AMD Instinct MI455X vs Intel Arc Pro B390 Comparison

Head-to-Head Benchmarks

The recorded database contains no head-to-head benchmark entries, no wins for either part, and no average benchmark scores for either GPU. Both the AMD Instinct MI455X and the Intel Arc Pro B390 show an average benchmark score of 0, with a percentile ranking of 50 against all GPUs in the database. This absence of direct measurements means the comparison must rely entirely on the architectural and specification data recorded for each part.

The AMD Instinct MI455X delivers 157.3 TFLOPS of FP32 compute and 157.3 TFLOPS of FP16 compute with a 1:1 ratio. The Intel Arc Pro B390 delivers 7.680 TFLOPS of FP32 and 15.36 TFLOPS of FP16 with a 2:1 ratio. The AMD part provides approximately 20.5 times the FP32 throughput of the Intel part, based on the recorded figures of 157.3 divided by 7.680. For FP16, the AMD part provides approximately 10.2 times the throughput, calculated from 157.3 divided by 15.36. These ratios come directly from the data, and without actual benchmark runs, they represent the clearest computational gap between the two.

Texture rate also diverges sharply. The AMD Instinct MI455X records 2,457.6 GTexel/s, while the Intel Arc Pro B390 records 120.0 GTexel/s. That is a 20.5-fold difference, matching the FP32 ratio because both parts process one texture operation per shader per clock in their respective architectures. Pixel rate, however, tells a different story. The AMD part records 0 MPixel/s with 0 ROPs, while the Intel part records 60.00 GPixel/s with 24 ROPs. The AMD accelerator has no raster output stage, a deliberate design choice for compute workloads, whereas the Intel integrated graphics part includes a full raster pipeline.

Memory bandwidth shows the largest absolute gap. The AMD Instinct MI455X uses 432 GB of HBM4 memory across a 24576-bit bus, delivering 23.3 TB/s of bandwidth. The Intel Arc Pro B390 uses system shared memory with system-dependent bandwidth, so no fixed number exists for comparison. The AMD memory clock runs at 1900 MHz with 7.6 Gbps effective data rate. The Intel part has no dedicated memory clock because it shares system memory.

The AMD part operates at a base clock of 1000 MHz and a boost clock of 2400 MHz. The Intel part has a base clock of 300 MHz and a boost clock of 2500 MHz. Despite the Intel part's higher boost clock, its much smaller shader count of 1536 versus 32768 explains the massive compute deficit. The AMD part's boost clock is only 100 MHz lower, but it carries over 21 times the shader units.

FAQ

Q: Which GPU has higher FP32 compute performance?

A: The AMD Instinct MI455X records 157.3 TFLOPS of FP32, while the Intel Arc Pro B390 records 7.680 TFLOPS. The AMD part provides roughly 20.5 times the FP32 throughput.

Q: Does the Intel Arc Pro B390 support any graphics APIs?

A: Yes. The Intel part supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The AMD Instinct MI455X records N/A for DirectX, OpenGL, and Vulkan, indicating no graphics API support.

Q: What memory configurations do the two GPUs use?

A: The AMD Instinct MI455X uses 432 GB of HBM4 memory with a 24576-bit bus and 23.3 TB/s bandwidth. The Intel Arc Pro B390 uses system shared memory with system-dependent bandwidth and no dedicated memory clock.

Q: How do the process nodes compare?

A: The AMD Instinct MI455X is fabricated on a 2 nm process at TSMC, with 320,000 million transistors on a 2990 mm² die. The Intel Arc Pro B390 uses a 3 nm process at Intel, with transistor count and die size listed as unknown.

Q: Which GPU has a higher boost clock?

A: The Intel Arc Pro B390 has a boost clock of 2500 MHz, which is 100 MHz higher than the AMD Instinct MI455X's boost clock of 2400 MHz. The AMD part has a higher base clock at 1000 MHz versus 300 MHz.

