Pemalite said:
EpicRandy said:
Also it does not loose 60% of the ram 10/16 is 62 % which is 38% lower not 60%
|
Check your math on that. I am talking physical Ram not usable Ram.
|
Think my math is correct on this. Series S is 10GB and X 16GB 10/16 = 0.625 meaning a loss of 37.5%. 16GB*0.375 = 6GB (the diff between X and S)
Pemalite said:
EpicRandy said:
Also it does not loose 60% of the ram 10/16 is 62 % which is 38% lower not 60%
|
Check your math on that. I am talking physical Ram not usable Ram.
EpicRandy said:
Memory bandwith need to be huge because at this point all assets are uncompressed. On memory bandwith a 4k texture will literally and exactly required 4x the bandwith of a 1080p one and texture are literaly what is bandwith hungry when rendering a game. So normaly the series S could get away with about 25% bandwith compare to the X but it have 40% meaning it has headroom when targeting 1/4 of the res. The lower bandwith ram part should be mostly used by the os reserved memory so no impact on game. You also find all other specs to have similar headroom to what they would required in this scenario. Dev can get away with minimum effort here by targeting 1/4 the res and having assets accordingly but great port to the S will try to tap into this headroom to target slightly higher res like 1200p when 4k on the series X.
So again the series S will be just fine for the entire gen.
|
The Xbox 360 had games that used 4k textures. I.E. 4096x4096 textures. Some games even higher resolution textures than that.
Xbox 360 has 0.5GB of Ram verses the Series S at 10GB. (8GB for games).
I think people get confused with texture resolution and output resolution of a rendered scene. They are actually independent.
And no, a 4k texture will not always require 4x the bandwidth of a 1080P one, delta colour compression will compress certain maps better than others based upon the frequency of predictable patterning in the map itself and can save a considerable amount of bandwidth.
|
True but for simplicity sake I was referring to the target output of the title, texture can be any resolution but if you target a 4k output for a specific title and use a 4096x4096 texture for a specific meshes and you want to output @1080P on series S you will more than likely target 2048*2048 (a 4 time reduction) texture as it should bring you the same amount of intended detail. Unless you want to use the same texture or something in between in a bid to bake in some Super Sampling effect which is entirely up to the dev to decide.
And while 4096*4096 should do compress better than a scaled down 2048*2048 it's variable base on the actual texture and the gain in compression ratio become more and more marginal as the size grows not to mention the fact you are pretty much bound to use less effective lossless compression algorithm not to introduce compression artefact.
Last edited by EpicRandy - on 13 March 2023
