|
| ▲ | reverius42 a day ago | parent | next [-] |
| You still probably don't want to use 10% of your memory bandwidth at all times for a windowing effect? |
| |
| ▲ | oblio a day ago | parent [-] | | Ergo my point about compression :-) Desktop images are generally large blobs of pixels next to each other. I haven't verified but I would be surprised if 10x compression is not possible. So maybe 2-5GB/s? Also it's not like the GPU doesn't use up energy itself while doing this. | | |
| ▲ | darkmighty 14 hours ago | parent [-] | | GPU can clock a lot lower and use much less energy by focusing on parallelism. |
|
|
|
| ▲ | nixon_why69 a day ago | parent | prev | next [-] |
| Dual channel DDR5 is like 60, tops, 100s is on big many-channel servers. |
|
| ▲ | happymellon a day ago | parent | prev | next [-] |
| I don't really want 10% of my memory bandwidth being taken up by rendering my glassy title bars. Now wobbly windows is something I can get behind. |
|
| ▲ | com2kid 13 hours ago | parent | prev [-] |
| DDR5 you are looking at around 40 gigabits / second for single channel setups. Oh wait someone stacked 7 full screen transparent terminal windows on top of each other and now you've exceeded memory bandwidth. Meanwhile any GPU from the last decade (15+?) years will handle dozens of transparent objects on top of each other. I'm 100% in love with old school but sometimes I just have to admit newer HW is better. |
