Nvidia confirms GT-200 with 1 billion transistors

"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

Finally something worth upgrading to!

When are they out?

Quote:

---

Originally Posted by Monkeywoman

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.

Quote:

---

Originally Posted by Jon Boy Deluxe

Finally something worth upgrading to!

When are they out?

---

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

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.

Quote:

---

Originally Posted by 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.

---

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

Quote:

---

Originally Posted by Monkeywoman

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! :rolleyes:

Quote:

---

Originally Posted by DilTech

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 :p: )

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.

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...

Quote:

---

Originally Posted by 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.

---

By the way, this isn't the 9900XXx ;)

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.

I'm waiting for trillion transistors though

Quote:

---

Originally Posted by BenchZowner

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 :p: )

---

Quote:

---

Originally Posted by BenchZowner

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? :D

Quote:

---

Originally Posted by Don_Dan

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? :D

---

Second'd :up:

Quote:

---

Originally Posted by Sly Fox

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.

Quote:

---

Originally Posted by Don_Dan

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? :D

---

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

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.

Quote:

---

Originally Posted by 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.

---

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)

Quote:

---

Originally Posted by BenchZowner

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

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!

Quote:

---

Originally Posted by BenchZowner

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

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?

geforce 1080 gtx!

Quote:

---

Originally Posted by [cTx]Warboy

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. :yepp: :up:

Quote:

---

Originally Posted by brinox

geforce 1080 gtx!

---

10800GTX you mean? ;)

How's about X800GTX :D

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.

Quote:

---

Originally Posted by Frank M

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.

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!

Quote:

---

Originally Posted by DilTech

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

---

http://img528.imageshack.us/img528/4716/gddr5pu0.png

256bits + GDDR5 gives you more bandwidth.

Look to R600 at 512bits. The formule didn´t work.
The PCB is much more complex, the core is bigger, you have more bus stops inside the chip.

Quote:

---

Originally Posted by Jon Boy Deluxe

Finally something worth upgrading to!

When are they out?

---

lol i so agree.:up:

i scrapped the 9800gtx from 8800gtx upgrade idea when i read lesser specs:shakes:


@ v rr "The PCB is much more complex'

the pcb is merely for heat i thought being thicker\thinner,nothing more.

Compare to SLI: 256, 48/32, 1.28/1gb.

Quote:

---

Originally Posted by v_rr

http://img528.imageshack.us/img528/4716/gddr5pu0.png

256bits + GDDR5 gives you more bandwidth.

Look to R600 at 512bits. The formule didn´t work.
The PCB is much more complex, the core is bigger, you have more bus stops inside the chip.

---

R600's formula didn't work because it couldn't USE the bandwidth. Too few rops, inefficient AA, too few tmus for AF... The rest of the architecture held it back.

Also, you realize you'd need roughly 4ghz GDDR5 at 256bit to match 2ghz GDDR3 at 512bit, right? ;) Also, GDDR3 is likely to be lower latency than GDDR5 as well.

Quote:

---

Originally Posted by [cTx]Warboy

Awww only GDDR3?

---

No need for more expensive RAM...
GDDR3-2100/2200 + 512BIT BUS = 134.4Gbps / 140.8Gbps

Already more than good enough ;)

Quote:

---

Originally Posted by DilTech

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

---

GDDR5-4035 + 256BIT BUS = GDDR3-2200 + 512BIT BUS.
GDDR5-6000 + 256BIT BUS = 192Gbps
GDDR3-2600 + 512BIT BUS = 166.4Gbps

{ Edit: quoting your message, not disagreeing ;) }
Those numbers are just insanely high, and current GPUs won't benefit from a point... and that point is under those figures ;)

And plain bandwidth wont give you a thing... if you haven't equipped the card with efficient sub-units ( ROPs/TMUs/TFUs ).

Quote:

---

Originally Posted by keiths

Compare to SLI: 256, 48/32, 1.28/1gb.

---

Care to elaborate ?
256 you mean SPs.
that 1.28/1GB ???

Time to get a second job. My 8800 gtx is looking to retire, its getting tired.

Quote:

---

Originally Posted by DilTech

R600's formula didn't work because it couldn't USE the bandwidth. Too few rops, inefficient AA, too few tmus for AF... The rest of the architecture held it back.

Also, you realize you'd need roughly 4ghz GDDR5 at 256bit to match 2ghz GDDR3 at 512bit, right? ;) Also, GDDR3 is likely to be lower latency than GDDR5 as well.

