The NEXT LAST R600 Rumours & Speculation Thread

Well. Usually hybrid AA modes get developed for professional hardware first. Then end up being trickled down into the registry of the desktop hardware. However I somewhat misunderstood the post here. So leave it at that. As far as Nvidia giving up on supersampling. I can assure that my conversation with them have not led me to believe that they are "dropping" SS permanently. But its simply not a priority feature for them right now. Things that are certainly coming first will be Vista stability, further XP tweaks, and SLIAA. I am still a little vexed that SLIAA hasnt made it public yet. For the moment best we can do is wait. Believe me when I say I've made sure the guys over at Nvidia have seen the lack of supersampling complaints. ((specially yours )) So they know the niche enthusiast wants them back.

Chris

 

Tommorw is the day where DICE and ATi demo Bad Company on R600.

 

You'll have to explain exactly what that entails to me. Quite honestly what I would like is a bunch of a fast stream processors and a rasterizer (not dissimilar to the G80 hardware, but not exposed like that in software)... the rest can be built efficiently on that base. Oh, and I wouldn't complain about a good chunk of cell-like local memory for the processors

On a completely different note, please pardon my ignorance: isn't EATM just alpha-to-coverage? Why is ATI getting so much attention for this when IIRC NVIDIA had it quite a while before ATI (i.e. Geforce 3 era). Has it not been usable (at least in GL) for quite a long time now? Why are we waiting on a new architecture/drivers?

 

I don't mean to herald EATM as the second coming of Jesus or anything. It's just a nice feature that's new to those who currently have an ATI card.

God knows why they're holding on to it, probably to slap an extra entry on to the R600 list of tricks.

 

Some new Fu(a)dZilla rumor/FUD:

Link

And looks like Hector Ruiz not happy anymore about the R600 (Orton days counted in AMD?)

Link
I belive when i see

 

Why read Fuad -- when you can listen to it yourself?

http://www.corporate-ir.net/ireye/confLobby.zhtml?ticker=AMD&item_id=1473686

 

Just listened to it - he says exactly:

"
Well, any time you're late with a product you are disappointed, and you do obviously suffer from it. But it's such an outstanding product, you can see when people begin to see the very high end capability of that technology - I think it's going to find a home pretty readily, and we're very bullish about what it can do, and we're pretty excited about what we're going to be able to offer our customers and partners.
"

 

Doesn't exactly inspire confidence. R600 will have to speak for itself.

 

Yup. I'm tired of all this pre-release talk with words like beast, monster and what not excessively used until the product is released and then that beast/monster always seems to run back to its closet and wait for the next generation.

 

So..we hated Faud so much at INQ now we chase him to his new site in the hunt for more FUD?

Please, let it die.

 

Something like programmable tesselation/HOS for instance. I recall a presentation from NV where the question rose whether "we're there yet" in D3D10 with procedural geometry and the answer was "not quite yet". I don't recall the exact text/phrasing, but I'm sure there were specific geometry related functionalities scratched through early "DX-Next" drafts.

I'm merely wondering what the next step might look like.

 

Link

 

Left to right: rv410 110nm, r520 90nm, r580 90nm, r600 80nm, rv630 65nm.
Link

 

That is a big chip.

Interestingly, there are fewer (though larger) resistors on the top of the package compared to the R5xx cores.

RV630 looks almost as clean on top as RV410.

 

Larger resistors are used because of the higher current/power consumption, I guess.

 

So that image of the R600 core was infact not fake as it was assumed quite some time ago when it first popped up in the interweb?

 

Huh? As far as I can remember, it was judged as legit R600 almost immediately?

 

You are confusing it with the image with a 45° rotated die-pin shot.

 

Oh.. i must be suffering from short term memory loss

 

I was wondering who will win ATI Randeon X2900XTX or upcoming Nvidia 8900GTX.

If you look at Nvidia 8800GTX which has 575MHz GPU-clock and 900MHz-GDDR3 memory. Based on the rumorer information that 8900GTX will be on 80nm tech and be clocked at 700MHz GPU and 1100MHz GDDR4 memory and it will still be at 384bit memory interface. "Plus 25% more shaders on G81"

By just increasing core clock on G80 575Mhz to 700MHz, we are actually talking example like "7800GTX 256MB" against "7900GTX 512MB" similar performance increase.

Since R600 only has 16 ROP's and G80 has 24 ROP's. Giving the fact that high clock speed on 24 ROP's will be significant performance increase.

I was wondering how R600 will compete against G81 Nvidia 8900GTX.