Intel Arc Graphics 32EU vs Intel Arc Pro B390 Comparison

FAQ

Q: What is the performance difference between the Intel Arc Graphics 32EU and the Intel Arc Pro B390?

A: The Arc Graphics 32EU has a recorded average benchmark score of 733 points (3DMark Steel Nomad DX12), placing it at the 3rd percentile of all GPUs. The Arc Pro B390 has no recorded benchmark scores in the database, and its average benchmark score is 0, though it sits at the 50th percentile based on its specifications.

Q: How do the shading unit counts compare between the two?

A: The Arc Graphics 32EU has 256 shading units, while the Arc Pro B390 has 1536 shading units. That is a 6x difference in raw shading hardware, which directly explains the large gap in theoretical FP32 throughput.

Q: What are the boost clock speeds of each GPU?

A: The Arc Graphics 32EU boosts to 1950 MHz, while the Arc Pro B390 boosts to 2500 MHz. Both have a base clock of 300 MHz.

Q: Which GPU supports hardware ray tracing?

A: The Arc Pro B390 includes 12 ray tracing cores. The Arc Graphics 32EU does not list any ray tracing cores in its specifications.

Q: What are the TDP ratings for these two processors?

A: The Arc Graphics 32EU has a TDP of 65 W, while the Arc Pro B390 has a TDP of 80 W.

Q: Are both GPUs based on the same architecture?

A: No. The Arc Graphics 32EU uses the Xe-LPG architecture (part of the Arrow Lake-S chip), while the Arc Pro B390 uses the newer Xe3-LPG architecture (part of the Panther Lake chip).

Architecture Differences

The Intel Arc Graphics 32EU and the Intel Arc Pro B390 represent two distinct generations of Intel's integrated graphics technology, built on the same 3 nm process node but with different foundries and fundamentally different architectures.

The Arc Graphics 32EU is based on the Xe-LPG architecture, manufactured by TSMC, and sits within the Arrow Lake-S chip. It belongs to the Arc Graphics-M generation. The Arc Pro B390, by contrast, uses the Xe3-LPG architecture, is manufactured by Intel's own foundry, and is part of the Panther Lake chip in the Arc Graphics-WM generation. This is a notable shift: the older part relies on an external foundry, while the newer part uses Intel's internal fabrication.

The transistor counts differ dramatically, though the data is incomplete for the B390. The 32EU contains 17,800 million transistors on a 243 mm² die, giving a transistor density of 73.3 million transistors per square millimeter. The B390's transistor count and die size are listed as unknown. This means the database cannot quantify the physical scaling between the two parts, but the architectural jump from Xe-LPG to Xe3-LPG indicates a substantial redesign.

The shading architecture scales significantly. The 32EU has 256 shading units, 16 texture mapping units, and 8 raster operation units. The B390 has 1536 shading units, 48 TMUs, and 24 ROPs. That is a 6x increase in shaders, a 3x increase in TMUs, and a 3x increase in ROPs. The B390 also adds 12 ray tracing cores, which the 32EU lacks entirely. This makes the B390 a much more complete graphics processor for modern workloads that use ray-traced effects.

Both GPUs share the same API support: DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Both also use system-shared memory, meaning they do not have dedicated VRAM but rather borrow from the main system memory. The memory bus width and bandwidth are both listed as system dependent for each part, so the actual memory performance depends entirely on the host platform.

The B390 has a higher boost clock at 2500 MHz versus 1950 MHz for the 32EU. That is a 28% higher boost frequency, which compounds with the much larger shading array to produce a very large theoretical performance gap. The base clocks are identical at 300 MHz for both.

Head-to-Head Benchmarks

The head-to-head benchmark table between these two GPUs is empty, and the recorded win counts are zero for both sides. This means the database contains no direct comparative benchmark results pairing the Arc Graphics 32EU against the Arc Pro B390. However, the available data for the 32EU and the specification-based percentile for the B390 allow for a clear analysis.

The Arc Graphics 32EU has one recorded benchmark: 3DMark Steel Nomad DX12, with a score of 733. This places it at the 3rd percentile of all GPUs, meaning it sits in the bottom 3% of all recorded graphics hardware. Its nearest rivals in the database are the Intel Arc Graphics 24EU and the Intel Arc Graphics 64EU, both with identical average scores of 733 and a delta of 0%. The AMD Radeon HD 6470M scores 723, which is 1.4% lower, and the NVIDIA GeForce GT 415M scores 751, which is 2.4% higher. This places the 32EU in a very low-performance tier, roughly on par with decade-old discrete mobile GPUs.

The Arc Pro B390 has no recorded benchmark scores. Its average benchmark score is 0, but its percentile versus all GPUs is listed at 50, which is a specification-based estimate rather than a measured result. That 50th percentile placement suggests the B390 is expected to perform around the median of all GPUs, a massive jump from the 3rd percentile of the 32EU.

The performance delta between these two parts is not directly measurable from the head-to-head data, but the specification differences point to a very large gap. The B390 has 6x the shading units, 3x the TMUs, 3x the ROPs, a 28% higher boost clock, and dedicated ray tracing hardware. Using the FP32 throughput figures as a proxy, the 32EU delivers 998.4 GFLOPS while the B390 delivers 7.680 TFLOPS. That is a 7.7x difference in raw single-precision compute. The pixel rate tells a similar story: 15.60 GPixel/s for the 32EU versus 60.00 GPixel/s for the B390, a 3.8x difference. The texture rate is 31.20 GTexel/s versus 120.0 GTexel/s, a 3.8x difference.

No benchmark data confirms these theoretical numbers, but the specification sheet strongly indicates that the B390 will outperform the 32EU by a wide margin in almost every graphics workload.

