What other hardware/Technology is on the horizon?
- Thread starter Fuz
- Start date
Image quality article
Dave, IIRC, after you published the 9700 Pro review you said you would later write an additional article about image quality. What has happended to that article? Would be mighty interesting for me. Do you plan to also cover topics like alpha texture antialiasing, 16bit FSAA and supersampling on R300?
Dave, IIRC, after you published the 9700 Pro review you said you would later write an additional article about image quality. What has happended to that article? Would be mighty interesting for me. Do you plan to also cover topics like alpha texture antialiasing, 16bit FSAA and supersampling on R300?
Re: Image quality article
I've still got a ton of stuff to work my way through, WRT reviews, so I've not had much chance to sit down and think about anything else at the moment. Time constraints mad me miss out on stuff like videoshader and hyravision for the main 9700 PRO review, but some of these were looked at in the 9000 PRO Review a little later on.
As for Alpha texture aliasing, to my knowledge ATI haven't implemented anything to sort that out, so thats the same as any other MSAA solution, 16-bit FSAA still doesn't work (and I'll be surprised if it ever does) and SuperSampling isn't an option. If thse change then perhaps there's grounds for more of a look.
(ATI guys can step in and tell me otherwise if thats the case).
madshi said:Dave, IIRC, after you published the 9700 Pro review you said you would later write an additional article about image quality. What has happended to that article? Would be mighty interesting for me. Do you plan to also cover topics like alpha texture antialiasing, 16bit FSAA and supersampling on R300?
I've still got a ton of stuff to work my way through, WRT reviews, so I've not had much chance to sit down and think about anything else at the moment. Time constraints mad me miss out on stuff like videoshader and hyravision for the main 9700 PRO review, but some of these were looked at in the 9000 PRO Review a little later on.
As for Alpha texture aliasing, to my knowledge ATI haven't implemented anything to sort that out, so thats the same as any other MSAA solution, 16-bit FSAA still doesn't work (and I'll be surprised if it ever does) and SuperSampling isn't an option. If thse change then perhaps there's grounds for more of a look.
(ATI guys can step in and tell me otherwise if thats the case).
Jerry Cornelius
Newcomer
apples for oranges, i evenly matched tiler puts about twice the pixels to the pavement as an IMR. As far as I'm concerned all this pixels shader stuff and high polygon count is a red herring. It's still about fillrate (baby).
If they could get a 4 piped part clocked to 300 mhz with 300 mhz DDR it could be a close match for something like the 9700, except that IMG don't like mixing up texture fetches between mip levels, so chances are it will bilinear filter with the best of them and fall back a bit when you dial in HQ filtering.
My prediction. I expect them to stick to their guns with the laborous filtering approaches of the past but offer some wicked AA modes. I'm also betting on a highly clocked (350) 4 pipe part with excellent T&L&Stencil. SHould make for a decent Doom III card.
Doug
If they could get a 4 piped part clocked to 300 mhz with 300 mhz DDR it could be a close match for something like the 9700, except that IMG don't like mixing up texture fetches between mip levels, so chances are it will bilinear filter with the best of them and fall back a bit when you dial in HQ filtering.
My prediction. I expect them to stick to their guns with the laborous filtering approaches of the past but offer some wicked AA modes. I'm also betting on a highly clocked (350) 4 pipe part with excellent T&L&Stencil. SHould make for a decent Doom III card.
Doug
Doesn't sound unreasonable to me; exept that I don't see much reason not keeping the design on .15um and have it released by now, instead of opting for the much riskier at the moment .13um process.
PS: the above scheme pictures more IMHO a STG5500 refresh@ 300/300MHz than anything else.
PS: the above scheme pictures more IMHO a STG5500 refresh@ 300/300MHz than anything else.
Jerry Cornelius said:
Not in the least. High polycounts are an absolute necessity for good images. Thankfully, these are increasing, but not as quickly as I would like. Of course, they're currently limited significantly by the fact that high polycounts need lots of data to be stored in system RAM and managed by the CPU.
This is where shaders come in, and where HOS will come in. Shaders will allow the GPU to do more and more of the vertex data management. This is definitely a good thing, and will allow for much higher polycounts, independent of CPU speed increases (though still dependent upon system memory bandwidth). HOS will help by reducing the amount of data that needs to be stored/managed on the system-side for the same output. If everything had gone well with the GeForce3, HOS would be more than just Truform by now. It's really too bad that nVidia's implementation just wasn't good enough for use in games (Truform=bad HOS because it doesn't allow as much developer control as other forms of HOS).
And that's just vertex shaders.
Pixel shaders (or fragment programs, as they should now be called) are exceedingly important because they offer a vast new array of new processing capabilities. For the most part, what we'll see are speed improvements from these new shaders. But as time goes on, the improvements will become more drastic as hardware expects the shader support to be in place.
Said another way, expect JC's next project after DOOM3 to be simply amazing (He has stated it will be based upon the NV30's capabilities). DOOM3 shows what was made possible with GeForce-class hardware, and you will see the massive performance improvements that shaders can make with this game. The quality improvements won't be so obvious, yet.
I'd be quite unhappy if ISV's just chose spend their time re-doing HOS on any programmable primitive generator that might be provided in the future just for the sake of being able to say "hey look I've just spent 6 months reproducing the work that the IHV has already done"! I'd rather see them spending time, say, procedurally generating hair (which may or may not include the use of HOS), or cloth, or water etc etc.
Basically HOS are a tiny part of the potentual uses.
John.
Basically HOS are a tiny part of the potentual uses.
John.
