NVIDIA GeForce RTX 5090 D vs NVIDIA Tesla P100 PCIe 12 GB Comparison

NVIDIA
GEFORCE

NVIDIA GeForce RTX 5090 D

CORE STATE GB202
VRAM 32 GB
CLOCK SPEED 2407 MHz
TDP 575 W
BUS WIDTH 512 bit
ARCHITECTURE Blackwell 2.0
nm
PROCESS 5 nm
LAUNCH DATE 2025
VS
NVIDIA
GEFORCE

Tesla P100 PCIe 12 GB

CORE STATE GP100
VRAM 12 GB
CLOCK SPEED 1329 MHz
TDP 250 W
BUS WIDTH 3072 bit
ARCHITECTURE Pascal
nm
PROCESS 16 nm
LAUNCH DATE 2016

PERFORMANCE BENCHMARKS

3dmark_3dmark_steel_nomad_dx12
14,326
N/A
geekbench_opencl
310,674
79,396
geekbench_vulkan
376,915
N/A
passmark_directx_10
231
N/A
passmark_directx_11
371
N/A
passmark_directx_12
219
N/A
passmark_directx_9
434
N/A
passmark_g2d
1,487
N/A
passmark_g3d
44,065
N/A
passmark_gpu_compute
28,396
N/A

Analysis: NVIDIA GeForce RTX 5090 D vs NVIDIA Tesla P100 PCIe 12 GB

The NVIDIA Tesla P100 PCIe 12 GB and the NVIDIA GeForce RTX 5090 D represent two vastly different eras of GPU design, separated by nearly a decade of architectural evolution. The data shows that in the single shared benchmark, the GeForce RTX 5090 D delivers a decisive victory, but the Tesla P100 remains a relevant data point in the database’s historical context, with both cards occupying the same 92nd percentile among all GPUs.

Head-to-Head Benchmarks

The only direct benchmark comparison available between these two cards is the Geekbench OpenCL test, and the results are stark. The NVIDIA GeForce RTX 5090 D scores 310,674, while the NVIDIA Tesla P100 PCIe 12 GB scores 79,396. This represents a delta of -74.4% from the perspective of the Tesla P100, meaning the RTX 5090 D is roughly 3.9 times faster in this specific compute workload. The margin is so large that it dwarfs any other comparison in the data set, underscoring the generational leap in raw compute power.

Looking at the broader benchmark landscape, the RTX 5090 D’s average benchmark score of 77,712 is actually pulled down by a diverse set of tests. In the Passmark suite, it scores 44,065 in G3D, 28,396 in GPU Compute, and much lower in legacy DirectX tests, with scores of 434 in DirectX 9, 371 in DirectX 11, and just 219 in DirectX 12. Its Vulkan score in Geekbench is even higher than OpenCL at 376,915, while the 3DMark Steel Nomad DX12 test yields 14,326. The Tesla P100, by contrast, has a single benchmark entry — the same Geekbench OpenCL score of 79,396 — with an average that matches it exactly.

In the context of their nearest rivals, the Tesla P100’s 79,396 score places it 0.3% behind the Tesla P100 PCIe 16 GB (which scores 79,605) and 0.6% behind the GeForce RTX 5090 (79,842), while edging out the AMD Radeon RX 6850M XT by 0.6% and trailing the Radeon Pro Vega 64X by 1.9%. The RTX 5090 D’s average of 77,712 tells a different story: it is 1.1% ahead of the AMD Radeon RX 6650M XT (76,904), 1.6% behind the RX 6850M XT (78,940), 2.1% behind the Tesla P100 12 GB, and 2.4% behind the Tesla P100 16 GB. This creates a peculiar situation where the RTX 5090 D’s average score is lower than the Tesla P100’s, despite winning the head-to-head OpenCL test by a massive margin — a discrepancy explained by the RTX 5090 D’s inclusion of additional, less favorable benchmark suites.

The GPU compute results further illustrate the divide. The RTX 5090 D’s Passmark GPU Compute score of 28,396 is substantial, but the Geekbench OpenCL result remains the single most telling metric for cross-generational comparison. The Tesla P100’s best-case scenario is its 92nd percentile ranking, which it shares with the RTX 5090 D, indicating that both are exceptional performers relative to the entire database of GPUs, even if the raw scores have diverged enormously.

FAQ

Q: How much faster is the RTX 5090 D than the Tesla P100 in the shared benchmark?

A: In the Geekbench OpenCL test, the RTX 5090 D scores 310,674 versus the Tesla P100’s 79,396, a delta of -74.4% from the Tesla P100’s perspective, making the RTX 5090 D approximately 3.9 times faster.

Q: Do both GPUs hold the same performance percentile?

A: Yes, both the NVIDIA Tesla P100 PCIe 12 GB and the NVIDIA GeForce RTX 5090 D are ranked in the 92nd percentile among all GPUs in the database, despite the significant raw score difference in the head-to-head test.

Q: Which GPU has a higher average benchmark score?

