…
A10 vs L4
Cloud rental price, compute per dollar and specs, side by side.
Open in the interactive comparer →
A10$0.85lowest on-demand per GPU·h
$6.82 per PFLOP·h · 5 providersL4$0.27lowest on-demand per GPU·h
$2.26 per PFLOP·h · 13 providersWhat actually differs
- The A10 has 1.0× more dense FP16/BF16 throughput than the L4 (0.125 PF vs 0.121 PF).
- The A10 has 2.0× more memory bandwidth than the L4 (600 vs 300 GB/s), which usually limits LLM token generation. Memory: 24 GB vs 24 GB.
- Only the L4 has native FP8 tensor cores.
- The L4 is 68% cheaper per hour than the A10 at the lowest on-demand prices ($0.85 vs $0.27).
- Per unit of compute, the L4 is the better deal: $2.26 vs $6.82 per PFLOP-hour of dense FP16/BF16.
| Spec / price | A10 | L4 |
|---|---|---|
| Memory | 24 GB GDDR6 | 24 GB GDDR6 |
| Memory bandwidth | 600 GB/sbetter | 300 GB/s |
| FP16 / BF16 (dense) | 0.125 PFbetter | 0.121 PF |
| FP8 (dense) | — | 0.242 PF |
| FP4 (dense) | — | — |
| INT8 (dense) | 0.25 PFbetter | 0.242 PF |
| TF32 (dense) | 0.0625 PFbetter | 0.06 PF |
| FP32 (dense) | 0.0312 PFbetter | 0.0303 PF |
| FP64 (dense) | 0.0005 PF | 0.0005 PF |
| Power | 150 W | 72 W |
| Architecture | Ampere (2021) | Ada Lovelace (2023) |
| Lowest on-demand | $0.85 | $0.27better |
| Median on-demand | $1.42 | $0.70better |
| Lowest spot | $0.59 | $0.12better |
| FP16/BF16 $ per PFLOP·h | $6.82 | $2.26better |
| Providers | 5 | 13 |
A10 vs L4: questions
Is the A10 faster than the L4?
The A10 has 1.0× more dense FP16/BF16 compute than the L4. The A10 has 2.0× more memory bandwidth than the L4. Real-world speedups depend on the workload: training and prefill scale with compute, while LLM decoding scales mostly with memory bandwidth.
Which is cheaper per FLOP, the A10 or the L4?
Right now the L4 is cheaper per unit of compute: $2.26 vs $6.82 per PFLOP-hour at the lowest on-demand prices.
How much does it cost to rent an A10 vs an L4 for a month?
Running one GPU around the clock at the lowest on-demand price costs about $622 per month for the A10 and $200 for the L4.
Which has more memory, the A10 or the L4?
Both have 24 GB of VRAM; the A10 has more bandwidth.