Hey all.
Was looking at purchasing a Titan X, but was concerned about its single-precision. Does anyone have any experience out there with single-precision rendering vs. double, for Octane tasks?
GPU Single precision vs. double-precision
Forum rules
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
Octane doesn't use double precision with the small exception of hair segments. Everything else is single precision only.treddie wrote:Hey all.
Was looking at purchasing a Titan X, but was concerned about its single-precision. Does anyone have any experience out there with single-precision rendering vs. double, for Octane tasks?
In theory there is no difference between theory and practice. In practice there is. - Yogi Berra
-
stratified
- Posts: 945
- Joined: Wed Aug 15, 2012 6:32 am
- Location: Auckland, New Zealand
Most of the calculations in Octane are done in single precision. Only a few features use double precision math (IIRC displacement and hair). I'm not sure what the exact question is, but Octane there's no way to "switch" to double precision rendering.
cheers,
Thomas
cheers,
Thomas
Thank you for the excelellent answers! What happens when you use a single-precision card with hair? Is it noticeable in the render?
Win7 | Geforce TitanX w/ 12Gb | Geforce GTX-560 w/ 2Gb | 6-Core 3.5GHz | 32Gb | Cinema4D w RipTide Importer and OctaneExporter Plugs.
all You need for Octane (& majority of other GPU based engines on the market now) is single precision (SP) performance. So, don't worry 'bout TitanX not having too much of it - that's the feature You don't feel any effect of not having it
to illustrate this 780 had way less of DP compared to Titan cards based on same chip - but performance using Octane is not influenced at all =) the same goes for QUADRO/TESLAs vs TITANs apart from sometimes more vRAM, those will not give any better performance..as nor DP, nor ECC could be used..-feature You would pay significant extra amount =)
to illustrate this 780 had way less of DP compared to Titan cards based on same chip - but performance using Octane is not influenced at all =) the same goes for QUADRO/TESLAs vs TITANs apart from sometimes more vRAM, those will not give any better performance..as nor DP, nor ECC could be used..-feature You would pay significant extra amount =)
but this interest me a bit..as Marcus mentioned in other threads too about this DP & hairs issuetreddie wrote:Thank you for the excelellent answers! What happens when you use a single-precision card with hair? Is it noticeable in the render?
could You Guys talk more about this? any tests You've done that would reveal the hit on performance?abstrax wrote: Octane doesn't use double precision with the small exception of hair segments. Everything is single precision only.
No, we didn't notice any real difference.treddie wrote:Thank you for the excelellent answers! What happens when you use a single-precision card with hair? Is it noticeable in the render?
In theory there is no difference between theory and practice. In practice there is. - Yogi Berra