Intel Arc Pro B390 vs NVIDIA N1 20SM Comparison

The Verdict

The Intel Arc Pro B390 and NVIDIA N1 20SM occupy the same performance percentile in the database, both scoring at the 50th percentile against all GPUs. Neither part has recorded benchmark scores, and the head-to-head benchmark table is empty, so the comparison rests entirely on architectural and specification data. The Intel part delivers 7.680 TFLOPS of FP32 compute, while the NVIDIA part delivers 12.01 TFLOPS, a 56% advantage in raw shader throughput. The NVIDIA N1 20SM also fields 2560 shading units against 1536 for the Intel part, 160 texture mapping units against 48, and 20 ray tracing cores against 12. The Intel Arc Pro B390 counters with a higher boost clock at 2500 MHz versus 2346 MHz, a faster pixel rate at 60.00 GPixel/s versus 56.30 GPixel/s, and a more efficient process node at 3 nm versus 5 nm.

The data indicates the NVIDIA part is the stronger compute device on paper, with higher FP32, more shading units, more texture units, and more ray tracing cores. The Intel part wins in pixel throughput and clock speed, and it carries a full DirectX 12 Ultimate feature set, while the NVIDIA part lists no DirectX, OpenGL, or Vulkan support in the database. The NVIDIA N1 20SM uses a 128 GB LPDDR5X memory pool with 273.2 GB/s bandwidth over a 256-bit bus, whereas the Intel part relies on system shared memory with system dependent bandwidth. The Intel Arc Pro B390 draws 80 W according to the TDP field, while the NVIDIA part has no recorded TDP.

Buyers seeking maximum FP32 compute, large dedicated memory, and extensive ray tracing hardware would favor the NVIDIA N1 20SM. Buyers needing a fixed power envelope, a smaller process node, and conventional graphics API support would favor the Intel Arc Pro B390. The Intel part is an integrated graphics processor with an IGP bus interface, while the NVIDIA part also uses an IGP slot width but connects via PCIe 5.0 x16. The database shows no benchmark wins for either side, so the verdict rests on specification analysis rather than measured performance.

Where Each One Wins

The NVIDIA N1 20SM wins in raw compute density. Its FP32 throughput of 12.01 TFLOPS exceeds the Intel Arc Pro B390's 7.680 TFLOPS by 4.33 TFLOPS, a 56% margin. The shading unit count of 2560 versus 1536 gives the NVIDIA part 67% more shader processors. Texture throughput favors NVIDIA heavily: 375.4 GTexel/s against 120.0 GTexel/s, a 3.1x advantage. The NVIDIA part also has 160 TMUs versus 48, and 80 tensor cores versus none listed for the Intel part. Ray tracing hardware favors NVIDIA with 20 RT cores versus 12.

The Intel Arc Pro B390 wins in pixel fill rate, posting 60.00 GPixel/s against 56.30 GPixel/s for the NVIDIA part, a 6.6% margin. The Intel boost clock of 2500 MHz exceeds the NVIDIA boost of 2346 MHz by 154 MHz. The Intel base clock of 300 MHz is lower than the NVIDIA base of 741 MHz, but the Intel part achieves its higher boost from a 3 nm process node versus 5 nm for NVIDIA. The Intel part lists a TDP of 80 W, giving it a defined power target, while the NVIDIA part has no TDP recorded.

Memory configuration splits the two clearly. The NVIDIA N1 20SM pairs 128 GB of LPDDR5X with a 256-bit bus and 273.2 GB/s bandwidth, while the Intel Arc Pro B390 uses system shared memory with the bus width, size, and bandwidth all marked system dependent. The NVIDIA memory runs at 1067 MHz with 8.5 Gbps effective data rate. The Intel part has no dedicated memory clock because it shares system memory.

API support also separates them. The Intel Arc Pro B390 lists DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The NVIDIA N1 20SM lists N/A for DirectX, OpenGL, and Vulkan. This means the Intel part supports conventional graphics APIs, while the NVIDIA part does not record any such support in the database.

Architecture Differences

The Intel Arc Pro B390 uses the Panther Lake chip with Xe3-LPG architecture, part of the Arc Graphics-WM (Panther Lake) generation. It is fabricated on a 3 nm process at Intel's foundry. The NVIDIA N1 20SM uses the GB20B chip with Blackwell 2.0 architecture, part of the Blackwell IGP (N1x) generation. It is fabricated on a 5 nm process at TSMC. The die size for the NVIDIA part is 382 mm², while the Intel die size is unknown.

