I'd love to hear what you think! Please drop me a line and let me know what you like and what could be better. 🙏
NVIDIA Tesla P100 16GB Specs, Benchmarks & Pricing
The NVIDIA Tesla P100 was announced in 2016 as the first of their Pascal architecture powered Tesla cards. The Tesla P100 came in PCIe and SXM versions and was the first GPU accelerator to use High Bandwidth Memory 2 (HBM2). It did not have Tensor Cores, so it isn't usually preferred for machine learning. However, it's FP32 and FP16 FLOPs are still quite good.
- Release Date: June 20, 2016
- MSRP: $5,699 USD
- GPU Architecture: pascal
- Hardware-Accelerated GEMM Operations:FP64 FP32 TF32 BF16 FP16 FP8 FP4 INT8 INT4
- CUDA Compute Capability : 6
Considerations
- Lower FP32 compute performance compared to other GPUs
- Lower FP16 compute performance compared to other GPUs
Specifications for NVIDIA Tesla P100
| Specification | Performance Ranking |
|---|---|
| FP32 TFLOPs | |
| FP16 TFLOPs | |
| Tensor Core Count | Data not available |
| Memory Capacity (GB) | |
| Memory Bandwidth (GB/s) | |
| Int8 TOPs | Data not available |
| FP8 TFLOPs | Data not available |
| FP4 TFLOPs | Data not available |
Real-time NVIDIA Tesla P100 GPU Prices
We're tracking 93 of the NVIDIA Tesla P100 GPUs currently available for sale. The lowest average price is $75 (what's this?)
Buy NowCompare Price/Performance to other GPUs
We track real-time prices of other GPUs too so that you can compare the price/performance of the NVIDIA Tesla P100 GPU to other GPUs.
Compare GPU Price/PerformanceCompare NVIDIA Tesla P100 to Another GPU
Compare the NVIDIA Tesla P100 directly to another GPU to see specs, benchmarks, and prices side-by-side.
Compare GPUs Side-by-SidePrice History
NVIDIA Tesla P100 Price History
This chart tracks NVIDIA Tesla P100 prices over the past 6 months based on eBay listings. The yellow line shows the average of the three lowest-priced listings each day—a useful indicator for finding deals.
Trend: Over this period, the lowest average price has been down 6.8%. The current lowest average is $73. The lowest average price is calculated as the average of the three lowest-priced listings each day.
gpupoet.com
Product Identifiers
Manufacturer Part Numbers (9)
- NVIDIA Part Number
- 699-2H400-0201-XXX
- NVIDIA Part Number
- 699-2H400-0202-XXX
- Product SKU
- 900-2H400-0000-000
- Product SKU
- 900-2H400-0010-000
- Product SKU
- 900-2H400-0300
- Product SKU
- 900-2H403-0000-000
- GPU CHIP
- GP100-890-A1
- GPU CHIP
- GP100-892-A1
- GPU CHIP
- GP100-893-A1
Available from 6 Partners (17 products)
Dell
- NVIDIA Tesla P100 Pascal GPU Accelerator 16GB 3584 CUDA Core PCI-E 3.0 x16
- H7WFC(part number)
- Nvidia Tesla P100 SXM2 16GB GPU Accelerator Card
- 7TGM9(part number)
- NVIDIA Tesla P100 12GB PCI Express Video Graphics Card
- PH400(part number)
- NVIDIA Tesla P100 12GB 250W DW x16 PCI-e Full Height
- 489-BBCH(part number)
HPE
- HPE Nvidia Tesla P100 PCI-e 16GB Computational Accelerator
- Q0E21A(part number)
- HPE NVIDIA Tesla P100 16GB PCIe GPU
- 868585-001(part number)
- HPE NVIDIA Tesla P100 16GB PCIe GPU
- 868199-001(part number)
- HPE NVIDIA Tesla P100 16GB SXM2 GPU Accelerator
- 876911-001(part number)
- HPE Nvidia Tesla P100 PCIe 12GB Graphics Card
- Q2S42A(part number)
Lenovo
- Lenovo NVIDIA Tesla P100 16GB Graphic Card
- 7X67A00068(part number)
Supermicro
- NVIDIA Tesla P100 12GB CoWoS HBM2 PCIe 3.0 - Passive Cooling
- GPU-NVTP100-12(part number)
- NVIDIA Tesla P100 16GB CoWoS HBM2 PCIe 3.0 - Passive Cooling
- GPU-NVTP100-16(part number)
- NVIDIA Tesla P100 SXM2 16GB CoWoS HBM2, NVLink
- GPU-NVTP100-SXM(part number)
PNY
- PNY Tesla P100 16GB HBM2 PCIe GPU
- TCSP100M-16GB-PB(sku)
- PNY Tesla P100 12GB HBM2 PCIe GPU
- TCSP100M-12GB-PB(sku)
Cisco
- Cisco NVIDIA P100 Graphics Card, 16GB
- UCSC-GPU-P100-16G(part number)
- Cisco NVIDIA P100 Graphics Card, 12GB
- UCSC-GPU-P100-12G(part number)
