I realy don't have a clue but i can't seem to find a reason why no tech site made these obvious remarks about the upcoming ATI products.(maybe they are obvious).
1)ATI is trying to push the 0.15 micron technology to the max but I certainly doubt that they are going to reach 425mhz (unless they want to have the availability of the AMD XP 2800+).The absolute top to have a logical product launch and availability is 400mhz.(I think 400MHZ core/memory is the most realistic figure and certainly will be enough to beat FX 5800ulta by 10% (ATI statements).Personaly I doubt even for the memory clock figure since is not the easiest thing in the world to find sufficient quantities of 400mhz mem modules in the price they want.We will analyze why later.
2)Of cource the majority of R350 production will not meet the 400mhz stable operation, so ATI is going to through the rest (majority) to the sea.Just kidding.They will be 2 versions of R350.One the FX5800 ultra beater (call me 400mhz version) and a lower clocked version.Possibly the lower clocked version is going to be 325 core/325 memory (this is just a speculation).But if you analize carefully what ATI needs to beat the simple version of the FX5800 and you check also the difference between the higher and lower clocked versions of a GPU historically and also the memories possibilities ATI has today 325/325 is a good guess.
I don't want to be extemely technical about the different aproach ATI & NVIDIA has over DX9 hardware implementation because I can't.(This is a joke).But if you check all the benchmark scores between R300 and NV30 and you made the right calculations you can easily tell why ATI needs the above aproach.
ofcource maybe you want to check also the following:
R300 4geometry procesors vs 3 (equal more or less) geometry procesors NV30
So 400x4=1600 and 3x500=1500.Just kidding.This is unaxeptable mathematical aproach between apple and oranges.With the same unaxeptable aproach you could check issues between 96Bit (R300) and 64bit & 128BIT (NV30) and clock cycles between 32bit, 64bit, 128bit implementations and ofcource ALOT othe techical staf between them.
I want to say something that is very obvious and nobody mention it.Nvidia has way better software division and driver team than ATI.ATI is also good but no in the same level us Nvidia.You should try checking 3d Workstations benchmarks (open gl 1.2 1.3) between nvidia quadro 4 900 and ATI FIRE GL X1 (R300 workstation version) without crying.Ofcource Nvidia is having for the first time a completely new architecture and I don't know how much they can relly on the old good way of doing things and ATI for the first time has the opportunity to develop good drivers since they are already 6months ahead of Nvidia and the R350 is fundamentaly a similar aproach of R300 so they have more time to mature the drivers.But it is obvious from the benchmarks that so carefully saw(especialy the 3d mark based) that the minimum increase Nvidia is able to achive through a new driver is 15%.I think Nvidia intetionally did this
in order to release new driver when the R350 is out and close the gap.Ofcource ATI knows that and I don't think They are sitting idle.
Also to me it is obvious that Nvidia has problems maxing out the potentials of 0.15-micron technology in relation with GPU clock speed.Also from what I saw from the power consumption and heat that the NV30 is producing they have also problems with 0.13.Certainly in relation with ATI they are distant second.Something note so obvious is that they having problems with the analog part of the design process but again maybe I am wrong.
From what I think ATI doesn't need 8 pixel/16 texel units to beat nvdia so I don't think they 'll use this aproach.(I pray but I don't think).Since ATI claims that with the upcoming R350 they will beat nvidia by only 10% (I guess they mean without FSAA & Anisotropic 10%) it is 8-pixel pipe also.But there must be changes also elseware in order not be called R305.Most of this tommorow.I think the most logical aproach in the naming of R350 is indeed Radeon 9900pro (400/400) and Radeon 9900 (325/325).I am already tired so tommorow I will post about RV350 (Big suprices everyone) and what I forgot today.
1)ATI is trying to push the 0.15 micron technology to the max but I certainly doubt that they are going to reach 425mhz (unless they want to have the availability of the AMD XP 2800+).The absolute top to have a logical product launch and availability is 400mhz.(I think 400MHZ core/memory is the most realistic figure and certainly will be enough to beat FX 5800ulta by 10% (ATI statements).Personaly I doubt even for the memory clock figure since is not the easiest thing in the world to find sufficient quantities of 400mhz mem modules in the price they want.We will analyze why later.
