AMD Instinct MI308X vs Intel Arc Pro B390 Comparison

Where Each One Wins

The AMD Instinct MI308X and Intel Arc Pro B390 occupy completely different positions in the recorded database, with no shared benchmark wins because the two parts never appear in the same workload category. The MI308X is a compute accelerator aimed at data center scale workloads, while the Arc Pro B390 is an integrated graphics processor designed for portable devices. The data shows that the MI308X wins in raw compute throughput, memory capacity, and bandwidth, while the Arc Pro B390 wins in power efficiency, API compatibility, and display functionality.

The MI308X delivers 81.72 TFLOPS of FP32 performance and 81.72 TFLOPS of FP16 performance at a 1:1 ratio. The Arc Pro B390 delivers 7.680 TFLOPS of FP32 and 15.36 TFLOPS of FP16 at a 2:1 ratio. The MI308X leads in absolute FP32 compute by a factor of roughly 10.6, and in FP16 by a factor of roughly 5.3. However, the Arc Pro B390 provides a much lower thermal envelope at 80 W TDP versus 750 W TDP for the MI308X, making it suitable for systems where power draw is the primary constraint.

In terms of memory, the MI308X offers 192 GB of HBM3 memory across an 8192 bit bus, yielding 5.32 TB/s of bandwidth. The Arc Pro B390 uses system shared memory with system dependent bandwidth, which means it has no dedicated memory resources of its own. The MI308X clearly wins in any memory-heavy workload that requires large datasets or high throughput.

The Arc Pro B390 wins in flexibility of output, with display outputs described as portable device dependent, while the MI308X has no display outputs at all. The Arc Pro B390 also supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, whereas the MI308X reports N/A for all three APIs. For graphics workloads that rely on standard consumer APIs, the Arc Pro B390 is the only option between the two.

Architecture Differences

The MI308X uses the Aqua Vanjaram chip built on CDNA 3.0 architecture, manufactured on a 5 nm process at TSMC. It integrates 153,000 million transistors on a 1017 mm² die, achieving a transistor density of 150.4M per mm². The Arc Pro B390 uses the Panther Lake chip built on Xe3-LPG architecture, manufactured on a 3 nm process at Intel, with transistor count and die size listed as unknown in the database.

The MI308X belongs to the Instinct (MIx) generation and ships as an OAM Module with a PCIe 5.0 x16 bus interface. The Arc Pro B390 belongs to the Arc Graphics-WM (Panther Lake) generation and ships as an IGP with an IGP bus interface. The MI308X requires no power connectors but carries a suggested PSU of 1150 W, while the Arc Pro B390 also has no power connectors but has no suggested PSU listed.

The MI308X has 19,456 shading units, 1,216 texture mapping units, and 0 ROPs. It reports 0 MPixel/s pixel rate and 2,553.6 GTexel/s texture rate. The Arc Pro B390 has 1,536 shading units, 48 texture mapping units, 24 ROPs, and 12 ray tracing cores. Its pixel rate is 60.00 GPixel/s and texture rate is 120.0 GTexel/s.

Clock behavior differs substantially. The MI308X runs at a 1000 MHz base clock and 2100 MHz boost clock, with memory clock at 1300 MHz or 5.2 Gbps effective. The Arc Pro B390 runs at a 300 MHz base clock and 2500 MHz boost clock, with system shared memory. The MI308X has a lower boost clock but a much higher base clock, reflecting its sustained compute design. The Arc Pro B390 has a very low base clock but a higher boost ceiling, typical of an integrated part that ramps under load.

The MI308X supports no display outputs, no consumer graphics APIs, and has no ray tracing cores listed. The Arc Pro B390 is production active, with a release date of 2026-01-26, while the MI308X has no production status listed and a release date of 2023-12-05. The MI308X lists Radeon Instinct as its predecessor, while the Arc Pro B390 lists HD Graphics-WM as its predecessor.

Head-to-Head Benchmarks

The database records no direct head-to-head benchmark entries for these two parts, and neither part has individual benchmark scores or nearest rivals listed. The winsA and winsB fields both report 0, and the headToHeadBenchmarks array is empty. The available comparison therefore relies on the specification data recorded in the database.

The largest compute gap appears in FP32 throughput. The MI308X records 81.72 TFLOPS, while the Arc Pro B390 records 7.680 TFLOPS. That is a 74.04 TFLOPS difference in favor of the MI308X. In FP16, the MI308X again records 81.72 TFLOPS at a 1:1 ratio, while the Arc Pro B390 records 15.36 TFLOPS at a 2:1 ratio, a difference of 66.36 TFLOPS.

Memory capacity and bandwidth also show a decisive split. The MI308X provides 192 GB of HBM3 memory with 5.32 TB/s bandwidth across an 8192 bit bus. The Arc Pro B390 provides no dedicated memory, relying on system shared memory with system dependent bandwidth. For workloads that require holding large models or datasets on the accelerator, the MI308X is the only viable choice.

Texture throughput favors the MI308X as well. The MI308X records 2,553.6 GTexel/s, while the Arc Pro B390 records 120.0 GTexel/s. The MI308X leads by 2,433.6 GTexel/s. Pixel rate, however, favors the Arc Pro B390, which records 60.00 GPixel/s, while the MI308X records 0 MPixel/s due to having no ROPs.

Power consumption shows the inverse relationship. The MI308X has a 750 W TDP and a suggested PSU of 1150 W. The Arc Pro B390 has an 80 W TDP, which is 670 W lower. The Arc Pro B390 also has a much higher boost clock at 2500 MHz compared to 2100 MHz for the MI308X, and a lower base clock at 300 MHz compared to 1000 MHz.

