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 A100 PCIe 80GB 80GB Specs, Benchmarks & Pricing
The NVIDIA A100 PCIe 80GB was announced in May 2020, is a formidable accelerator in the field of machine learning and artificial intelligence. Built on NVIDIA's advanced Ampere architecture, this accelerator is designed for high-performance computing, deep learning training, and inference tasks. With its massive 80 GB of HBM2e memory and superior memory bandwidth of 1,935 GB/s, it caters to the most demanding AI workloads. The inclusion of 432 tensor cores significantly accelerates machine learning applications, making it a go-to choice for researchers and data scientists. Operating at a base clock of 1065 MHz and a boost clock up to 1410 MHz, it delivers impressive computational power, capped at a maximum power consumption of 300 watts. The A100-PCIE-80GB is notable for its high FP32 performance of 19.5 TFLOPS, emphasizing its capability in handling floating-point operations efficiently.
- Release Date: June 28, 2021
- MSRP: $15,000 USD†
- GPU Architecture: Ampere
- Hardware-Accelerated GEMM Operations:FP64 FP32 TF32 BF16 FP16 FP8 FP4 INT8 INT4
- CUDA Compute Capability : 8
† Estimated MSRP of $15,000 USD for the A100 PCIe 80GB based on launch-time market pricing. NVIDIA does not publish official MSRP for data center GPUs. Sources from around the 2020-2021 launch period cite A100 40GB PCIe at ~$10,000-$12,000 and 80GB PCIe at ~$15,000-$20,000. Note: A100 pricing varies significantly by form factor (PCIe vs SXM) and memory size (40GB vs 80GB). SXM variants are typically more expensive than PCIe. Sources: CNBC, Modal.com, LoveChip, Wikipedia.
Strengths
- Excellent FP32 compute performance (top 13% of GPUs)
- Excellent FP16 compute performance (top 18% of GPUs)
Specifications for NVIDIA A100 PCIe 80GB
| Specification | Performance Ranking |
|---|---|
| FP32 TFLOPs | |
| FP16 TFLOPs | |
| Tensor Core Count | |
| Memory Capacity (GB) | |
| Memory Bandwidth (GB/s) | |
| Int8 TOPs | |
| FP8 TFLOPs | Data not available |
| FP4 TFLOPs | Data not available |
Real-time NVIDIA A100 PCIe 80GB GPU Prices
We're tracking 24 of the NVIDIA A100 PCIe 80GB GPUs currently available for sale. The lowest average price is $14,433 (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 A100 PCIe 80GB GPU to other GPUs.
Compare GPU Price/PerformanceCompare NVIDIA A100 PCIe 80GB to Another GPU
Compare the NVIDIA A100 PCIe 80GB directly to another GPU to see specs, benchmarks, and prices side-by-side.
Compare GPUs Side-by-SidePrice History
NVIDIA A100 PCIe 80GB Price History
This chart tracks NVIDIA A100 PCIe 80GB 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 up 54.1%. The current lowest average is $14,433, compared to a typical $11,323 over the period. The lowest average price is calculated as the average of the three lowest-priced listings each day.
