Intel Arc A380E vs Intel Arc Pro B370 Comparison

Head-to-Head Benchmarks

The database contains no direct benchmark scores for either the Intel Arc A380E or the Intel Arc Pro B370. Both GPUs show an average benchmark score of zero and no recorded wins in head-to-head testing. This absence of measured data means the comparison must be built entirely from the recorded specifications, which differ substantially in almost every category.

The arithmetic rates from the specification sheet provide the clearest separation. The Arc A380E delivers 4.096 TFLOPS of FP32 throughput, while the Arc Pro B370 reaches 6.144 TFLOPS. That is a 50% advantage for the B370 in raw compute. The FP16 figures follow the same pattern: 8.192 TFLOPS for the A380E versus 12.29 TFLOPS for the B370, again a 50% gap. The B370 also holds a lead in shading units, with 1280 versus 1024, a 25% increase in that count.

The A380E strikes back in the render output and texture departments. Its pixel rate is 64.00 GPixel/s, compared to 48.00 GPixel/s for the B370. That is a 33% advantage for the older card. The texture rate also favors the A380E at 128.0 GTexel/s versus 96.00 GTexel/s, a 33% lead. The reason is visible in the TMU and ROP counts: the A380E has 64 TMUs and 32 ROPs, while the B370 has 40 TMUs and 20 ROPs. Clock speeds explain part of this. The A380E runs at a flat 2000 MHz for both base and boost, whereas the B370 has a 300 MHz base and a 2400 MHz boost. The B370's boost clock is 20% higher, but its lower fixed-function unit counts cap its pixel and texture throughput.

Ray tracing hardware also differs. The B370 carries 10 RT cores, the A380E has 8. The A380E offsets this with a dedicated 6 GB GDDR6 memory pool on a 96 bit bus delivering 186.0 GB/s of bandwidth. The B370 uses system shared memory with system dependent bandwidth, so its memory performance cannot be quantified from the database.

Both GPUs share the same API support: DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The percentile ranking against all GPUs is identical at 50 for both, which places them in the middle of the historical database distribution. With no benchmark data, the percentile is a placeholder rather than a measured result.

FAQ

Q: Which GPU has higher raw compute performance?

A: The Intel Arc Pro B370. It records 6.144 TFLOPS FP32 and 12.29 TFLOPS FP16, compared to 4.096 TFLOPS FP32 and 8.192 TFLOPS FP16 for the Arc A380E.

Q: Which GPU has faster pixel and texture throughput?

A: The Intel Arc A380E. Its pixel rate is 64.00 GPixel/s versus 48.00 GPixel/s for the B370, and its texture rate is 128.0 GTexel/s versus 96.00 GTexel/s.

Q: What memory configurations do the two GPUs use?

A: The A380E uses 6 GB of dedicated GDDR6 on a 96 bit bus with 186.0 GB/s bandwidth. The B370 uses system shared memory with system dependent bus width and bandwidth.

Q: How do their power requirements compare?

A: The A380E has a 75 W TDP and a 250 W suggested PSU. The B370 has a 25 W TDP and no suggested PSU figure recorded. Both use no power connectors.

Q: Do they support the same graphics APIs?

A: Yes. Both list DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: What are the physical form factors?

A: The A380E is a single-slot card measuring 254 mm long, 127 mm high, and 20 mm wide. The B370 is an integrated graphics processor (IGP) with no recorded dimensions.

Where Each One Wins

The Intel Arc A380E wins in scenarios that depend on memory bandwidth and fixed-function throughput. Its 186.0 GB/s of dedicated GDDR6 bandwidth gives it a predictable memory subsystem, while its 64 TMUs and 32 ROPs drive a texture rate of 128.0 GTexel/s and a pixel rate of 64.00 GPixel/s. Workloads that are fill-rate bound, such as traditional rasterization at lower resolutions, will benefit from these figures. The card also has a standalone physical form factor, a 254 mm single-slot design with four DisplayPort 2.0 outputs, which suits desktop systems that need discrete video outputs.

The Intel Arc Pro B370 wins in compute-bound scenarios. Its 1280 shading units and 10 RT cores, combined with a 2400 MHz boost clock, produce 6.144 TFLOPS FP32 and 12.29 TFLOPS FP16. That is a 50% compute advantage over the A380E. The B370 also draws only 25 W TDP, one third of the A380E's 75 W, which makes it suitable for power constrained systems despite the fact that it has no standalone dimensions recorded and relies on shared system memory.

The B370 also wins on process technology and production status. It is built on a 3 nm Intel node and is marked as active in production. The A380E is on a 6 nm TSMC node and is marked end-of-life. The B370's release date is later, and it has a successor path from the HD Graphics-WM line, while the A380E's successor is listed as Battlemage.

Specification Differences

The two GPUs differ in nearly every recorded field. The A380E uses the DG2-128 chip with the Xe-HPG architecture and belongs to the Alchemist (Arc 3) generation. The B370 uses the Panther Lake chip with the Xe3-LPG architecture and belongs to the Arc Graphics-WM (Panther Lake) generation.

