Newer ASICs have more VRAM on average and allocating more GART as well can have advantages. Also see commit edcd26e8.
Ideally, we should scale GART size based on actual VRAM size, but that requires significant restructuring of initialization. --- drivers/gpu/drm/radeon/radeon_device.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index a7fdfa4..b6769b0 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -1098,7 +1098,9 @@ static void radeon_check_arguments(struct radeon_device *rdev)
if (radeon_gart_size == -1) { /* default to a larger gart size on newer asics */ - if (rdev->family >= CHIP_RV770) + if (rdev->family >= CHIP_TAHITI) + radeon_gart_size = 2048; + else if (rdev->family >= CHIP_RV770) radeon_gart_size = 1024; else radeon_gart_size = 512;