PerGB

NVIDIA A100 SXM4 40 GB

Live eBay prices, cheapest first. The best one right now is $2,195 (used).

Not a PCIe card. SXM module: needs an SXM-to-PCIe adapter board, its own cooling and power to run in a normal PC. The PCIe version: NVIDIA A100 PCIe 40 GB, from $5,588 → With its adapter board: NVIDIA A100 SXM4 40 GB mod kit, from $6,049 →

Other sizes: 80 GB, 64 GB

Specs

VRAM
40 GB HBM2e
Bandwidth
1,555 GB/s
FP32
19.5 TFLOPS
FP16
312 TFLOPS
Power
400 W
Connector
SXM4 (PCIe 4.0)
Released
2020
CUDA compute
8.0
Llama 3.1 8B at Q4 (est.)
160–240 tokens/s

Price right now

Live on eBay
Lowest now$2,195
Median ask$3,200
Per GB VRAM$54.88
vs launch price—
Live listings23

From the live listings below; shipping included.

Listings

Live on eBay
ListingConditionPriceChecked
Used$2,195−31%shipping not listed
Used$2,350−27%shipping not listed
Used$2,499or best offer−22%free shipping
Used$2,499or best offer−22%free shipping
Used$2,769or best offer−13%free shipping
Used$2,770or best offer−13%free shipping
Used$2,800−12%free shipping
Used$2,850or best offer−11%shipping not listed
Used$2,889or best offer−10%free shipping
Used$3,000or best offer−6%free shipping
Used$3,000or best offer−6%shipping not listed
Used$3,200or best offerfree shipping
Used$3,349or best offerfree shipping
Used$5,247or best offerfree shipping
Used$5,248or best offerfree shipping
Used$5,248or best offerfree shipping
Open box$5,263free shipping
Open box$5,263free shipping
Open box$5,280or best offerfree shipping
Open box$5,280or best offerfree shipping
Used$5,344or best offerfree shipping
Used$5,345or best offerfree shipping
Used$8,641+$10.00 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

Faster, for the least

More VRAM, for not much more

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 SXM4 40 GB neededCosts nowSpeed (est.)
Llama 3.1 8B1 card$2,195160–240 tokens/s
Qwen3 14B1 card$2,19587–130 tokens/s
gpt-oss 20B1 card$2,195214–321 tokens/s
Gemma 3 27B1 card$2,19547–70 tokens/s
Qwen3 32B1 card$2,19539–59 tokens/s
Llama 3.3 70B2 cards$4,54518–27 tokens/s
gpt-oss 120B2 cards$4,545151–226 tokens/s
Qwen3 235B-A22B4 cards$9,54335–52 tokens/s
DeepSeek R1 / V3 671Bmore 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 SXM4 40 GB neededCosts nowPer question (est.)
Julia-10.144B1 card$2,1952.9–6.1 ms
Laya0.421B1 card$2,1954.8–10 ms
Kev-4B4B1 card$2,19511–24 ms
lev4B1 card$2,19534–73 ms
OpenJev27B1 card$2,19541–88 ms
Clef27B1 card$2,19541–88 ms
Perplexity Decider v1.227B1 card$2,19541–88 ms
mammoth-127B1 card$2,19541–88 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 SXM4 40 GB. These open the shop's own search; we don't show their prices here.