AMD Instinct MI355X vs NVIDIA Jetson T5000 Comparison

Head-to-Head Benchmarks

The recorded database contains no direct head-to-head benchmark results for the AMD Instinct MI355X and the NVIDIA Jetson T5000. Both products have an empty benchmark array, and the head-to-head comparison table is likewise empty. Consequently, there are no measured wins for either side, and the win counts for both products stand at zero. The absence of benchmark data means that raw performance comparisons cannot be derived from our measurements. What the database does provide is a detailed specification profile for each accelerator, and from those specifications, the performance envelope of each product can be inferred, though not verified through direct testing.

The AMD Instinct MI355X is positioned as a high-throughput accelerator with 16,384 shading units, 1,024 texture mapping units, and zero raster operation units. Its FP32 throughput is recorded at 78.64 TFLOPS, with FP16 also at 78.64 TFLOPS on a 1:1 basis. Texture rate reaches 2,457.6 GTexel/s, while pixel rate is recorded as 0 MPixel/s due to the absence of ROPs. The NVIDIA Jetson T5000, by contrast, uses 2,560 shading units, 80 TMUs, and 32 ROPs, with FP32 and FP16 both at 8.064 TFLOPS. Texture rate is 126.0 GTexel/s and pixel rate is 50.40 GPixel/s. The FP32 delta is substantial: the MI355X delivers roughly 9.75 times the FP32 throughput of the Jetson T5000, a ratio that follows directly from the recorded values (78.64 divided by 8.064). In texture throughput, the MI355X leads by a similar magnitude, with 2,457.6 GTexel/s versus 126.0 GTexel/s, a factor of approximately 19.5.

The Jetson T5000 does hold a clear advantage in pixel processing. Its 50.40 GPixel/s pixel rate contrasts with the MI355X's 0 MPixel/s, a consequence of the latter having no ROPs. This is not a meaningful contest in the context of the MI355X's intended role, but it is a measurable difference in the specification sheet. Clock speeds also differ: the Jetson T5000 runs at a base of 1386 MHz and a boost of 1575 MHz, while the MI355X runs at 1000 MHz base and 2400 MHz boost. The MI355X's boost clock is 825 MHz higher, yet its FP32 advantage is driven primarily by the massive difference in shading unit count rather than clock speed alone.

Memory bandwidth is another area of decisive separation. The MI355X uses 288 GB of HBM3e across an 8192-bit bus, yielding 8.19 TB/s. The Jetson T5000 uses 128 GB of LPDDR5X across a 256-bit bus, yielding 273.2 GB/s. The bandwidth ratio is approximately 30 to 1 in favor of the MI355X. Capacity is also lopsided: 288 GB versus 128 GB, a 160 GB difference. These memory figures have direct implications for workloads that are bandwidth-bound or require large resident models.

FAQ

Q: Which product has higher FP32 compute throughput?

A: The AMD Instinct MI355X records 78.64 TFLOPS FP32, while the NVIDIA Jetson T5000 records 8.064 TFLOPS FP32. The MI355X delivers approximately 9.75 times the FP32 throughput.

Q: How do the memory systems compare?

A: The MI355X uses 288 GB of HBM3e with a bandwidth of 8.19 TB/s over an 8192-bit bus. The Jetson T5000 uses 128 GB of LPDDR5X with a bandwidth of 273.2 GB/s over a 256-bit bus. The MI355X has both higher capacity and roughly 30 times the bandwidth.

Q: Does the Jetson T5000 have any specification advantage?

A: Yes. The Jetson T5000 has 32 ROPs and a pixel rate of 50.40 GPixel/s, whereas the MI355X has 0 ROPs and a pixel rate of 0 MPixel/s. The Jetson T5000 also has a higher base clock (1386 MHz versus 1000 MHz).

Q: What are the power requirements of each product?

A: The MI355X has a TDP of 1400 W and a suggested PSU of 1800 W. The Jetson T5000 has a TDP of 120 W and a suggested PSU of 300 W. The MI355X draws substantially more power, consistent with its higher compute and memory specifications.

Q: What is the physical form factor of each product?

A: The MI355X is an OAM module measuring 102 mm in length and 165 mm in width. The Jetson T5000 is an IGP with dimensions of 87 mm length, 100 mm height, and 15 mm width. Neither product has display outputs.

Q: Are there any direct benchmark results available?

A: No. The database lists no benchmark scores for either product. Both have an average benchmark score of 0, and the head-to-head benchmark table is empty. All comparisons must be drawn from recorded specifications.

The Verdict

The recorded data points to two accelerators built for different operating envelopes. The AMD Instinct MI355X is a compute-dense accelerator with 78.64 TFLOPS FP32, 8.19 TB/s of memory bandwidth, and 288 GB of HBM3e. It uses a 3 nm process, has a 2380 mm² die, and is rated at 1400 W TDP. The NVIDIA Jetson T5000 is a low-power Blackwell part with 8.064 TFLOPS FP32, 273.2 GB/s of memory bandwidth, and 128 GB of LPDDR5X. It uses a 5 nm process, has a 391 mm² die, and is rated at 120 W TDP. The Jetson T5000 has a launch MSRP of 2,999 USD.

For workloads that demand maximum FP32 throughput, large memory capacity, and extreme bandwidth, the MI355X is the only choice supported by the data. Its FP32 figure is roughly an order of magnitude higher, its memory bandwidth is about 30 times higher, and its capacity is more than double. For workloads that require pixel processing, the Jetson T5000 is the only option, as the MI355X has no ROPs and a pixel rate of 0 MPixel/s. For power-constrained environments, the Jetson T5000's 120 W TDP and 300 W suggested PSU stand in sharp contrast to the MI355X's 1400 W TDP and 1800 W suggested PSU.

