Intel Arc A380M vs Intel Arc B390 Comparison

FAQ

Q: How do the two Intel Arc GPUs compare in raw compute performance?

A: The Arc B390 delivers 7.680 TFLOPS FP32 performance, which is 87.5% higher than the Arc A380M's 4.096 TFLOPS. The FP16 figures follow the same pattern, with the B390 at 15.36 TFLOPS versus 8.192 TFLOPS for the A380M.

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

A: The Arc A380M uses 6 GB of dedicated GDDR6 memory on a 96-bit bus with 186.0 GB/s bandwidth. The Arc B390 relies entirely on System Shared memory, with bandwidth listed as System Dependent.

Q: Which GPU has the higher boost clock?

A: The Arc B390 boosts to 2500 MHz, while the Arc A380M boosts to 2000 MHz. However, the A380M has a much higher base clock at 1550 MHz compared to the B390's 300 MHz base.

Q: What are the transistor and die size differences?

A: The Arc A380M uses 7,200 million transistors on a 157 mm² die with a density of 45.9M per mm². The Arc B390's transistor count and die size are listed as unknown in the database.

Q: How does the Arc B390 perform in its only recorded benchmark?

A: The Arc B390 scores 1482 in 3DMark Steel Nomad DX12. This places it at the 9th percentile among all GPUs, with its closest rival being the NVIDIA GeForce GT 520MX at 1463, a 1.3% advantage.

Q: What are the manufacturer and process node details?

A: Both are Intel products. The Arc A380M is built on TSMC's 6 nm process, while the Arc B390 uses Intel's 3 nm process.

Architecture Differences

The Intel Arc A380M and Intel Arc B390 represent two distinct architectural generations from Intel. The A380M is based on the DG2-128 chip using the Xe-HPG architecture, belonging to the Alchemist generation for Arc 3 Mobile. The B390 employs the Panther Lake chip with the Xe3-LPG architecture, part of the Arc Graphics-M Panther Lake generation.

The manufacturing processes differ substantially. The A380M uses a 6 nm process from TSMC with 7,200 million transistors packed into a 157 mm² die, achieving a transistor density of 45.9 million per mm². The B390 uses Intel's 3 nm process, though the database records its transistor count and die size as unknown.

Core configurations show notable divergence. The Arc B390 packs 1536 shading units, 50% more than the A380M's 1024. However, the A380M has more texture mapping units at 64 versus 48, and more raster operation pipelines at 32 versus 24. Ray tracing cores favor the B390 with 12 units against 8 for the A380M.

Clock behavior differs significantly between the two. The A380M runs at a 1550 MHz base clock and 2000 MHz boost. The B390 has a dramatically lower 300 MHz base but boosts to 2500 MHz, indicating a design that relies on aggressive dynamic clocking under load. Memory clocks also diverge: the A380M uses 1937 MHz memory with 15.5 Gbps effective speed, while the B390 shares system memory.

The physical form factors reflect different deployment targets. The A380M is an MXM Module with MXM-A 3.1 bus interface, a removable graphics card standard. The B390 is an integrated graphics processor with an IGP bus interface and no power connectors. Both have display outputs listed as Portable Device Dependent.

API support is identical across both products: DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Neither product lists tensor cores in the database.

Head-to-Head Benchmarks

The database contains only one recorded benchmark for the B390, with no head-to-head benchmark entries comparing the two directly. The A380M has no benchmark scores recorded at all, making direct performance comparisons limited to the B390's single result.

The Arc B390's 3DMark Steel Nomad DX12 score of 1482 places it at the 9th percentile among all GPUs. Its nearest rivals in the database are all NVIDIA products with very similar scores. The GeForce GT 520MX scores 1463, putting the B390 1.3% ahead. The GeForce 800M scores 1460, a 1.5% gap. The GeForce GT 625 OEM scores 1446, and the GeForce GT 710 scores 1443, representing 2.5% and 2.7% advantages respectively.

These narrow margins indicate that the B390 sits in a performance tier populated by very old entry-level discrete GPUs, despite its modern 3 nm process and newer architecture. The 9th percentile ranking confirms this positioning, meaning roughly 91% of all GPUs in the database outperform it.

The absence of benchmark data for the A380M means its percentile ranking of 50 cannot be validated against actual recorded scores. The database lists its average benchmark score as 0, which is consistent with no test results being available.

Given the FP32 compute figures, the B390 at 7.680 TFLOPS would theoretically outperform the A380M's 4.096 TFLOPS in compute-heavy workloads, but no direct benchmark confirms this. The B390 also has a 25% higher boost clock, which would contribute to better single-threaded graphics performance if thermal and power limits allow sustained operation at that frequency.

