Intel Graphics 24EU Mobile vs NVIDIA Jetson Orin Nano Super Comparison

Head-to-Head Benchmarks

The database does not contain any recorded benchmark scores for either the Intel Graphics 24EU Mobile or the NVIDIA Jetson Orin Nano Super. Both entries show an average benchmark score of 0, and no head-to-head benchmark results are listed. Consequently, there are no exact performance deltas to report from direct measurements. The absence of recorded data means the comparison must rely entirely on the architectural and specification fields captured in the database.

The only measurable performance indicators available are the theoretical throughput figures. The Intel part delivers 384.0 GFLOPS of FP32 compute and 768.0 GFLOPS of FP16 compute. The NVIDIA part delivers 2.089 TFLOPS of FP32 and 4.178 TFLOPS of FP16. Expressed as a ratio, the NVIDIA GPU provides roughly 5.4 times the FP32 throughput and roughly 5.4 times the FP16 throughput of the Intel GPU. This is the largest single numerical gap in the recorded data.

Pixel and texture rates follow the same pattern. The Intel GPU outputs 4.000 GPixel/s and 12.00 GTexel/s. The NVIDIA GPU outputs 16.32 GPixel/s and 32.64 GTexel/s. That places the NVIDIA part at 4.08 times the pixel throughput and 2.72 times the texture throughput of the Intel part. These figures represent peak theoretical rates, not measured application performance, but they establish the relative ceiling for each design.

Memory bandwidth is another decisive differentiator. The Intel GPU relies on System Shared memory with bandwidth described as System Dependent, meaning no fixed value exists in the database. The NVIDIA part specifies 102.4 GB/s from its 8 GB LPDDR5 configuration across a 128 bit bus. Without a concrete number for the Intel side, the database cannot compute a bandwidth ratio, but the presence of a fixed 102.4 GB/s figure versus a dependent, unspecified value indicates the NVIDIA module operates with a defined memory budget while the Intel solution varies by host platform.

Neither part shows any wins in the head-to-head benchmark section, as both winsA and winsB are 0. The percentile ranking for both is 50, placing them at the midpoint of the database's GPU distribution, though this is likely a default value given the absence of actual benchmark submissions.

FAQ

Q: Which GPU has higher FP32 compute?

A: The NVIDIA Jetson Orin Nano Super has 2.089 TFLOPS of FP32, which is approximately 5.4 times the 384.0 GFLOPS of the Intel Graphics 24EU Mobile.

Q: What memory configurations do these GPUs use?

A: The Intel GPU uses System Shared memory with a System Shared type, bus width, and System Dependent bandwidth. The NVIDIA GPU uses 8 GB of LPDDR5 with a 128 bit bus and 102.4 GB/s bandwidth.

Q: Do either of these GPUs include tensor cores?

A: The NVIDIA Jetson Orin Nano Super includes 32 tensor cores. The Intel Graphics 24EU Mobile lists no tensor cores.

Q: What are the power limits of each part?

A: The Intel GPU has a TDP of 6 W. The NVIDIA GPU has a TDP of 25 W.

Q: Which GPU supports a newer DirectX version?

A: The NVIDIA part supports DirectX 12 Ultimate (12_2), while the Intel part supports DirectX 12 (12_1). Both support OpenGL 4.6 and Vulkan 1.4.

Q: What is the physical size of the NVIDIA module?

A: The NVIDIA Jetson Orin Nano Super measures 70 mm in length and 45 mm in height. The Intel GPU is listed as an IGP with no dimensions recorded.

Architecture Differences

The two GPUs come from different manufacturers and use different process nodes. Intel builds the Graphics 24EU Mobile on a 10 nm process at its own foundry. NVIDIA builds the Jetson Orin Nano Super on an 8 nm process at Samsung. The Intel chip is named Twin Lake and belongs to the HD Graphics-T (Twin Lake) generation, while the NVIDIA chip is named GA10B and belongs to the Tegra (Ampere) generation. The underlying architectures are Xe-LP for Intel and Ampere for NVIDIA.

The Intel GPU uses 192 shading units, 12 texture mapping units, and 4 raster output units. The NVIDIA GPU uses 1024 shading units, 32 texture mapping units, and 16 raster output units. The NVIDIA part also includes 32 tensor cores, while the Intel part has none. Neither GPU includes dedicated ray tracing cores. The NVIDIA die is 200 mm², while the Intel die size is listed as unknown. Transistor counts are unknown for both parts.

Memory architecture differs substantially. The Intel GPU shares system memory with no dedicated VRAM, using a Ring Bus interface. The NVIDIA GPU has its own 8 GB LPDDR5 pool on a 128 bit bus with 102.4 GB/s bandwidth, and it connects via PCIe 4.0 x4. Clock behavior also differs: the Intel GPU has a base clock of 300 MHz and a boost clock of 1000 MHz, while the NVIDIA GPU lists no base or boost clock, only a memory clock of 800 MHz with 6.4 Gbps effective data rate.

API support is similar but not identical. Both support OpenGL 4.6 and Vulkan 1.4. DirectX support differs, with Intel at 12 (12_1) and NVIDIA at 12 Ultimate (12_2). Display outputs for both are described as Portable Device Dependent. Power delivery differs as well: the Intel part runs at 6 W, the NVIDIA part at 25 W. The NVIDIA module has defined physical dimensions of 70 mm by 45 mm, while the Intel IGP has no dimensions listed.

