Skip to content

Instantly share code, notes, and snippets.

@Alan-FGR
Last active July 24, 2018 13:19
Show Gist options
  • Save Alan-FGR/a9d5c37967822a1a0ab2d78686045539 to your computer and use it in GitHub Desktop.
Save Alan-FGR/a9d5c37967822a1a0ab2d78686045539 to your computer and use it in GitHub Desktop.
SIMD Frustum Culling Performance Results
processor simd naive (times in µs)
4.2ghz haswell: 320 360
3.75ghz haswell-e: 400 420
3.0ghz haswell: 580 610
3.3ghz sandy brid: 540 520 ! (i5 2500)
1.5ghz sandy brid:1500 3300 (pentium987)
3.0ghz haswell: 510 560 (g3220)
2.66ghz conroe: 640 620 ! (e6750)
4.0ghz skylake: 370 400 (i7 6700k)
3.2ghz ivy bridge: 475 540 (i5 3470)
3.0ghz haswell: 280 340 (g3220)
2.66ghz conroe: 370 400 (e6750)
4.0ghz skylake: 150 170 (i7 6700k)
3.3ghz sandy b EP: 180 220 (e5 1666)
@Alan-FGR
Copy link
Author

Alan-FGR commented Jul 23, 2018

GitHub Source Repo

How to report data (thank you):

  • Requirements: Windows 64 bit, version 7 or newer
  • Download the Win64 Binary (400kb), unzip and run it (MT Version)
  • Report the number of µs it's taking with and without the SIMD checkbox ticked (draw box always unticked) and the processor (just model is OK). It shouldn't be too jumpy but I guess depending on the system load it could get hard to read the numbers. Just report an estimation; this is no real science anyway :trollface:
  • Thank you ❤️

@Alan-FGR
Copy link
Author

Alan-FGR commented Jul 23, 2018

L5 result is reported by @nem0 so it's totally unreliable :trollface:

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment