AMD Instinct MI350P vs Intel Arc Pro B390 Comparison

FAQ

Q: What are the fundamental architectural differences between the AMD Instinct MI350P and the Intel Arc Pro B390?

A: The AMD Instinct MI350P is built on CDNA 4.0 architecture with an MI350 128CU chip, while the Intel Arc Pro B390 uses Xe3-LPG architecture with a Panther Lake chip. The AMD part is a discrete accelerator with 73,000 million transistors on a 1190 mm² die, whereas the Intel part is an integrated graphics processor (IGP) with unknown transistor count and die size.

Q: How do the memory subsystems compare?

A: The AMD Instinct MI350P features 144 GB of HBM3e memory on an 8192-bit bus, delivering 8.19 TB/s of bandwidth. The Intel Arc Pro B390 uses System Shared memory with System Dependent bandwidth, meaning its memory performance relies entirely on the host system.

Q: Which GPU has higher FP32 compute throughput?

A: The AMD Instinct MI350P delivers 36.04 TFLOPS of FP32 performance, which is approximately 4.7 times higher than the Intel Arc Pro B390's 7.680 TFLOPS. The AMD part also maintains a 1:1 FP16 to FP32 ratio, while Intel's FP16 is 2:1.

Q: What are the power requirements for each GPU?

A: The AMD Instinct MI350P has a 600 W TDP and requires a 1000 W suggested PSU with a single 16-pin power connector. The Intel Arc Pro B390 has an 80 W TDP and requires no power connectors, as it is an integrated processor.

Q: What display outputs do these GPUs provide?

A: The AMD Instinct MI350P has no display outputs, making it a compute-only accelerator. The Intel Arc Pro B390's display outputs are Portable Device Dependent, meaning they vary based on the host device.

Q: What API support does each GPU offer?

A: The AMD Instinct MI350P has no DirectX, OpenGL, or Vulkan support (all listed as N/A). The Intel Arc Pro B390 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Architecture Differences

The AMD Instinct MI350P and Intel Arc Pro B390 represent two fundamentally different approaches to GPU design. The AMD part is a discrete, dual-slot accelerator built on the CDNA 4.0 architecture, specifically designed for data center compute workloads. It uses a 3 nm process from TSMC and packs 73,000 million transistors into a 1190 mm² die, achieving a transistor density of 61.3M per mm². The chip is designated MI350 128CU, indicating 128 compute units.

The Intel Arc Pro B390, in contrast, is an integrated graphics processor built on the Xe3-LPG architecture with a Panther Lake chip. It also uses a 3 nm process, but from Intel's own foundry. The transistor count and die size are unknown, which is typical for integrated processors where the GPU shares the die with CPU cores. The Intel part is classified as an IGP with no slot width, no power connectors, and no dimensions listed, confirming its integration into a host processor package.

The compute resources differ dramatically. The AMD Instinct MI350P has 8192 shading units, 512 texture mapping units, and 0 ROPs, with a pixel rate of 0 MPixel/s and a texture rate of 1,126.4 GTexel/s. The zero ROP count and pixel rate confirm this is a pure compute accelerator with no rasterization pipeline. The Intel Arc Pro B390 has 1536 shading units, 48 TMUs, 24 ROPs, and 12 ray tracing cores, with a pixel rate of 60.00 GPixel/s and a texture rate of 120.0 GTexel/s.

Clock speeds also differ significantly. The AMD part has a 1000 MHz base clock and 2200 MHz boost clock, with memory running at 2000 MHz (8 Gbps effective). The Intel part has a much lower 300 MHz base clock but a higher 2500 MHz boost clock, reflecting its integrated nature where the GPU shares power and thermal budget with the host CPU.

The AMD Instinct MI350P uses PCIe 5.0 x16 for its bus interface, while the Intel Arc Pro B390 uses an IGP bus interface. The AMD part is 267 mm long, 111 mm high, and 40 mm wide, requiring a dual-slot form factor. The Intel part has no listed dimensions.

Head-to-Head Benchmarks

The head-to-head benchmark data contains no recorded entries for either GPU, and both parts show an average benchmark score of 0 with an equal percentile ranking of 50 against all GPUs. The wins count is 0 for both the AMD Instinct MI350P and the Intel Arc Pro B390.

Without direct benchmark comparisons, the compute specifications provide the basis for performance analysis. The AMD Instinct MI350P delivers 36.04 TFLOPS of FP32 performance, which is 4.7 times the Intel Arc Pro B390's 7.680 TFLOPS. In FP16, the AMD part again delivers 36.04 TFLOPS with a 1:1 ratio, while the Intel part delivers 15.36 TFLOPS with a 2:1 ratio, meaning the AMD part is 2.3 times faster in FP16.

Texture throughput heavily favors the AMD part. The MI350P achieves 1,126.4 GTexel/s compared to the Intel Arc Pro B390's 120.0 GTexel/s, a 9.4 times difference. This reflects the massive difference in texture mapping units, with 512 on the AMD part versus 48 on the Intel part.

