AMD: R8xx Speculation
- Thread starter Shtal
- Start date
I read the posts here about it (the frostbite engine) being converted to DX11 in a couple of hours, but I haven't read anything on 43 and BC2 actually being DX11.
Last edited by a moderator:
No, BattleForge doesn’t use tessellation.
We tried but with the RTS camera there wasn’t any noticeable visual improvement.
They would be going up against "twice" RV870 anyway aren't they?
Probably the NV design goals are a little different.
For example having a twice as wide bus also enables a twice as much maximum amount of memory.
For the HPC segment and medical volume processing/rendering market segment, or server based rendering having as much memory as possible can be critical.
For fluid / gas simulations or other advanced physics calculations having the memory bandwidth matching the available TFLOPS also can be crucial.
From my limited knowledge from an introductory class on IC design and optimization, the yield depends on: redundancy (optimized by the designers), process error density (rather high on new process) and die size (I may be forgetting something, sorry).
Redundancy of course can be done on the design and this would be different from ATI vs NVIDIA, also their die sizes are. But the yields due to process error density and die size are not linear, they increase exponentially with the increase of die size (I think in most cases a Poisson distribution is assumed on the wafer area).
So basically the bigger the die the bigger the chance of getting errors due to the process and if the error lands in some logic without redundancy you have a bad die. A die that is 50% bigger can have more than twice worse yield due to the exponential increase in error and even different strategy with regards to redundancy.
Redundancy of course can be done on the design and this would be different from ATI vs NVIDIA, also their die sizes are. But the yields due to process error density and die size are not linear, they increase exponentially with the increase of die size (I think in most cases a Poisson distribution is assumed on the wafer area).
So basically the bigger the die the bigger the chance of getting errors due to the process and if the error lands in some logic without redundancy you have a bad die. A die that is 50% bigger can have more than twice worse yield due to the exponential increase in error and even different strategy with regards to redundancy.
Nice hint there
Making a DX11 version without using at least one feature doesn’t make that much sense as you will get nothing from a direct port.
hoom
Veteran
Oh, I had gotten the impression that BattleForge used the 10.1 tessellator
Makes sense that it wouldn't make any notable visible difference from a typical RTS camera angle.
Did it make any performance difference positive or negative?
The Froblins demo doesn't exactly run at amazing fps but that wasn't even using the official DX 10.1 path was it?
Last edited by a moderator:
Why do you waste time reading and referring to such an idiot reporter?
I wonder if it took him more than 5 minutes to write that.
Do u know him or why r u insulting him like this? Fudo isnt an idiot and im shure hes much nicer than u. So before u talk about someone else like this make shure to look in a mirror!
Maybe, if Fuad came out and nicely tell us who is giving him all his "information" I'm sure you too would say this article shouldn't have an ATI Radeon logo on the top.
Just saying.
Hint: It's not someone with AMD stock!
CS5.0 only or was CS4.x good enough?
Fudo & his source?
Can I have my cookie?
Entertainment. I prefer charlie tho, his anti-nV stuff is vastly more amusing.
He is not an idiot I'm sure, on the other hand, I'm quite sure he is in bed with nV.
I mean srsly! Overpriced?? Compared to what? And what about millions that nV bribes devs with in TWIMTBP. He obviously has something to gain by writing this nonsense.
Similar threads
