Nvidia confirms GT-200 with 1 billion transistors

[Monkeywoman]

"Actually Nvidia wanted to talk on 10 April 2008 during a press meeting in Munich only about new initiatives approximately around its Quadro Grafikkarten. Rather casually one confirmed then however still some data approximately around the next GPU generation, which was to follow the G92-Chips.

Jeff Brown stated openly and without asking to have been that the next architecture actually "GT-200". The GPU will consist of approximately one billion transistors, and it was "pure logic, no memory as CPUs," said Nvidia Manager. This enormous effort circuit coincides with previous rumors, according to which the GT-200 about 200 shader units. Previous G80 and G92 GPUs bring a maximum of 128 of the computing engines.

Furthermore, said Brown nor that Nvidia itself in May 2008 closer to the GT-200 express will. Whether an official architecture-intentioned idea, or product confidentiality agreements with selected press representatives, including several vesting period presented, Jeff Brown, however, was still open."

source; http://translate.google.com/translat...hl=en&ie=UTF-8

[Jon Boy Deluxe]

Finally something worth upgrading to!

When are they out?

[Sly Fox]

Rather casually one confirmed then however still some data approximately around the next GPU generation, which was to follow the G92-Chips.

So does that mean none of this G94 nonsense they'd talked about doing?

I sure do hope so, I wouldn't mind dropping $500 on a card at this point if it would just blow the past series out of the water.

I remember when buying a brand new card from a brand new line meant that all those games you could barely play on your Ti4600... Suddenly ran at 120 FPS on your fancy new 9800 Pro. I miss those days.

**EDIT**

And you know, this is very off topic, but it's something I thought about last night that's been bugging me.

Why is it that Intel for example is so open about their research and future products, and yet Nvidia lets NOTHING out of the box until the damn cards are about ready to ship out?
I much prefer Intel's way myself... It's really nice to know roughly what's coming, roughly when, and roughly how. It makes planning your purchases much easier, and for me at least
it makes it appear that Intel cares at least somewhat about their customers. Nvidia seems content to just keep their mouthes shut and make our butts wait though. Seems to me
the only reason is they want people to make bad buys and purchase things like their 9800GX2 now, so they don't get a GT200 instead. Not exactly the kind of company attitude
that impresses me.

[Shintai]

Finally something worth upgrading to!

When are they out?

Whenever AMD is near or the G9x cores dont sell well more.

[Jimmer411]

Geforce X800.... cmon NV...


So this will pretty much be G80/G92 with more SP and GDDR5? I wonder if they will use the 384bit memory interface like they did with the G80 rather than castrate the new high end card with a 256bit


I really hope that NV allows tri sli on more than their top tier cards. And maybe they can stop using the sli chipset only BS.

[DilTech]

Geforce X800.... cmon NV...


So this will pretty much be G80/G92 with more SP and GDDR5? I wonder if they will use the 384bit memory interface like they did with the G80 rather than castrate the new high end card with a 256bit


I really hope that NV allows tri sli on more than their top tier cards. And maybe they can stop using the sli chipset only BS.

I'm actually expecting a 512bit bus, but I highly doubt it'll be less than 384-bit.

[saaya]

The GPU will consist of approximately one billion transistors, and it was "pure logic, no memory as CPUs,"

complete nonsense... gpus have less cache than cpus, but they DO have cache. nvidia even shows L1 and L2 caches in their own G80 and D92 design schematics!

[BenchZowner]

I'm actually expecting a 512bit bus, but I highly doubt it'll be less than 384-bit.

let's just say that you're in for the BUS
And you're right, we're talking about 32 ROPs
And 1GB framebuffer ( of course you could "math" this one out )

[korda]

So approximately 250 M more trannies than G92. Hopefully it doesn't mean that GT200 is only 25% faster. :p

Looks like we're gonna have fun summer after all.