---

I think his point is that if the two options can provide the same bandwidth (256+GDDR5 vs 512+GDDR3) it's cheaper and way less complex to design and use the first. And I agree with that. I loved the R600 PCB's visual, but looking at it from a practical POV, 256+GDDR5 is better. Also, 512bits will force NVIDIA to use 16 chips in one PCB, just like the 2900XT. If they plan to do a GX2 with the GT-200, I just can't imagine the monster PCB (two of them more than likely) it'll be with 32 GDDR ICs, that huge chip and all the power circuitry.

Of couse it'll be fast, but I don't want a second 2900XT talking about heat and power consumption.

BenchZowner really gave out a lot of information here, thanks. :)

More ROPs? Bigger memory bus?
Did they finally listen to the thousands of shouts?
Before or after they said "we are opening a can of whoop ass on intel"?

This going to be awesome, only thing I'm worried about is die size. The yields of this will be even worse than that of g80's, pricing will definitely be steep here. I have no doubt it'll definitely be a thorn in the r700's ass because hardly any of ati's xfire drivers are offering consistant above 75% scaling of the 3870x2 vs 3870. Not to mention I can't think of any games except for maybe CoJ that's coded to make use of all the shaders the r700 will have, let alone a single rv770. Thing is, I'd be willing to bet there's going to be a huge difference in price.

This battle is definitely going to be hot. I wouldn't be surprised if either side took the performance crown here (but I'm leaning more towards gt200 due to how inefficient the r600 design is in current games), but this will be more than just pure performance. Its going to be close, and pricing will mean a lot this time around.

Quote:

---

Originally Posted by Shintai

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

---

ha. qft

Quote:

---

Originally Posted by DilTech

R600's formula didn't work because it couldn't USE the bandwidth. Too few rops, inefficient AA, too few tmus for AF... The rest of the architecture held it back.

Also, you realize you'd need roughly 4ghz GDDR5 at 256bit to match 2ghz GDDR3 at 512bit, right? ;) Also, GDDR3 is likely to be lower latency than GDDR5 as well.

---

You forgot that GDDR5 is still not "available" so the original statement stands correct, for the time being.

^_^

65nm
240 SPs [ MADD + MUL more than likely ]
120 TMUs
120 TFUs (?) [ not so sure about that ]

Quote:

---

Originally Posted by 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!

---

192...

Quote:

---

Originally Posted by LordEC911

196...

---

240... ;)

Quote:

---

Originally Posted by BenchZowner

240... ;)

---

So "close to 200" becomes "256 with a cluster disabled."
This just shows(if true) what was obvious about the rumors, huge, hot and expensive due to poor yields.

Edit- Sorry I didn't see your post.

Quote:

---

Originally Posted by BenchZowner

^_^
65nm
240 SPs [ MADD + MUL more than likely ]
60 TMUs
120 TFUs (?) [ not so sure about that ]

---

Let me work through the numbers but they don't really make much sense at first glance.
Edit- Hmmm... So going back to the original and just buffing up G80 numbers.

Quote:

---

Originally Posted by LordEC911

So "close to 200" becomes "256 with a cluster disabled."
This just shows(if true) what was obvious about the rumors, huge, hot and expensive due to poor yields.

Edit- Sorry I didn't see your post.



Let me work through the numbers but they don't really make much sense at first glance.

---

Oh come on! Don't make me do that :D
10 clusters...

http://www.benchzone.com/images/helpfularchitecture.jpg

Quote:

---

Originally Posted by BenchZowner

^_^

65nm
240 SPs [ MADD + MUL more than likely ]
60 TMUs
120 TFUs (?) [ not so sure about that ]

---

how many STFUs? :D j/k

wow if that is true and this this is a single card, single core solutions
this will be crazy fast lol

Quote:

---

Originally Posted by dinos22

how many STFUs? :D j/k

---

One, a major one :p: [ sssh, be quiet! ]

Quote:

---

Originally Posted by dinos22

wow if that is true and this this is a single card, single core solutions
this will be crazy fast lol

---

Supposed to be single card, single die.

wow man
if your source is reliable AMD should be packing :banana::banana::banana::banana:
this is looking like a proper Ultra replacement that will murder it heheh

GT200 performance should be higher than 2* G92GTS in many cases.

