# CUDA Float64 Float32 difference

**URL:** https://discourse.julialang.org/t/cuda-float64-float32-difference/135922
**Category:** General Usage
**Created:** [February 28, 2026, 3:40pm UTC](https://discourse.julialang.org/t/cuda-float64-float32-difference/135922 "2026-02-28T15:40:58Z")
**Posts on this page:** 2
**Page:** 1

<div class="post-metadata">

### Author: ![andrey2185](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/andrey2185/32/9889_2.png) [@andrey2185](https://discourse.julialang.org/u/andrey2185)
#### Post date: [February 28, 2026, 3:40pm UTC](https://discourse.julialang.org/t/cuda-float64-float32-difference/135922/1 "2026-02-28T15:40:58Z")

</div>

Hello. I use RTX 3070. Why is there no difference?

```julia-auto
using CUDA, BenchmarkTools
function add!(X::AbstractVector{T}) where T
    return X .+= T(1)
end
A = CUDA.zeros(Float32, 10_000_000);
B = CUDA.zeros(Float64, 10_000_000);
@btime add!(A); # -> 5.250 μs (37 allocations: 1.09 KiB)
@btime add!(B); # -> 5.267 μs (37 allocations: 1.09 KiB)

```

---

<div class="post-metadata">

### Author: ![andrey2185](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/andrey2185/32/9889_2.png) [@andrey2185](https://discourse.julialang.org/u/andrey2185)
#### Post date: [February 28, 2026, 3:45pm UTC](https://discourse.julialang.org/t/cuda-float64-float32-difference/135922/2 "2026-02-28T15:45:28Z")

</div>

```julia-auto
using CUDA, BenchmarkTools
function add!(X::AbstractVector{T}) where T
    return X .+= T(1)
end
A = CUDA.zeros(Float32, 10_000_000);
B = CUDA.zeros(Float64, 10_000_000);
@btime CUDA.@sync add!(A); # -> 225.800 μs (91 allocations: 1.94 KiB)
@btime CUDA.@sync add!(B); # -> 437.800 μs (77 allocations: 1.72 KiB)

```