Q: What is the power requirement for each GPU?

A: The AMD Instinct MI455X has a TDP of 2300 W with a suggested PSU of 2700 W. The Intel Arc Pro B390 has a TDP of 80 W with no suggested PSU listed. Neither GPU uses external power connectors.

Architecture Differences

The AMD Instinct MI455X uses the CDNA 5.0 architecture, specifically the MI450 256CU chip, designed for data center compute acceleration. The Intel Arc Pro B390 uses the Xe3-LPG architecture on the Panther Lake chip, designed as an integrated graphics processor for mobile or portable devices. These architectural lineages target entirely different workloads.

The AMD part uses a 2 nm process at TSMC, packing 320,000 million transistors into a 2990 mm² die, yielding a transistor density of 107.0M per mm². The Intel part uses a 3 nm process at Intel, with unknown transistor count and die size. The AMD die is enormous, reflecting its 256 compute units and 32768 shading units. The Intel die, while unmeasured, supports only 1536 shading units and 48 TMUs.

The AMD Instinct MI455X has no ROPs, no pixel rate, no ray tracing cores, and no tensor cores listed. The Intel Arc Pro B390 has 24 ROPs, a pixel rate of 60.00 GPixel/s, and 12 ray tracing cores. The AMD part omits graphics-specific hardware entirely, consistent with CDNA's focus on compute. The Intel part includes ray tracing support and a full raster pipeline, consistent with an integrated GPU that must handle display and graphics workloads.

The AMD part uses HBM4 memory with a 24576-bit bus, a configuration optimized for massive bandwidth in compute workloads. The Intel part shares system memory, which simplifies integration but ties bandwidth to the host system's memory configuration. The AMD memory clock runs at 1900 MHz with 7.6 Gbps effective, while the Intel part has no memory clock because it uses system shared memory.

The AMD part supports no graphics APIs, recording N/A for DirectX, OpenGL, and Vulkan. The Intel part supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The AMD part has no display outputs, while the Intel part has display outputs described as portable device dependent. The AMD part uses a PCIe 6.0 x16 bus interface, while the Intel part uses an IGP bus interface, meaning it is integrated into a processor package rather than installed as a discrete card.

Specification Differences

The AMD Instinct MI455X records a base clock of 1000 MHz, while the Intel Arc Pro B390 records 300 MHz. The AMD boost clock is 2400 MHz, the Intel boost clock is 2500 MHz. The AMD memory clock is 1900 MHz with 7.6 Gbps effective, while the Intel memory clock is listed as system shared.

Memory capacity differs completely: AMD has 432 GB of HBM4, Intel has system shared memory. The AMD memory bus width is 24576 bit, Intel is system shared. AMD bandwidth is 23.3 TB/s, Intel is system dependent. The AMD part has 32768 shading units, Intel has 1536. AMD has 1024 TMUs, Intel has 48. AMD has 0 ROPs, Intel has 24. AMD has no ray tracing cores listed, Intel has 12.

Pixel rate: AMD records 0 MPixel/s, Intel records 60.00 GPixel/s. Texture rate: AMD records 2,457.6 GTexel/s, Intel records 120.0 GTexel/s. FP32: AMD 157.3 TFLOPS, Intel 7.680 TFLOPS. FP16: AMD 157.3 TFLOPS (1:1), Intel 15.36 TFLOPS (2:1). TDP: AMD 2300 W, Intel 80 W. Suggested PSU: AMD 2700 W, Intel none. Slot width: AMD EAM Module, Intel IGP. Power connectors: both none. Bus interface: AMD PCIe 6.0 x16, Intel IGP. Display outputs: AMD none, Intel portable device dependent.