The production status also differs. The Jetson T5000 is marked as Active, with a known successor (Server Rubin) and a predecessor (Server Hopper). The MI355X has no recorded production status and no successor. The release dates are close: the MI355X was released on 2025-06-11, and the Jetson T5000 on 2025-08-26. Neither product exposes display outputs or any graphics API support; both list DirectX, OpenGL, and Vulkan as N/A. The choice between them is therefore dictated by the workload's requirement for compute and memory scale versus power efficiency and pixel processing.

Specification Differences

The two products differ in nearly every recorded specification category. The MI355X uses the MI350 256CU chip, while the Jetson T5000 uses the GB10B chip. The MI355X is built on a 3 nm process at TSMC; the Jetson T5000 is built on a 5 nm process, also at TSMC. Transistor count is listed as 185,000 million for the MI355X and unknown for the Jetson T5000. Die size is 2380 mm² for the MI355X and 391 mm² for the Jetson T5000. Transistor density is 77.7M per mm² for the MI355X and not recorded for the Jetson T5000.

Clock speeds differ across the board. The MI355X has a base clock of 1000 MHz and a boost of 2400 MHz; the Jetson T5000 has a base of 1386 MHz and a boost of 1575 MHz. Memory clocks are 2000 MHz with 8 Gbps effective for the MI355X, and 1067 MHz with 8.5 Gbps effective for the Jetson T5000. Memory capacity, type, bus width, and bandwidth all differ: 288 GB HBM3e on an 8192-bit bus at 8.19 TB/s versus 128 GB LPDDR5X on a 256-bit bus at 273.2 GB/s.

Compute resources differ dramatically. The MI355X has 16,384 shading units, 1,024 TMUs, and 0 ROPs. The Jetson T5000 has 2,560 shading units, 80 TMUs, and 32 ROPs. The Jetson T5000 additionally lists 20 ray tracing cores and 96 tensor cores; the MI355X lists neither. Pixel rate is 0 MPixel/s for the MI355X and 50.40 GPixel/s for the Jetson T5000. Texture rate is 2,457.6 GTexel/s versus 126.0 GTexel/s. FP32 and FP16 are both 78.64 TFLOPS for the MI355X and 8.064 TFLOPS for the Jetson T5000.

Power and physical specifications also diverge. The MI355X has a TDP of 1400 W and a suggested PSU of 1800 W; the Jetson T5000 has a TDP of 120 W and a suggested PSU of 300 W. The MI355X is an OAM module with no power connectors listed; the Jetson T5000 is an IGP with no power connectors. The MI355X measures 102 mm by 165 mm; the Jetson T5000 measures 87 mm by 100 mm by 15 mm. The bus interface differs: PCIe 5.0 x16 for the MI355X and PCIe 5.0 x8 for the Jetson T5000. Neither has display outputs, and both have N/A API support.

Architecture Differences

The architectural separation is defined by the compute paradigm each product employs. The MI355X uses CDNA 4.0, AMD's compute-focused architecture, built for dense FP32 and FP16 throughput with a 1:1 ratio. Its 16,384 shading units and 1,024 TMUs are organized around a large 2380 mm² die with 185,000 million transistors. The Jetson T5000 uses Blackwell, NVIDIA's architecture for server and edge parts, with 2,560 shading units, 80 TMUs, 32 ROPs, 20 ray tracing cores, and 96 tensor cores. The presence of tensor cores and ray tracing cores on the Jetson T5000 indicates support for features that the MI355X does not list in the database.

The memory architecture is fundamentally different. The MI355X relies on HBM3e with a 8192-bit bus, which explains its 8.19 TB/s bandwidth. The Jetson T5000 uses LPDDR5X on a 256-bit bus, delivering 273.2 GB/s. The MI355X's memory subsystem is designed for large-scale data movement, while the Jetson T5000's is suited to lower-power, compact deployments. The MI355X has no ROPs, which aligns with its role as a compute accelerator rather than a graphics device. The Jetson T5000, despite being an IGP, includes 32 ROPs and a pixel rate of 50.40 GPixel/s.

The process technology also differs. The MI355X is fabricated on a 3 nm node, while the Jetson T5000 uses a 5 nm node, both at TSMC. The MI355X's smaller node and massive die allow for 185,000 million transistors, whereas the Jetson T5000's transistor count is unknown. The MI355X has a transistor density of 77.7M per mm²; no density figure is recorded for the Jetson T5000. The MI355X belongs to the Instinct (MIx) generation with a predecessor of Radeon Instinct. The Jetson T5000 belongs to the Server Blackwell (Bxx) generation, with a predecessor of Server Hopper and a successor of Server Rubin. The MI355X has no recorded successor.

The power architecture is another differentiator. The MI355X is rated at 1400 W TDP with a suggested PSU of 1800 W, reflecting its high-throughput design. The Jetson T5000 is rated at 120 W TDP with a suggested PSU of 300 W, reflecting its low-power design. The MI355X is an OAM module, while the Jetson T5000 is an IGP. Both use PCIe 5.0, but with different lane counts: x16 for the MI355X and x8 for the Jetson T5000. Neither product supports display outputs or graphics APIs, so both are strictly compute-focused in the database's recorded feature set.