The Intel architecture implements 1536 shading units, 48 TMUs, 24 ROPs, and 12 RT cores. It has no tensor cores listed. The NVIDIA architecture implements 2560 shading units, 160 TMUs, 24 ROPs, 20 RT cores, and 80 tensor cores. Both parts have 24 ROPs, so pixel output is close, but the Intel part edges ahead at 60.00 GPixel/s versus 56.30 GPixel/s due to its higher boost clock.

FP16 computation differs. The Intel Arc Pro B390 delivers 15.36 TFLOPS of FP16 with a 2:1 ratio relative to FP32, meaning it doubles its FP32 rate for half-precision work. The NVIDIA N1 20SM delivers 12.01 TFLOPS of FP16 with a 1:1 ratio, meaning its half-precision rate equals its single-precision rate. The Intel part offers higher peak FP16 throughput, but the NVIDIA part maintains consistent throughput across precisions.

The NVIDIA part integrates 80 tensor cores, which are absent from the Intel specification. Tensor cores are specialized for matrix operations, commonly used in AI and machine learning workloads. The Intel part does not list any equivalent hardware. The NVIDIA part also has a larger memory subsystem with 128 GB of LPDDR5X on a 256-bit bus, while the Intel part shares system memory with no dedicated allocation.

Clock behavior differs. The Intel base clock runs at 300 MHz and boosts to 2500 MHz, a wide dynamic range. The NVIDIA base clock runs at 741 MHz and boosts to 2346 MHz, a narrower range. The Intel part has a higher peak clock, which helps its pixel rate, while the NVIDIA part uses a higher base clock and more shader units to achieve its compute advantage.

The bus interface differs: the Intel part connects via IGP, while the NVIDIA part uses PCIe 5.0 x16. The display outputs also differ: the Intel part uses portable device dependent outputs, while the NVIDIA part lists 1x HDMI. The NVIDIA part has a known die size of 382 mm², while the Intel die size is unknown. Both parts have unknown transistor counts and densities.

FAQ

Q: Which GPU has higher FP32 compute?

A: The NVIDIA N1 20SM delivers 12.01 TFLOPS of FP32, which is 56% higher than the Intel Arc Pro B390's 7.680 TFLOPS.

Q: Does the Intel Arc Pro B390 support DirectX?

A: Yes, the Intel part lists DirectX 12 Ultimate (12_2), along with OpenGL 4.6 and Vulkan 1.4. The NVIDIA N1 20SM lists N/A for DirectX, OpenGL, and Vulkan.

Q: How much memory does each GPU use?

A: The NVIDIA N1 20SM has 128 GB of LPDDR5X on a 256-bit bus with 273.2 GB/s bandwidth. The Intel Arc Pro B390 uses system shared memory with system dependent size, type, bus width, and bandwidth.

Q: Which GPU has more texture mapping units?

A: The NVIDIA N1 20SM has 160 TMUs, more than three times the 48 TMUs on the Intel Arc Pro B390. Its texture rate is 375.4 GTexel/s versus 120.0 GTexel/s.

Q: What is the pixel fill rate for each GPU?

A: The Intel Arc Pro B390 achieves 60.00 GPixel/s, while the NVIDIA N1 20SM achieves 56.30 GPixel/s. The Intel part is ahead by 6.6%.

Q: Which GPU has tensor cores?

A: The NVIDIA N1 20SM has 80 tensor cores. The Intel Arc Pro B390 lists no tensor cores in its specification.

Head-to-Head Benchmarks

The database contains no recorded benchmark scores for either part, and the head-to-head benchmark table is empty. Both GPUs have an average benchmark score of 0 and zero wins each. The percentile ranking for both is identical at the 50th percentile against all GPUs. This means no measured performance data exists to compare directly, so the analysis must rely on the recorded specifications.

The largest numerical advantage belongs to the NVIDIA N1 20SM in FP32 compute. At 12.01 TFLOPS versus 7.680 TFLOPS, the NVIDIA part is 4.33 TFLOPS ahead, a 56% improvement. Shader throughput follows the same pattern: 2560 shading units versus 1536, giving NVIDIA a 1024-unit lead. Texture rate shows the widest gap: 375.4 GTexel/s versus 120.0 GTexel/s, a 255.4 GTexel/s difference that translates to a 3.1x advantage for NVIDIA. The TMU count of 160 versus 48 reinforces this dominance in texture-heavy workloads.

