GPU MODELS / RTX 3090

RTX 3090 cloud GPU pricing

52 listings across 25 providers. Cheapest published rate: $0.090/GPU/hr at Salad. Sorted by per-GPU hourly price. Every row links to the provider's own published price.

WHAT IS THE RTX 3090

The RTX 3090 remains the budget multi-GPU favorite: 24 GB per card, and uniquely among consumer cards its generation supports NVLink bridge pairs - 48 GB linked for under a dollar an hour on parts of the panel. It is the cheapest credible path to fine-tuning 13B-34B models.

MAKER
NVIDIA
ARCHITECTURE
Ampere
LAUNCHED
2020
MEMORY
24 GB GDDR6X
MEMORY BANDWIDTH
936 GB/s
TENSOR COMPUTE
142 TFLOPS FP16 tensor dense (285 sparse)
POWER DRAW
350 W
INTERCONNECT
PCIe 4.0 x16 (NVLink bridge on pairs)

WHAT IT IS GOOD AT

Budget fine-tuning
13B QLoRA per card; 34B across an NVLink pair.
Small-model inference
936 GB/s keeps quantized serving snappy.

WHO SHOULD NOT RENT THE RTX 3090

Single-card speed - a 4090 is ~2.3x faster for similar money per card in many markets.
Anything needing FP8 or modern media engines.
ProviderConfigVRAMLocation$/hr$/GPU/hrSource
SpotGPUsSpotGPUsRTX 3090 × 124 GBNot published$0.080$0.080STALEreceipt
SaladSaladRTX 3090 × 124 GBGlobal (distributed)$0.090$0.090receipt
CloreCloreRTX 3090 × 124 GBRU$0.110$0.110receipt
CloreCloreRTX 3090 × 224 GBRU$0.221$0.110receipt
CloreCloreRTX 3090 × 124 GBRS$0.120$0.120receipt
CloreCloreRTX 3090 × 124 GBUS$0.136$0.136receipt
Theta EdgeCloudTheta EdgeCloudRTX 3090 × 124 GBNot published$0.140$0.140receipt
CloreCloreRTX 3090 × 124 GBAR$0.142$0.142receipt
CloreCloreRTX 3090 × 124 GBBR$0.146$0.146receipt
AkashAkashRTX 3090 × 124 GBGlobal$0.150$0.150STALEreceipt
AtmosComputeAtmosComputeRTX 3090 (from) × 124 GBNot published$0.150$0.150receipt
CloreCloreRTX 3090 × 124 GBCA$0.158$0.158receipt
LiumLiumRTX 3090 × 124 GBNot published$0.160$0.160receipt
AI GalaxyAI GalaxyRTX 3090 × 124 GBChina$0.165$0.165receipt
CloreCloreRTX 3090 × 124 GBPL$0.169$0.169receipt
CloreCloreRTX 3090 × 124 GBBH$0.175$0.175receipt
CloreCloreRTX 3090 × 124 GBVN$0.179$0.179receipt
QuickPodQuickPodRTX 3090 × 124 GBNot published$0.180$0.180STALEreceipt
AutoDLAutoDLRTX 3090 × 124 GBChina$0.208$0.208receipt
