# Global memory bandwidth profiling?

**URL:** <https://forums.developer.nvidia.com/t/global-memory-bandwidth-profiling/24784>\
**Category:** CUDA Programming and Performance\
**Created:** [November 1, 2011, 11:56am UTC](https://forums.developer.nvidia.com/t/global-memory-bandwidth-profiling/24784 "2011-11-01T11:56:04Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![enot](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@enot](https://forums.developer.nvidia.com/u/enot)\
**Post date:** [November 1, 2011, 11:56am UTC](https://forums.developer.nvidia.com/t/global-memory-bandwidth-profiling/24784/1 "2011-11-01T11:56:04Z")

</div>

I don’t understand why in my WebCL tests float4 copy is 3 times faster than float one on GTX?  
[http://www.ibiblio.org/e-notes/webcl/gtx560.html](http://www.ibiblio.org/e-notes/webcl/gtx560.html)

Evgeny

---

<div class="post-metadata">

**Author:** ![enot](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@enot](https://forums.developer.nvidia.com/u/enot)\
**Post date:** [November 14, 2011, 5:44am UTC](https://forums.developer.nvidia.com/t/global-memory-bandwidth-profiling/24784/2 "2011-11-14T05:44:47Z")

</div>

“Global memory is accessed via 32-, 64-, or 128-byte memory transactions”. For (8×8) thread blocks a warp reads 8×4 bytes = 32 bytes in one float transaction but for (32×8) thread blocks it reads 32×4 bytes = 128 bytes (as like as in float4 transactions for (8×8) thread blocks). 128-byte transactions are ~3-4 times faster on GTX than-32 byte ones.