rent-a-gpu.eu
Menu

NVIDIA · Ampere

Rent RTX 3090

NVIDIA GeForce RTX 3090, 24 GB.LLM inferenceRenderinggeneral ml
Full price from
€0.17
in stock, per GPU-hour, as 1×
Spot from
—
in stock, per GPU-hour
Offers in stock
1 of 1
full price or spot
Providers in stock
1 of 1
full price or spot

Offers (1)

Cheapest in stock
NosanaIn stock · 9 freeStart right away

1× RTX 309024 GB VRAM

Nosana network, hosts worldwide

Details

Stock and price checked 15 min ago.

Self-service: you start it yourself, with no order or sales contact.

nvidia-3090

Full price
€0.17
per hour, whole instance, 1 GPU
View offer

Price and availability history

Price per GPU-hour

Excluding VAT.

Providers with stock

Each provider counts once. Full price and spot are separate.

Share of offers in stock

Full price. Offers we can't say are really sold there are left out until they've been in stock once.

Units free

Only providers that report a number. Each line is one provider and one server size, because their counts cover different things and can't be added up. Providers that only say yes or no are under Stock by provider.

Stock by provider

In stock, yes or no, for every provider selling this GPU. A filled band means in stock: green at full price, orange as spot.

NosanaIn stock
Reports a number

In stock as

The server sizes you can rent right now.

Nosana
1×

Using several together

Can be linked in pairs with an NVLink 3 bridge at up to 112.5 GB/s, if the provider has fitted one. Most providers don't say, so ask before relying on it. Software still has to split the work, so the memory doesn't act as one big pool.

Specifications

Consumer card: memory bandwidth and FLOPS figures come from NVIDIA's Ampere whitepaper (Table 9), not a product datasheet. FP16, BF16 and INT8 are with sparsity (FP16 with FP16 accumulate, BF16 with FP32 accumulate), following the RTX 3080 Ti entry. NVIDIA's Blackwell whitepaper gives more precise tensor figures (FP16 284.6, BF16 142.4, INT8 569.4); the Ampere whitepaper is used. fp64Tflops is derived: NVIDIA states GA102 FP64 is 1/64th of FP32 (35.6 / 64). NVLink figures are NVIDIA's GA102 interface description (four x4 links, 14.0625 GB/s per direction per link, 112.5 GB/s total), the same convention as the RTX A6000 entry; NVIDIA states two RTX 3090 cards can be connected for SLI using NVLink, and 3-way and 4-way are not supported. nvSwitchSupport stays null.

Memory

VRAM
24 GB
Memory type
GDDR6X
Memory bandwidth
How fast the GPU reaches its memory. Often limits LLM speed
936 GB/s
Bus width
384 bit

Compute

FP64
Speed at 64-bit precision, used for scientific simulation
0.56 TFLOPS
FP32
Speed at 32-bit precision
35.6 TFLOPS
FP16
Speed at 16-bit precision
284 TFLOPS
BF16
16-bit AI format. Standard for training and full-quality inference
142 TFLOPS
INT8
8-bit integer speed, used by quantized models
568 TOPS
FP8 support
No
FP4 support
No
Transformer Engine
Hardware that speeds up FP8 and FP4
No
Compute cores
10,496

Multi-GPU and connectivity

Interconnect
NVLink 3
Interconnect bandwidth
How fast linked GPUs exchange data
112.5 GB/s
Links
4
Per link
28.125 GB/s
Max GPUs in fabric
2
PCIe generation
4

Sharing and security

MIG partitioning
Splits one GPU into several isolated smaller GPUs
No
Max MIG instances
0
Confidential computing
No

Hardware

Architecture
Ampere
Launch year
2020
Process node
Samsung 8N
Transistors
28.3 billion
Die size
628.4 mm²
Power (TDP)
Maximum power draw
350 W
Cooling
active

Video

NVENC engines
1
NVDEC engines
1
AV1 encode
No
AV1 decode
Yes
Sources (43)