*Clocks are estimated, theoritically easy to reach*

G92GTS - GT200
648MHz/1620MHz/GDDR3-1944 - 600MHz/1350MHz/GDDR3-2200
256BIT BUS Width - 512BIT BUS Width
512MB FrameBuffer - 1024MB FrameBuffer
64 TFUs - 120 TFUs
64 TMUs - 120 TMUs
16 ROPs - 32 ROPs
128 SPs - 240 SPs

http://www.youtube.com/watch?v=TBtpyeLxVkI

:D

Quote:

---

Originally Posted by BenchZowner

Oh come on! Don't make me do that :D
10 clusters...

http://www.benchzone.com/images/helpfularchitecture.jpg

---

Interesting approach...

Quote:

---

Originally Posted by BenchZowner

GT200 performance should be higher than 2* G92GTS in many cases.

*Clocks are estimated, theoritically easy to reach*

G92GTS - GT200
648MHz/1620MHz/GDDR3-1944 - 600MHz/1350MHz/GDDR3-2200
256BIT BUS Width - 512BIT BUS Width
512MB FrameBuffer - 1024MB FrameBuffer
64 TFUs - 120 TFUs
64 TMUs - 120 TMUs
16 ROPs - 32 ROPs
128 SPs - 240 SPs

---

You know they are gunning for at least 1500mhz on the shader domain, they can't stop just shy of 1TFlop.

All I can say is it better come out within 3 months. I just bought an eVGA 9800GTX since my 320MB GTS wasn't hacking it on my 24" monitor.

If this comes out in June or early early July then I'll be very happy.

Quote:

---

Originally Posted by Lightning_Rider

All I can say is it better come out within 3 months. I just bought an eVGA 9800GTX since my 320MB GTS wasn't hacking it on my 24" monitor.

If this comes out in June or early early July then I'll be very happy.

---

don't hold ya horses.

just add 5% on to whatever ati brings out.:yawn2: (leaves office to play golf)

212 SP @ 2000 MHZ ----> 9900GTX

160 SP @ 1800 MHZ ----> 9900GT (I want this )

if this or close to it happen my Dream come true

Quote:

---

Originally Posted by nemsis

212 SP @ 2000 MHZ ----> 9900GTX

160 SP @ 1800 MHZ ----> 9900GT (I want this )

if this or close to it happen my Dream come true

---

Don't hold your breath.
9900's will more than likely be what some people call G92b.
55nm & some enhancements.
Aiming to hit default clocks like 800/1890/2000.

Quote:

---

Originally Posted by Sly Fox

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?

---

Ummm maybe because chipset and motherboard manufacturers and developers need long lead times in order to support the chips when they come out? x86 is also an industry standard so Intel can't just work in the shadows and come out at the last minute with what they were doing.

A GPU is vastly different. You don't need to know anything about its inner workings in order to support it. You have a standardized instruction set (DirectX/OpenGL) and standardized interfaces (PCIe). It has nothing to do with respect for customers :shakes:

Sweet Jesus... 1 Billion!

Quote:

---

Originally Posted by STaRGaZeR

I think his point is that if the two options can provide the same bandwidth (256+GDDR5 vs 512+GDDR3) it's cheaper and way less complex to design and use the first.

---

GDDR3 is far cheaper and much more widely available than GDDR5. Also, Nvidia is concerned about the cost of their chips not the cost of the cards. If they use GDDR5 they don't get any benefit from using more expensive memory. Expensive chip + cheap memory for the same net performance and price to the consumer is an overall win for the chip maker. It's the AIB's that get squeezed.

cool news. Wonder if it will need pcie 2.0 out the door for full performance.

and where are the games to take advantage of a gpu like this? lot's of card's but no new games lately...sloww

Quote:

---

Originally Posted by strange|ife

cool news. Wonder if it will need pcie 2.0 out the door for full performance.

and where are the games to take advantage of a gpu like this? lot's of card's but no new games lately...sloww

---

Whether we NEED it or not, bet your ass that we'll be told its the only way to get full functionality.

Who remembers nVidia say the G80 would only work in PCI-E x16 slots? <- interesting wording!

I also saw an e-note from a GPU companys tech support to an 8800GT owner stating "Hi, thanks for your message, yes these cards are not compatible with PCI-E GEN 1.0 0r 1.1. They both will only run on PCI-E GEN 2.0 mainboards."

Quote:

---

Originally Posted by trinibwoy

GDDR3 is far cheaper and much more widely available than GDDR5. Also, Nvidia is concerned about the cost of their chips not the cost of the cards. If they use GDDR5 they don't get any benefit from using more expensive memory. Expensive chip + cheap memory for the same net performance and price to the consumer is an overall win for the chip maker. It's the AIB's that get squeezed.