Specification Differences

The two GPUs differ across nearly every major specification category.

Process and Chip:

• 32EU: 3 nm process, TSMC foundry, Arrow Lake-S chip, Xe-LPG architecture, Arc Graphics-M generation
• B390: 3 nm process, Intel foundry, Panther Lake chip, Xe3-LPG architecture, Arc Graphics-WM generation

Transistors and Die:

• 32EU: 17,800 million transistors, 243 mm² die, 73.3M / mm² density
• B390: Unknown transistor count, unknown die size, no density data

Clocks:

• 32EU: 300 MHz base, 1950 MHz boost
• B390: 300 MHz base, 2500 MHz boost

Compute Units:

• 32EU: 256 shading units, 16 TMUs, 8 ROPs, no ray tracing cores
• B390: 1536 shading units, 48 TMUs, 24 ROPs, 12 ray tracing cores

Rates:

• 32EU: 15.60 GPixel/s pixel rate, 31.20 GTexel/s texture rate, 998.4 GFLOPS FP32, 1.997 TFLOPS FP16 (2:1)
• B390: 60.00 GPixel/s pixel rate, 120.0 GTexel/s texture rate, 7.680 TFLOPS FP32, 15.36 TFLOPS FP16 (2:1)

Power and Interface:

• 32EU: 65 W TDP, IGP slot width, no power connectors listed, Ring Bus interface
• B390: 80 W TDP, IGP slot width, no power connectors (explicitly listed as None), IGP bus interface

Display Outputs:

• 32EU: Motherboard Dependent
• B390: Portable Device Dependent

Release Dates:

• 32EU: 2024-10-23
• B390: 2026-01-26

APIs:

• Both: DirectX 12 Ultimate (12_2), OpenGL 4.6, Vulkan 1.4

Memory:

• Both: System Shared size, type, and bus width; bandwidth is System Dependent

Benchmark Data:

• 32EU: One recorded benchmark (733 in 3DMark Steel Nomad DX12), 3rd percentile, average score 733
• B390: No recorded benchmarks, 50th percentile estimate, average score 0

The Verdict

The data shows a clear separation in intended roles. The Intel Arc Graphics 32EU is a low-power integrated graphics solution with a 65 W TDP, 256 shading units, and a recorded benchmark score of 733 that places it at the 3rd percentile of all GPUs. Its nearest rivals are other Intel integrated parts with identical scores, plus a pair of very old discrete mobile GPUs from AMD and NVIDIA. This is entry-level integrated graphics, suitable for basic display output and light 2D workloads.

The Intel Arc Pro B390 is a substantially more capable processor. Its 1536 shading units, 12 ray tracing cores, 2500 MHz boost clock, and 80 W TDP place it at the 50th percentile of all GPUs in the database, even without any measured benchmark scores. The specification-based percentile estimate puts it at the median of all recorded graphics hardware, which is a completely different performance class from the 32EU.

The release dates reinforce this separation. The 32EU launched on 2024-10-23, while the B390 launched on 2026-01-26, roughly 15 months later. The B390 is a newer, more advanced part built on a different architecture generation (Xe3-LPG versus Xe-LPG) and fabricated by a different foundry (Intel versus TSMC).

For anyone building a system with an Arrow Lake-S processor, the 32EU is whatever integrated graphics ships with that chip. It handles basic tasks and light gaming at very low settings. For anyone using a Panther Lake-based platform, the B390 offers a much larger graphics subsystem with ray tracing support and roughly 7.7x the FP32 compute throughput. The data does not support any scenario where the 32EU outperforms the B390; every measurable specification favors the B390.

Where Each One Wins

Intel Arc Graphics 32EU wins in power efficiency and simplicity. The 65 W TDP is 15 W lower than the B390's 80 W. It uses a Ring Bus interface, which is a standard on-die interconnect, and it has no power connectors listed, meaning it draws entirely from the motherboard. The 32EU also has a smaller shading array, which translates to lower silicon complexity and potentially lower manufacturing cost for the host chip. For ultra-low-power systems where graphics performance is secondary, the 32EU is the more economical choice. Its release date of 2024 also makes it available earlier in the market.

Intel Arc Pro B390 wins in every performance category. The 1536 shading units versus 256 is a 6x advantage. The 48 TMUs versus 16 is a 3x advantage. The 24 ROPs versus 8 is a 3x advantage. The 2500 MHz boost clock versus 1950 MHz is a 28% advantage. The FP32 throughput of 7.680 TFLOPS versus 998.4 GFLOPS is a 7.7x advantage. The pixel rate of 60.00 GPixel/s versus 15.60 GPixel/s is a 3.8x advantage. The texture rate of 120.0 GTexel/s versus 31.20 GTexel/s is also a 3.8x advantage. The B390 adds 12 ray tracing cores, which the 32EU lacks entirely. The B390's 50th percentile placement versus the 32EU's 3rd percentile encapsulates the overall performance gap.

The B390 also has a higher FP16 throughput at 15.36 TFLOPS versus 1.997 TFLOPS, a 7.7x difference that matters for compute workloads using half precision. The B390 uses an IGP bus interface and lists no power connectors, consistent with its integrated nature, but its higher TDP of 80 W indicates it expects a more robust power delivery system.

For gaming, the B390's ray tracing cores and 6x shading unit count make it the only viable option for modern 3D titles. For professional or compute-oriented tasks, the B390's FP32 and FP16 throughput dominate. The 32EU remains suitable only for basic display output, office productivity, and video playback. The recorded benchmark data confirms the 32EU sits in the bottom 3% of all GPUs, while the B390's specification-based estimate places it at the median. There is no workload category in the data where the 32EU beats the B390 on raw performance.