NVIDIA GeForce RTX 5090 vs NVIDIA Tesla P100 PCIe 16 GB Comparison

The NVIDIA GeForce RTX 5090 and the NVIDIA Tesla P100 PCIe 16 GB occupy opposite ends of the GPU lifecycle, yet their aggregate benchmark scores place them in adjacent percentile rankings. The RTX 5090, built on the Blackwell 2.0 architecture, posts an average benchmark score of 79,842, while the Pascal-based Tesla P100 achieves 79,605. The delta between them is a razor-thin 0.3%, placing the RTX 5090 marginally ahead. This near-parity in overall standing, however, masks a dramatic divergence in raw compute performance, as the single head-to-head benchmark available demonstrates a 320% victory for the newer card.

Head-to-Head Benchmarks

The only direct comparison in the data is the Geekbench OpenCL test, and the result is decisively one-sided. The NVIDIA GeForce RTX 5090 scores 334,370 points, while the NVIDIA Tesla P100 PCIe 16 GB scores 79,605 points. This represents a 320% delta in favor of the RTX 5090, meaning the newer card delivers more than four times the compute throughput in this workload. The data shows a complete sweep for the RTX 5090, with one win and zero losses in the head-to-head comparison. While the average benchmark scores across all tests suggest the two cards are close rivals, that aggregate figure is heavily influenced by the Tesla P100 having only a single benchmark entry. The Geekbench OpenCL result is the only point of direct comparison, and it reveals a generational chasm in execution capability.

The RTX 5090’s dominance in this test is consistent with its internal specifications. Its FP32 throughput is listed at 104.8 TFLOPS, a figure that dwarfs the Tesla P100’s 9.526 TFLOPS. Similarly, the texture rate of 1,636.8 GTexel/s on the RTX 5090 is over five times higher than the 297.7 GTexel/s delivered by the Tesla P100. The pixel rate also favors the RTX 5090 at 423.6 GPixel/s versus 127.6 GPixel/s. These raw throughput advantages translate directly into the observed benchmark margin, confirming that the RTX 5090 is not merely incrementally faster but categorically superior in compute-intensive tasks.

FAQ

Q: How does the average benchmark score of the RTX 5090 compare to the Tesla P100?

A: The RTX 5090 holds a slight edge with an average benchmark score of 79,842, compared to the Tesla P100’s 79,605. This is a 0.3% difference, placing the RTX 5090 ahead in the aggregate ranking, though both cards share the same 92nd percentile among all GPUs.

Q: Which card wins in the Geekbench OpenCL test, and by how much?

A: The RTX 5090 wins decisively with a score of 334,370 versus the Tesla P100’s 79,605. The delta is 320%, indicating the RTX 5090 delivers over four times the performance in this specific OpenCL workload.

Q: What are the competing cards around these two GPUs in the benchmark database?

A: For the RTX 5090, the nearest rivals include the Tesla P100 PCIe 16 GB (0.3% behind), the Tesla P100 PCIe 12 GB (0.6% behind), and the AMD Radeon RX 6850M XT (1.1% behind). The AMD Radeon Pro Vega 64X sits 1.4% ahead. The Tesla P100 has a similar field, with the RTX 5090 0.3% ahead, the P100 12 GB 0.3% behind, and the RX 6850M XT 0.8% behind.

Q: Does the Tesla P100 have any benchmark entries beyond Geekbench OpenCL?

A: No, the data lists only one benchmark score for the Tesla P100: the Geekbench OpenCL result of 79,605. In contrast, the RTX 5090 has ten benchmark entries, including scores for 3DMark Steel Nomad, PassMark DirectX tests, and Geekbench Vulkan, contributing to its higher average.

Q: How do the memory subsystems differ between the two cards?

A: The RTX 5090 uses 32 GB of GDDR7 memory on a 512-bit bus, providing 1.79 TB/s of bandwidth. The Tesla P100 uses 16 GB of HBM2 memory on a 4096-bit bus, with a bandwidth of 732.2 GB/s. The RTX 5090’s bandwidth is more than double that of the Tesla P100.

Q: Which card has a higher boost clock speed?

A: The RTX 5090 boosts to 2407 MHz, while the Tesla P100 boosts to 1329 MHz. The base clocks are 2017 MHz for the RTX 5090 and 1190 MHz for the Tesla P100, reflecting the newer process node and architecture.

Architecture Differences

The architectural gap between these two GPUs spans two distinct generations of NVIDIA design. The RTX 5090 is built on the Blackwell 2.0 architecture using the GB202 chip, fabricated on a 5 nm process at TSMC. The Tesla P100, by contrast, uses the Pascal architecture with the GP100 chip, manufactured on a 16 nm process, also at TSMC. This process shrink from 16 nm to 5 nm enables the RTX 5090 to pack 92,200 million transistors onto a 750 mm² die, resulting in a transistor density of 122.9M per mm². The Tesla P100 contains 15,300 million transistors on a 610 mm² die, giving it a density of just 25.1M per mm². The RTX 5090 thus achieves nearly five times the transistor density, a fundamental enabler of its performance advantage.

