NVIDIA GeForce GTX 960A vs NVIDIA P106-090 Comparison

Where Each One Wins

The benchmark data paints an unusually clear picture: the NVIDIA P106-090 wins the only head-to-head test recorded in the database, and it wins decisively. In Geekbench OpenCL, the P106-090 scores 21304 against 11998 for the GeForce GTX 960A, a delta of 77.6%. That is not a marginal edge, it is a category-level gap. The P106-090 does not just outperform the GTX 960A in this workload, it nearly doubles the score.

The GeForce GTX 960A has no recorded wins in the head-to-head comparison. The database lists winsA as 1 and winsB as 0. The GTX 960A's only benchmark entry is that same Geekbench OpenCL test, where it falls behind. For any user choosing between these two solely on measured compute performance, the P106-090 is the only option with a victory on the board.

Looking beyond the direct comparison, the percentile data reinforces the split. The P106-090 sits at the 54th percentile of all GPUs in the database, with an average benchmark score of 13470. The GTX 960A sits at the 51st percentile, with an average score of 11998. The percentile gap is modest, but the average score difference of roughly 12% shows that the P106-090 holds a consistent edge across the broader database, not just in one synthetic test.

The use-case split, then, is straightforward. The P106-090 wins where raw compute throughput matters, particularly in OpenCL-accelerated workloads. The GTX 960A, by contrast, has no benchmark category where it leads. Its role is defined by what it is not: it is not faster in the recorded tests, it is not more efficient per the measured scores, and it does not occupy a higher percentile rank. The data does not support any scenario where the GTX 960A is the preferred choice on performance grounds.

Architecture Differences

The two cards come from different NVIDIA architectures and different manufacturing generations. The P106-090 uses the GP106 chip built on the Pascal architecture at TSMC's 16 nm process. The GTX 960A uses the GM107 chip built on the Maxwell architecture at TSMC's 28 nm process. That process node difference is substantial: 16 nm versus 28 nm, which helps explain the transistor density gap. The P106-090 packs 4,400 million transistors into a 200 mm² die, for a density of 22.0M transistors per mm². The GTX 960A contains 1,870 million transistors on a 148 mm² die, for a density of 12.6M per mm².

The compute resources also differ. The P106-090 has 768 shading units, 48 texture mapping units, and 48 ROPs. The GTX 960A has 640 shading units, 40 TMUs, and only 16 ROPs. The ROP discrepancy is particularly large, and it shows up in the pixel rates: the P106-090 delivers 73.49 GPixel/s, while the GTX 960A manages 18.82 GPixel/s. Texture rate follows a similar pattern, 73.49 GTexel/s for the P106-090 versus 47.04 GTexel/s for the GTX 960A, though that gap is smaller.

Clock speeds favor the newer card as well. The P106-090 runs at a 1354 MHz base and 1531 MHz boost. The GTX 960A runs at 1097 MHz base and 1176 MHz boost. Combined with the higher shader count, the P106-090 reaches 2.352 TFLOPS of FP32 throughput, while the GTX 960A tops out at 1.505 TFLOPS. The P106-090 also lists FP16 performance at 36.74 GFLOPS, though at a 1:64 ratio that indicates minimal consumer relevance; the GTX 960A lists no FP16 figure at all.

Memory is another clear separation point. The P106-090 offers 3 GB of GDDR5 on a 192-bit bus, yielding 192.2 GB/s of bandwidth. The GTX 960A offers 2 GB of GDDR5 on a 128-bit bus, yielding 80.19 GB/s. Memory clocks are 2002 MHz (8 Gbps effective) for the P106-090 versus 1253 MHz (5 Gbps effective) for the GTX 960A. The bandwidth advantage for the P106-090 is roughly 2.4x, which matters in memory-bound workloads even if the synthetic tests do not isolate it.

Form factor and connectivity differ sharply. The P106-090 is a dual-slot card, 250 mm or 9.8 inches long, with a 6-pin power connector and a suggested 250 W power supply. It has no display outputs, which is consistent with its mining-oriented generation. The GTX 960A is an MXM module using the MXM-B (3.0) interface, with no power connectors and display outputs described as portable device dependent. The P106-090's bus interface is listed as PCIe 1.0 x1, a limitation that could constrain data transfer in some systems, while the GTX 960A uses the MXM standard.

Both cards support DirectX 12, OpenGL 4.6, and Vulkan 1.4 in the API listings, though the P106-090 lists DirectX 12 (12_1) while the GTX 960A lists DirectX 12 (11_0). The feature level difference, 12_1 versus 11_0, indicates a higher DirectX feature tier on the Pascal card. Both cards are end-of-life. The P106-090 released in 2017, while the GTX 960A released in 2015 and lists the GeForce 800A as its predecessor.

The Verdict

The recorded data is unambiguous. The NVIDIA P106-090 outperforms the NVIDIA GeForce GTX 960A in the only shared benchmark, Geekbench OpenCL, by 77.6%. It also holds a higher average benchmark score, 13470 versus 11998, and a higher percentile rank, 54 versus 51. On every direct performance metric in the database, the P106-090 leads.

