-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD gets back in the development game with OpenCL tools and drivers
- Thread starter Final8y
- Start date
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
What a douchebag that writer is. Complete and fundamental lack of understanding, lack of understanding where AMD are, where they are going and whats going on.
The sisoft benchmark is fantastic on AMD cards because it was writen with their architecture in mind, the previous benchmarks Nvidia won hugely in were designed with their architecture in mind, neither is better nor worse in reality, though AMD's has a massive advantage when it comes to manufacturing.
To suggest AMD will move back towards less efficient design with more basic shaders, rather than move towards better optimised software shows his lack of understanding.
More basic, and larger shaders would mean a larger core, = less cores per waifer, higher clock speeds needed and, well, worse production. AMD shifted to the current style of architecture for one reason, make a shedload of cheap cores that are less effected by difficulty in manufacturing, its been an unparalleled success, moving away from that would lose them all price/performance advantage they have and thats whats been winning them market share and sales left right and centre.
Likewise, opencl has pretty much smeg all to do with gaming in the future. Within one generation of cpu's we will have a exponential increase in FPU processing power on the cpu die. Considering Nvidia when using physx on die are only using a VERY minor portion of the gpu, remember most of it is doing normal graphics data processing, we're looking at moving a low end gpu's amount of FPU power on die, within a year.
WAY before more fpu is needed in every game released every last person on earth will have a CPU with compared to today, 100x more fpu power onboard, it would be lunacy to offload this data to a gpu, when you have lower latency access to the same fpu architecture on die.
In the next few years physx will be offloaded from the GPU back to the CPU die, more specifically the gpu on the cpu die. Well it won't be physx, it will be havok and a couple others.
AS manufacturing gets harder Nvidia will be dropping more all encompassing shaders for more efficient AMD style shaders because they HAVE to reduce core size and frequencies, manufacturing is getting harder and harder to shrink processes with more problems with each process.
The sisoft benchmark is fantastic on AMD cards because it was writen with their architecture in mind, the previous benchmarks Nvidia won hugely in were designed with their architecture in mind, neither is better nor worse in reality, though AMD's has a massive advantage when it comes to manufacturing.
To suggest AMD will move back towards less efficient design with more basic shaders, rather than move towards better optimised software shows his lack of understanding.
More basic, and larger shaders would mean a larger core, = less cores per waifer, higher clock speeds needed and, well, worse production. AMD shifted to the current style of architecture for one reason, make a shedload of cheap cores that are less effected by difficulty in manufacturing, its been an unparalleled success, moving away from that would lose them all price/performance advantage they have and thats whats been winning them market share and sales left right and centre.
Likewise, opencl has pretty much smeg all to do with gaming in the future. Within one generation of cpu's we will have a exponential increase in FPU processing power on the cpu die. Considering Nvidia when using physx on die are only using a VERY minor portion of the gpu, remember most of it is doing normal graphics data processing, we're looking at moving a low end gpu's amount of FPU power on die, within a year.
WAY before more fpu is needed in every game released every last person on earth will have a CPU with compared to today, 100x more fpu power onboard, it would be lunacy to offload this data to a gpu, when you have lower latency access to the same fpu architecture on die.
In the next few years physx will be offloaded from the GPU back to the CPU die, more specifically the gpu on the cpu die. Well it won't be physx, it will be havok and a couple others.
AS manufacturing gets harder Nvidia will be dropping more all encompassing shaders for more efficient AMD style shaders because they HAVE to reduce core size and frequencies, manufacturing is getting harder and harder to shrink processes with more problems with each process.
Associate
- Joined
- 3 Sep 2005
- Posts
- 1,670
- Location
- Glasgow, Scotland
The guy that wrote that article doesn't have a clue.