So, do we know anything about RV670 yet?
- Thread starter nicolasb
- Start date
I'm not saying anything about RV670. I'm just saying that D3D10 has inertia that I don't think 10.1 can really defeat, and I doubt conusmers are going to really care.
Unless AMD pulls a "Chuck patch" out of their collective arse for every DX10 game out there
And where is the data that shows "rather bad utilization in comparison"? (Lest we forget that we can actually access our peak rates as well, where other designs still appear to have "MULs-missing" for much of the time...)
Pffft, you kidding? AMD will come out with some magic DX10.1 drivers....
... which you may want to rethink with everything DX10.
DX10 is one of the few times where a number of features at a certain point in time was really essential. Who showed up on time?Otherwise, first to implement or even first to market (which is surely what you meant) wrt features is really not as important as "hitting the window of opportunity with a new product."
One of the hardest things to do in the tech business is to cut features to get the product out. It may be that the tesselator of R600 is almost a copy from Xenos, but even then it still takes 2 man-years to integrate, test, and write a demo for. Maybe those resources had been better spent somewhere else?
It's a really delicate trade-off, of course, and Nvidia has it easier because they more influence over developers, but it's the curse of an engineering driven company to add more at the expense of schedule slip. In the GPU world, that's a bad idea: there will always be a refresh a few months down the road.
Bouncing Zabaglione Bros.
Legend
Doesn't matter as the subject in hand was getting AMD's hardware into developers' machines, even if that only happens because devs want an early start on the next iteration and capabilities of DX (even if they don't end up using them). To have devs using AMDs products as their base programming platform would still a valuable advantage to AMD, much as it is to Nvidia now.
It will only happen if the developer's desire to start work on future DX capabilities outweighs Nvidia's good dev-rel. AMD can only capitalise on it's more advanced hardware in this particular instance if it can balance the scales in their favour by also stepping up and improving their dev-rel as well as offering advanced DX10/DC10.1 hardware.
If RV670 has really swept the board at the OEMS, consumers will have that hardware in their new PCs whether they care about it or not.
This is incorrect and stems from knowing little about the actual architectural details or process.
But 7950GX2 /2 > 7900GT
What does you mean? You should see what I was responding to.
chavvdarrr
Veteran
Then why G80 wipes the floor with R600 ?
Less ALUs, much lower bandwidth, yet R600 manages to win... where? Except in 3dmark? which just comes to show how "valuable" 3dmark is ATM.
It doesn't matter if the weak spot is in shader core or, rather, in AA performance (who cares about speed, we have Features) or TMUsj. Fact is, 1 year after G80 came, AMD/ATi still have nothing in the high-end, They go for cheaper cards, not for faster ones.
I really hope R6xx won't be the next Savage2000 ...
Because of double-clocked shader core and contextual benefits?
chavvdarrr
Veteran
So? Even if the crappy "feature above everything" AA implementation and unproportionally low TMU power are not the reason why R6xx architecture is slower, does that makes things better? Being unable to make faster chip for 1 year, struggling with drivers, after laughing at NV, hell breaking drivers for 1gen old hardware... does all this looks like heading in the right way?
S3 made lots of money from Virge, too, even Savage3&4 kept them up and running, but next fiasko killed them.
Forget the math power, it's the raw pushing up rate here -- G80 have an excess of bilerp fill rate, while R600 is exactly the opposite (any title with heavy vertex texturing or Fetch4 around?); an overwhelming in colour sampling and depth/stencil rate for G80 again, despite the fact that Hi-S addition to the R600 effectively doubles the stencil test rate in some cases, it's just not enough... (Dave, don't bite me)
To make it simple, their is not much ATI can do with R6xx series GPU's, just like Nvidia with Geforce FX (NV30/NV35)
I hope R700 will see weakness of R600 and don't follow the same direction.
Actually Voodoo5 5000/5500 was not so bad.
I hope R700 will see weakness of R600 and don't follow the same direction.
Actually Voodoo5 5000/5500 was not so bad.
Last edited by a moderator:
Well, we'll see about that, tho I certainly don't expect them to suddenly pull ahead in the single-gpu beauty pageant. If that's your standard, then yeah R7xx is likely the next reasonable opportunity.
But there's room for improvement, and performance improvement has been made on the driver side as even [H] felt the need to note on Monday. Narrowing the gap is still a useful thing to be doing.
But there's room for improvement, and performance improvement has been made on the driver side as even [H] felt the need to note on Monday. Narrowing the gap is still a useful thing to be doing.