The architecture comparison supports the benchmark results. The P106-090 uses a newer 16 nm process, has more transistors, more shading units, more TMUs, triple the ROP count, higher clocks, more memory, a wider bus, and more than double the memory bandwidth. Its FP32 throughput is 2.352 TFLOPS versus 1.505 TFLOPS. Its pixel rate is 73.49 GPixel/s versus 18.82 GPixel/s. There is no category in the recorded data where the GTX 960A closes the gap.

Who should pick the P106-090? Anyone running compute-oriented OpenCL workloads where the benchmark score matters. The card's lack of display outputs means it is not suitable as a primary graphics solution, but the data does not assess that role. For raw measured compute performance, the P106-090 is the clear choice.

Who should pick the GTX 960A? The data offers no performance-based reason. It is a Maxwell-generation MXM module with a lower score, lower percentile, and no recorded wins. Its only potential advantage would come from the MXM form factor, which suits portable or proprietary systems, but the benchmark database does not quantify that. On measured performance, the GTX 960A loses every comparison available.

The verdict, strictly from the data: choose the P106-090 for higher compute scores, higher bandwidth, and a more advanced architecture. Choose the GTX 960A only if the MXM form factor is a hard requirement, since no performance metric in the database favors it.

FAQ

Q: Which GPU is faster in Geekbench OpenCL?

A: The NVIDIA P106-090 scores 21304, while the NVIDIA GeForce GTX 960A scores 11998. The P106-090 leads by 77.6%.

Q: How do the two cards compare in average benchmark score?

A: The P106-090 has an average benchmark score of 13470, while the GTX 960A has an average score of 11998. The P106-090 also ranks at the 54th percentile of all GPUs, versus the 51st percentile for the GTX 960A.

Q: What are the memory specifications of each card?

A: The P106-090 has 3 GB of GDDR5 on a 192-bit bus with 192.2 GB/s bandwidth. The GTX 960A has 2 GB of GDDR5 on a 128-bit bus with 80.19 GB/s bandwidth.

Q: Which card has a higher pixel rate?

A: The P106-090 delivers 73.49 GPixel/s. The GTX 960A delivers 18.82 GPixel/s. The P106-090 is nearly 4x higher.

Q: Do both cards support the same APIs?

A: Both support OpenGL 4.6 and Vulkan 1.4. The P106-090 lists DirectX 12 (12_1), while the GTX 960A lists DirectX 12 (11_0).

Q: What process nodes are used by each GPU?

A: The P106-090 uses TSMC's 16 nm process. The GTX 960A uses TSMC's 28 nm process.

Head-to-Head Benchmarks

The database records exactly one head-to-head benchmark between these two cards: Geekbench OpenCL. The P106-090 scores 21304, and the GTX 960A scores 11998. The delta is 77.6% in favor of the P106-090. That is the single largest measurable separation in the comparison, and it aligns with the architectural gap between the two.

Breaking down why the OpenCL result is so lopsided requires looking at the compute resources. The P106-090 has 768 shading units against 640 for the GTX 960A, a 20% shader advantage. Its boost clock is 1531 MHz versus 1176 MHz, a 30% advantage. Its FP32 throughput is 2.352 TFLOPS versus 1.505 TFLOPS, a 56% advantage. Together, those factors explain most of the 77.6% score gap, with memory bandwidth likely contributing the rest. The P106-090's 192.2 GB/s versus 80.19 GB/s is a 140% advantage, which matters for OpenCL workloads that stream data.

The pixel rate comparison is even more one-sided. The P106-090 hits 73.49 GPixel/s, while the GTX 960A manages 18.82 GPixel/s. The ROP count difference, 48 versus 16, drives this. Texture rate is closer but still favors the P106-090: 73.49 GTexel/s versus 47.04 GTexel/s, a 56% lead that mirrors the FP32 difference.

The GTX 960A's nearest rivals in the database illustrate its standing. It sits near the NVIDIA GeForce GTX 1080, which has an average score of 11960, a delta of 0.3% in the GTX 960A's favor. It also sits near the AMD Radeon RX 6500 XT at 11842, a 1.3% lead, and the NVIDIA GeForce GTX 1660 at 11680, a 2.7% lead. These are not competitive pairings in real-world terms, but they show that the GTX 960A's single benchmark score places it in a cluster of cards that are themselves far below the P106-090's level.

The P106-090's nearest rivals tell a different story. It is nearly tied with the NVIDIA GeForce GTX 570 at 13515, a delta of -0.3%, and the AMD Radeon HD 7770M at 13536, a delta of -0.5%. It edges the AMD Radeon Pro 555 at 13407 by 0.5%. The P106-090's average score of 13470 places it in a midrange cluster, well above the GTX 960A's 11998.

The only head-to-head result, therefore, is not an outlier. It matches the overall average score gap, the architecture differences, and the raw specification comparison. The P106-090 wins the benchmark by a wide margin, and the supporting data shows why: more shaders, higher clocks, more memory bandwidth, and a newer manufacturing process. The GTX 960A, despite sharing the NVIDIA brand and a 75 W TDP, is in a different performance class according to every recorded measurement. The verdict is clear, and no further benchmark data is needed to settle the comparison.