---

Do you think the difference in cost will be enough to compensate double the chips, double the PCB complexity? Maybe in the first weeks/months of GDDR5, but not later.

Quote:

---

Originally Posted by STaRGaZeR

Do you think the difference in cost will be enough to compensate double the chips, double the PCB complexity? Maybe in the first weeks/months of GDDR5, but not later.

---

Maybe not, but 512-bit GDDR3 is a far safer bet from Nvidia's perspective as they have more control. If AMD could do it with R600 I don't see why Nvidia couldn't do it now. Going with GDDR5 puts your fate in the hands of the memory manufacturers. Personally, I would prefer GDDR5 on 256-bit for the obvious reasons.

1 billion sounds great, but its only ~250 million more than g92 has.

So when will this monster be out?

70 to 80&#37; faster than a 8800gtx in demanding games,retail for $500?

So a single card that is 30% than 9800gx2?

Exactly what I need

Quote:

---

Originally Posted by Don_Dan

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? :D

---

Oh, benchzowner is in the know but he just likes to tease - he did the same thing when the G90 was supposed to come out last year before it got delayed.

Quote:

---

Originally Posted by EternityZX9

Oh, benchzowner is in the know but he just likes to tease - he did the same thing when the G90 was supposed to come out last year before it got delayed.

---

Sorry, wasn't intended m8.
Original info was what I gave you guys.
But... with the help of ATi the plans changed...dramatically & drastically.

Quote:

---

Originally Posted by BenchZowner

Sorry, wasn't intended m8.
Original info was what I gave you guys.
But... with the help of ATi the plans changed...dramatically & drastically.

---

I know, I'm not upset, I know how the business goes :)

I'm just saying you're a big tease! :p:

Quote:

---

Originally Posted by 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...

---

You should read the original, which, indeed, states that it's about consumer cards ;)

Btw will we see both a high-end (GTX) and mid-high-end card (GTS) at launch? Or only a high-end card?

Quote:

---

Originally Posted by Jacky

Btw will we see both a high-end (GTX) and mid-high-end card (GTS) at launch? Or only a high-end card?

---

Don't know about a high end and mid-high launch altogether.
However, the "GTS" will be something like 7 or 8 or 9 clusters.

I'd like to go ahead and confirm that I've been hearing similar from my sources as bench. Big surprise, huh bench? We always seem to have the same info on nvidia parts.

Anyway, I'm hearing a late may-early june release, just thought I'd add that.

BTW, for those who forget..one of the number one reason for high-end shortages in the past has been because companies go for high-end ram near it's initial launch, and stuck with not being able to get said ram. Remember the 6800Ultra and X800XT/PE? Both had countless supply problems because of going for what was the best ram at the time.

Sticking with GDDR3 means ram will be in high supply, meaning NVidia can pump out as many cards as they can make gpus.

Quote:

---

Originally Posted by BenchZowner

Don't know about a high end and mid-high launch altogether.
However, the "GTS" will be something like 7 or 8 or 9 clusters.

---

Some rumours say that the "GTS" branded cards will be terminated .

really? I mean geforce 9 is pretty screwed up right now, but victor wang has gotten a hold of a 9800gts (whatever the hell it is) and a 9800gt is supposed to coming on the way. I think the big reason for geforce 9 being so screwed up is because of the 8800gt being too powerful eliminating the old performance gap between the gtx and gts cards

Quote:

---

Originally Posted by BenchZowner

Don't hold your breath.
9900's will more than likely be what some people call G92b.
55nm & some enhancements.
Aiming to hit default clocks like 800/1890/2000.

---

I don't think so

the new GT200 cards will be Completely new cards

Quote:

---

Originally Posted by w0mbat

1 billion sounds great, but its only ~250 million more than g92 has.

---

"only 250 million"? it's 25% more than the 9800gtx has... a 8600gts got a bit more than 250 million transistors - and a 8600 gts isn't that bad. so with 250 million transistors there's a lot than can be improved.

Quote:

---

Originally Posted by nemsis

I don't think so

the new GT200 cards will be Completely new cards

---

umm...

erm...

hmm...

how to put this...


You made a mistake:p:

G92b isn't a new design, its just a 55nm version of g92 and has nothing to do with gt200. G92b will be high end mobile and desktop performance with g96b and g98 being performance+mainstream mobile and mainstream+low end desktop. GT200 will be high end desktop and probably will have its own naming scheme, which is exactly what benchzowner said so I don't see what the problem is

Quote:

---

Originally Posted by DilTech

Big surprise, huh bench?