Process node and foundry are different. The A380E is on a 6 nm process from TSMC with 7,200 million transistors on a 157 mm² die. The B370 is on a 3 nm process from Intel, with transistor count and die size listed as unknown.

Clock behavior differs sharply. The A380E has a flat 2000 MHz base and boost. The B370 has a 300 MHz base and 2400 MHz boost, which is a 20% higher peak clock but a much lower idle clock.

Memory is completely different. The A380E has 6 GB GDDR6, a 96 bit bus, and 186.0 GB/s bandwidth. The B370 uses system shared memory with system dependent bandwidth.

Power figures differ. The A380E has a 75 W TDP and a 250 W suggested PSU. The B370 has a 25 W TDP and no suggested PSU. The A380E is single-slot; the B370 is an IGP. The A380E uses PCIe 4.0 x8; the B370 uses an integrated graphics connection. Display outputs are four DisplayPort 2.0 on the A380E, while the B370 is portable device dependent.

The A380E measures 254 mm by 127 mm by 20 mm. The B370 has no recorded dimensions. The A380E is end-of-life with a release date in 2024; the B370 is active with a release date in 2026.

Architecture Differences

The architecture gap is the core of this comparison. The A380E is built on Xe-HPG, Intel's high performance graphics architecture from the Alchemist generation. It uses a 6 nm TSMC process and packs 7,200 million transistors into a 157 mm² die. That gives a transistor density of 45.9 million per square millimeter. The B370 moves to Xe3-LPG, a low power graphics architecture from the Panther Lake generation, fabricated on Intel's own 3 nm process. The database records no transistor count or die size for the B370, so a density comparison is impossible.

The compute arrays differ. The A380E has 1024 shading units, 64 TMUs, 32 ROPs, and 8 RT cores. The B370 has 1280 shading units, 40 TMUs, 20 ROPs, and 10 RT cores. The B370 leads in shading units and RT cores, the A380E leads in TMUs and ROPs. This split explains the benchmark-free specification story: the B370 is built for compute, the A380E for fill rate.

Memory architecture is another major divergence. The A380E has a dedicated GDDR6 subsystem, which is standard for discrete graphics. The B370 shares system memory, a characteristic of integrated GPUs. The B370's bandwidth is system dependent, meaning its performance scales with the host platform's memory configuration, while the A380E has a fixed 186.0 GB/s.

The A380E has a flat clock curve at 2000 MHz, which suggests a stable operating point for sustained workloads. The B370's 300 MHz base and 2400 MHz boost indicate a wide dynamic range, typical of integrated parts that idle low and boost when needed. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, so the API feature set is identical.

The production status also signals different lifecycles. The A380E is end-of-life, with its predecessor listed as Xe Graphics and its successor as Battlemage. The B370 is active, with its predecessor listed as HD Graphics-WM and no successor recorded. The B370 is the newer design by roughly two years.

The Verdict

The data splits the decision cleanly. The Intel Arc A380E is the choice for systems that need a discrete, single-slot card with dedicated memory. Its 6 GB GDDR6 pool, 186.0 GB/s bandwidth, 64 TMUs, and 32 ROPs make it the stronger part for fill-rate bound rendering. It also provides four DisplayPort 2.0 outputs and a fixed 2000 MHz operating point, which suits a stable desktop environment. The card has a documented physical footprint of 254 mm by 127 mm by 20 mm, so it fits in standard slots with no power connector requirement.

The Intel Arc Pro B370 is the choice for compute-focused workloads. Its 1280 shading units, 10 RT cores, and 2400 MHz boost clock produce 6.144 TFLOPS FP32, a 50% advantage over the A380E. The 25 W TDP makes it far more power efficient, and its 3 nm Intel process represents a newer manufacturing generation. The tradeoff is memory: system shared with system dependent bandwidth, which introduces platform dependency. The B370 also has no recorded dimensions, reflecting its integrated nature.

The A380E wins on pixel rate (64.00 GPixel/s versus 48.00 GPixel/s), texture rate (128.0 GTexel/s versus 96.00 GTexel/s), TMU count (64 versus 40), ROP count (32 versus 20), and memory bandwidth (186.0 GB/s versus system dependent). The B370 wins on FP32 (6.144 TFLOPS versus 4.096 TFLOPS), FP16 (12.29 TFLOPS versus 8.192 TFLOPS), shading units (1280 versus 1024), RT cores (10 versus 8), boost clock (2400 MHz versus 2000 MHz), TDP (25 W versus 75 W), and production status (active versus end-of-life).

Neither part has recorded benchmark scores, so the percentile ranking of 50 for both is not a measured outcome. The practical guidance from the database is straightforward: pick the A380E for discrete rendering tasks with predictable memory, pick the B370 for compute throughput on an integrated platform. The A380E's end-of-life status is a consideration for long-term availability, while the B370's active status suggests ongoing supply. The shared API support means software compatibility is not a differentiator. The decision rests on memory architecture and compute versus fill rate, and the specifications make that tradeoff explicit.