NVIDIA L40S
Live eBay prices, cheapest first. The best one right now is $6,501 (used).
Specs
- VRAM
- 48 GB GDDR6
- Bandwidth
- 864 GB/s
- FP32
- 91.6 TFLOPS
- FP16
- 366 TFLOPS
- Power
- 350 W
- Connector
- PCIe 4.0
- Released
- 2023
- Launch price
- $8,000 (est.)
- CUDA compute
- 8.9
- Llama 3.1 8B at Q4 (est.)
- 89–134 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 16-pin (12VHPWR / 12V-2x6) power input. Rated 350 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.
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 |
|---|---|---|---|
| Used | $6,501or best offerCheck carefully+$5.98 shipping | ||
| Used | $7,800or best offer−34%shipping not listed | ||
| Used | $7,900or best offer−34%Check carefullyfree shipping | ||
| Open box | $8,799or best offer−26%free shipping | ||
| Used | $8,950−25%shipping not listed | ||
| New | $8,950−25%free shipping | ||
| New | $8,950−25%free shipping | ||
| New | $8,950−25%free shipping | ||
| Used | $8,999−24%shipping not listed | ||
| New | $9,000−24%free shipping | ||
| New | $9,249or best offer−22%free shipping | ||
| New | $9,299or best offer−22%+$50.00 shipping | ||
| New | $9,299or best offer−22%+$50.00 shipping | ||
| New | $9,388or best offer−21%free shipping | ||
| New | $9,418or best offer−21%+$5.51 shipping | ||
| Used | $10,065or best offer−15%+$77.16 shipping | ||
| New | $10,127−15%free shipping | ||
| Open box | $10,150or best offer−15%free shipping | ||
| Used | $10,349or best offer−13%free shipping | ||
| Refurbished | $10,771−9%+$98.10 shipping | ||
| New | $11,849or best offerfree shipping | ||
| New | $11,850or best offerfree shipping | ||
| New | $11,850free shipping | ||
| New | $11,888+$58.00 shipping | ||
| Used | $11,995free shipping | ||
| New | $12,190or best offerfree shipping | ||
| New | $12,190or best offerfree shipping | ||
| New | $12,190or best offerfree shipping | ||
| New | $12,190or best offerfree shipping | ||
| New | $12,190or best offerfree shipping | ||
| New | $12,334or best offerfree shipping | ||
| Refurbished | $12,499or best offerfree shipping | ||
| Open box | $12,499or best offerfree shipping | ||
| Open box | $12,995free shipping | ||
| Open box | $12,995free shipping | ||
| Open box | $12,995free shipping | ||
| Open box | $12,995free shipping | ||
| New | $14,046or best offerfree shipping | ||
| New | $14,608or best offerfree shipping | ||
| New | $14,699or best offerfree shipping | ||
| New | $14,699or best offerfree shipping | ||
| New | $14,749or best offerfree shipping | ||
| New | $15,030or best offerfree shipping | ||
| New | $16,500or best offerfree shipping | ||
| New | $16,500or best offerfree shipping | ||
| New | $16,500or best offerfree shipping | ||
| New | $16,500or 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- #8 of 19 PCIe GPUs with 48 GB or more for the money: $2,980 per GeForce RTX 3090 of performance.
- Software: CUDA 11.8 and newer.
As much VRAM, about as fast, for less
- AMD Instinct MI210: $3,799, 211% of the GeForce RTX 3090, 64 GB
Faster, for the least
- NVIDIA A100 PCIe 40 GB: $5,588, 270% of the GeForce RTX 3090, 40 GB
- NVIDIA A100 PCIe 80 GB: $10,999, 301% of the GeForce RTX 3090, 80 GB
- NVIDIA RTX PRO 6000 Blackwell: $13,100, 367% of the GeForce RTX 3090, 96 GB
More VRAM, for not much more
- NVIDIA A16: $2,200, 93% of the GeForce RTX 3090, 64 GB
- AMD Instinct MI210: $3,799, 211% 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) | L40S needed | Costs now | Speed (est.) |
|---|---|---|---|
| Llama 3.1 8B | 1 card | $6,501 | 89–134 tokens/s |
| Qwen3 14B | 1 card | $6,501 | 48–72 tokens/s |
| gpt-oss 20B | 1 card | $6,501 | 119–178 tokens/s |
| Gemma 3 27B | 1 card | $6,501 | 26–39 tokens/s |
| Qwen3 32B | 1 card | $6,501 | 22–33 tokens/s |
| Llama 3.3 70B | 1 card | $6,501 | 10–15 tokens/s |
| gpt-oss 120B | 2 cards | $14,301 | 84–126 tokens/s |
| Qwen3 235B-A22B | 4 cards | $31,000 | 19–29 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) | L40S needed | Costs now | Per question (est.) |
|---|---|---|---|
| Julia-10.144B | 1 card | $6,501 | 3.5–7.6 ms |
| Laya0.421B | 1 card | $6,501 | 5.9–13 ms |
| Kev-4B4B | 1 card | $6,501 | 14–30 ms |
| lev4B | 1 card | $6,501 | 42–91 ms |
| OpenJev27B | 1 card | $6,501 | 51–109 ms |
| Clef27B | 1 card | $6,501 | 51–109 ms |
| Perplexity Decider v1.227B | 1 card | $6,501 | 51–109 ms |
| mammoth-127B | 1 card | $6,501 | 51–109 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 L40S. These open the shop's own search; we don't show their prices here.














































