Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
Technology means nothing, design is EVERYTHING, physics/physx are NOTHING without a guy there behind the scenes programming things that interact with each other. You can't put a basic wall thats designed as one block and a physx engine magically makes it breakable, a guy has to design a wall in a dozen pieces and texture every edge and make it breakable, same goes for everything else.
You don't write code and it will work differently under a dx11 wrapper, you have to program tesselation effects, or anything else.
You can usually make a game look great under ANY DX version, its just about design time and overall end performance, giving you shortcuts and often just removing overheads enables you to use more effects, but effects that you could already make.
DX11 itself should make it easier for programmers to add MORE effects into a game but doesn't really make anything particularly new available.
Unfortunately we went through a period where games had to be designed for dx9 and dx10 at the same time and as the bulk of users were on dx9 and dx10 was such a slim upgrade(thanks Nvidia) and Vista was a bit of a hulking beast of bloatware rubbish(thanks MS) that usual push for everyone to move forward, just wasn't there. The uptake to Windows 7 has been FAR more successful than the uptake to Vista, and this dramatic change in numbers of people with them, plus the amounts of people with DX10 cards at least, mean designing for dx11 and ignoring dx9 is starting to happen and THATS where the real push forwards comes from, not new stuff being available, but programmers being able to not have to think about the old stuff.
Thats why DX10 is such a good thing, a break from DX9, a distinct level of support you have to give.
As for the cards themselves, theres NEVER been a reason to look at if a new card is good or not EXCEPT based on its raw performance. THe 5xxx series has been great with a large improvement over last gen(remember in general the performance difference for the past decade is going for a 70-90% performance bump over the normal last gen cards, ie 4870 to 5870, x800xt to x1800xt.
In some games, Metro 2033, which is hard on cards and looks great, the 5770 is faster than the 4890, and the 5870 is double the fps of a 4890, thats simply as good scaling as you could possibly hope with double the rops, double the shader power and similar clocks, infact, its ruddy amazing performance.
The Fermi's, well, less impressive, expensive, hot, hungry and due to missing clock speed targets, well a 465gtx should NOT be fighting alongside a 275gtx a 465gtx should be comftably ahead of a 285gtx, even if not by a huge amount, it just should be, always. They aren't bad cards, but if they were 50% smaller, on time, cheaper and offered the same power usage/noise/heat levels as AMD's cards it would be a whole lot better.
Both cards are roughly speaking, about where they should be. Some generations you'll move from a crappy to a fantastic process, a large to a smaller core and get a fantastic speed boost. Some gen's you'll move from the peak of design efficiency in terms of die size and process quality, to an awful process, the fact a 5870 can offer double the performanec of a 4890 in some games is, honestly, astonishing considering the process each card is on. Prices aren't fantastic, and thats the downside of the process AMD haven't been able to avoid by far enough, but if they didn't have a couple of fixes to raise the yields, prices would be even higher again.
AMD have made the most of a terrible process and poor year for production to produce cards that have the performance increase we wanted. Nvidia have made very little of the process and managed to make a card 20% slower than it should have been, but its still a decent card and considering the TSMC process, its actually not half bad and could have been a LOT worse.
Raw performance, these cards are fine, next gen(another 40nm card) people will be very dissappointed, theres no way in hell AMD will hit a 70-90% performance bump on the same process(nor should they be able to). THe jump from 40nm -28nm should bring hopefully be a bigger than usual performance jump to compensate for a very small jump between 5870-6870.
This gen, not dissappointing, except for late and hungry if you waited for Nvidia, next gen(staying on 40nm) will appear to suck, and the one after might be the biggest jump for the past decade.
DX11 is neither here nor there, you judge new cards on their performance, dx11, display port, Shader model 58, ieee7584, Cuda 68, Stream 45, physx, opencl, these are buzzwords, they mean nothing to anyone but a programmer but are used as advertising spin. Game quality/graphics are down to design not what features the cards support(which for the past decade have been so close to each other as to never matter in more than 1-2 games a generation and really completely ignorable SM2c and 3, FarCry, was that the only game that mattered and it barely looked different, woo).
