AMD Instinct MI308X vs Intel UHD Graphics 770 Mobile Comparison

Where Each One Wins

The AMD Instinct MI308X and Intel UHD Graphics 770 Mobile occupy opposite ends of the GPU spectrum, and the recorded data confirms they are not competitors in any practical sense. The Instinct MI308X is a compute accelerator designed for data center workloads, while the UHD Graphics 770 Mobile is an integrated graphics processor embedded in Raptor Lake mobile chips.

In raw compute throughput, the MI308X dominates every measurable category. Its FP32 performance is recorded at 81.72 TFLOPS, which is exactly 100 times the 819.2 GFLOPS delivered by the Intel part. The FP16 comparison is similarly lopsided: the MI308X outputs 81.72 TFLOPS with a 1:1 ratio, while the Intel UHD 770 Mobile reaches 1.638 TFLOPS with a 2:1 ratio. That places the AMD accelerator at roughly 49.9 times the FP16 throughput of the Intel integrated solution.

Texture processing follows the same pattern. The MI308X achieves 2,553.6 GTexel/s, whereas the UHD 770 Mobile manages 25.60 GTexel/s. This is a 99.0% deficit for the Intel part relative to the AMD accelerator. Pixel rate is the only category where the Intel GPU posts a non-zero figure: 12.80 GPixel/s versus 0 MPixel/s for the MI308X, because the AMD accelerator has no display outputs and its ROP count is recorded as zero. The MI308X is not designed to render pixels to a screen; it processes data arrays for compute workloads.

Memory capacity and bandwidth further separate the two. The MI308X carries 192 GB of HBM3 on an 8192-bit bus, delivering 5.32 TB/s of bandwidth. The UHD 770 Mobile uses system shared memory with system dependent bandwidth, so no fixed capacity or bus width is recorded. The AMD part has 19,456 shading units and 1,216 texture mapping units, while the Intel part has 256 shading units and 16 TMUs. The MI308X uses a PCIe 5.0 x16 interface; the Intel IGP connects through a Ring Bus.

The percentile data places both parts at the 50th percentile against all GPUs in the database, but this is a reflection of the database's scoring structure, not an indication of equivalent performance. Neither part has recorded benchmark scores or nearest rivals in the database, which limits direct comparative scoring. The architectural and specification differences are so vast that they define entirely different product categories.

Architecture Differences

The MI308X is built on TSMC's 5 nm process with 153,000 million transistors on a 1017 mm² die. The transistor density is recorded at 150.4M per mm². The chip is named Aqua Vanjaram and uses the CDNA 3.0 architecture from AMD's Instinct generation. It is an OAM module with no display outputs, no power connectors listed, and a suggested PSU of 1150 W. Its TDP is 750 W. The release date is recorded as 2023-12-05, and its predecessor is listed as Radeon Instinct.

The UHD Graphics 770 Mobile uses Intel's Generation 12.2 architecture on a 10 nm process, fabricated by Intel. The chip is Raptor Lake, and the generation field is listed as HD Graphics-M (Raptor Lake). Its TDP is 15 W, which is 50 times lower than the MI308X. The Intel part is an IGP with portable device dependent display outputs. Its release date is 2023-01-03, making it roughly 11 months older than the AMD accelerator. Its successor is listed as Arc Graphics-M. No transistor count, die size, or transistor density is recorded for the Intel part.

The MI308X lists no supported APIs: DirectX, OpenGL, and Vulkan are all recorded as N/A. The Intel UHD 770 Mobile supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. This reflects their intended roles: the MI308X is a compute accelerator with no graphics pipeline, while the Intel part is a full graphics solution for mobile displays.

Memory clocks differ fundamentally. The MI308X runs memory at 1300 MHz with 5.2 Gbps effective data rate, while the Intel part uses system shared memory with no dedicated clock. The AMD part's base clock is 1000 MHz with a boost of 2100 MHz. The Intel part's base clock is 300 MHz with a boost of 1600 MHz. The boost ratio is 2.1x for the AMD part and 5.33x for the Intel part, though the absolute clocks remain far apart.

The production status for the MI308X is not recorded, while the Intel part is listed as Active. The MI308X has no successor listed, and the Intel part lists Arc Graphics-M as its successor. Neither part has a launch MSRP recorded in the database.

The Verdict

The data supports a single unambiguous conclusion: these two products serve completely different markets, and no user would realistically choose between them. The MI308X is a 750 W OAM compute module with 192 GB of HBM3, 81.72 TFLOPS of FP32 throughput, and no display outputs. It is built for server racks and high-performance compute clusters. The UHD Graphics 770 Mobile is a 15 W integrated GPU inside a mobile processor, with 256 shading units, system shared memory, and portable device dependent display outputs. It exists to drive laptop screens and handle basic graphics acceleration.

For compute workloads that demand massive memory bandwidth and FP32 or FP16 throughput, the MI308X is the only viable option in this comparison. Its 5.32 TB/s memory bandwidth and 192 GB capacity provide a data handling envelope that the Intel IGP cannot approach, because the Intel part has no dedicated memory at all. For mobile graphics output, direct rendering, and API support, the UHD Graphics 770 Mobile is the only option, since the MI308X has no display outputs and no graphics API support.

The 50x difference in TDP (750 W versus 15 W) and the 100x difference in FP32 throughput define the product boundaries. The MI308X uses PCIe 5.0 x16 for host connectivity; the Intel part uses a Ring Bus. The MI308X is fabricated on a 5 nm process by TSMC; the Intel part uses Intel's 10 nm process. The MI308X has 19,456 shading units; the Intel part has 256. Any workload that fits the Intel IGP's capabilities would be massively overprovisioned on the MI308X, and any workload that requires the MI308X's memory and compute scale cannot run on the Intel part at all.