A: The Tesla P100 has a higher average benchmark score of 79,396, while the RTX 5090 D’s average is 77,712. This is because the RTX 5090 D is evaluated across multiple test suites, some of which yield lower scores, whereas the Tesla P100 has only the OpenCL result.

Q: How does the RTX 5090 D compare to its nearest rival, the RTX 5090?

A: The RTX 5090 D’s nearest rival data shows it is 2.1% behind the Tesla P100 12 GB and 2.4% behind the Tesla P100 16 GB, but it does not list the RTX 5090 as a direct rival — the RTX 5090 itself scores 79,842, which is 0.6% ahead of the Tesla P100 12 GB.

Q: What is the most impressive non-OpenCL score for the RTX 5090 D?

A: The RTX 5090 D achieves its highest score in Geekbench Vulkan at 376,915, followed by its OpenCL score of 310,674 and Passmark G3D at 44,065.

Q: Does the Tesla P100 have any benchmark wins over the RTX 5090 D?

A: No. The head-to-head data records zero wins for the Tesla P100 and one win for the RTX 5090 D, with the only compared test being Geekbench OpenCL.

Architecture Differences

The NVIDIA Tesla P100 PCIe 12 GB is built on the Pascal architecture, specifically the GP100 chip, fabricated on a 16 nm process at TSMC. It houses 15,300 million transistors on a 610 mm² die, yielding a transistor density of 25.1 million per square millimeter. This is a compute-focused design with 3,584 shading units, 224 texture mapping units, and 96 raster output units, but it lacks dedicated ray tracing and tensor cores entirely. The P100’s FP32 throughput is 9.526 TFLOPS, while its FP16 performance doubles to 19.05 TFLOPS via a 2:1 ratio. It supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.3, and has no display outputs, reflecting its server-oriented role.

In contrast, the NVIDIA GeForce RTX 5090 D uses the Blackwell 2.0 architecture with the GB202 chip, produced on a 5 nm process, also at TSMC. The transistor count jumps to 92,200 million, spread across a larger 750 mm² die, resulting in a much higher density of 122.9 million transistors per square millimeter. The RTX 5090 D features 21,760 shading units, 680 TMUs, and 176 ROPs, and it introduces 170 ray tracing cores and 680 tensor cores, enabling hardware-accelerated ray tracing and AI workloads that the Pascal-based P100 cannot perform. Its FP32 and FP16 performance are both rated at 104.8 TFLOPS, indicating a 1:1 ratio, and it supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and the newer Vulkan 1.4 API.

The architectural gulf is also evident in memory technology. The P100 uses HBM2 with a 3072-bit bus and 549.1 GB/s bandwidth, while the RTX 5090 D employs GDDR7 on a 512-bit bus, delivering 1.79 TB/s — more than three times the bandwidth. The process node shrink from 16 nm to 5 nm, combined with a near-sixfold increase in transistor count, explains the RTX 5090 D’s raw compute advantage. The presence of tensor cores in the RTX 5090 D is a fundamental feature difference, as the P100 has none, meaning AI-specific workloads would see an even larger disparity than the OpenCL benchmark suggests.

Specification Differences

The two cards diverge on nearly every measurable specification. The Tesla P100 has a base clock of 1190 MHz and a boost clock of 1329 MHz, while the RTX 5090 D starts at 2017 MHz and boosts to 2407 MHz. Memory configurations differ completely: the P100 offers 12 GB of HBM2 on a 3072-bit bus, whereas the RTX 5090 D provides 32 GB of GDDR7 on a 512-bit bus. Bandwidth scales from 549.1 GB/s to 1.79 TB/s. The P100’s memory clock is 715 MHz (1430 Mbps effective), while the RTX 5090 D runs at 1750 MHz (28 Gbps effective).

The compute units show a similar pattern. The P100 has 3,584 shading units, 224 TMUs, and 96 ROPs, compared to the RTX 5090 D’s 21,760 shading units, 680 TMUs, and 176 ROPs. Pixel rate jumps from 127.6 GPixel/s to 423.6 GPixel/s, and texture rate from 297.7 GTexel/s to 1,636.8 GTexel/s. Power requirements escalate from 250 W TDP with a single 8-pin connector and a 600 W suggested PSU, to 575 W TDP with a 16-pin connector and a 950 W suggested PSU. The bus interface advances from PCIe 3.0 x16 to PCIe 5.0 x16. The P100 has no display outputs, while the RTX 5090 D includes 1x HDMI 2.1b and 3x DisplayPort 2.1b. Physical dimensions also grow: the P100 is 267 mm (10.5 inches) long, while the RTX 5090 D measures 304 mm (12 inches) in length, 137 mm (5.4 inches) in height, and 48 mm (1.9 inches) in width. Both are dual-slot cards.