The shader configurations further illustrate the generational leap. The RTX 5090 features 21,760 shading units, 680 texture mapping units, and 176 render output units. It also includes 170 ray tracing cores and 680 tensor cores, reflecting its focus on modern graphics workloads like real-time ray tracing and AI acceleration. The Tesla P100 has 3,584 shading units, 224 TMUs, and 96 ROPs, with no ray tracing or tensor cores listed. This absence of dedicated RT and tensor hardware means the Pascal card is purely a compute-oriented part, lacking the specialized silicon found in the Blackwell design. The FP16 performance ratio also differs: the RTX 5090 achieves FP16 at a 1:1 ratio with FP32 (104.8 TFLOPS), while the Tesla P100 achieves FP16 at a 2:1 ratio (19.05 TFLOPS versus 9.526 FP32), indicating a different throughput balance.

The API support also evolves between generations. The RTX 5090 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, while the Tesla P100 supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.3. The newer card’s higher DirectX feature level and newer Vulkan version enable access to more advanced rendering techniques. The Tesla P100 offers no display outputs, confirming its role as a server or compute accelerator without video connectivity, whereas the RTX 5090 includes one HDMI 2.1b and three DisplayPort 2.1b outputs. This difference underscores the divergent target markets: one for datacenter compute, the other for consumer and prosumer graphics.

Specification Differences

The two cards differ across nearly every measurable specification. The RTX 5090 has a process node of 5 nm versus 16 nm for the Tesla P100, and the transistor count is 92,200 million versus 15,300 million. The die size is 750 mm² for the RTX 5090 and 610 mm² for the Tesla P100. Base clocks are 2017 MHz versus 1190 MHz, with boost clocks of 2407 MHz versus 1329 MHz. Memory configuration is a major split: the RTX 5090 offers 32 GB of GDDR7 on a 512-bit bus with 1.79 TB/s bandwidth, while the Tesla P100 offers 16 GB of HBM2 on a 4096-bit bus with 732.2 GB/s bandwidth. The shading units number 21,760 versus 3,584, TMUs are 680 versus 224, and ROPs are 176 versus 96. The RTX 5090 has 170 RT cores and 680 tensor cores, while the Tesla P100 lists none.

Pixel rate is 423.6 GPixel/s for the RTX 5090 compared to 127.6 GPixel/s for the Tesla P100, and texture rate is 1,636.8 GTexel/s versus 297.7 GTexel/s. FP32 performance is 104.8 TFLOPS versus 9.526 TFLOPS, and FP16 is 104.8 TFLOPS versus 19.05 TFLOPS. Power consumption is nearly doubled: the RTX 5090 has a TDP of 575 W with a suggested PSU of 950 W and a single 16-pin connector, while the Tesla P100 has a TDP of 250 W, a suggested PSU of 600 W, and a single 8-pin connector. The bus interface is PCIe 5.0 x16 for the RTX 5090 versus PCIe 3.0 x16 for the Tesla P100. Physical dimensions also differ: the RTX 5090 measures 304 mm in length, 137 mm in height, and 40 mm in width, while the Tesla P100 is 267 mm long with unspecified height and width. The launch MSRP for the RTX 5090 is 1,999 USD, and for the Tesla P100 it is 5,699 USD.

Where Each One Wins

The RTX 5090 wins in every directly comparable benchmark and every raw performance metric listed. It dominates in compute throughput, memory bandwidth, pixel fill rate, and texture rate. Its advantage in FP32 performance is an order of magnitude, and its FP16 performance is similarly superior. The card is also the only one of the two with modern graphics features like ray tracing cores and tensor cores, making it suitable for gaming, content creation, and AI workloads that leverage these accelerators. Its dual-slot design with display outputs and PCIe 5.0 interface positions it as a high-end consumer graphics card, capable of driving multiple high-resolution displays with HDMI 2.1b and DisplayPort 2.1b outputs. The 32 GB of GDDR7 memory provides substantial capacity for large datasets and high-resolution textures, and the 1.79 TB/s bandwidth supports demanding workloads.

The Tesla P100 PCIe 16 GB, despite being decisively outclassed in the Geekbench OpenCL test, retains its niche as a compute-focused accelerator. Its 16 GB of HBM2 memory on a 4096-bit bus offers a memory architecture optimized for bandwidth, even if the absolute bandwidth is lower. The card’s 250 W TDP is less than half that of the RTX 5090, making it more power-efficient in absolute terms, and its lower power draw and single 8-pin connector may be advantageous in systems with limited power delivery. The lack of display outputs and the PCIe 3.0 interface reflect its intended use in servers or workstations where graphics output is unnecessary. For legacy compute environments that rely on Pascal-era CUDA capabilities or that require a passive, low-power accelerator, the Tesla P100 remains a functional option, though its end-of-life production status limits its appeal. In summary, the RTX 5090 is the clear winner for any modern workload, while the Tesla P100 serves only as a specialized, older compute part with a specific power and memory profile.