在5080显卡上使用testbuild进行insight profile,可以看到GPU Compute的信息
[Image Removed]
但是2060显卡没有
[Image Removed]
硬件信息
[Image Removed]
[Image Removed]
在5080显卡上使用testbuild进行insight profile,可以看到GPU Compute的信息
[Image Removed]
但是2060显卡没有
[Image Removed]
硬件信息
[Image Removed]
[Image Removed]
Hi,
能否debug一下 GSupportsEfficientAsyncCompute是否为true?[Image Removed]
感谢回复。通过日志输出
#if PLATFORM_WINDOWS
else
{
D3D12_FEATURE_DATA_D3D12_OPTIONS6 D3D12Caps6{};
HRESULT Options6HR = RootDevice->CheckFeatureSupport(D3D12_FEATURE_D3D12_OPTIONS6, &D3D12Caps6, sizeof(D3D12Caps6));
// Allow async compute by default on nVidia cards which support PerPrimitiveShadingRateSupportedWithViewportIndexing
// this should be a good metric according to nVidia itself (this is set for Ampere and newer cards)
if (IsRHIDeviceNVIDIA() && Options6HR == S_OK && D3D12Caps6.PerPrimitiveShadingRateSupportedWithViewportIndexing)
{
GSupportsEfficientAsyncCompute = true;
UE_LOG(LogD3D12RHI, Warning, TEXT("GSupportsEfficientAsyncCompute is supported"));
}
else
{
UE_LOG(LogD3D12RHI, Warning, TEXT("GSupportsEfficientAsyncCompute is not supported"));
}
}
#endif
打包测试。
2060设备输出
[2025.11.04-11.20.42:811][ 0]LogD3D12RHI: Warning: GSupportsEfficientAsyncCompute is not supported
5080设备输出的是
[2025.11.04-09.56.07:597][ 0]LogD3D12RHI: Warning: GSupportsEfficientAsyncCompute is supported
好的,那应该是这个原因了。
感谢回复,请问GSupportsEfficientAsyncCompute 这个不支持的话,是导致Insight抓不到computeshader执行的流程还是computeshader不能异步执行?这种情况下如果Profile compute shader的性能情况呢?
是不支持compute shader异步执行,理论上性能比开启async compute肯定会稍微慢一点,但是不同平台对async compute支持的兼容性各异,我们都是根据实际测试情况来判断是否可以开启。