[Lestat]

you boys did read this is QUATTRO CARDS not geforce..
quatro first, geforce last....

and reading that translation was like watching the video of the dingbat girl from the mis-teen usa pagent totally blow her speech b sounding like a retard...
And such as, and such as, as in, such as...

[BenchZowner]

So approximately 250 M more trannies than G92. Hopefully it doesn't mean that GT200 is only 25% faster. :p

Looks like we're gonna have fun summer after all.

By the way, this isn't the 9900XXx

[Monkeywoman]

anyone notice that they changed the name of the drivers to GeForce from Forceware? this is a sign that the GT-200 will still have GeForce name, but who knows what numbering version they will use.

[Shift]

I'm waiting for trillion transistors though

[Don_Dan]

let's just say that you're in for the BUS
And you're right, we're talking about 32 ROPs
And 1GB framebuffer ( of course you could "math" this one out )

By the way, this isn't the 9900XXx

From your post in this thread and in the GPU-Z thread one could get the impression that you have some insight... Would you like to share it?

[Sly Fox]

From your post in this thread and in the GPU-Z thread one could get the impression that you have some insight... Would you like to share it?

Second'd

[halo112358]

Why is it that Intel for example is so open about their research and future products, and yet Nvidia lets NOTHING out of the box until the damn cards are about ready to ship out?
I much prefer Intel's way myself... It's really nice to know roughly what's coming, roughly when, and roughly how. It makes planning your purchases much easier

Nvidia wants you to buy whatever is on the shelf _now_ so their resellers aren't stuck with a bunch of stale/devalued product.

[BenchZowner]

From your post in this thread and in the GPU-Z thread one could get the impression that you have some insight... Would you like to share it?

Yeap, his name is Nick... Ha! Gotcha!

In fact, I have someone with info directly from nVIDIA, but I'm not going to give out things that can cause troubles to him/me/others.

Last spec for the night... GDDR3 somewhere in between 1000-1100MHz.

[Warboy]

anyone notice that they changed the name of the drivers to GeForce from Forceware? this is a sign that the GT-200 will still have GeForce name, but who knows what numbering version they will use.

Well, its not really Geforce, I just think that make sure people get the right idea, its still forceware.

Also, The only name change has been like so.
Detonator -> Forceware -> Forceware X(started after build 10x.xx)

[Don_Dan]

Yeap, his name is Nick... Ha! Gotcha!

In fact, I have someone with info directly from nVIDIA, but I'm not going to give out things that can cause troubles to him/me/others.

Last spec for the night... GDDR3 somewhere in between 1000-1100MHz.

Sure, I understand that of course!

But the info you already gave are very valuable indeed!

[Warboy]

Yeap, his name is Nick... Ha! Gotcha!

In fact, I have someone with info directly from nVIDIA, but I'm not going to give out things that can cause troubles to him/me/others.

Last spec for the night... GDDR3 somewhere in between 1000-1100MHz.

Awww only GDDR3?

[brinox]

geforce 1080 gtx!

[DilTech]

Awww only GDDR3?

512bit bus with GDDR3 at 2000-2200 would be faster than any available ram on a 256bit memory bus....

So, even if they stick with GDDR3, it'd still have a ton of bandwidth.

So 512bit bus with 32rops...This thing is going to be untouchable when it comes to post processing, and unlike the R600 it'll actually have the power to take advantage of the memory bandwidth it'll have.

[Frank M]

geforce 1080 gtx!

10800GTX you mean?

How's about X800GTX

Just DDR3? While that's a bit disappointing, it can also be a good sign
that this will come out sooner than waiting for GDDR5 to get to mass
production.

[Don_Dan]

Just DDR3? While that's a bit disappointing, it can also be a good sign that this will come out sooner than waiting for GDDR5 to get to mass production.

Using GDDR3 makes it cheaper I guess and availability should be much better.

[Scimitar]

1100 MHz on a 512 bit bus would be 140 GB / second of memory bandwidth. 200 shaders, 32 ROPs, and 1 GB or Ram, damm. Sounds amazing!