AMD Instinct MI300 vs Intel UHD Graphics 710 Mobile Comparison

FAQ

Q: What are the core architectural identities of the AMD Instinct MI300 and the Intel UHD Graphics 710 Mobile?

A: The AMD Instinct MI300 uses the CDNA 3.0 architecture with the Aqua Vanjaram chip, fabricated on a 5 nm process by TSMC. The Intel UHD Graphics 710 Mobile uses the Generation 12.2 architecture with a Raptor Lake chip, fabricated on a 10 nm process by Intel.

Q: How do the memory subsystems compare?

A: The AMD Instinct MI300 has 128 GB of HBM3 memory on an 8192-bit bus, delivering 5.32 TB/s of bandwidth. The Intel UHD Graphics 710 Mobile uses System Shared memory, with a System Shared bus width and System Dependent bandwidth.

Q: What is the difference in shading unit counts?

A: The AMD Instinct MI300 has 14,080 shading units, 880 texture mapping units, and 0 ROPs. The Intel UHD Graphics 710 Mobile has 128 shading units, 8 texture mapping units, and 4 ROPs.

Q: What are the thermal design power (TDP) figures?

A: The AMD Instinct MI300 has a TDP of 600 W and requires a suggested power supply of 1000 W. The Intel UHD Graphics 710 Mobile has a TDP of 15 W.

Q: What display outputs does each part provide?

A: The AMD Instinct MI300 provides no display outputs. The Intel UHD Graphics 710 Mobile provides outputs that are Portable Device Dependent.

Q: What is the release date for both products?

A: Both the AMD Instinct MI300 and the Intel UHD Graphics 710 Mobile have a release date of 2023-01-03.

Architecture Differences

The AMD Instinct MI300 and the Intel UHD Graphics 710 Mobile represent opposite poles of the GPU design spectrum. The MI300 is built on a 5 nm process at TSMC with 153,000 million transistors on a 1017 mm² die, yielding a transistor density of 150.4M per mm². The Intel part uses a 10 nm process at Intel with no transistor count or die size recorded in the database.

The MI300 uses the CDNA 3.0 architecture, which is designed for compute acceleration. It has 14,080 shading units and 880 TMUs, but critically it has 0 ROPs and a pixel rate of 0 MPixel/s. This indicates a pure compute processor with no rasterization pipeline. Its texture rate is 1,496.0 GTexel/s. The Intel UHD Graphics 710 Mobile uses the Generation 12.2 architecture, a conventional graphics design with 128 shading units, 8 TMUs, and 4 ROPs. Its pixel rate is 4.800 GPixel/s and its texture rate is 9.600 GTexel/s.

The MI300’s FP32 throughput is 47.87 TFLOPS, with FP16 also at 47.87 TFLOPS at a 1:1 ratio. The Intel part delivers 307.2 GFLOPS FP32 and 614.4 GFLOPS FP16 at a 2:1 ratio. The MI300’s FP16 capability equals its FP32, while the Intel part doubles its FP16 rate relative to FP32.

The MI300’s memory is 128 GB of HBM3 on an 8192-bit bus, with a memory clock of 1300 MHz and 5.2 Gbps effective, producing 5.32 TB/s of bandwidth. The Intel part uses System Shared memory with System Dependent bandwidth, meaning its performance is tied to the host system’s memory configuration. The MI300 uses a PCIe 5.0 x16 bus interface, while the Intel part uses a Ring Bus. The MI300 has no API support listed for DirectX, OpenGL, or Vulkan, while the Intel part supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4.

The MI300 uses 2x 8-pin power connectors, has a length of 267 mm (10.5 inches) and a height of 111 mm (4.4 inches), and is an add-in card. The Intel part is an IGP with no dimensions recorded and no power connectors.

Head-to-Head Benchmarks

The database records no head-to-head benchmark results between these two products, and both have zero wins in the winsA and winsB fields. Their average benchmark scores are both 0, and both sit at the 50th percentile versus all GPUs. This is unsurprising given the fundamental difference in their roles: the MI300 is a 600 W compute accelerator with no display outputs, while the Intel UHD Graphics 710 Mobile is a 15 W integrated graphics processor for portable devices.

The compute gap is enormous. The MI300’s FP32 throughput of 47.87 TFLOPS is 156 times the Intel part’s 307.2 GFLOPS. Its texture rate of 1,496.0 GTexel/s is about 156 times the Intel part’s 9.600 GTexel/s. The MI300’s memory bandwidth of 5.32 TB/s is not directly comparable to the Intel part’s System Dependent bandwidth, but the MI300’s 128 GB of dedicated HBM3 memory contrasts sharply with the Intel part’s System Shared allocation.

The Intel part does have advantages in specific areas. It has 4 ROPs versus the MI300’s 0, and a pixel rate of 4.800 GPixel/s versus the MI300’s 0 MPixel/s. It also supports the DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4 APIs, whereas the MI300 lists no graphics API support. The Intel part is also a functional display output device, while the MI300 has no outputs.

In terms of power efficiency, the Intel part operates at 15 W, which is 40 times lower than the MI300’s 600 W. The MI300 requires a 1000 W suggested power supply, while the Intel part has no such requirement. The Intel part is also physically integrated into the host processor via a Ring Bus, while the MI300 is a discrete card measuring 267 mm in length.

The data shows that these products are not competitors in any meaningful sense. Their performance profiles are too divergent, and their target platforms are entirely different. The MI300 is a data center accelerator, while the Intel part is an entry-level integrated GPU for mobile systems.

