Mais uma achas para a fogueira, lançadas pelo Inquirer.
Vamos ver se a realidade confirma estes dados...
"NV 30 "30% faster" than Radeon 9700 Pro
Details emerge of 48GB effective memory bandwidth
By Fuad Abazovic: Friday 15 November 2002, 09:30
WE EXPECT NVIDIA TO not only present the NV28M but probably also the NV30 at Comdex next week.
The British press gig is being held on the 18th, anyroad.
If you remember, Nvidia decided to opt for a .13µ (micron) process but faced some obstacles because shrinking the die and increasing the frequency greater than 400MHz is technically tough.
0.15µ marchitecture with its aluminum interconnections is limited to 400 MHz while Nvidia plans to clock its chips at over 400 MHz and even to 500MHz.
Will it get to 500MHz? It will have a nice old try.
Nvidia will use DDR II memory running at 500MHz or 1GHz effective that will be the fastest memory used so far in this type of semiconductor, as we reported before.
The NV30 chip will feature eight pipelines and two TMUs – that's one more than ATI – and this will give it quite a boost since it will be able to render 16 textures per pass.
Documents seen in the Vienna equivalent of the Porcupine pub reveal that an Nvidia card will be able to deliver 200 million triangles per second and will be "the most efficient architecture in the world".
We can also confirm that with 1.0 GHz data rate plus Lightspeed Memory Architecture 3, a card will be able to reach an amazng 48 GB/s, double what the Radeon 9700 PRO can do with its DDR 256 bit memory. Remember Nvidia will use 128 bit memory this time and will save its breath for 256 bit DDR II for future designs.
As for anti-aliasing, you will be able to use FSAA 8X now and anisotropic filtering will go up to 128 tap anisotropic filtering leaving ATI's methods in the dust.
Here, briefly, is what that means:
ATI method Nvidia Method
0 AF 0 AF
1x AF 4Tap AF
2x AF 8Tap AF
4x AF 16Tap AF
8x AF 32Tap AF
16x AF 64Tap AF
128Tap AF
As for performance, the Nvidia NV30 card will be about 30 per cent faster than Radeon 9700 PRO in FSAA, Anisotropic filtering is set for maximum settings.
Performance wise you should be able to see 20000 to 22000 3dmarks 2001 SE on the fastest processors but we will keep our powder dry on those claims until we've seen it in action.
NV30 will use higher precision 64 and 128 bit colour which is "film-quality color precision – higher precision than used for Toy Story2, Floating point or integer", Nvidia said.
As for the known things it will use CineFX engine for cinematic rendering and we assume we will see some nice demos to follow these cards.
Some figures seen on the Nvidia official documents in Vienna show that the NV30 will be 2.5 times faster in quake 3 than TI 4600, 3.5 times faster in Doom 3, and three times faster than 3dmark 2001 SE but this remains to be seen.
We already know the NV30 will use 125 millions of transistors, will use AGP 8X and will have support for DirectX 9 and even beyond.
As for the name we wont spoil a big plans to Nvidia to surprise the world and will keep you waiting for that for sometime next week, when NV30 will debut with a new mobile baby NV28M. It might not be the same day, OK?
This will surely be the fastest offering once Nvidia manage to have it in shelves and the company still claims that some cards will be available for Christmas in shops. Let's wait and see.
Nvidia will wrench the performance crown for 3D performance if its NV30 claims are true. Why should it boast if they're not?"
Fonte : The Inquirer
Vamos ver se a realidade confirma estes dados...
"NV 30 "30% faster" than Radeon 9700 Pro
Details emerge of 48GB effective memory bandwidth
By Fuad Abazovic: Friday 15 November 2002, 09:30
WE EXPECT NVIDIA TO not only present the NV28M but probably also the NV30 at Comdex next week.
The British press gig is being held on the 18th, anyroad.
If you remember, Nvidia decided to opt for a .13µ (micron) process but faced some obstacles because shrinking the die and increasing the frequency greater than 400MHz is technically tough.
0.15µ marchitecture with its aluminum interconnections is limited to 400 MHz while Nvidia plans to clock its chips at over 400 MHz and even to 500MHz.
Will it get to 500MHz? It will have a nice old try.
Nvidia will use DDR II memory running at 500MHz or 1GHz effective that will be the fastest memory used so far in this type of semiconductor, as we reported before.
The NV30 chip will feature eight pipelines and two TMUs – that's one more than ATI – and this will give it quite a boost since it will be able to render 16 textures per pass.
Documents seen in the Vienna equivalent of the Porcupine pub reveal that an Nvidia card will be able to deliver 200 million triangles per second and will be "the most efficient architecture in the world".
We can also confirm that with 1.0 GHz data rate plus Lightspeed Memory Architecture 3, a card will be able to reach an amazng 48 GB/s, double what the Radeon 9700 PRO can do with its DDR 256 bit memory. Remember Nvidia will use 128 bit memory this time and will save its breath for 256 bit DDR II for future designs.
As for anti-aliasing, you will be able to use FSAA 8X now and anisotropic filtering will go up to 128 tap anisotropic filtering leaving ATI's methods in the dust.
Here, briefly, is what that means:
ATI method Nvidia Method
0 AF 0 AF
1x AF 4Tap AF
2x AF 8Tap AF
4x AF 16Tap AF
8x AF 32Tap AF
16x AF 64Tap AF
128Tap AF
As for performance, the Nvidia NV30 card will be about 30 per cent faster than Radeon 9700 PRO in FSAA, Anisotropic filtering is set for maximum settings.
Performance wise you should be able to see 20000 to 22000 3dmarks 2001 SE on the fastest processors but we will keep our powder dry on those claims until we've seen it in action.
NV30 will use higher precision 64 and 128 bit colour which is "film-quality color precision – higher precision than used for Toy Story2, Floating point or integer", Nvidia said.
As for the known things it will use CineFX engine for cinematic rendering and we assume we will see some nice demos to follow these cards.
Some figures seen on the Nvidia official documents in Vienna show that the NV30 will be 2.5 times faster in quake 3 than TI 4600, 3.5 times faster in Doom 3, and three times faster than 3dmark 2001 SE but this remains to be seen.
We already know the NV30 will use 125 millions of transistors, will use AGP 8X and will have support for DirectX 9 and even beyond.
As for the name we wont spoil a big plans to Nvidia to surprise the world and will keep you waiting for that for sometime next week, when NV30 will debut with a new mobile baby NV28M. It might not be the same day, OK?
This will surely be the fastest offering once Nvidia manage to have it in shelves and the company still claims that some cards will be available for Christmas in shops. Let's wait and see.
Nvidia will wrench the performance crown for 3D performance if its NV30 claims are true. Why should it boast if they're not?"
Fonte : The Inquirer