PerGB

NVIDIA H200 NVL

Live eBay prices, cheapest first. The best one right now is $32,300 (new).

Specs

VRAM
141 GB HBM3e
Bandwidth
4,800 GB/s
FP32
60 TFLOPS
FP16
835 TFLOPS
Power
600 W
Connector
PCIe 5.0
Released
2024
Launch price
$32,000 (est.)
CUDA compute
9.0
Llama 3.1 8B at Q4 (est.)
495–742 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.
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 eBay
Lowest now$32,300
Median ask$77,450
Per GB VRAM$229.08
vs launch price1% over
Live listings16

From the live listings below; shipping included.

Listings

Live on eBay
ListingConditionPriceChecked
New$32,300Check carefullyshipping not listed
New$60,000−23%free shipping
Open box$62,999or best offer−19%free shipping
Used$64,020or best offer−17%+$20.00 shipping
New$65,000or best offer−16%free shipping
Open box$68,000−12%free shipping
Open box$72,000or best offer−7%free shipping
Open box$76,999free shipping
Open box$77,900or best offerfree shipping
New$87,456or best offerfree shipping
Open box$92,000or best offerfree shipping
New$92,000or best offerfree shipping
New$92,000or best offerfree shipping
New$92,000or best offerfree shipping
Open box$92,200or best offerfree shipping
Open box$92,200or 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

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)H200 NVL neededCosts nowSpeed (est.)
Llama 3.1 8B1 card$32,300495–742 tokens/s
Qwen3 14B1 card$32,300267–401 tokens/s
gpt-oss 20B1 card$32,300660–990 tokens/s
Gemma 3 27B1 card$32,300144–217 tokens/s
Qwen3 32B1 card$32,300121–181 tokens/s
Llama 3.3 70B1 card$32,30056–84 tokens/s
gpt-oss 120B1 card$32,300466–699 tokens/s
Qwen3 235B-A22B2 cards$92,300108–162 tokens/s
DeepSeek R1 / V3 671B3 cards$155,29964–96 tokens/s

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)H200 NVL neededCosts nowPer question (est.)
Julia-10.144B1 card$32,3001.0–2.1 ms
Laya0.421B1 card$32,3001.7–3.5 ms
Kev-4B4B1 card$32,3004.0–8.5 ms
lev4B1 card$32,30012–26 ms
OpenJev27B1 card$32,30014–30 ms
Clef27B1 card$32,30014–30 ms
Perplexity Decider v1.227B1 card$32,30014–30 ms
mammoth-127B1 card$32,30014–30 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 H200 NVL. These open the shop's own search; we don't show their prices here.