Memory bandwidth also favors NVIDIA decisively. The 273.2 GB/s of LPDDR5X bandwidth over a 256-bit bus contrasts with the Intel part's system dependent bandwidth, which has no fixed value in the database. The 128 GB memory capacity on NVIDIA dwarfs the Intel part's shared memory allocation, although the Intel part's memory size is also system dependent. The NVIDIA memory clock of 1067 MHz with 8.5 Gbps effective data rate is a fixed specification, while the Intel memory clock is marked system shared.

The Intel Arc Pro B390 wins in two categories. Pixel rate of 60.00 GPixel/s exceeds the NVIDIA 56.30 GPixel/s by 3.70 GPixel/s. Boost clock of 2500 MHz exceeds the NVIDIA 2346 MHz by 154 MHz. The Intel part also has a higher FP16 throughput at 15.36 TFLOPS versus 12.01 TFLOPS, although this comes with a 2:1 ratio for Intel versus 1:1 for NVIDIA, meaning the NVIDIA part sustains its FP16 rate without the ratio penalty.

The NVIDIA part has 20 RT cores versus 12 for Intel, an 8-core advantage. The NVIDIA part has 80 tensor cores, while the Intel part has none listed. The NVIDIA die size is 382 mm², while the Intel die size is unknown. The process node favors Intel at 3 nm versus 5 nm, which likely contributes to the Intel part's higher boost clock and lower TDP of 80 W, though the NVIDIA TDP is unknown.

Specification Differences

The two GPUs differ across nearly every field in the database. The Intel Arc Pro B390 uses the Panther Lake chip, while the NVIDIA N1 20SM uses the GB20B chip. The architecture differs: Xe3-LPG for Intel versus Blackwell 2.0 for NVIDIA. The generation names also differ: Arc Graphics-WM (Panther Lake) versus Blackwell IGP (N1x). The process node is 3 nm for Intel and 5 nm for NVIDIA, and the foundry is Intel versus TSMC.

Clock speeds differ in both base and boost. The Intel base is 300 MHz, the NVIDIA base is 741 MHz. The Intel boost is 2500 MHz, the NVIDIA boost is 2346 MHz. The Intel memory clock is system shared, while the NVIDIA memory clock is 1067 MHz with 8.5 Gbps effective.

Memory configuration differs completely. The Intel memory size is system shared, while the NVIDIA memory size is 128 GB. The Intel memory type is system shared, the NVIDIA type is LPDDR5X. The Intel bus width is system shared, the NVIDIA bus width is 256 bit. The Intel bandwidth is system dependent, the NVIDIA bandwidth is 273.2 GB/s.

Compute unit counts differ. The Intel shading units number 1536, the NVIDIA shading units number 2560. The Intel TMUs number 48, the NVIDIA TMUs number 160. Both have 24 ROPs. The Intel RT cores number 12, the NVIDIA RT cores number 20. The Intel tensor cores are null, the NVIDIA tensor cores number 80.

Performance rates differ. The Intel pixel rate is 60.00 GPixel/s, the NVIDIA pixel rate is 56.30 GPixel/s. The Intel texture rate is 120.0 GTexel/s, the NVIDIA texture rate is 375.4 GTexel/s. The Intel FP32 is 7.680 TFLOPS, the NVIDIA FP32 is 12.01 TFLOPS. The Intel FP16 is 15.36 TFLOPS with a 2:1 ratio, the NVIDIA FP16 is 12.01 TFLOPS with a 1:1 ratio.

The TDP field shows 80 W for Intel and unknown for NVIDIA. The bus interface is IGP for Intel and PCIe 5.0 x16 for NVIDIA. Display outputs are portable device dependent for Intel and 1x HDMI for NVIDIA. The API support differs: Intel lists DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, while NVIDIA lists N/A for all three. The die size is unknown for Intel and 382 mm² for NVIDIA. The release dates differ: the Intel part released on 2026-01-26, the NVIDIA part on 2026-05-31. The Intel part has a predecessor listed as HD Graphics-WM, while the NVIDIA part has no predecessor. Both parts have no launch MSRP, no successors, and unknown transistor counts.