AMD Instinct MI455X vs AMD Radeon 890M Comparison

AMD
RADEON

AMD Instinct MI455X

CORE STATE MI450 256CU
VRAM 432 GB
CLOCK SPEED 2400 MHz
TDP 2300 W
BUS WIDTH 24576 bit
ARCHITECTURE CDNA 5.0
nm
PROCESS 2 nm
LAUNCH DATE 2026
VS
AMD
RADEON

Radeon 890M

CORE STATE Strix Point
VRAM System Shared
CLOCK SPEED 2900 MHz
TDP 15 W
BUS WIDTH System Shared
ARCHITECTURE RDNA 3.5
nm
PROCESS 4 nm
LAUNCH DATE 2024

PERFORMANCE BENCHMARKS

3dmark_3dmark_steel_nomad_dx12
N/A
590
geekbench_opencl
N/A
37,254
geekbench_vulkan
N/A
40,808
passmark_directx_10
N/A
33
passmark_directx_11
N/A
73
passmark_directx_12
N/A
37
passmark_directx_9
N/A
97
passmark_g2d
N/A
979
passmark_g3d
N/A
8,076
passmark_gpu_compute
N/A
4,157

Analysis: AMD Instinct MI455X vs AMD Radeon 890M

Where Each One Wins

The recorded data presents a stark asymmetry between these two AMD parts. The AMD Instinct MI455X is a compute accelerator with no benchmark entries, while the AMD Radeon 890M, an integrated graphics processor, has ten recorded benchmark results. Consequently, the 890M wins every measurable benchmark category by default, but the nature of those wins matters more than the simple count.

The 890M delivers its strongest relative performance in compute-oriented workloads. Its Geekbench OpenCL score of 37,254 and Vulkan score of 40,808 indicate that the shader array, despite having only 1,024 shading units, is capable of substantial parallel throughput. The PassMark GPU Compute score of 4,157 reinforces this pattern. In contrast, the 890M's DirectX 10, 11, and 12 scores (33, 73, and 37 respectively) are far lower, which suggests that the part is optimized for modern compute APIs and Vulkan rather than legacy DirectX paths.

The MI455X, with no recorded benchmarks, cannot claim any measured victory. The database lists its percentile versus all GPUs at 50 and its average benchmark score at 0, meaning it has not been exercised by the standard test suite. The 890M holds a 45th percentile ranking with an average score of 9,210. The MI455X's wins, if any, would have to come from architectural capabilities that the current benchmark suite does not measure, such as its 157.3 TFLOPS FP32 rate, which is a theoretical specification rather than a tested result.

For users requiring immediate, verified performance data, the 890M is the only option with evidence. Its nearest rivals in the database are the AMD Radeon Vega 8 (average score 9,221, delta -0.1%), the NVIDIA GeForce GTX 960 (9,273, delta -0.7%), the NVIDIA GeForce GTX 465 (9,294, delta -0.9%), and the NVIDIA GeForce GTX 850M (9,302, delta -1%). The 890M trails each of these by less than a full percentage point, placing it in a tightly contested band of mid-range legacy discrete GPUs. The MI455X has no nearest rivals listed, so no comparative positioning is possible.

Architecture Differences

The architectural gulf between the two chips is fundamental. The MI455X uses CDNA 5.0 architecture, built for data-center compute, while the 890M uses RDNA 3.5, designed for graphics rendering in mobile integrated systems. The MI455X is fabricated on a 2 nm process at TSMC, whereas the 890M uses a 4 nm process at the same foundry. The MI455X packs 320,000 million transistors on a 2,990 mm² die, giving a density of 107.0 million transistors per mm². The 890M contains 34,000 million transistors on a 233 mm² die, achieving a higher density of 145.9 million per mm². Despite the MI455X's sheer scale, the 890M's smaller design is more transistor-dense.

The MI455X is built around the MI450 256CU chip, a 256 compute unit configuration that yields 32,768 shading units, 1,024 texture mapping units, and zero ROPs. It has no ray tracing cores and no tensor cores listed. The 890M, based on the Strix Point chip, features 1,024 shading units, 64 TMUs, 32 ROPs, and 16 ray tracing cores. The MI455X cannot rasterize at all, with a pixel rate of 0 MPixel/s, while the 890M renders at 92.80 GPixel/s. The texture rates tell a similar story: the MI455X reaches 2,457.6 GTexel/s versus the 890M's 185.6 GTexel/s.

