NVIDIA A100 PCIe 40 GB
Live eBay prices, cheapest first. The best one right now is $5,588 (open box).
Other sizes: 80 GB
Specs
- VRAM
- 40 GB HBM2e
- Bandwidth
- 1,555 GB/s
- FP32
- 19.5 TFLOPS
- FP16
- 312 TFLOPS
- Power
- 300 W
- Connector
- PCIe 4.0
- Released
- 2020
- Launch price
- $10,000 (est.)
- CUDA compute
- 8.0
- Llama 3.1 8B at Q4 (est.)
- 160–240 tokens/s
Running it at home
- Cooling
- Passive: it has no fan. In a desktop it needs air pushed through it, front to back: a fan shroud on its end (3D-printed ones with a 40 mm blower or a 92 mm fan are sold for most of these cards) or a server chassis. Without one it overheats within minutes under load.
- Power
- One 8-pin CPU (EPS-12V) power input, not a PCIe one: it needs an adapter cable (two PCIe 8-pin to one EPS 8-pin, sold for these cards). A PCIe plug fits but is wired differently: never force one in. Rated 300 W: allow that for each card on top of the rest of the PC.
- Size
- Full height, two slots.
- Motherboard
- Turn on "Above 4G decoding" in the BIOS: without it many boards won't start with a card with this much memory.
- Screen
- No display outputs: use the processor's graphics or a small second card for a monitor.
A server card: made for a rack server's airflow and power. With these sorted it works in a normal PC.
Price right now
Live on eBayFrom the live listings below; shipping included.
Listings
Live on eBay| Listing | Condition | Price | Checked |
|---|---|---|---|
| Open box | $5,588or best offer−28%free shipping | ||
| Used | $5,800or best offer−25%free shipping | ||
| Refurbished | $5,850or best offer−25%free shipping | ||
| Used | $6,000−23%shipping not listed | ||
| Refurbished | $6,120or best offer−21%free shipping | ||
| Refurbished | $6,499−16%free shipping | ||
| Used | $6,890or best offer−11%+$40.00 shipping | ||
| Used | $6,900or best offer−11%shipping not listed | ||
| Used | $6,995or best offer−10%free shipping | ||
| Used | $6,995or best offer−10%free shipping | ||
| Used | $6,999or best offer−10%free shipping | ||
| Used | $7,500or best offerfree shipping | ||
| Used | $7,500or best offerfree shipping | ||
| Used | $7,500or best offerfree shipping | ||
| Used | $7,750or best offerfree shipping | ||
| Used | $7,750or best offerfree shipping | ||
| Used | $7,750or best offerfree shipping | ||
| Used | $7,900or best offerfree shipping | ||
| Used | $7,995or best offerfree shipping | ||
| Used | $8,055or best offerfree shipping | ||
| Used | $8,064or best offerfree shipping | ||
| Used | $8,067or best offerfree shipping | ||
| Used | $8,069or best offerfree shipping | ||
| Used | $8,069or best offerfree shipping | ||
| Used | $8,085or best offerfree shipping | ||
| Used | $8,939or best offerfree shipping | ||
| Used | $9,500or best offerfree shipping | ||
| Open box | $11,200or best offerfree shipping | ||
| Open box | $12,500free shipping | ||
| Used | $12,995or best offerfree shipping |
Auctions ending soon
No auctions ending soon right now.
Prices include shipping and were read from eBay at the time shown. They can change: check the listing before you buy.
Owner reviews
No owner reviews yet. Own one? Tell others how it runs.
From signed-in owners, checked before they're published. PerGB doesn't pay for or edit reviews.
How it compares
Live on eBay- #3 of 20 PCIe GPUs with 40 GB or more for the money: $2,068 per GeForce RTX 3090 of performance.
- Software: CUDA 11.0 and newer.
Faster, for the least
- NVIDIA RTX PRO 6000 Blackwell: $13,100, 367% of the GeForce RTX 3090, 96 GB
- NVIDIA H100 PCIe: $26,935, 477% of the GeForce RTX 3090, 80 GB
- NVIDIA H200 NVL: $32,300, 777% of the GeForce RTX 3090, 141 GB
More VRAM, for not much more
- NVIDIA Quadro RTX 8000: $2,000, 115% of the GeForce RTX 3090, 48 GB
- Intel Arc Pro B60 48 GB: $2,049, 82% of the GeForce RTX 3090, 48 GB
- NVIDIA A16: $2,200, 93% of the GeForce RTX 3090, 64 GB
From the live listings of every GPU, shipping included; performance against the GeForce RTX 3090 (FP16 or FP32 compute and memory bandwidth). Updated with every build.
Local AI: what it runs
| Model (Q4, 8K context) | A100 PCIe 40 GB needed | Costs now | Speed (est.) |
|---|---|---|---|
| Llama 3.1 8B | 1 card | $5,588 | 160–240 tokens/s |
| Qwen3 14B | 1 card | $5,588 | 87–130 tokens/s |
| gpt-oss 20B | 1 card | $5,588 | 214–321 tokens/s |
| Gemma 3 27B | 1 card | $5,588 | 47–70 tokens/s |
| Qwen3 32B | 1 card | $5,588 | 39–59 tokens/s |
| Llama 3.3 70B | 2 cards | $11,388 | 18–27 tokens/s |
| gpt-oss 120B | 2 cards | $11,388 | 151–226 tokens/s |
| Qwen3 235B-A22B | 4 cards | $23,238 | 35–52 tokens/s |
| DeepSeek R1 / V3 671B | more than 8 | — | — |
Cost: that many of the cheapest live listings. Speed: an estimate from the memory bandwidth for one person chatting. Try other models, sizes and contexts.
Decision models
| Model (Q4, 8K context) | A100 PCIe 40 GB needed | Costs now | Per question (est.) |
|---|---|---|---|
| Julia-10.144B | 1 card | $5,588 | 2.9–6.1 ms |
| Laya0.421B | 1 card | $5,588 | 4.8–10 ms |
| Kev-4B4B | 1 card | $5,588 | 11–24 ms |
| lev4B | 1 card | $5,588 | 34–73 ms |
| OpenJev27B | 1 card | $5,588 | 41–88 ms |
| Clef27B | 1 card | $5,588 | 41–88 ms |
| Perplexity Decider v1.227B | 1 card | $5,588 | 41–88 ms |
| mammoth-127B | 1 card | $5,588 | 41–88 ms |
Decision models score the options you give them instead of writing text, so they're timed per question. The time is scaled from each model's reported speed on an RTX PRO 6000 by this card's compute and memory bandwidth, within about half either way. About the models.
Compare elsewhere
Search other shops for NVIDIA A100 PCIe 40 GB. These open the shop's own search; we don't show their prices here.



