gpupoet.com
Product Identifiers
Manufacturer Part Numbers (4)
- BOARD NUMBER
- P1001 SKU 200
- NVIDIA Part Number
- 900-21001-0020-100
- NVPN (WITH CEC)
- 900-21001-0020-000
- NVPN (LIQUID COOLED)
- 900-21001-0060-000
Available from 7 Partners (16 products)
Dell
- NVIDIA Ampere A100, PCIe, 300W, 80GB Passive, Double Wide, Full Height GPU
- 490-BHBK(part number)
- NVIDIA A100 80GB PCIe GPU
- 575TK(part number)
- NVIDIA A100 80GB PCIe GPU
- WGWR4(part number)
- NVIDIA A100 80GB PCIe GPU (OEM)
- 900-21001-0120-030-DELL(part number)
HPE
- NVIDIA A100 80GB PCIe Non-CEC Accelerator
- R9P49C(part number)
- NVIDIA A100 80GB PCIe Non-CEC Accelerator
- P49082-001(part number)
- SPS-PCA NVIDIA A100 80GB PCIe GPU
- P54869-001(part number)
Lenovo
- ThinkSystem NVIDIA A100 80GB PCIe Gen4 Passive GPU w/o CEC
- 4X67A76715(part number)
PNY
- NVIDIA A100 80GB PCIe GPU
- NVA100TCGPU80-KIT(sku)
- NVIDIA A100 80GB PCIe GPU (Non-CEC)
- NVA100TCGPU80NC-KIT(sku)
- NVIDIA A100 80GB PCIe GPU
- TCSA100M-80GB-PB(sku)
Supermicro
- NVIDIA A100 80GB CoWoS HBM2 PCIe w/o CEC
- GPU-NVA100-80-NC(part number)
Cisco
- NVIDIA Tesla A100 80GB PCIe GPU
- UCSC-GPU-A100-80(part number)
- NVIDIA Tesla A100 80GB PCIe GPU (Spare/Orderable)
- UCSC-GPU-A100-80=(part number)
- Tesla A100, Passive, 300W, 80GB
- UCSC-GPUA100-80-D(part number)
Advantech
- Tesla A100 80GB PCIe DSFH w/ATX BKT (Bundle Sale)
- SKY-TESL-A100-80P(part number)
References
- https://www.nvidia.com/en-us/data-center/a100/
- https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/a100/pdf/nvidia-a100-datasheet-nvidia-us-2188504-web.pdf
- https://developer.nvidia.com/blog/nvidia-ampere-architecture-in-depth/
- https://www.techpowerup.com/gpu-specs/a100-pcie-80-gb.c3821
- https://www.cnbc.com/2023/02/23/nvidias-a100-is-the-10000-chip-powering-the-race-for-ai-.html
- https://modal.com/blog/nvidia-a100-price-article
- https://www.lovechip.com/blog/the-nvidia-a100-gpu-the-ai-workhorse-that-changed-everything
- https://www.nvidia.com/en-us/data-center/tensor-cores/
- https://insidesystems.com/product/575tk/
- https://www.dell.com/en-us/shop/nvidia-ampere-a100-pcie-300w-80gb-passive-double-wide-full-height-gpu-customer-install/apd/490-bhbk/graphic-video-cards
- https://expresscomputersystems.com/products/hpe-nvidia-a100-80gb-pcie-non-cec-accelerator-r9p49c
- https://www.hpe.com/us/en/product-catalog/compute/accelerators/pip.specifications.nvidia-accelerators.1014651806.html
- https://lenovopress.lenovo.com/lp1734-thinksystem-nvidia-a100-pcie-40-gpu
- https://www.pny.com/nvidia-a100-80gb
- https://store.supermicro.com/us_en/supermicro-nvidia-a100-ampere-40-gb-graphics-card.html
- https://www.expresscomputersystems.com/products/cisco-nvidia-tesla-a100-80gb-dw-ucsc-gpu-a100-80
- https://www.advantech.com/en-us/products/8d6a8886-b57e-493f-87c8-6f1e02d0608a/sky-tesl-a100-80p/mod_7af76fc9-a36d-4e4b-8ab8-b91d0248507a
- https://devicehunt.com/view/type/pci/vendor/10DE/device/20B5
- https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/a100/pdf/nvidia-a100-datasheet-us-nvidia-1758950-r4-web.pdf
Notes
- fp32TFLOPS of 156 is TF32 Tensor Core throughput without sparsity, per the NVIDIA A100 datasheet (nvidia-a100-datasheet-us-nvidia-1758950-r4-web.pdf), which lists TF32 Tensor Core as 156 dense and 312 with sparsity. FP32 vector (CUDA core) throughput is 19.5 TFLOPS.