Specification Differences

The two products differ in nearly every recorded specification field. The MI300 has a process node of 5 nm, while the Intel part uses 10 nm. The MI300 has 153,000 million transistors on a 1017 mm² die, while the Intel part has no recorded transistor count or die size. The transistor density for the MI300 is 150.4M per mm², with no equivalent for the Intel part.

Clock speeds differ substantially. The MI300 has a base clock of 1000 MHz and a boost clock of 1700 MHz, with a memory clock of 1300 MHz (5.2 Gbps effective). The Intel part has a base clock of 300 MHz and a boost clock of 1200 MHz, with System Shared memory clock.

Memory configurations are fundamentally different. The MI300 has 128 GB of HBM3 on an 8192-bit bus with 5.32 TB/s bandwidth. The Intel part has System Shared memory, type, and bus width, with System Dependent bandwidth.

Compute unit counts differ: the MI300 has 14,080 shading units, 880 TMUs, and 0 ROPs, while the Intel part has 128 shading units, 8 TMUs, and 4 ROPs. Pixel rates are 0 MPixel/s for the MI300 and 4.800 GPixel/s for the Intel part. Texture rates are 1,496.0 GTexel/s versus 9.600 GTexel/s.

FP32 performance is 47.87 TFLOPS for the MI300 and 307.2 GFLOPS for the Intel part. FP16 performance is 47.87 TFLOPS (1:1) for the MI300 and 614.4 GFLOPS (2:1) for the Intel part.

TDP is 600 W for the MI300 and 15 W for the Intel part. The MI300 uses 2x 8-pin power connectors and has a suggested PSU of 1000 W; the Intel part has no power connectors and no suggested PSU. The MI300 uses PCIe 5.0 x16, while the Intel part uses a Ring Bus. The MI300 has no display outputs; the Intel part has Portable Device Dependent outputs.

API support is absent for the MI300 across DirectX, OpenGL, and Vulkan, while the Intel part supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The MI300 has dimensions of 267 mm length and 111 mm height; the Intel part has no recorded dimensions. The MI300’s production status is not recorded, while the Intel part is Active. The MI300’s predecessor is Radeon Instinct, while the Intel part’s successor is Arc Graphics-M.

The Verdict

The data indicates that the AMD Instinct MI300 is a high-power compute accelerator with no graphics output capability, while the Intel UHD Graphics 710 Mobile is a low-power integrated GPU for portable devices. The MI300 delivers 47.87 TFLOPS of FP32 compute and 5.32 TB/s of memory bandwidth, making it suitable for workloads that demand massive parallel throughput. The Intel part delivers 307.2 GFLOPS of FP32 compute and relies on System Shared memory, making it suitable for basic display output and light graphics tasks.

The MI300’s lack of ROPs and display outputs confirms it is not intended for rendering to a screen. Its 0 MPixel/s pixel rate and absence of DirectX, OpenGL, and Vulkan support reinforce this. The Intel part’s 4 ROPs, 4.800 GPixel/s pixel rate, and full API support confirm its role as a conventional graphics processor.

The MI300 requires a 600 W TDP and a 1000 W suggested power supply, while the Intel part operates at 15 W with no external power requirement. This makes the Intel part appropriate for battery-powered mobile devices, while the MI300 is appropriate for systems with substantial power delivery infrastructure.

Both products share a release date of 2023-01-03, but their architectures, memory systems, and power profiles place them in entirely different market segments. The MI300 is a compute-first accelerator, and the Intel part is a display-oriented integrated GPU.

Where Each One Wins

The AMD Instinct MI300 wins in raw compute throughput. Its FP32 performance of 47.87 TFLOPS and FP16 performance of 47.87 TFLOPS are orders of magnitude higher than the Intel part’s 307.2 GFLOPS and 614.4 GFLOPS. Its texture rate of 1,496.0 GTexel/s is far beyond the Intel part’s 9.600 GTexel/s. Its 128 GB of HBM3 memory with 5.32 TB/s bandwidth provides a memory subsystem that the Intel part cannot match with System Shared memory.

The Intel UHD Graphics 710 Mobile wins in graphics output capability. It has 4 ROPs and a pixel rate of 4.800 GPixel/s, while the MI300 has 0 ROPs and a 0 MPixel/s pixel rate. The Intel part supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4, while the MI300 lists no graphics APIs. The Intel part also provides display outputs, albeit Portable Device Dependent, while the MI300 has no outputs.

The Intel part wins in power efficiency. Its 15 W TDP is dramatically lower than the MI300’s 600 W, and it requires no power connectors or suggested PSU. The MI300 needs 2x 8-pin power connectors and a 1000 W suggested power supply.

The Intel part wins in integration. It uses a Ring Bus interface and is an IGP, meaning it is built into the host processor with no separate card dimensions. The MI300 is a discrete card measuring 267 mm in length and 111 mm in height, using a PCIe 5.0 x16 connection.

The MI300 wins in memory capacity and bandwidth. Its 128 GB of HBM3 on an 8192-bit bus is a dedicated, high-bandwidth pool, whereas the Intel part’s System Shared memory is dependent on the host system’s memory controller and capacity.

The Intel part wins in production status, recorded as Active, while the MI300’s production status is not recorded in the database. The Intel part also has a recorded successor, Arc Graphics-M, while the MI300 has no successor listed.