Anand has the details about r520,rv530,rv515
- Thread starter tEd
- Start date
http://bbs.gzeasy.com/index.php?showtopic=460367&st=20
From what some guy (Robin B at hardware.no forum) could make of it:
From what some guy (Robin B at hardware.no forum) could make of it:
Rys said:
So you're confirming R520 has two 'full' ALUs per pipe?
R481 has one full and one 'mini' according to Dave's tables.
But even with 32 full ALUs per pipe, I was trying to say that G70 still has more at 48.
Edit: obviously, I'm drastically oversimplifying the capabilities, but to do it right gets just too complicated.
Edit: Oops! not ALUs per pipe, but total.
Last edited by a moderator:
Ati Tray Tools - 1.0.3.750
* Added support for Adaptive anti-aliasing mode. Substantially improves visual image quality by enabling supersampling anti-aliasing on transparent textures, and multi-sampling anti-aliasing on all other textures. This delivers exceptional levels of image quality, while maintaining very high levels of performance. (Only Catalyst 5.9 and up and only if supported by hardware)
* Added support for Adaptive anti-aliasing mode. Substantially improves visual image quality by enabling supersampling anti-aliasing on transparent textures, and multi-sampling anti-aliasing on all other textures. This delivers exceptional levels of image quality, while maintaining very high levels of performance. (Only Catalyst 5.9 and up and only if supported by hardware)
no-X said:Ati Tray Tools - 1.0.3.750
* Added support for Adaptive anti-aliasing mode. Substantially improves visual image quality by enabling supersampling anti-aliasing on transparent textures, and multi-sampling anti-aliasing on all other textures. This delivers exceptional levels of image quality, while maintaining very high levels of performance. (Only Catalyst 5.9 and up and only if supported by hardware)
Hmmm nice. I was looking for something more revolutionary though. Still don't know what the big "secret" is that Rys was carrying on about earlier.
Edit: I completely agree with Ratchet that "Adaptive" is not the proper term for this technology if it works like Nvidia's TRSAA.
Last edited by a moderator:
geo said:Me likey the sound of that. . .
Particularly the hardware part. Goodness, did both of them have the same great idea at the same time? Surely ATI couldn't have stuffed this in just since June. . .
Perhaps they decided that Humus' dxoverride stuff was worth investing some time in and got the grid-pattern problems with some of the transparent textures taken care of.
Using dxoverride with world of warcraft is amazing on my 9700pro. Outdoor scenes look incredible now, and at basically the same framerate!
Nite_Hawk
"only if supported by hardware". Could they have stuffed in hardware that quickly? I would think not, even tho it seems that the hardware support for NV's solution is relatively minimal as such things go. Certainly there was another spin in there, by all reports, but wouldn't it be unprecedented to cram hardware for a new feature in like that on the last spin you're planning to go to production with?
But maybe next Thursday or so we can get Humus to tell us. . .
But maybe next Thursday or so we can get Humus to tell us. . .
Last edited by a moderator:
Hellbinder
Banned
This is Ridiculous Rubbish.
Typical Nvidia slanting spin machine stuff.
If its Fully SM3 and has a new Ring bus and supports MSAA in HDR how the *heck* is it just a R420 with stuff tacked on? This is just obvious Nvidia influnced spin imo. This kind of logic says that everything ever made is just the same old 1 pipeline chip with "Stuff" tacked on.
This is not an R420 with "stuff" tacked on. I wonder how many review sites are going to fall for this kind of spin.
geo said:
I thought that was because Nvidia already had very flexible support for SSAA and hybrid modes? What spurred ATi to go this route all of a sudden - I'm betting it's nothing unique to R520 and R300 supports it too.
On the other hand, it would really suck for their marketing division if this was going to be their killer IQ feature before they saw the G70.
trinibwoy said:
No kidding! "We would have had it in May, we swear, if it wasn't for the damn soft ground!" isn't going to cut any ice with the masses, even if the cognoscenti might buy it.
Why would you make that kind of distinction re hardware if it is supported by R3/4xx? Unless the RVxxx line is missing something that is required. . .
Gibbo has some "information from sources in the industry" - nothing substantial, athough he confirms that X1800XL will be available on launch day (in the UK at least).
Slightly OT, but has there been explaination/discussion as to why the softground issue was discovered so late in the sampling/testing phase? If the first samples were back in December 2004, why was this only picked up in May/June 2005 (if indeed that was when it was identified), was it only detected when production was ramped?
Slightly OT, but has there been explaination/discussion as to why the softground issue was discovered so late in the sampling/testing phase? If the first samples were back in December 2004, why was this only picked up in May/June 2005 (if indeed that was when it was identified), was it only detected when production was ramped?