The Verdict

The recorded data points to a clear separation in capability. The NVIDIA Jetson Orin Nano Super exceeds the Intel Graphics 24EU Mobile in every quantified specification field: shading units, texture units, raster units, FP32 throughput, FP16 throughput, pixel rate, texture rate, memory bandwidth, and tensor core presence. The NVIDIA part consumes more power, 25 W versus 6 W, and costs a launch MSRP of 249 USD, but its raw compute figures are uniformly higher.

The Intel part has one advantage in the data: lower power consumption. At 6 W, it draws less than a quarter of the NVIDIA module's 25 W budget. It also uses system memory, which removes the need for a dedicated VRAM pool and may simplify integration into a host platform. For workloads that fit within the Intel GPU's modest 384.0 GFLOPS FP32 ceiling, the power efficiency could be attractive.

The NVIDIA part, however, is the only one of the two with tensor cores, a fixed memory subsystem, and a DirectX 12 Ultimate feature set. Its 2.089 TFLOPS FP32 and 4.178 TFLOPS FP16 figures place it in a different performance class. The 102.4 GB/s memory bandwidth is a concrete figure, whereas the Intel part's bandwidth is unspecified and varies by host.

Neither part has recorded benchmark submissions in the database, so the verdict rests on architectural specifications rather than measured application performance. Based on those specifications, the NVIDIA Jetson Orin Nano Super is the higher-performing GPU. The Intel Graphics 24EU Mobile is a lower-power integrated option with no dedicated memory and no tensor acceleration.

Specification Differences

The database records the following fields where the two parts differ:

• Process Node: Intel 10 nm versus NVIDIA 8 nm
• Foundry: Intel versus Samsung
• Die Size: Unknown versus 200 mm²
• Base Clock: 300 MHz versus none listed
• Boost Clock: 1000 MHz versus none listed
• Memory Clock: System Shared versus 800 MHz 6.4 Gbps effective
• Memory Size: System Shared versus 8 GB
• Memory Type: System Shared versus LPDDR5
• Memory Bus Width: System Shared versus 128 bit
• Memory Bandwidth: System Dependent versus 102.4 GB/s
• Shading Units: 192 versus 1024
• TMUs: 12 versus 32
• ROPs: 4 versus 16
• Tensor Cores: None versus 32
• Pixel Rate: 4.000 GPixel/s versus 16.32 GPixel/s
• Texture Rate: 12.00 GTexel/s versus 32.64 GTexel/s
• FP32 Compute: 384.0 GFLOPS versus 2.089 TFLOPS
• FP16 Compute: 768.0 GFLOPS versus 4.178 TFLOPS
• TDP: 6 W versus 25 W
• Bus Interface: Ring Bus versus PCIe 4.0 x4
• DirectX Support: 12 (12_1) versus 12 Ultimate (12_2)
• Dimensions: Not listed versus 70 mm by 45 mm
• Release Date: 2024-12-31 versus 2024-12-16
• Launch MSRP: None listed versus 249 USD

The two parts share several fields: both are listed as IGP slot width, both have Portable Device Dependent display outputs, both support OpenGL 4.6 and Vulkan 1.4, both lack ray tracing cores, both have unknown transistor counts, and both have null predecessors and successors. Both are marked Active in production status.

Where Each One Wins

The Intel Graphics 24EU Mobile wins on power consumption. Its 6 W TDP is the lowest recorded figure among the two, making it suitable for tightly power-constrained embedded or portable designs where the 25 W NVIDIA module would exceed the budget. The Intel part also wins on integration simplicity, as it uses System Shared memory and a Ring Bus interface, eliminating the need for dedicated VRAM allocation. Its lower shading unit count and throughput figures suggest it targets basic display output, lightweight 2D workloads, or simple video tasks rather than compute-heavy applications.

The NVIDIA Jetson Orin Nano Super wins on every raw performance metric in the database. Its 1024 shading units, 32 TMUs, and 16 ROPs provide the structural basis for substantially higher throughput. The 32 tensor cores add a capability the Intel part lacks entirely, which matters for any workload that can use tensor operations. The 8 GB LPDDR5 memory with 102.4 GB/s bandwidth provides a fixed, high-bandwidth memory pool, whereas the Intel part's bandwidth is System Dependent and therefore variable. The DirectX 12 Ultimate support gives it access to a newer graphics feature level than the Intel part's DirectX 12 (12_1).

The FP32 and FP16 figures reinforce the split. The NVIDIA part's 2.089 TFLOPS FP32 and 4.178 TFLOPS FP16 are roughly 5.4 times the Intel part's corresponding 384.0 GFLOPS and 768.0 GFLOPS. Pixel rate and texture rate follow suit, with the NVIDIA part at 16.32 GPixel/s and 32.64 GTexel/s versus the Intel part's 4.000 GPixel/s and 12.00 GTexel/s. The physical size of the NVIDIA module, 70 mm by 45 mm, is recorded, while the Intel IGP has no dimensions, so no size comparison is possible.

The use-case split is straightforward from the data. The Intel part suits low-power integrated graphics roles where the host system provides memory and the workload stays within 384.0 GFLOPS. The NVIDIA part suits compute-oriented tasks that benefit from tensor cores, high memory bandwidth, and the 2.089 TFLOPS FP32 ceiling. The 25 W power draw is the trade-off for that capability.