Hardware
Our GPUs vs RTX
Same Ada Lovelace generation as the RTX 40 series. More VRAM, ECC memory, and you rent it by the minute.
| NVIDIA L4 | NVIDIA L40S | RTX 4080 | RTX 4090 | |
|---|---|---|---|---|
| Generation | Ada Lovelace | Ada Lovelace | Ada Lovelace | Ada Lovelace |
| VRAM | 24 GB | 48 GB | 16 GB | 24 GB |
| CUDA cores | 7,424 | 18,176 | 9,728 | 16,384 |
| ECC memory | Yes | Yes | No | No |
| Max per machine | 8× | 8× | 1× | 1× |
| How you get one | from $1.29/hr | from $2.79/hr | $1,199 + a tower | $1,599 + a tower |
A 4090 clocks higher for gaming-style throughput. The L4 and L40S trade peak clocks for VRAM, ECC stability and server-grade cooling, and you only pay while they run.
Which one
Which one is for you?
NVIDIA L4
24 GB VRAM
from $1.29/hr
Modelling, editing, real-time viz and everyday creative work.
More VRAM than an RTX 4080, for a fraction of one day's rent.
NVIDIA L40S
48 GB VRAM
from $2.79/hr
Final renders, simulation, AI training and heavy scenes.
Twice the VRAM of an RTX 4090, up to 8 GPUs in one Studio.
Start on the L4. Move to the L40S when VRAM or render times bite. Switching keeps everything installed.