Process node: AMD 2 nm, Intel 3 nm. Foundry: AMD TSMC, Intel Intel. Transistors: AMD 320,000 million, Intel unknown. Die size: AMD 2990 mm², Intel unknown. Transistor density: AMD 107.0M / mm², Intel none listed. Release date: AMD 2026-07-22, Intel 2026-01-26. Production status: AMD none listed, Intel active. Predecessor: AMD Radeon Instinct, Intel HD Graphics-WM. The AMD part has no launch MSRP recorded, and neither does the Intel part.

The Verdict

The data distinguishes these two GPUs as fundamentally different products. The AMD Instinct MI455X is a compute accelerator with enormous FP32 and FP16 throughput, massive HBM4 memory, and no graphics output capabilities. The Intel Arc Pro B390 is an integrated graphics processor with ray tracing support, graphics API compatibility, and modest compute performance relative to the AMD part.

Benchmark results indicate the AMD part should be selected for workloads requiring the highest raw compute throughput. Its 157.3 TFLOPS FP32 and 23.3 TB/s memory bandwidth, combined with 432 GB of HBM4, position it for large-scale compute tasks that fit entirely in on-package memory. The absence of ROPs and display outputs confirms it is not intended for rendering to a screen.

The Intel Arc Pro B390 should be selected for systems needing integrated graphics with modern API support. Its DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4 compatibility, along with 12 ray tracing cores and 24 ROPs, make it suitable for graphics workloads on portable devices. Its 80 W TDP and system shared memory simplify integration into a processor package.

The AMD part's 2300 W TDP and suggested 2700 W PSU indicate a server-class power envelope, while the Intel part's 80 W TDP operates within typical mobile power budgets. The release date for the AMD part is approximately six months after the Intel part, with the Intel part listed as active in production and the AMD part lacking a production status.

Where Each One Wins

The AMD Instinct MI455X wins decisively in compute throughput. Its FP32 performance of 157.3 TFLOPS is roughly 20.5 times the Intel part's 7.680 TFLOPS. Its FP16 performance of 157.3 TFLOPS is roughly 10.2 times the Intel part's 15.36 TFLOPS. The texture rate of 2,457.6 GTexel/s is 20.5 times the Intel part's 120.0 GTexel/s. Memory bandwidth of 23.3 TB/s, while the Intel part has system dependent bandwidth, represents a fixed high-speed resource that does not contend with other system traffic.

The AMD part also wins on memory capacity with 432 GB of HBM4, a configuration that eliminates the need to move data over a system bus. Its 24576-bit bus width and 7.6 Gbps effective memory clock provide a dedicated memory subsystem. The PCIe 6.0 x16 interface, while not a performance number in itself, indicates a discrete accelerator slot designed for data center servers.

The Intel Arc Pro B390 wins in graphics features. It supports DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4, while the AMD part records N/A for all three. It has 12 ray tracing cores and 24 ROPs, delivering a pixel rate of 60.00 GPixel/s, while the AMD part delivers 0 MPixel/s. The Intel part has display outputs, described as portable device dependent, while the AMD part has none.

The Intel part wins on power efficiency. Its 80 W TDP is dramatically lower than the AMD part's 2300 W, and it requires no suggested PSU rating. The AMD part suggests a 2700 W PSU, indicating a system power delivery requirement far beyond the Intel part's integrated design. The Intel part's 3 nm process at Intel and 2500 MHz boost clock show a design optimized for low power and high clock speeds, whereas the AMD part's 2 nm process at TSMC, 2400 MHz boost clock, and 2300 W TDP show a design optimized for absolute throughput at any power cost.

The release timing favors the Intel part for earlier availability. The Intel Arc Pro B390 has a release date of 2026-01-26 and active production status. The AMD Instinct MI455X has a release date of 2026-07-22 with no production status listed. The Intel part's predecessor is HD Graphics-WM, indicating a lineage of integrated graphics, while the AMD part's predecessor is Radeon Instinct, indicating a lineage of discrete compute accelerators.

The use-case split is clear: the AMD part delivers maximum compute density for data center acceleration, while the Intel part delivers integrated graphics capability with modern API support for portable devices.