The A380M does hold advantages in texture and pixel fill rates. Its 128.0 GTexel/s texture rate exceeds the B390's 120.0 GTexel/s, and its 64.00 GPixel/s pixel rate surpasses the B390's 60.00 GPixel/s. These figures suggest the A380M may handle certain fill-rate-bound scenarios better despite its lower compute throughput.

Specification Differences

The two Intel Arc GPUs differ across nearly every specification field in the database. Process node distinguishes them immediately: 6 nm TSMC for the A380M versus 3 nm Intel for the B390. Transistor count is 7,200 million for the A380M while the B390's is unknown. Die size similarly shows 157 mm² for the A380M versus unknown for the B390.

Clock speeds differ in both base and boost frequencies. The A380M has a 1550 MHz base clock against the B390's 300 MHz. Boost clocks reverse the relationship: 2000 MHz for the A380M versus 2500 MHz for the B390. Memory clock is 1937 MHz with 15.5 Gbps effective for the A380M, while the B390 uses system shared memory with no dedicated clock.

Memory configuration shows the starkest contrast. The A380M has 6 GB of GDDR6 on a 96-bit bus delivering 186.0 GB/s bandwidth. The B390 has System Shared memory with System Shared type, bus width, and System Dependent bandwidth.

Core counts vary across different unit types. Shading units number 1024 for the A380M and 1536 for the B390. Texture mapping units are 64 versus 48. Raster operation pipelines are 32 versus 24. Ray tracing cores are 8 versus 12. Neither lists tensor cores.

Fill rates and compute throughput both favor different products. The A380M records 64.00 GPixel/s pixel rate and 128.0 GTexel/s texture rate. The B390 records 60.00 GPixel/s and 120.0 GTexel/s respectively. FP32 performance favors the B390 at 7.680 TFLOPS versus 4.096 TFLOPS. FP16 follows at 15.36 TFLOPS versus 8.192 TFLOPS, both with 2:1 ratios.

Thermal design power differs by 45 W: 35 W for the A380M and 80 W for the B390. Slot width is MXM Module for the A380M and IGP for the B390. Power connectors are null for the A380M and None for the B390. Bus interface is MXM-A 3.1 for the A380M and IGP for the B390.

Release dates show a three-year gap: 2023-01-23 for the A380M and 2026-01-26 for the B390. Both are listed as Active in production status. Neither has a recorded launch MSRP.

Where Each One Wins

The Arc A380M establishes dominance in memory bandwidth and fill-rate metrics. Its 186.0 GB/s dedicated GDDR6 bandwidth far exceeds the system-dependent memory of the B390. The A380M also leads in pixel rate at 64.00 GPixel/s and texture rate at 128.0 GTexel/s, meaning it can sustain higher output in resolution-heavy and texture-heavy rendering scenarios.

Power efficiency favors the A380M. Its 35 W TDP is less than half the B390's 80 W, making it a more suitable option for thermally constrained mobile platforms. The MXM module form factor also allows for discrete card replacement and upgrade, whereas the B390 is integrated into the processor package.

The Arc B390 wins decisively in compute throughput. Its 7.680 TFLOPS FP32 performance is 87.5% higher than the A380M, and FP16 performance doubles that advantage at 15.36 TFLOPS versus 8.192 TFLOPS. The B390 also has 50% more shading units and 50% more ray tracing cores, positioning it better for compute workloads and ray-traced effects.

Clock flexibility favors the B390. Its 2500 MHz boost clock is 25% higher than the A380M's 2000 MHz, though the 300 MHz base clock indicates a design that scales up aggressively under load. The B390's 3 nm Intel process represents a newer manufacturing generation that could enable better power scaling at higher frequencies.

The B390's benchmark result, while modest at the 9th percentile, does confirm functional performance in 3DMark Steel Nomad DX12. The A380M has no recorded benchmarks, so its actual gaming performance cannot be verified from the database. The B390's narrow wins over very old NVIDIA parts suggest it targets low-end integrated graphics duty rather than discrete-class performance.

Form factor differences determine deployment scenarios. The A380M suits modular laptops with MXM slots and dedicated memory. The B390 integrates into Panther Lake processors, eliminating the need for separate graphics memory and reducing board complexity. The B390's lack of power connectors reflects its integrated nature, while the A380M requires the MXM slot's power delivery.

In summary, the A380M wins on memory bandwidth, fill rates, and power consumption. The B390 wins on compute throughput, shading unit count, ray tracing capability, and boost clock headroom. The absence of direct head-to-head benchmarks means these conclusions derive from the recorded specification data rather than comparative testing.