FAQ

Q: Which GPU has higher FP32 performance?

A: The AMD Instinct MI308X delivers 81.72 TFLOPS of FP32 throughput, which is exactly 100 times the 819.2 GFLOPS recorded for the Intel UHD Graphics 770 Mobile.

Q: Does the Intel UHD Graphics 770 Mobile support modern graphics APIs?

A: Yes. The database records DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4 support for the Intel part. The AMD Instinct MI308X lists all three APIs as N/A.

Q: How much memory does each GPU have?

A: The AMD Instinct MI308X has 192 GB of HBM3 on an 8192-bit bus with 5.32 TB/s bandwidth. The Intel UHD Graphics 770 Mobile uses system shared memory with system dependent bandwidth, so no dedicated capacity is recorded.

Q: What is the power consumption difference?

A: The AMD Instinct MI308X has a TDP of 750 W, while the Intel UHD Graphics 770 Mobile has a TDP of 15 W. The AMD part also lists a suggested PSU of 1150 W, while the Intel part has no suggested PSU recorded.

Q: Can the AMD Instinct MI308X output video to a display?

A: No. The MI308X records no display outputs and a pixel rate of 0 MPixel/s. The Intel UHD Graphics 770 Mobile lists portable device dependent display outputs and a pixel rate of 12.80 GPixel/s.

Q: Which GPU has more shading units?

A: The AMD Instinct MI308X has 19,456 shading units and 1,216 TMUs. The Intel UHD Graphics 770 Mobile has 256 shading units and 16 TMUs.

Head-to-Head Benchmarks

The database contains no recorded head-to-head benchmark entries for these two parts, and neither has an average benchmark score or nearest rivals listed. The comparison therefore rests entirely on the specification data. The largest single delta appears in FP32 throughput: the MI308X's 81.72 TFLOPS versus the Intel part's 819.2 GFLOPS is a 100x difference. In FP16, the MI308X's 81.72 TFLOPS (1:1) versus 1.638 TFLOPS (2:1) represents a 49.9x gap.

Texture rate shows a 99.0x advantage for the AMD part: 2,553.6 GTexel/s versus 25.60 GTexel/s. Pixel rate is the only metric where the Intel part leads, because the MI308X records 0 MPixel/s while the UHD 770 Mobile records 12.80 GPixel/s. This is not a performance advantage in any practical sense; it reflects the absence of a rasterization pipeline on the compute-focused MI308X.

Memory bandwidth is another decisive gap. The MI308X delivers 5.32 TB/s from 192 GB of HBM3 over an 8192-bit interface. The Intel part uses system shared memory with no recorded fixed bandwidth. The memory clock for the AMD part is 1300 MHz with 5.2 Gbps effective, while the Intel part has no dedicated memory clock.

Clock speeds show the AMD part running a 1000 MHz base and 2100 MHz boost, versus 300 MHz base and 1600 MHz boost for the Intel part. The MI308X has a 2.1x boost multiplier; the Intel part has a 5.33x boost multiplier. Absolute boost clocks still favor the AMD part by 500 MHz.

Shading resources reinforce the compute divide. The MI308X packs 19,456 shading units and 1,216 TMUs, while the Intel part has 256 shading units and 16 TMUs. The ROP count is zero for the AMD part and 8 for the Intel part. The transistor count for the MI308X is 153,000 million on a 1017 mm² die; no transistor data is recorded for the Intel part.

Specification Differences

The two parts differ in nearly every recorded specification field. The MI308X uses a 5 nm TSMC process with 153,000 million transistors and a 1017 mm² die. The Intel part uses Intel's 10 nm process with no transistor, die size, or density data recorded. The MI308X's architecture is CDNA 3.0 from the Instinct (MIx) generation; the Intel part uses Generation 12.2 from the HD Graphics-M (Raptor Lake) generation.

Memory specifications diverge completely. The MI308X has 192 GB of HBM3, an 8192-bit bus, 5.32 TB/s bandwidth, and a 1300 MHz memory clock with 5.2 Gbps effective rate. The Intel part uses system shared memory for size, type, bus width, and bandwidth, with system dependent bandwidth.

Compute unit counts favor the AMD part overwhelmingly: 19,456 shading units versus 256, 1,216 TMUs versus 16, and 0 ROPs versus 8. Pixel rate is 0 MPixel/s for the MI308X and 12.80 GPixel/s for the Intel part. Texture rate is 2,553.6 GTexel/s versus 25.60 GTexel/s. FP32 is 81.72 TFLOPS versus 819.2 GFLOPS, and FP16 is 81.72 TFLOPS (1:1) versus 1.638 TFLOPS (2:1).

Power and physical configuration differ by design intent. The MI308X has a 750 W TDP, is an OAM Module, lists no power connectors, and suggests a 1150 W PSU. The Intel part has a 15 W TDP, is an IGP, lists no power connectors, and has no suggested PSU. The MI308X connects via PCIe 5.0 x16; the Intel part uses a Ring Bus. Display outputs are absent on the MI308X and portable device dependent on the Intel part. API support is N/A for the AMD part and DirectX 12 (12_1), OpenGL 4.6, Vulkan 1.4 for the Intel part. Release dates are 2023-12-05 for the MI308X and 2023-01-03 for the Intel part. The MI308X's predecessor is Radeon Instinct; the Intel part's successor is Arc Graphics-M. Production status is unrecorded for the AMD part and Active for the Intel part.