References
- https://developer.nvidia.com/blog/mixed-precision-programming-cuda-8/
- https://technical.city/en/video/Tesla-P100-PCIe-16-GB
- https://www.techpowerup.com/gpu-specs/tesla-p100-pcie-16-gb.c2888
- https://www.bhphotovideo.com/c/product/1368043-REG/nvidia_nvidia_900_2h400_0000_000_tesla_p100_16gb_cowos.html
- https://www.disctech.com/NVIDIA-Tesla-P100-900-2H400-0300-GPU-Accelerator-Graphics-Card
- https://www.microway.com/hpc-tech-tips/nvidia-tesla-p100-pci-e-16gb-gpu-accelerator-pascal-gp100-close/
- https://www.serversupply.com/GPU/GDDR5/16GB/CISCO/UCSC-GPU-P100-16G_401845.htm
- https://www.connection.com/product/cisco-nvidia-p100-graphics-card-16gb/ucsc-gpu-p100-16g/33795137
- https://www.wiredzone.com/shop/product/10027278-supermicro-gpu-nvtp100-12-nvidia-tesla-p100-12gb-cowos-hbm2-pcie-3-0-passive-cooling-1878
- https://www.pny.eu/en/professional/explore-all-products/nvidia-tesla/747-tesla-p100
- https://www.sabrepc.com/7X67A00068-Lenovo-S729687
- https://support.hpe.com/hpesc/public/docDisplay?docId=a00069881en_us&page=GUID-ADF9D45D-2BBD-459F-9B46-951204022110.html&docLocale=en_US
- https://pcserverandparts.com/dell-nvidia-tesla-p100-pascal-16gb-hbm2-graphics-accelerator-used-h7wfc/
- https://images.nvidia.com/content/pdf/tesla/whitepaper/pascal-architecture-whitepaper.pdf
Notes
- tensorCoreCount: Tesla P100 only has CUDA cores
- supportedHardwareOperations corrected 2026-09-02 to the canonical Pascal baseline (empty array) documented in .claude/agents/gpu-researcher.md, sourced from the NVIDIA Pascal Architecture whitepaper for GP100 (images.nvidia.com/content/pdf/tesla/whitepaper/pascal-architecture-whitepaper.pdf), which never mentions "Tensor Core" — Tensor Cores were introduced one generation later, in Volta. Under the locked GEMM-only definition of supportedHardwareOperations (hardware-accelerated matrix/tensor-core path support only, not "the ALU can execute this precision"), FP16 and FP32 were removed: P100's 2x-rate FP16 ("Using FP16 computation improves performance up to 2x compared to FP32 arithmetic") is described under standard SM/CUDA-core arithmetic in the whitepaper, not a separate matrix engine, and plain FP32 is likewise a CUDA-core ALU path with no tensor-accelerated row in any Pascal documentation.
- supportedHardwareOperations: It appears that hardware-accelerated GEMM operations require "Warp matrix functions" supported in Compute Capability >=7.x: https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#compute-capabilities
- supportedHardwareOperations: The 8-bit and 16-bit DP4A and DP2A dot product instructions are supported on GP102-GP106, but not on GP100.
- fp8TFLOPS and fp4TFLOPS are null: Pascal predates both FP8 (introduced in Ada) and FP4 (introduced in Blackwell). Confirmed hardware negative per the NVIDIA Pascal Architecture whitepaper (GP100), which documents no matrix/tensor engine at all — only standard CUDA-core FP16/FP32/FP64 SIMD throughput. Not added to unverifiedSpecs since these are confirmed negatives, not data gaps.
- MSRP of $5,699 USD for the 16GB PCIe model sourced from technical.city and TechPowerUp GPU Database, which both cite the official NVIDIA launch price on June 20, 2016.
- Manufacturer identifiers: NVPN (NVIDIA Part Number) format 699-2H400-0201-XXX is for 16GB PCIe variants, 699-2H400-0202-XXX is for 12GB PCIe variants. The 900-2H400-XXXX numbers are product SKUs. The 900-2H403-0000-000 SKU is for the SXM2 16GB variant.
- GPU chip variants: GP100-890-A1 (SXM2), GP100-892-A1 (PCIe 12GB), GP100-893-A1 (PCIe 16GB) as documented in NVIDIA datasheets and TechPowerUp GPU Database.
- Third-party products: Tesla P100 was widely adopted across datacenter OEMs. Multiple variants exist with different memory capacities (12GB/16GB) and form factors (PCIe/SXM2). Part numbers vary by vendor, memory configuration, and whether heatsinks are included.