The Intel Arc Pro B390 holds one clear advantage: pixel rate. It achieves 60.00 GPixel/s with its 24 ROPs, while the AMD Instinct MI350P has 0 ROPs and a 0 MPixel/s pixel rate. This difference is expected, as the AMD part omits the rasterization pipeline entirely.

Memory bandwidth is another area of massive divergence. The AMD Instinct MI350P has 8.19 TB/s of bandwidth from its 8192-bit HBM3e interface, while the Intel Arc Pro B390's bandwidth is System Dependent, providing no fixed figure. The AMD part also has 144 GB of dedicated memory, whereas the Intel part relies on System Shared memory.

Specification Differences

The two GPUs differ across nearly every specification category:

• Architecture: CDNA 4.0 versus Xe3-LPG
• Process node: Both use 3 nm, but from different foundries (TSMC versus Intel)
• Transistors: 73,000 million versus unknown
• Die size: 1190 mm² versus unknown
• Base clock: 1000 MHz versus 300 MHz
• Boost clock: 2200 MHz versus 2500 MHz
• Memory size: 144 GB versus System Shared
• Memory type: HBM3e versus System Shared
• Memory bus: 8192 bit versus System Shared
• Memory bandwidth: 8.19 TB/s versus System Dependent
• Shading units: 8192 versus 1536
• TMUs: 512 versus 48
• ROPs: 0 versus 24
• Ray tracing cores: None listed versus 12
• Pixel rate: 0 MPixel/s versus 60.00 GPixel/s
• Texture rate: 1,126.4 GTexel/s versus 120.0 GTexel/s
• FP32: 36.04 TFLOPS versus 7.680 TFLOPS
• FP16: 36.04 TFLOPS (1:1) versus 15.36 TFLOPS (2:1)
• TDP: 600 W versus 80 W
• Slot width: Dual-slot versus IGP
• Power connectors: 1x 16-pin versus None
• Suggested PSU: 1000 W versus not listed
• Bus interface: PCIe 5.0 x16 versus IGP
• Display outputs: No outputs versus Portable Device Dependent
• API support: N/A for DirectX, OpenGL, Vulkan versus DirectX 12 Ultimate (12_2), OpenGL 4.6, Vulkan 1.4
• Dimensions: 267 mm x 111 mm x 40 mm versus not listed
• Release date: 2026-05-06 versus 2026-01-26
• Predecessor: Radeon Instinct versus HD Graphics-WM

The Verdict

The recorded data indicates these two GPUs serve entirely different purposes. The AMD Instinct MI350P is a data center accelerator with no display outputs, no graphics API support, and a 600 W power envelope. Its specifications are oriented toward massive parallel compute: 36.04 TFLOPS of FP32, 8.19 TB/s of memory bandwidth, and 144 GB of HBM3e memory. The zero ROP count and 0 MPixel/s pixel rate confirm it cannot render graphics.

The Intel Arc Pro B390 is an integrated graphics processor with a 80 W TDP, support for DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4, plus 12 ray tracing cores. Its 60.00 GPixel/s pixel rate and 24 ROPs indicate a functional graphics pipeline. The base clock of 300 MHz with a boost of 2500 MHz suggests power management behavior typical of integrated parts.

Neither GPU has recorded benchmark scores in the database, and both sit at the 50th percentile among all GPUs. The wins tally shows 0 for each side. The analysis therefore rests on the architectural and specification data.

Where Each One Wins

The AMD Instinct MI350P wins decisively in compute throughput. Its 36.04 TFLOPS FP32 performance is 4.7 times the Intel part's 7.680 TFLOPS. In FP16, the AMD part's 36.04 TFLOPS at 1:1 ratio provides 2.3 times the Intel part's 15.36 TFLOPS at 2:1 ratio. The 1,126.4 GTexel/s texture rate is 9.4 times the Intel part's 120.0 GTexel/s. The 8.19 TB/s memory bandwidth and 144 GB capacity create a clear advantage for large data sets and memory-bound workloads. The PCIe 5.0 x16 interface provides a high-bandwidth host connection.

The Intel Arc Pro B390 wins in graphics-oriented tasks. Its 60.00 GPixel/s pixel rate and 24 ROPs enable rasterization, while the 12 ray tracing cores provide hardware-accelerated ray tracing. The DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4 support make it compatible with graphics applications. The 80 W TDP requires no additional power connectors and no dedicated PSU, simplifying integration. The portable device dependent display outputs allow flexible display configuration.

The AMD part wins for compute-intensive workloads such as large-scale matrix operations, where FP32 and FP16 throughput and memory bandwidth are primary factors. The Intel part wins for integrated graphics scenarios where power efficiency, display output, and graphics API compatibility matter. The AMD part's 2026-05-06 release date follows the Intel part's 2026-01-26 release by several months. The AMD part succeeds the Radeon Instinct line, while the Intel part succeeds HD Graphics-WM.