The LAST R600 Rumours & Speculation Thread

Status
Not open for further replies.
Geo said:
Catalyst 6.10 vs Catalyst 6.12, perhaps?



Since the author of that piece was the current editor of Beyond3D, you may wish to reconsider that question, which is certainly unnecessarily confrontational for "step 1" of getting to the bottom of any question no matter who the other party is.
Click to expand...

I really don't care who the author is and it shouldn't matter for my question's sake. Why was there not AA or AF on that CoH test? It does look suspisious. Perhaps Ryszard Sommefeldt could answer that for me. ;)
 
R300King! said:
I really don't care who the author is and it shouldn't matter for my question's sake. Why was there not AA or AF on that CoH test? It does look suspisious. Perhaps Ryszard Sommefeldt could answer that for me. ;)
Click to expand...

Ratchet deigned to skip the AA test with CoH as well due to issues he encountered with it. Could have been something fixed in a later patch.

Company of Heroes results were gathered using the games built in Performance test. Antialiasing with Company of Heroes was not run due to problems encountered during testing (which appears to be a game issue).
Click to expand...
 
Sometimes drivers don't just improve performance. Rys is gone for the day, but I suspect if he checks in tomorrow he's going to say that there was an issue with Cat 6.10 and AA in their built-in benchmark, and that in retrospect he should have mentioned it on the page. But you'll have to wait for him to do so.
 
trinibwoy said:
Ratchet deigned to skip the AA test with CoH as well due to issues he encountered with it. Could have been something fixed in a later patch.
Click to expand...

So it could be a game problem? Ok, thanks for your answer. That could be a possiblity. I wonder how Digit-life got theirs to work and if it really showed up in the game.
 
kaigai03l.gif


found this swinging around on the web...
 
If I'm not mistaken that exact same diagram could be used to describe G80 architecture too. If you're doing an unified shader architecture for D3D10, then you're gonna need these units in that configuration. At least I can't imagine what you'd do differently.

So I'd say, it's a bit vague.
 
stepz said:
If I'm not mistaken that exact same diagram could be used to describe G80 architecture too. If you're doing an unified shader architecture for D3D10, then you're gonna need these units in that configuration. At least I can't imagine what you'd do differently.

So I'd say, it's a bit vague.
Click to expand...


I figured that as much, but what about the first pic of the chip I posted...

Different?:p
 
I really hope that second pic is the right one for the sake of all those guys who spent countless hours trying to count pins and guessing why the die was rotated :LOL:
 
trinibwoy said:
I really hope that second pic is the right one for the sake of all those guys who spent countless hours trying to count pins and guessing why the die was rotated :LOL:
Click to expand...

The First one could still have all those pins. Hell, maybe more. I always thought that having the die rotated helps get rid of electrical noise.:smile:

I will say the second one looks sexy, more so than the first...:D But the First one has the ATI and PCI express badges like it is ready to go for retail....:oops:
 
trinibwoy said:
Yeah but the 4xAA hit is pretty significant even in your figures - unless I'm the only one who considers 30-40% as significant.

All I'm saying is that a 30% AA hit on G80 leaves a lot of room for R600 to shine given its bandwidth advantage.
Click to expand...

Did anyone tell us that AA is for "free" and we believed it?

Bandwidth isn't the only thing you need for faster AA; one of the most significant factors most definitely yes on IMRs ( :p ).

How can you know what kind of room it leaves on R600 anyway? The 1xAA performance (yes it's 1x and not 0xAA dagnabit) could be by say 10% lower and the 4xAA score by 10% higher; that way you might have a significantly smaller drop from 1x to 4xAA, but the final result not being all that impressive either.

This is again just pure speculation, but it's just to show that any of your assumptions really don't have much foothold at this point of time without knowing how the R600 behaves exactly. I had also added another link about memory consumption; given the fact that the highest model will most likely have 25% more onboard memory you would have to factor that also in for some corner cases where the demands on memory is high enough. If you kill texture compression in SS2 and use 16xQ on a G80 in 2048, you've just consumed all your onboard memory; the fact that using TC in that one is nonsense is irrelevant it was merely to show that if you want you can find scenarios to bring the card to the edge.

Would there be a high end TBDR in the works with a R600 alike arithmetic throughput yet exactly half it's bandwidth, would you say that it would lose against the latter because it has way less bandwidth? Weird example I know, but for me it's more important to see how each architecture utilizes A or B and how high the given efficiency is, than just meaningless raw figures.
 
Sound_Card said:
This R600 looks a little bit different than the first chip we saw....:smile:

hmm.....



04845fi.jpg


which one is the real one....?

r600kn2.jpg
Click to expand...

You could have known right away that the non-rotated one is a fake: it was packaged end of 2005, week 45 to be exact. The rotated die was packaged week 32 of 2006.

Yes, this is a convention that is followed universally across all chip manufacturers: a 4 digit YYWW time stamp... Clearly the photoshoppers didn't know about this. ;)
 
Last edited by a moderator:
Status
Not open for further replies.

Similar threads

A
The Official G84/G86 Rumours & Speculation Thread
22 23 24
Replies
460
Views
68K
Twinkie
T
Back
Top