L4 vs Tesla T4
Manufacturer specifications of the L4 and the Tesla T4, side by side, per GPU. A value the manufacturer does not publish for the exact model is marked as such, never estimated.
NVIDIA L4 24GB
Ada Lovelace
- GPU memory
- 24 GB
- Memory bandwidth
- 300 GB/s
Live price on EverXCloud · from $0.44 per GPU-hour
Suited workloads
NVIDIA Tesla T4 16GB
Turing
- GPU memory
- 16 GB
- Memory bandwidth
- 320 GB/s
Live price on EverXCloud · from $0.18 per GPU-hour
Suited workloads
Key differences
- The L4 has 1.5× the GPU memory of the Tesla T4 (24 GB vs 16 GB).
- The Tesla T4 has 1.1× the memory bandwidth of the L4 (320 vs 300 GB/s), which usually sets the pace of LLM token generation.
Specifications
| Specification | L4 | Tesla T4 |
|---|---|---|
| GPU memory | 24 GB (higher) | 16 GB |
| Memory type | GDDR6 | GDDR6 |
| Memory bandwidth | 300 GB/s | 320 GB/s (higher) |
| FP16 tensor | 121 TFLOPS (higher) | 65 TFLOPS |
| BF16 tensor | 121 TFLOPS | Not published |
| FP8 tensor | 242 TFLOPS | Not published |
| Max power | 72 W | 70 W |
| Architecture | Ada Lovelace | Turing |
| Interconnect | PCIe 4.0 | PCIe 3.0 |
The higher value of each measured row is shown in bold. Manufacturer specifications, per GPU. Tensor throughput is shown without sparsity. Values the manufacturer does not publish for this exact model are left out.
What fits in memory, on one GPU
| Workload | L4 | Tesla T4 |
|---|---|---|
| LLM inference, 16-bit | ≈ 9B parameters | ≈ 6B parameters |
| LLM inference, 4-bit | ≈ 30B parameters | ≈ 20B parameters |
| Full training, mixed precision | ≈ 1B parameters | Under 1B parameters |
Cautious estimates counting 80% of memory for model weights; context length, batch size and framework change the real limit. Several GPUs together hold larger models.
L4 vs Tesla T4: frequently asked questions
Which has more memory, the L4 or the Tesla T4?
The L4 has 24 GB of GPU memory, against 16 GB for the Tesla T4.
Which is faster for LLM inference, the L4 or the Tesla T4?
Generating tokens is usually bound by memory bandwidth, and the Tesla T4 has 1.1× the bandwidth of the L4. Memory size then decides which models fit. Measure your own model before a large run.
Can I rent the L4 and the Tesla T4 by the hour?
Yes, when sellers list them on the EverXCloud spot exchange: you buy GPU hours at the live price from a prepaid balance in euros, US dollars or pounds. Prices are shown excluding tax and EverXCloud is the merchant of record for every rental.
Be ready on launch day
Create your account now: you can top up your balance and rent the GPU that fits your workload as soon as the market opens.
Have GPU servers?
Businesses can join the EverXCloud Network and sell hours of their GPU servers on the spot exchange.