Compare · prices checked 2026-09-03
RTX 5090 vs RTX 4090: specs, price per hour, which to rent
NVIDIA RTX 5090 (32 GB, $0.318/hr) against NVIDIA RTX 4090 (24 GB, $0.262/hr): public specs side by side, live fixed prices from our sheet, what fits in each card's VRAM, and a verdict written for real workloads — both are rentable right now.
RTX 5090 vs RTX 4090 specifications
Public NVIDIA figures (dense, non-sparsity). The last column is RTX 5090 relative to RTX 4090.
| Spec | RTX 5090 | RTX 4090 | Difference |
|---|---|---|---|
| Architecture | Blackwell (2025) | Ada Lovelace (2022) | — |
| VRAM | 32 GB GDDR7 | 24 GB GDDR6X | +33% |
| Memory bandwidth | 1,792 GB/s | 1,008 GB/s | +78% |
| FP16 tensor (dense) | 419 TFLOPS | 330 TFLOPS | +27% |
| FP32 | 104.8 TFLOPS | 82.6 TFLOPS | +27% |
| CUDA cores | 21,760 | 16,384 | +33% |
| TDP | 575 W | 450 W | +28% |
| PCIe · NVLink | Gen 5.0 · no NVLink | Gen 4.0 · no NVLink | — |
| PowerScore (RTX 3090 = 100) | 295 | 232 | +27% |
| Max GPUs per machine | 8× | 8× | — |
RTX 5090 vs RTX 4090 price per hour
Fixed rates from our sheet — every on-demand price is the marketplace median × 0.70, rounded down. Monthly = 730 hours.
| Rate | RTX 5090 | RTX 4090 | Cheaper |
|---|---|---|---|
| On-demand, per GPU-hour | $0.318 | $0.262 | RTX 4090 (−18%) |
| Interruptible, per GPU-hour | $0.159 | $0.131 | RTX 4090 |
| Reserved (3 mo), per GPU-hour | $0.206 | $0.170 | RTX 4090 |
| On-demand, per month | $232 | $191 | RTX 4090 |
| Market median (reference) | $0.45 | $0.37 | — |
| $ per 1,000 FP16 TFLOP-hours | $0.76 | $0.79 | RTX 5090 (better value) |
| $ per GB of VRAM per hour | $0.0099 | $0.0109 | RTX 5090 (better value) |
Try a full month with storage and bandwidth in the GPU cost calculator.
What fits in VRAM: 32 GB vs 24 GB
| Workload | RTX 5090 | RTX 4090 |
|---|---|---|
| Largest LLM in FP16, one card | ~13B | ~8B |
| Largest LLM at 4-bit, one card | ~49B | ~32B |
| Flux dev (FP8, ~17 GB) | fits | fits |
| Wan 2.x 14B video (offloaded) | yes | yes |
| 70B 4-bit LLM on one card | no | no |
Rules of thumb: ~2.4 GB per billion parameters in FP16 all-in, ~0.62 GB in 4-bit. Full tables in the VRAM guide.
Verdict: which should you rent?
The RTX 5090 adds 8 GB (32 vs 24 GB), 78% more memory bandwidth and FP4 tensor cores, landing 30–70% faster on LLM inference; the RTX 4090 is cheaper per hour with the deeper software ecosystem. Rent the 5090 for memory-bound serving and video models, the 4090 for image generation and budget LoRA runs.
- Cheaper per hour: RTX 4090 ($0.262 vs $0.318, −18%).
- More VRAM: RTX 5090 (32 GB vs 24 GB).
- More FP16 throughput: RTX 5090 (about 1.3×).
- Best value per TFLOP-hour: RTX 5090.
- Best value per GB of VRAM: RTX 5090.
- Multi-GPU: RTX 5090 over PCIe · RTX 4090 over PCIe.
Related comparisons
- RTX 4090 vs RTX 3090 $0.262 vs $0.104 per hour Compare
- RTX 4090 vs A100 PCIE $0.262 vs $0.662 per hour Compare
- RTX 4090 vs L40S $0.262 vs $0.466 per hour Compare
- RTX 5090 vs L40S $0.318 vs $0.466 per hour Compare
- RTX 5090 vs A100 SXM4 $0.318 vs $0.583 per hour Compare
- RTX PRO 6000 WS vs RTX 5090 $1.097 vs $0.318 per hour Compare
RTX 5090 vs RTX 4090: FAQ
Deeper reading: H100 vs H200 vs B200, RTX 4090 vs RTX 5090, how cloud GPU pricing works.
Is the RTX 5090 faster than the RTX 4090?
On dense FP16 tensor throughput the RTX 5090 leads by about 1.3× (419 vs 330 TFLOPS). Memory bandwidth matters as much for inference: RTX 5090 1,792 GB/s vs RTX 4090 1,008 GB/s.
Which is cheaper to rent, the RTX 5090 or the RTX 4090?
The RTX 4090: $0.262/hr on-demand versus $0.318/hr — 18% less. Interruptible rates are $0.159 (RTX 5090) and $0.131 (RTX 4090). Per TFLOP-hour the better value is the RTX 5090.
Which has more VRAM and what does that change?
The RTX 5090 has 32 GB versus 24 GB. In LLM terms that is roughly a 13B FP16 model (or ~49B in 4-bit) on one card against 8B FP16 (~32B 4-bit). If the model does not fit, speed is irrelevant.
Can I rent both on PowerGPU right now?
Yes — 516 × RTX 5090 and 439 × RTX 4090 are online as this page renders, deployable in about 30 seconds, billed per second, paid in crypto with no KYC.