- Estimated MSRP of $15,000 USD for the A100 PCIe 80GB based on launch-time market pricing. NVIDIA does not publish official MSRP for data center GPUs. Sources from around the 2020-2021 launch period cite A100 40GB PCIe at ~$10,000-$12,000 and 80GB PCIe at ~$15,000-$20,000. Note: A100 pricing varies significantly by form factor (PCIe vs SXM) and memory size (40GB vs 80GB). SXM variants are typically more expensive than PCIe. Sources: CNBC, Modal.com, LoveChip, Wikipedia.
- Board Number P1001 SKU 200 sourced from TechPowerUp GPU Database.
- NVIDIA part number 900-21001-0020-100 (non-CEC version) is the primary NVPN for A100 PCIe 80GB. Part number scheme: 900-21001-XXXX-1XX indicates non-CEC (without secondary root of trust), while 900-21001-XXXX-0XX indicates CEC variant (with secondary root of trust). Sources: harddiskdirect.com, acmemicro.com, bitworks.io.
- PCI Device ID 10DE:20B5 sourced from Device Hunt and NVIDIA driver documentation. Vendor ID 10DE is NVIDIA Corporation, Device ID 20B5 is GA100 [A100 PCIe 80GB].
- Dell part numbers verified from multiple reseller sources including Inside Systems, Express Computer Systems, and Dell official product page.
- HPE part numbers R9P49C (product number), P49082-001 (part number), and P54869-001 (spare part number) verified from Express Computer Systems, HPE official product catalog, and ShopBLT.com.
- Lenovo part number 4X67A76715 verified from Lenovo Press product guide and multiple reseller sources. Note: Lenovo A100 PCIe products are now withdrawn, but SXM variants remain available.
- PNY part numbers NVA100TCGPU80-KIT (standard), NVA100TCGPU80NC-KIT (non-CEC), and TCSA100M-80GB-PB verified from PNY official website and multiple reseller sources.
- Supermicro part number GPU-NVA100-80-NC verified from Supermicro official store and authorized resellers. This part corresponds to NVIDIA base part 900-21001-0020-100.
- Cisco part numbers UCSC-GPU-A100-80, UCSC-GPU-A100-80=, and UCSC-GPUA100-80-D verified from Express Computer Systems, Provantage, and Cisco documentation. Note: Cisco requires unique SBIOS ID for GPU cards, so all A100 cards must be procured from Cisco.
- Advantech part number SKY-TESL-A100-80P verified from Advantech official product page and pf-electronic.pl. This is a bundle sale item including ATX bracket.
- Phase 5 audit (2026-09-02): gpuArchitecture normalized from lowercase 'ampere' to canonical 'Ampere' capitalization per the canonical table in .claude/agents/gpu-researcher.md.
- Phase 5 audit (2026-09-02): supportedHardwareOperations reviewed against the canonical Ampere baseline in .claude/agents/gpu-researcher.md (TF32, BF16, FP16, INT8, INT4) — file already matched. FP64 is retained because the A100 is GA100-based (confirmed by PCI Device ID 20B5 = GA100 per the Device Hunt reference above), and per NVIDIA's A100 architecture whitepaper the GA100 die's 3rd-gen Tensor Cores add a dedicated FP64 Tensor Core path not present on GA10x consumer/professional dies. No change made to the array.
- fp8TFLOPS and fp4TFLOPS are left null (field omitted) and not added to unverifiedSpecs — this is a confirmed hardware negative, not a research gap. The A100's Ampere 3rd-generation Tensor Cores predate both FP8 (introduced in Ada Lovelace/Hopper) and FP4 (introduced in Blackwell) tensor formats, per the canonical architecture table in .claude/agents/gpu-researcher.md.
- fp16TFLOPS of 624 is FP16 Tensor Core throughput with 2:4 structured sparsity, as the NVIDIA A100 datasheet (nvidia-a100-datasheet-us-nvidia-1758950-r4-web.pdf) publishes it: "312 | 624*", footnoted "With sparsity".