The RTX 5090 D’s launch MSRP is 2,299 USD, while the Tesla P100’s launch MSRP was 4,599 USD. Production status differs, with the P100 marked end-of-life and the RTX 5090 D active. Release dates are 2016-06-19 for the P100 and 2025-01-29 for the RTX 5090 D. The P100’s predecessor is Tesla Maxwell and successor is Tesla Volta, while the RTX 5090 D’s predecessor is GeForce 40 and successor is GeForce 60.

Where Each One Wins

The NVIDIA GeForce RTX 5090 D is the clear winner in raw computational performance, as evidenced by its 310,674 OpenCL score versus the Tesla P100’s 79,396. It is also the only card of the two with any display outputs, making it suitable for direct graphics workloads, gaming, and content creation. The presence of ray tracing and tensor cores gives it a functional edge in modern applications that leverage these features, and its 32 GB of GDDR7 memory with 1.79 TB/s bandwidth is far superior for large datasets and high-resolution textures. The 104.8 TFLOPS FP32 and FP16 performance, both at a 1:1 ratio, indicate balanced compute for a wide range of tasks.

The NVIDIA Tesla P100 PCIe 12 GB, despite losing the head-to-head, still holds a niche in legacy compute environments. Its average benchmark score of 79,396 is higher than the RTX 5090 D’s 77,712, which suggests that in the specific OpenCL workload where both are measured, the P100’s performance is more consistent relative to its single test. Its 92nd percentile ranking ties it with the RTX 5090 D, meaning it remains a top-tier performer even by modern standards. The P100’s HBM2 memory, while slower in bandwidth, offers a 3072-bit bus that was state-of-the-art for its time, and its 250 W TDP is significantly lower, making it more power-efficient in absolute terms. For institutions running legacy Pascal-optimized code or requiring a compute card without display output, the P100 retains relevance, though its end-of-life status and lack of modern features like ray tracing limit its future-proofing.

The RTX 5090 D wins decisively in every head-to-head metric, but the Tesla P100’s historical significance and its higher average score in the database’s limited comparison create a nuanced picture. The RTX 5090 D is for users needing maximum performance today, while the P100 remains a benchmark reference point for Pascal-era compute.

DETAILED SPECIFICATIONS

SPECIFICATION
RTX 5090 D
Tesla P100 PCIe 12 GB
Core Specs
Shading Units
21,760
3,584 -83.5%
Shaders
21,760
3,584 -83.5%
TMUs
680
224 -67.1%
ROPs
176
96 -45.5%
SM Count
170
56 -67.1%
Clocks
Base Clock
2017 MHz
1190 MHz
Boost Clock
2407 MHz
1329 MHz
Memory Clock
1750 MHz 28 Gbps effective
715 MHz 1430 Mbps effective
Memory
Memory Size
32 GB
12 GB
VRAM (MB)
32,768
12,288 -62.5%
Memory Type
GDDR7
HBM2
Memory Bus
512 bit
3072 bit
Bandwidth
1.79 TB/s
549.1 GB/s
Cache
L1 Cache
128 KB (per SM)
24 KB (per SM)
L2 Cache
96 MB
3 MB
Performance
Pixel Rate
423.6 GPixel/s
127.6 GPixel/s
Texture Rate
1,636.8 GTexel/s
297.7 GTexel/s
FP32 (TFLOPS)
104.8 TFLOPS
9.526 TFLOPS
FP64 (TFLOPS)
1.637 TFLOPS (1:64)
4.763 TFLOPS (1:2)
FP16 (TFLOPS)
104.8 TFLOPS (1:1)
19.05 TFLOPS (2:1)
AI/RT
RT Cores
170
—
Tensor Cores
680
—
Power
TDP
575 W
250 W
TDP (W)
575
250 -56.5%
Suggested PSU
950 W
600 W
Power Connectors
1x 16-pin
1x 8-pin
Architecture
Architecture
Blackwell 2.0
Pascal
GPU Name
GB202
GP100
Generation
GeForce 50
Tesla Pascal (Pxx)
Process Size
5 nm
16 nm
Transistors
92,200 million
15,300 million
Die Size
750 mm²
610 mm²
Foundry
TSMC
TSMC
Density
122.9M / mm²
25.1M / mm²
API Support
DirectX
12 Ultimate (12_2)
12 (12_1)
OpenGL
4.6
4.6
Vulkan
1.4
1.3
OpenCL
3.0
3.0
CUDA
12.0
6.0
Shader Model
6.9
6.0
Physical
Slot Width
Dual-slot
Dual-slot
Length
304 mm 12 inches
267 mm 10.5 inches
Height
137 mm 5.4 inches
—
Outputs
1x HDMI 2.1b3x DisplayPort 2.1b
No outputs
Bus Interface
PCIe 5.0 x16
PCIe 3.0 x16
Other
Launch Price
2,299 USD
4,599 USD
Production
Active
End-of-life
Predecessor
GeForce 40
Tesla Maxwell
Successor
GeForce 60
Tesla Volta
View GeForce RTX 5090 D Details View Tesla P100 PCIe 12 GB Details