TensorDockTensorDockRTX 3090 × 124 GBMultiple regions$0.210$0.210receipt
gpu.aigpu.aiRTX 3090 × 124 GBNot published$0.210$0.210receipt
RunPodRunPodRTX 3090 × 124 GBGlobal$0.220$0.220receipt
GPUBrazilGPUBrazilRTX 3090 × 124 GBAsia-Pacific$0.238$0.238receipt
GPUBrazilGPUBrazilRTX 3090 × 124 GBGlobal$0.238$0.238receipt
HydraHostHydraHostRTX 3090 × 124 GBNot published$0.240$0.240receipt
Hostnot GPUHostnot GPURTX 3090 × 224 GBAsia-Pacific$0.480$0.240receipt
FeaturizeFeaturizeRTX 3090 × 125.4 GBChina$0.242$0.242receipt
Hostnot GPUHostnot GPURTX 3090 × 824 GBUS East$1.980$0.247receipt
Hostnot GPUHostnot GPURTX 3090 × 124 GBAsia-Pacific$0.250$0.250receipt
Hostnot GPUHostnot GPURTX 3090 × 124 GBUS East$0.260$0.260receipt
Hostnot GPUHostnot GPURTX 3090 × 124 GBUS West$0.260$0.260receipt
Hostnot GPUHostnot GPURTX 3090 × 124 GBEU East$0.270$0.270receipt
Hostnot GPUHostnot GPURTX 3090 × 224 GBEU East$0.540$0.270receipt
MatPoolMatPoolRTX 3090 × 124 GBChina$0.277$0.277receipt
CloudGPUCloudGPURTX 3090 × 124 GBNot published$0.290$0.290receipt
Hostnot GPUHostnot GPURTX 3090 × 124 GBCanada Central$0.290$0.290receipt
Hostnot GPUHostnot GPURTX 3090 × 124 GBEU West$0.290$0.290receipt
Hostnot GPUHostnot GPURTX 3090 × 224 GBEU West$0.580$0.290receipt
Hostnot GPUHostnot GPURTX 3090 × 424 GBAsia-Pacific$1.270$0.318receipt
Hostnot GPUHostnot GPURTX 3090 × 224 GBCanada Central$0.660$0.330receipt
GPUBrazilGPUBrazilRTX 3090 × 124 GBEurope$0.384$0.384receipt
GPUBrazilGPUBrazilRTX 3090 × 124 GBNorth America$0.397$0.397receipt
SynexaSynexaRTX 3090 × 124 GBGlobal$0.430$0.430receipt
NovaGPU (IP ServerOne)NovaGPU (IP ServerOne)RTX 3090 × 124 GBMalaysia$0.447$0.447receipt
NovaGPU (IP ServerOne)NovaGPU (IP ServerOne)RTX 3090 × 496 GBMalaysia$1.791$0.448receipt
NovaGPU (IP ServerOne)NovaGPU (IP ServerOne)RTX 3090 × 248 GBMalaysia$0.897$0.448receipt
TeamCloudTeamCloudRTX 3090 (from) × 124 GBMalaysia$0.482$0.482receipt
ParateraParateraRTX 3090 × 124 GBChina$0.494$0.494receipt
GPUBrazilGPUBrazilRTX 3090 × 124 GBSouth America$0.546$0.546receipt
Hostnot GPUHostnot GPURTX 3090 × 124 GBEU East$0.660$0.660receipt
DCPDCPRTX 3090 (from) × 124 GBSaudi Arabia$0.667$0.667receipt
Safari CloudSafari CloudRTX 3090 × 124 GBAfrica$1.111$1.111receipt