Memory configurations could not be more different. The MI455X carries 432 GB of HBM4 memory across a 24,576-bit bus, delivering 23.3 TB/s of bandwidth. The 890M uses system-shared memory, with its bus width and bandwidth marked as system-dependent. The MI455X's memory clock is 1900 MHz (7.6 Gbps effective), while the 890M's memory clock is simply listed as system shared. The MI455X draws 2,300 W with a suggested PSU of 2,700 W; the 890M operates at 15 W. The MI455X is an EAM module with no power connectors and no display outputs, while the 890M is an IGP with portable-device-dependent display outputs.

API support splits them cleanly. The MI455X has no DirectX, OpenGL, or Vulkan support listed. The 890M supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The MI455X uses a PCIe 6.0 x16 interface; the 890M uses PCIe 4.0 x8. Release dates differ by two years: the 890M launched on 2024-07-14, and the MI455X on 2026-07-22. The 890M is marked as active in production; the MI455X has no production status recorded.

Head-to-Head Benchmarks

Because the MI455X has no benchmark scores in the database, every head-to-head comparison defaults to the 890M. The absence of MI455X data is itself a finding: the standard suite produces zero results for that accelerator, so no direct numerical comparison is possible.

The 890M's strongest recorded result is its Geekbench Vulkan score of 40,808, which exceeds its OpenCL score of 37,254. This gap of 3,554 points indicates that the RDNA 3.5 architecture responds well to Vulkan's explicit graphics and compute model. The PassMark suite shows a different profile: G3D score of 8,076, G2D score of 979, and GPU Compute score of 4,157. The legacy DirectX tests are uniformly low, with DirectX 9 at 97, DirectX 11 at 73, DirectX 12 at 37, and DirectX 10 at 33. These scores cluster far below the compute results, suggesting the 890M's drivers and hardware favor modern APIs.

The MI455X's theoretical FP32 throughput of 157.3 TFLOPS is roughly 26.5 times the 890M's 5.939 TFLOPS, but this is a specification, not a benchmark. The database records no measured performance for the MI455X, so any claim about its real-world speed must rely on those theoretical figures. The 890M's FP32 rate of 5.939 TFLOPS matches its FP16 rate at a 1:1 ratio, a pattern shared by the MI455X, which also lists 157.3 TFLOPS for both FP32 and FP16.

The 890M's nearest rival comparisons show a tight cluster. It sits 0.1% behind the AMD Radeon Vega 8, 0.7% behind the GTX 960, 0.9% behind the GTX 465, and 1% behind the GTX 850M. These deltas are small enough that run-to-run variance could reorder the standings. The MI455X has no rivals listed, so it cannot be placed relative to any other accelerator in the database.

FAQ

Q: Does the AMD Instinct MI455X have any benchmark scores in the database?

A: No. The MI455X has an empty benchmarks array, an average benchmark score of 0, and a 50th percentile ranking. The 890M has ten recorded scores with an average of 9,210.

Q: Which GPU has more shading units?

A: The MI455X has 32,768 shading units, while the 890M has 1,024. The MI455X also has 1,024 TMUs versus the 890M's 64, and 32,768 versus 1,024 in compute units (256 versus an implied smaller count based on the 890M's chip).

Q: What is the memory setup of each part?

A: The MI455X uses 432 GB of HBM4 on a 24,576-bit bus with 23.3 TB/s bandwidth. The 890M uses system shared memory with system-dependent bandwidth and bus width.

Q: Can the MI455X output video?

A: No. The MI455X has no display outputs. The 890M has display outputs that are portable device dependent.

Q: What APIs does each support?

A: The MI455X lists no DirectX, OpenGL, or Vulkan support. The 890M supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: How does the 890M compare to its nearest rivals?

A: The 890M trails the AMD Radeon Vega 8 by 0.1%, the NVIDIA GeForce GTX 960 by 0.7%, the NVIDIA GeForce GTX 465 by 0.9%, and the NVIDIA GeForce GTX 850M by 1% in average benchmark score.

The Verdict

The data supports only one choice for anyone needing verified performance: the AMD Radeon 890M. It has ten benchmark scores, an average of 9,210, and a 45th percentile ranking. Its nearest rivals are all within a single percentage point, so it delivers competitive integrated graphics performance. The AMD Instinct MI455X has no measured results, so it cannot be recommended on the basis of database evidence. Its theoretical specifications, including 157.3 TFLOPS FP32 and 23.3 TB/s bandwidth, suggest enormous compute capability, but the absence of any benchmark data means those numbers remain unverified.

