NVIDIA · Turing
T4 16 GB
NVIDIA T4 with 16 GB GDDR6. Manufacturer DEC18 datasheet lists 300 GB/s bandwidth and mixed FP16/FP32 tensor throughput. Confirm usable memory and system resources for each rental.
Profile checked · Primary source ↗
- Memory
- 16 GB
- Bandwidth
- 300 GB/s
- Offers
- 1
- FP16 dense
- 65 TFLOPS
- BF16 dense
- Not published
- FP8 dense
- Not published / unsupported
- Interconnect
- PCIe Gen3 x16, 32 GB/s.
DEC18 datasheet: FP16 inputs with FP32 accumulation, no structural sparsity. BF16/FP8 peaks are not published in this source. Bandwidth follows the datasheet specification. · Specification source 1 ↗
Available capacity
Observed rental offers
| GPU / provider | Billing & commitment | Per GPU | Instance total | Stock & region | Checked / actions |
|---|---|---|---|---|---|
| T4 16 GBAmazon Web Servicesg4dn.xlarge Linux shared · 16 GB per GPU | HourlyOn-demandNo monthly commitmentOn-demand Linux shared tenancy; no monthly commitment. 60-second minimum. Capacity is not verified. | USD 0.526/hr per GPU | USD 0.526/hr · 1 GPU | Stock not verifiedus-east-1AWS EC2 | Browse provider Details & price history |
Prices and availability can change. Confirm the current offer with the provider.
Measured performance
Inference benchmarks
| Model | GPU setup | Engine | Quant. | Workload | Reported tok/s | Evidence | Max-load capacity scenarioUSD / million output tokens |
|---|---|---|---|---|---|---|---|
| No benchmark records have been supplied yet. | |||||||