You don't write code and it will work differently under a dx11 wrapper, you have to program tesselation effects, or anything else.
You can usually make a game look great under ANY DX version, its just about design time and overall end performance, giving you shortcuts and often just removing overheads enables you to use more effects, but effects that you could already make.
DX11 itself should make it easier for programmers to add MORE effects into a game but doesn't really make anything particularly new available.
Unfortunately we went through a period where games had to be designed for dx9 and dx10 at the same time and as the bulk of users were on dx9 and dx10 was such a slim upgrade(thanks Nvidia) and Vista was a bit of a hulking beast of bloatware rubbish(thanks MS) that usual push for everyone to move forward, just wasn't there. The uptake to Windows 7 has been FAR more successful than the uptake to Vista, and this dramatic change in numbers of people with them, plus the amounts of people with DX10 cards at least, mean designing for dx11 and ignoring dx9 is starting to happen and THATS where the real push forwards comes from, not new stuff being available, but programmers being able to not have to think about the old stuff.
Thats why DX10 is such a good thing, a break from DX9, a distinct level of support you have to give.
As for the cards themselves, theres NEVER been a reason to look at if a new card is good or not EXCEPT based on its raw performance. THe 5xxx series has been great with a large improvement over last gen(remember in general the performance difference for the past decade is going for a 70-90% performance bump over the normal last gen cards, ie 4870 to 5870, x800xt to x1800xt.
In some games, Metro 2033, which is hard on cards and looks great, the 5770 is faster than the 4890, and the 5870 is double the fps of a 4890, thats simply as good scaling as you could possibly hope with double the rops, double the shader power and similar clocks, infact, its ruddy amazing performance.
The Fermi's, well, less impressive, expensive, hot, hungry and due to missing clock speed targets, well a 465gtx should NOT be fighting alongside a 275gtx a 465gtx should be comftably ahead of a 285gtx, even if not by a huge amount, it just should be, always. They aren't bad cards, but if they were 50% smaller, on time, cheaper and offered the same power usage/noise/heat levels as AMD's cards it would be a whole lot better.
Both cards are roughly speaking, about where they should be. Some generations you'll move from a crappy to a fantastic process, a large to a smaller core and get a fantastic speed boost. Some gen's you'll move from the peak of design efficiency in terms of die size and process quality, to an awful process, the fact a 5870 can offer double the performanec of a 4890 in some games is, honestly, astonishing considering the process each card is on. Prices aren't fantastic, and thats the downside of the process AMD haven't been able to avoid by far enough, but if they didn't have a couple of fixes to raise the yields, prices would be even higher again.
AMD have made the most of a terrible process and poor year for production to produce cards that have the performance increase we wanted. Nvidia have made very little of the process and managed to make a card 20% slower than it should have been, but its still a decent card and considering the TSMC process, its actually not half bad and could have been a LOT worse.
Raw performance, these cards are fine, next gen(another 40nm card) people will be very dissappointed, theres no way in hell AMD will hit a 70-90% performance bump on the same process(nor should they be able to). THe jump from 40nm -28nm should bring hopefully be a bigger than usual performance jump to compensate for a very small jump between 5870-6870.
This gen, not dissappointing, except for late and hungry if you waited for Nvidia, next gen(staying on 40nm) will appear to suck, and the one after might be the biggest jump for the past decade.
DX11 is neither here nor there, you judge new cards on their performance, dx11, display port, Shader model 58, ieee7584, Cuda 68, Stream 45, physx, opencl, these are buzzwords, they mean nothing to anyone but a programmer but are used as advertising spin. Game quality/graphics are down to design not what features the cards support(which for the past decade have been so close to each other as to never matter in more than 1-2 games a generation and really completely ignorable SM2c and 3, FarCry, was that the only game that mattered and it barely looked different, woo).