The Arc Pro B390 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The MI308X lists N/A for all three. This makes the Arc Pro B390 the only part in this comparison that can run standard graphics workloads through consumer APIs.

FAQ

Q: Which part has higher FP32 compute performance?

A: The AMD Instinct MI308X records 81.72 TFLOPS of FP32 performance, while the Intel Arc Pro B390 records 7.680 TFLOPS. The MI308X leads by 74.04 TFLOPS.

Q: Does the Intel Arc Pro B390 have dedicated memory?

A: No. The Arc Pro B390 uses system shared memory for both size and type, with system dependent bandwidth. The MI308X, by contrast, has 192 GB of HBM3 memory with 5.32 TB/s bandwidth.

Q: Which part supports ray tracing?

A: The Intel Arc Pro B390 has 12 ray tracing cores. The AMD Instinct MI308X has no ray tracing cores listed in the database.

Q: What is the TDP difference between the two parts?

A: The MI308X has a 750 W TDP with a suggested PSU of 1150 W. The Arc Pro B390 has an 80 W TDP with no suggested PSU listed. The difference is 670 W.

Q: Which part has display outputs?

A: The Arc Pro B390 has display outputs described as portable device dependent. The MI308X has no display outputs.

Q: What process nodes do the two parts use?

A: The MI308X uses a 5 nm process at TSMC. The Arc Pro B390 uses a 3 nm process at Intel.

Specification Differences

The two parts differ across nearly every recorded specification field.

Process node: the MI308X uses 5 nm at TSMC, while the Arc Pro B390 uses 3 nm at Intel.

Transistors: the MI308X integrates 153,000 million transistors, while the Arc Pro B390 lists unknown.

Die size: the MI308X measures 1017 mm², while the Arc Pro B390 lists unknown.

Base clock: the MI308X runs at 1000 MHz, while the Arc Pro B390 runs at 300 MHz.

Boost clock: the MI308X runs at 2100 MHz, while the Arc Pro B390 runs at 2500 MHz.

Memory clock: the MI308X runs at 1300 MHz or 5.2 Gbps effective, while the Arc Pro B390 uses system shared memory.

Memory size: the MI308X has 192 GB of HBM3, while the Arc Pro B390 has system shared memory.

Memory bus: the MI308X uses an 8192 bit bus, while the Arc Pro B390 uses system shared.

Memory bandwidth: the MI308X provides 5.32 TB/s, while the Arc Pro B390 is system dependent.

Shading units: the MI308X has 19,456, while the Arc Pro B390 has 1,536.

Texture mapping units: the MI308X has 1,216, while the Arc Pro B390 has 48.

ROPs: the MI308X has 0, while the Arc Pro B390 has 24.

Ray tracing cores: the MI308X has none listed, while the Arc Pro B390 has 12.

Pixel rate: the MI308X records 0 MPixel/s, while the Arc Pro B390 records 60.00 GPixel/s.

Texture rate: the MI308X records 2,553.6 GTexel/s, while the Arc Pro B390 records 120.0 GTexel/s.

FP32: the MI308X records 81.72 TFLOPS, while the Arc Pro B390 records 7.680 TFLOPS.

FP16: the MI308X records 81.72 TFLOPS at 1:1, while the Arc Pro B390 records 15.36 TFLOPS at 2:1.

TDP: the MI308X has 750 W, while the Arc Pro B390 has 80 W.

Slot width: the MI308X is an OAM Module, while the Arc Pro B390 is an IGP.

Suggested PSU: the MI308X lists 1150 W, while the Arc Pro B390 lists none.

Bus interface: the MI308X uses PCIe 5.0 x16, while the Arc Pro B390 uses IGP.

Display outputs: the MI308X has none, while the Arc Pro B390 is portable device dependent.

APIs: the MI308X lists N/A for DirectX, OpenGL, and Vulkan, while the Arc Pro B390 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Production status: the MI308X lists none, while the Arc Pro B390 lists Active.

Release date: the MI308X is 2023-12-05, while the Arc Pro B390 is 2026-01-26.

Predecessor: the MI308X lists Radeon Instinct, while the Arc Pro B390 lists HD Graphics-WM.

The Verdict

The AMD Instinct MI308X is the choice for compute-heavy workloads that demand massive memory capacity and throughput. Its 192 GB of HBM3 memory, 5.32 TB/s bandwidth, and 81.72 TFLOPS of FP32 and FP16 performance position it as a data center accelerator for large-scale processing tasks. The absence of display outputs and consumer API support confirms that it is not intended for graphics rendering or end-user display workloads.

The Intel Arc Pro B390 is the choice for portable systems that need graphics output and standard API support within a low power envelope. Its 80 W TDP, 12 ray tracing cores, DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4 support, plus portable device dependent display outputs, make it a functional integrated graphics solution. Its 7.680 TFLOPS of FP32 and 15.36 TFLOPS of FP16 are far below the MI308X, but its 2500 MHz boost clock and 60.00 GPixel/s pixel rate show it is built for rendering efficiency rather than raw compute.

The data supports a clear separation: the MI308X delivers roughly 10.6 times the FP32 throughput and 5.3 times the FP16 throughput of the Arc Pro B390, but consumes 670 W more power and lacks all display functionality. The Arc Pro B390 delivers the only ray tracing, the only API support, and the only display outputs in this comparison. Neither part can substitute for the other in its intended role. A system requiring sustained high-bandwidth compute should use the MI308X, while a portable device requiring graphics output and low power should use the Arc Pro B390.