if this is related at all to their vulkan efforts, then hardware detection are meant to be a critical part. Games are meant to decide which graphics card to use, and which cards can be used for off-loading. Optimizations for recognized cards could be about optimized settings (disable 32x ani on AMD) as well as hardware specific codepaths.
Announcement
Collapse
No announcement yet.
Unigine Needs Your Help Testing Out Their New Hardware Detection
Collapse
X
-
Originally posted by mdias View Post
How is it problematic? Unless you're talking about benchmarks, it's pretty normal for an engine to choose different code-paths for different hardware to achieve higher performance.
Perhaps some shader instructions execute faster on card A, but another different set of instructions (that achieve the same, or close to the same effect) will execute better on card B? Or maybe an API extension is better suited for vendor A, while the other extension that provides similar capabilities will be better of vendor B, even though both vendors support both extensions...
Remember we're talking about a game engine; a benchmark should of course execute exacly the same code eveywhere, even though hardware vendors often detect the benchmarking software and replace shaders with their own hand-crafter ones for speed.
Comment
-
Originally posted by cj.wijtmans View Post
Because a codepath can be really well optimized for one card but horrible for another even though the latter one has more potential due to its hardware. So you are not testing just your hardware but how well the code is optimized to your hardware. Ofcourse optimizations are desirable to get good performance but this can cause a lot of problems especially as your game ages and is not updated. Vulkan is meant to be abstracted by a dynamic library so it can be constantly updated and optimized even if your game ages and isnt updated. Otherwise the benefits of vulkan can be its downfall.
I was under the impression Vulkan drivers are meant to become a small shim between the hardware and applications. Thus, there's little left to optimize outside the application.
Comment
-
Originally posted by bug77 View Post
I'm really curious where did you read that.
I was under the impression Vulkan drivers are meant to become a small shim between the hardware and applications. Thus, there's little left to optimize outside the application.
Comment
-
Welp, this is the response I got when I tried to send it:
UNIGINE Mail Gateway: Your message was rejected by our spam filter. Please contact us at http://unigine.com/company/contacts/
I of couse don't have Facebook so I can't add it as a comment there, so if anyone else wants to send my report their way, feel free to do so: https://bpaste.net/raw/da8c9702874d
Looks like their detector completely failed to find most of the information about my GTX 660 with proprietary NVIDIA drivers.
Comment
-
Originally posted by mdias View Post
How is it problematic? Unless you're talking about benchmarks, it's pretty normal for an engine to choose different code-paths for different hardware to achieve higher performance.
Perhaps some shader instructions execute faster on card A, but another different set of instructions (that achieve the same, or close to the same effect) will execute better on card B? Or maybe an API extension is better suited for vendor A, while the other extension that provides similar capabilities will be better of vendor B, even though both vendors support both extensions...
Remember we're talking about a game engine; a benchmark should of course execute exacly the same code eveywhere, even though hardware vendors often detect the benchmarking software and replace shaders with their own hand-crafter ones for speed.
Comment
-
Originally posted by cj.wijtmans View Post
a dynamic library that uses vulkan not vulkan itself.
I simply have no idea what you are talking about.
Comment
Comment