SXM6 reports N/A for DirectX, OpenGL, and Vulkan, so it has no conventional graphics API support.

Q: What are the power requirements?

A: The Intel Arc Pro B370 has a 25 W TDP and needs no power connectors. The NVIDIA B200 SXM6 has a 1000 W TDP and a suggested PSU of 1400 W.

Q: What tensor core counts do these parts have?

A: The NVIDIA B200 SXM6 has 592 tensor cores. The Intel Arc Pro B370 lists no tensor cores in the database.

The Verdict

The data describes two devices that share almost no intended use case. The Intel Arc Pro B370 is a 25 W integrated GPU from the Panther Lake generation, built on a 3 nm process, with 1280 shading units and support for modern graphics APIs. The NVIDIA B200 SXM6 is a 1000 W server module with 18944 shading units, 592 tensor cores, 180 GB of HBM3e, and no display outputs or graphics API support.

For anyone selecting a component for graphics rendering, gaming, or portable device integration, the Intel Arc Pro B370 is the only one of the two that supports DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4. Its 48.00 GPixel/s pixel rate exceeds the B200 SXM6. Its 25 W power envelope and lack of power connectors make it suitable for integrated designs.

For any workload that involves FP32 or FP16 compute at scale, texture-heavy operations, or large memory footprints, the B200 SXM6 is the clear choice. Its 69.34 TFLOPS FP32 and FP16 performance, 1,083.4 GTexel/s texture rate, and 8.19 TB/s memory bandwidth place it in a performance class that the Intel part cannot approach. The 592 tensor cores also enable dedicated AI acceleration that the Arc Pro B370 lacks entirely.

The database places both parts at the 50th percentile versus all GPUs, but this reflects the absence of benchmark data rather than comparable real-world performance. The recorded specifications show a 40x TDP difference and an 11.3x FP32 throughput difference, which suggests the percentile ranking will change significantly once benchmark scores are populated.

The B200 SXM6 carries a launch MSRP of 34,999 USD, and its release date of October 31, 2024 places it earlier in the product cycle. The Intel part launched on January 26, 2026. The B200 SXM6's successor is listed as Server Rubin, while the Intel part's predecessor is HD Graphics-WM.

The choice between these two parts is dictated entirely by workload class. The Intel Arc Pro B370 serves integrated graphics and portable device scenarios where power is constrained and graphics API support is required. The NVIDIA B200 SXM6 serves server and AI compute scenarios where raw throughput, tensor operations, and massive memory bandwidth are the priorities. The recorded data shows no overlap in those design goals.