---

Big coincidence :p:

Wanna try our "luck" on the other side ? ( ATi ) :D

Quote:

---

Originally Posted by whocaresbg

Some rumours say that the "GTS" branded cards will be terminated .

---

They're talking about 8800GTS going EOL.
Not the naming scheme generally ;)

Quote:

---

Originally Posted by nemsis

I don't think so

the new GT200 cards will be Completely new cards

---

That's what I said. And as Ali_G clarified ;)
The G92b will be a follow up to the G92 cards that we have now, but built on TSMC's 55nm Low-K process.
The GT200 will be the "real new" thing.

Quote:

---

Originally Posted by RaZz!

"only 250 million"? it's 25% more than the 9800gtx has... a 8600gts got a bit more than 250 million transistors - and a 8600 gts isn't that bad. so with 250 million transistors there's a lot than can be improved.

---

Sources are indicating a transistor count near 1.3bil
Even if it ends up being 1bil... still it'd be:
319m over the G80 [ I'm only counting the G80 core, not with the nV-IO ]
246m over the G92

Quote:

---

Originally Posted by AliG

umm...

erm...

hmm...

how to put this...


You made a mistake:p:

G92b isn't a new design, its just a 55nm version of g92 and has nothing to do with gt200. G92b will be high end mobile and desktop performance with g96b and g98 being performance+mainstream mobile and mainstream+low end desktop. GT200 will be high end desktop and probably will have its own naming scheme, which is exactly what benchzowner said so I don't see what the problem is

---

:up:

wait

wait

the GT200 cards come in june.. right ???

and there name is 9900GT and 9900GTX ???

source fudzilla.com

if this wrong i think i will have heart attack

fudzilla what i can say anymore than that :)

Quote:

---

Originally Posted by nemsis

wait

wait

the GT200 cards come in june.. right ???

and there name is 9900GT and 9900GTX ???

source fudzilla.com

if this wrong i think i will have heart attack

---

Unfortunately my info will depress you.
june = G92b ( 9900XXXs )
GT200 = q4 08 more than likely.

Quote:

---

Originally Posted by BenchZowner

Unfortunately my info will depress you.
june = G92b ( 9900XXXs )
GT200 = q4 08 more than likely.

---

damnittttttttttt....oh well looks like i'll just have to stick it out for a little bit more

Quote:

---

Originally Posted by BenchZowner

Unfortunately my info will depress you.
june = G92b ( 9900XXXs )
GT200 = q4 08 more than likely.

---

I kinda expected this, Nvidia loves launching cards right a month or so before Christmas since pretty much everyone goes on a shopping spree that time of the year.

Quote:

---

Originally Posted by NeGe0

damnittttttttttt....oh well looks like i'll just have to stick it out for a little bit more

---

Were you expecting something different?

actually I was hoping for late q3, but looks more like late q4 instead. How ironic, especially with how the r600 launch went. Makes you wonder if we have another r600 on our hands because nvidia's been so quiet about it compared to how they were about g80 (especially now that ati's giving out info it really makes you wonder) not to mention trying to debug a core this complex must be killer (think about it, the current r600 is a cut down of the original, the original was too complex and was full of bugs)

Quote:

---

Originally Posted by BenchZowner

Unfortunately my info will depress you.
june = G92b ( 9900XXXs )
GT200 = q4 08 more than likely.

---

that sounds like a legit info!

Quote:

---

Originally Posted by DilTech

I'd like to go ahead and confirm that I've been hearing similar from my sources as bench. Big surprise, huh bench? We always seem to have the same info on nvidia parts.

Anyway, I'm hearing a late may-early june release, just thought I'd add that.

BTW, for those who forget..one of the number one reason for high-end shortages in the past has been because companies go for high-end ram near it's initial launch, and stuck with not being able to get said ram. Remember the 6800Ultra and X800XT/PE? Both had countless supply problems because of going for what was the best ram at the time.

Sticking with GDDR3 means ram will be in high supply, meaning NVidia can pump out as many cards as they can make gpus.

---

Doesn`t seem strange you have said that Benchzowner and you have similar info about GT200 and then he is saying GT200=Q4 release and you are saying it`s about May-June:) So who is right?:confused:

Quote:

---

Originally Posted by BenchZowner

Unfortunately my info will depress you.
june = G92b ( 9900XXXs )
GT200 = q4 08 more than likely.