2)Of cource the majority of R350 production will not meet the 400mhz stable operation, so ATI is going to through the rest (majority) to the sea.Just kidding.They will be 2 versions of R350.One the FX5800 ultra beater (call me 400mhz version) and a lower clocked version.Possibly the lower clocked version is going to be 325 core/325 memory (this is just a speculation).But if you analize carefully what ATI needs to beat the simple version of the FX5800 and you check also the difference between the higher and lower clocked versions of a GPU historically and also the memories possibilities ATI has today 325/325 is a good guess.
I don't want to be extemely technical about the different aproach ATI & NVIDIA has over DX9 hardware implementation because I can't.(This is a joke).But if you check all the benchmark scores between R300 and NV30 and you made the right calculations you can easily tell why ATI needs the above aproach.
ofcource maybe you want to check also the following:
R300 4geometry procesors vs 3 (equal more or less) geometry procesors NV30
So 400x4=1600 and 3x500=1500.Just kidding.This is unaxeptable mathematical aproach between apple and oranges.With the same unaxeptable aproach you could check issues between 96Bit (R300) and 64bit & 128BIT (NV30) and clock cycles between 32bit, 64bit, 128bit implementations and ofcource ALOT othe techical staf between them.
I want to say something that is very obvious and nobody mention it.Nvidia has way better software division and driver team than ATI.ATI is also good but no in the same level us Nvidia.You should try checking 3d Workstations benchmarks (open gl 1.2 1.3) between nvidia quadro 4 900 and ATI FIRE GL X1 (R300 workstation version) without crying.Ofcource Nvidia is having for the first time a completely new architecture and I don't know how much they can relly on the old good way of doing things and ATI for the first time has the opportunity to develop good drivers since they are already 6months ahead of Nvidia and the R350 is fundamentaly a similar aproach of R300 so they have more time to mature the drivers.But it is obvious from the benchmarks that so carefully saw(especialy the 3d mark based) that the minimum increase Nvidia is able to achive through a new driver is 15%.I think Nvidia intetionally did this
in order to release new driver when the R350 is out and close the gap.Ofcource ATI knows that and I don't think They are sitting idle.
Also to me it is obvious that Nvidia has problems maxing out the potentials of 0.15-micron technology in relation with GPU clock speed.Also from what I saw from the power consumption and heat that the NV30 is producing they have also problems with 0.13.Certainly in relation with ATI they are distant second.Something note so obvious is that they having problems with the analog part of the design process but again maybe I am wrong.
From what I think ATI doesn't need 8 pixel/16 texel units to beat nvdia so I don't think they 'll use this aproach.(I pray but I don't think).Since ATI claims that with the upcoming R350 they will beat nvidia by only 10% (I guess they mean without FSAA & Anisotropic 10%) it is 8-pixel pipe also.But there must be changes also elseware in order not be called R305.Most of this tommorow.I think the most logical aproach in the naming of R350 is indeed Radeon 9900pro (400/400) and Radeon 9900 (325/325).I am already tired so tommorow I will post about RV350 (Big suprices everyone) and what I forgot today.
The new Rads are expected to use DDR-II and we already know the cores CAN hit 400mhz although 350mhz is almost certain. They'll keep 9700PRO and below so these will use up the cores which can't hit 350-400mhz. By using DDR-II they should be hitting at least 800mhz if not 1000mhz and on a 256bit bus that's a LOT of bandwidth. Considering GF-FX is only very slightly faster than Rad9700PRO even the cheaper R350 @ 350/800 should beat it comfortably let alone the 400/1000 version (speculation of course). Perhaps these 2 new R350 cards will be 9800 and 9900, maybe ATI will hit the 10000 or go nVidia-esch and call it Radeon-SpecialFX (sounds like a mixture of washing powder and breakfast cereal), Radeon9700PRO-Platinum (9700-PI?) or possibly even come up with a new name to replace the aging detergent sounding 'Radeon'.
Facebook
Twitter