THE RTX 3090 MARKET, BY THE NUMBERS

PROVIDERS
25
CONFIGURATIONS
52
MEDIAN $/GPU/HR
$0.25
SPREAD
12.3x

25 providers publish 52 RTX 3090 configurations on the panel today. Published per-GPU hourly rates run from $0.090 (Salad) to $1.11 (Safari Cloud), a 12x spread between the cheapest and the most expensive published rate for the same chip. The median listing sits at $0.25/GPU/hr. 29 rows were reported in stock by the provider at last observation; the rest are published catalog rates. All rates are on-demand - committed-use and reserved pricing is excluded.

CHEAPEST RTX 3090 BY GEOGRAPHY
united statesClore$0.136/hrbrazilClore$0.146/hrcanadaClore$0.158/hrchinaAI Galaxy$0.165/hrpolandClore$0.169/hrvietnamClore$0.179/hrmalaysiaNovaGPU (IP ServerOne)$0.447/hr
RTX 3090 PRICE TREND, DAILY PANEL SNAPSHOTS

Across every RTX 3090 row on the panel, the daily floor held flat at $0.090/GPU/hr over 53 days of tracking. The median continuously-tracked listing held flat at $0.16. Snapshots run daily; the full history is public in the repo.

RTX 3090 PROVIDER NOTES
Salad · 1 config · 1 region
Sets the panel floor for this model.
from $0.090/hr
Clore · 10 configs · 9 regions · 10 in stock
Widest regional footprint here - 9 regions.
from $0.110/hr
Theta EdgeCloud · 1 config
Published catalog rates.
from $0.140/hr
AtmosCompute · 1 config
Published catalog rates.
from $0.150/hr
Lium · 1 config
Published catalog rates.
from $0.160/hr
AI Galaxy · 1 config · 1 region
Published catalog rates.
from $0.165/hr
AutoDL · 1 config · 1 region
Published catalog rates.
from $0.208/hr
TensorDock · 1 config · 1 region
Published catalog rates.
from $0.210/hr
BUYING RTX 3090: WHAT THE PANEL SAYS

At the current floor, one RTX 3090 running around the clock costs about $66 a month (730 hours of on-demand arithmetic). Node shape matters: the single-GPU floor is $0.090/GPU/hr against $0.247/GPU/hr on an 8-GPU node (+175.0%). Per-GPU floors by config size: 1x $0.09 · 2x $0.11 · 4x $0.32 · 8x $0.25. Compare the alternatives below before committing - the same budget often buys more than one chip class.

RTX 3090 VS THE ALTERNATIVES

RTX 3090 vs RTX 4090
4090 is ~2.3x faster per card; 3090 has NVLink pairs and lower panel prices. Single card: 4090. Two-card 48 GB jobs: 3090 pair.
RTX 3090 vs RTX A5000
Same 24 GB and NVLink support with workstation drivers and lower power - the datacenter-friendly sibling.

Compare live rates: RTX 3090 vs RTX 4090 · RTX 3090 vs RTX A5000 · RTX 3090 vs RTX 3080 Ti · RTX 3090 vs RTX 3090 Ti

RTX 3090 PRICING FAQ

What is the cheapest RTX 3090 cloud GPU right now?
The cheapest published RTX 3090 rate on the panel is $0.090 per GPU-hour at Salad (RTX 3090 x 1, Global (distributed)). Every price links to the provider's own published rate.
How much does a RTX 3090 cost per hour?
Across 25 providers tracked today, published per-GPU hourly rates for RTX 3090 run from $0.09 to $1.11, with a median of $0.25. Committed-use and reserved rates are excluded; everything shown is on-demand.
How many providers rent RTX 3090 GPUs?
25 providers publish 52 distinct RTX 3090 configurations on Compute Cafe. 29 of those rows were reported in stock by the provider's live endpoint at last observation.
Where is RTX 3090 cheapest?
By published per-GPU hourly rate, the cheapest geography for RTX 3090 today is united states at $0.14/hr (Clore). The spread across 7 tracked geographies runs to $0.45/hr in malaysia.
Are RTX 3090 prices going up or down?
Over the 53 days Compute Cafe has tracked the panel, the RTX 3090 floor is holding flat: $0.090 to $0.090/GPU/hr (+0.0%). The median continuously-tracked listing moved from $0.16 to $0.16 (+0.0%). Daily snapshots are public in the repo, so the trend is auditable.
What does a RTX 3090 cost per month?
At the current floor of $0.090/GPU/hr, a RTX 3090 running 24/7 costs about $66 per month (730 hours of on-demand arithmetic on the cheapest published rate). Committed-use contracts can price lower; the panel tracks on-demand rates only.
Is a full 8-GPU RTX 3090 node cheaper than a single card?
On today's panel: the single-GPU floor is $0.090/GPU/hr and the 8-GPU node floor is $0.247/GPU/hr - more expensive per GPU on a full node (+175.0%). Check the table for the exact configs behind each floor.
Why do people still rent 3090s?
NVLink pairs and price. Two 3090s link into a 48 GB pool for fine-tuning 30B-class models at consumer-card rates - nothing newer in the consumer line offers the bridge.

See also alternatives to the RTX 3090 · cheapest GPU-hour per model · inference chips · training chips · by location