GPU Comparison

NVIDIA RTX 5090 vs NVIDIA RTX 4080.

Side-by-side comparison of specs, performance, and cloud rental pricing. NVIDIA RTX 4080 is the cheaper option at $0.35/hr.

TL;DR

Choose NVIDIA RTX 5090 if you need 32GB+ on a single GPU (16GB more than NVIDIA RTX 4080).

Choose NVIDIA RTX 5090 if you need raw FP16 throughput (157% faster than NVIDIA RTX 4080).

Switch comparison

vs
NVIDIA

Blackwell

NVIDIA RTX 5090

$0.55/hr

$0.45-$0.65/hr range · 3 regions

VRAM32GB GDDR7
FP16250 TFLOPS
Bandwidth1.8 TB/s
Deploy NVIDIA RTX 5090Full pricing details
NVIDIA

Ada Lovelace

NVIDIA RTX 4080

$0.35/hrCheaper

$0.25-$0.45/hr range · 4 regions

VRAM16GB GDDR6X
FP1697.4 TFLOPS
Bandwidth736 GB/s
Deploy NVIDIA RTX 4080Full pricing details

At a glance

Spec by spec.

Visual comparison across the four specs that matter most for AI workloads. Green bar marks the winner.

VRAM

NVIDIA RTX 5090
32 GB
NVIDIA RTX 4080
16 GB

Memory Bandwidth

NVIDIA RTX 5090
1.8 TB/s
NVIDIA RTX 4080
736 GB/s

FP16 Performance

NVIDIA RTX 5090
250 TFLOPS
NVIDIA RTX 4080
97.4 TFLOPS

Hourly Price

NVIDIA RTX 5090
$0.55/hr
NVIDIA RTX 4080
$0.35/hr

Full specs

Complete spec sheet.

SpecNVIDIA RTX 5090NVIDIA RTX 4080
ArchitectureBlackwellAda Lovelace
VRAM32 GB16 GB
Memory TypeGDDR7GDDR6X
Memory Bandwidth1.8 TB/s736 GB/s
FP32 Performance125 TFLOPS48.7 TFLOPS
FP16 Performance250 TFLOPS97.4 TFLOPS
INT8 Performance1000 TOPS390 TOPS
TDP575 W320 W
Form FactorConsumerConsumer
NVLinkNoNo
Max GPU/Node22
Hourly Price$0.55/hr$0.35/hr

FAQ

Frequently asked.

Pick a GPU and deploy.

Both GPUs available on Runcrate · per-second billing · no commitments

Other comparisons