Richteralan

引用:

Originally posted by dom at 2007-7-11 04:36:


真心話 定係 知道某d 事之前走人先

remember  Dave Orton is THE MAN behind R300

Richteralan

Well,
1. He for sure wouldn\'t stay because AMD simply don\'t need double CEO. His resignation is destined on the day AMD merged with ATI.
2. Also, the merged AMD simply won\'t let ATI do their own business like before. ATI now is AMD and they already replanned their future R&D structure.
3. Of course he earnt enough. I guess no-one even need to doubt that.
4. It is a good time to resign (or leave) AMD because the future of AMD won\'t be as bright as ATI was.

So He\'s smart.

Richteralan

引用:

Originally posted by Tommi_Vercetti at 2007-7-11 20:13:


the way of designin AA VGA chips (e.g usin high-cost materials...) may be changed...

This is not really a big problem.

The big problem is the chip architecture design philosophy will be changed.

AMD won\'t let ATI continue to go on the road it did before.

Richteralan

引用:

Originally posted by Tommi_Vercetti at 2007-7-16 23:41:


for e.g?? I think AA won\'t continue to make chips which not 2 consider the need of present usage...

AMD now almost cannot compete with Intel in performance arena.
They sustaining their market by slashing price.

They need to open a new front to battle Intel.

ATI, merged by AMD facing the same situation. They are losing development pace from NVIDIA. ATI also need to open a new front, e.g: GP-GPU (Unfortunately nothing from ATI can compete with Tesla), low-power GPU/integrated GPUs.

These are just my prediction. There are no solid examples since it\'s just the beginning.

Richteralan

The "present usage", I guess you mean present market-target, is of course important for AA. But unfortunately can this market section support their costs and relief their debt?

As I just mentioned, AA need more money, that means a more competent product line, a more loyal fan base, a more lateral/spreaded user base.
If any of these is conflicted with their present market-target, they will make some sacrifice.

Richteralan

引用:

Originally posted by Tommi_Vercetti at 2007-7-20 13:45:


Well, I mean they didn\'t consider the present situation, e.g R600 --> Poor in DX9 performance, but better in DX10....
u r rite, the need of AA is startin a new battlefront, e.g multi-core GPU.

GPU is already extensively parallel when compared with CPU.
So a multi-core GPU won\'t benefit as much as a multi-core CPU.

And the better DX10 performance of R600 is not certain yet because we don\'t know if it\'s gain from 512-bit memory bus or very efficient VLIW compiler(which i seriously doubt that).

Richteralan

引用:

Originally posted by Tommi_Vercetti at 2007-7-22 02:38:


I guess multi-core GPU would not be beneficial IF 2 cores linked in sth like crossfire mode...

even it\'s linked with something more efficient on die level, it won\'t be more beneficial either.
It probably will be less efficient since the 2 cores has to have 2 independent cache/ROP/memory interface, and resulting larger die size.

If using 2 cores with single set of cache/ROP/memory interface, how\'s it gonna be different by adding pipelines within a single core?

Richteralan

Also,
remember why during 3DFX era that their Voodoo2 SLi was more efficient than present day\'s SLI/CF?

This is because GPU at that time was relatively simple. They had single pixel pipeline. It was more "serial" than present day GPU.

By doing SLi, they were able to achieve Single-core to Dual-core CPU performance increase.