The 890M suits systems that need graphics output, API support, and low power draw (15 W). The MI455X suits workloads that can use its 2,300 W envelope, 432 GB memory pool, and PCIe 6.0 interface, but only if its unmeasured performance meets expectations. The database cannot confirm that. For a buyer choosing between these two, the 890M is the only part with recorded evidence of functioning under the standard test suite.

Specification Differences

| Field | AMD Instinct MI455X | AMD Radeon 890M |

|---|---|---|

| Architecture | CDNA 5.0 | RDNA 3.5 |

| Process node | 2 nm | 4 nm |

| Transistors | 320,000 million | 34,000 million |

| Die size | 2990 mm² | 233 mm² |

| Transistor density | 107.0M / mm² | 145.9M / mm² |

| Base clock | 1000 MHz | 400 MHz |

| Boost clock | 2400 MHz | 2900 MHz |

| Memory clock | 1900 MHz, 7.6 Gbps effective | System Shared |

| Memory size | 432 GB | System Shared |

| Memory type | HBM4 | System Shared |

| Memory bus width | 24576 bit | System Shared |

| Memory bandwidth | 23.3 TB/s | System Dependent |

| Shading units | 32768 | 1024 |

| TMUs | 1024 | 64 |

| ROPs | 0 | 32 |

| RT cores | None listed | 16 |

| Pixel rate | 0 MPixel/s | 92.80 GPixel/s |

| Texture rate | 2,457.6 GTexel/s | 185.6 GTexel/s |

| FP32 | 157.3 TFLOPS | 5.939 TFLOPS |

| FP16 | 157.3 TFLOPS (1:1) | 5.939 TFLOPS (1:1) |

| TDP | 2300 W | 15 W |

| Slot width | EAM Module | IGP |

| Power connectors | None | None |

| Suggested PSU | 2700 W | None listed |

| Bus interface | PCIe 6.0 x16 | PCIe 4.0 x8 |

| Display outputs | No outputs | Portable Device Dependent |

| DirectX | N/A | 12 Ultimate (12_2) |

| OpenGL | N/A | 4.6 |

| Vulkan | N/A | 1.4 |

| Release date | 2026-07-22 | 2024-07-14 |

| Production status | None listed | Active |

DETAILED SPECIFICATIONS

SPECIFICATION
Instinct MI455X
890M
Core Specs
Shading Units
32,768
1,024 -96.9%
Shaders
32,768
1,024 -96.9%
TMUs
1,024
64 -93.8%
ROPs
0
32 +∞%
Compute Units
256
16 -93.8%
Clocks
Base Clock
1000 MHz
400 MHz
Boost Clock
2400 MHz
2900 MHz
Memory Clock
1900 MHz 7.6 Gbps effective
System Shared
Memory
Memory Size
432 GB
System Shared
VRAM (MB)
442,368
Memory Type
HBM4
System Shared
Memory Bus
24576 bit
System Shared
Bandwidth
23.3 TB/s
System Dependent
Cache
L1 Cache
32 KB (per CU)
128 KB per Array
L2 Cache
192 MB
2 MB
L0 Cache
32 KB per WGP
Performance
Pixel Rate
0 MPixel/s
92.80 GPixel/s
Texture Rate
2,457.6 GTexel/s
185.6 GTexel/s
FP32 (TFLOPS)
157.3 TFLOPS
5.939 TFLOPS
FP64 (TFLOPS)
2.458 TFLOPS (1:64)
371.2 GFLOPS (1:16)
FP16 (TFLOPS)
157.3 TFLOPS (1:1)
5.939 TFLOPS (1:1)
AI/RT
RT Cores
16
Matrix Cores
1,024
Power
TDP
2300 W
15 W
TDP (W)
2,300
15 -99.3%
Suggested PSU
2700 W
Power Connectors
None
None
Architecture
Architecture
CDNA 5.0
RDNA 3.5
GPU Name
MI450 256CU
Strix Point
Generation
Instinct (MIx)
Navi III IGP (Strix Point Mobile)
Process Size
2 nm
4 nm
Transistors
320,000 million
34,000 million
Die Size
2990 mm²
233 mm²
Foundry
TSMC
TSMC
Density
107.0M / mm²
145.9M / mm²
API Support
DirectX
12 Ultimate (12_2)
OpenGL
4.6
Vulkan
1.4
OpenCL
3.0
2.1
Shader Model
6.8
Physical
Slot Width
EAM Module
IGP
Outputs
No outputs
Portable Device Dependent
Bus Interface
PCIe 6.0 x16
PCIe 4.0 x8
Other
Production
Active
Predecessor
Radeon Instinct
Navi II IGP
View Instinct MI455X Details View Radeon 890M Details