Welcome to od|forum

Register now to gain access to all of our features. Once registered and logged in, you will be able to contribute to this site by submitting your own content or replying to existing content. You'll be able to customize your profile, receive reputation points as a reward for submitting content, while also communicating with other members via your own private inbox, plus much more! This message will be removed once you have signed in.

kiaspid

OpenCL simulations on ADM GPU

hi guys, sorry can't speak English well. I planning to buy new hardware for OpenCL simulations on GPU. I study a little about pro-GPUs computing performance and find some information in https://en.wikipedia.org/wiki/FLOPS "Built using commercially available parts. AMD Ryzen 7 1700 CPU combined with AMD Radeon Vega FE cards in CrossFire tops out at over 50 TFLOPS (OMGOD!!!) at just under US$3,000 for the complete system." is it right? Can combine two AMD Radeon Vega FE cards in CrossFire with CPU? Does it help to speed up simulations in Houdini?

Which Single-precision or Double-precision floating-point is more important in Houdini? is it depend on 32bit or 64bit process in software?

 

https://pro.radeon.com/en-us/product/radeon-vega-frontier-edition/

OpenCL™ 2.0 Support

Enables professionals to tap into the parallel computing power of modern GPUs and multi-core CPUs to accelerate compute-intensive tasks in leading CAD/CAM/CAE and Media & Entertainment applications that support OpenCL.

Does Houdini support OpenCL 2ver?

Edited by kiaspid

Share this post


Link to post
Share on other sites

Currently Houdini 16 can only use 1 OpenCL device at a time. This may change in the future.

1 person likes this

Share this post


Link to post
Share on other sites

Houdini either uses  OpenCL 1.1 or 1.2. OpenCL 2.0 doesn't expose any features which Houdini needs right now. Single precision is also more important at the moment, and as Atom says, only 1 compute device is currently supported.

1 person likes this

Share this post


Link to post
Share on other sites
12 hours ago, kiaspid said:

AMD Radeon Vega FE cards in CrossFire tops out at over 50 TFLOPS (OMGOD!!!)

that's only possible for fp16, Houdini uses more precision with fp32+

1 person likes this

Share this post


Link to post
Share on other sites
21 hours ago, kiaspid said:

Which Single-precision or Double-precision floating-point is more important in Houdini? is it depend on 32bit or 64bit process in software?

Since almost five years, Houdini only support 64 bit platforms:

https://www.sidefx.com/forum/topic/27117/

 

About Single Precision or Double Precision, it depends on what kind of data you want to store. For example, in the FBX import doc, it is specified that some vex datas can be single or double precision. http://www.sidefx.com/docs/houdini/commands/fbximport

Maybe this other page of the doc will help you, it explains that you can changes the precision and amount of memory used to store an attribute. Warning:  Not all operations in Houdini maintain 64-bit precision.

http://www.sidefx.com/docs/houdini/nodes/sop/attribcast

Edited by dudenay
1 person likes this

Share this post


Link to post
Share on other sites
On 01/08/2017 at 10:20 PM, kiaspid said:

Which Single-precision or Double-precision floating-point is more important in Houdini? is it depend on 32bit or 64bit process in software?

 

For OpenCL sims, single-precision is most important - double is used as an accumulation buffer to help avoid rounding errors afaik. This may be different for the OpenCL SOP.

1 person likes this

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now