---

So it means FUdos info about GT200-300 tapeed out is BS? If it`s truth then we should see GT200 in 2-3 months not later.

The second thing - is there ANY possibility (if these specs are true of course) to push 120TMU, 240SP, 32 ROPs and 512-MC :eek: in "only" 1B trannies? It`s only about 250M more than G92 which is supposed to have about 2x less TMU, 2x less SP, 2x lessROPs etc. With these rumoured specs (i hope they are really true) GT200 should have about 1.2B trannies at least.


Hmm i can only say i really wait for NVIDIAs "next BIG thing" (worthy G80 successor) and i hope that after crappy GF9 series they will release very powerful GPU (single die).

PS. You have said performance is better than 2xG92 (GF8800GTS) in many cases? But it`s only rumoured performance or someone already has working samples of it and this performance level is what this GPU can really do?:rolleyes:

Quote:

---

Originally Posted by Barys

then he is saying GT200=Q4 release and you are saying it`s about May-June:) So who is right?:confused:

---

Both timeframes can be wrong as well.
Release dates ain't firm, and of course depend on various things.

Quote:

---

Originally Posted by Barys

So it means FUdos info about GT200-300 tapeed out is BS? If it`s truth then we should see GT200 in 2-3 months not later.

---

The fact that it has been taped out ( if it's true ), doesn't mean that it's going to be released soon.
You can always keep the goodies for the future.
Releasing it now, can kill the prices & sales of the G92 and especially the 9800GX2 & the G92b.

Quote:

---

Originally Posted by Barys

The second thing - is there ANY possibility (if these specs are true of course) to push 120TMU, 240SP, 32 ROPs and 512-MC :eek: in "only" 1B trannies? It`s only about 250M more than G92 which is supposed to have about 2x less TMU, 2x less SP, 2x lessROPs etc. With these rumoured specs (i hope they are really true) GT200 should have about 1.2B trannies at least.

---

Check my posts... people & rumors on the web talk about 1bil, but I'm talking about 1.3b [ info from my source ]

Quote:

---

Originally Posted by Barys

PS. You have said performance is better than 2xG92 (GF8800GTS) in many cases? But it`s only rumoured performance or someone already has working samples of it and this performance level is what this GPU can really do?:rolleyes:

---

That's my performance estimate, if the specs I've posted are true ;)

Quote:

---

Originally Posted by BenchZowner

That's my performance estimate, if the specs I've posted are true ;)

---

Bit disappointing then isnt it?

9800gx2 or 9900gx2 should be very affordable by the time the GT-200 launches with similar performance..

Nvidia have made sli too good lol, killing there high end single card...

Quote:

---

Originally Posted by y2kbos

Bit disappointing then isnt it?

9800gx2 or 9900gx2 should be very affordable by the time the GT-200 launches with similar performance..

Nvidia have made sli too good lol, killing there high end single card...

---

Well, I always drop my estimates down, underneath the average.
The performance can be higher than that... depends on various factors.
It will be faster than a 9800GX2/9900GX2... but by how much ? 15&#37; average or so (?)

The interesting part will be the high resolution & details with AA/AF, where the card will have a greater advantage over resolutions lower than 180x1050 ( when compared to the GX2 )

I can't really say for sure whether its taped out or not, but reason could support either answer. For one just because the design is taped out doesn't mean its 100&#37; bug free, just means the design is pretty much finalized and the debugging is in its final stages (especially driver debugging). But also look at the r600's launch, the original r600 rumors talked about a monster card (but all we got was monster power consumption lolz) but ati kept quiet and it turned out to demolish everything previously released by ati, but still no performance crown stealer. So far gt200 is following the same steps, extremely complex and large die, 512 bit memory, rumors of monster. Thing is, usually nvidia gives out a lot more information than this on projects they are confident on. So either they are worried about the rv770 stealing the performance section where the largest sales have been taking place recently (most profit:costs comes from high end though) or something's wrong with gt200 if its not taped out yet. But don't forget, g80 was in volume production months before it launched, so that could mean there are driver issues or samples are still being given to coders, or simply yields are crap so nvidia is stockpiling them. You never know until we actually see a real preview.

As I said, if you think about it rationally based off past events, you could support either arguement.

Quote:

---

Originally Posted by BenchZowner

Both timeframes can be wrong as well.
Release dates ain't firm, and of course depend on various things.



The fact that it has been taped out ( if it's true ), doesn't mean that it's going to be released soon.
You can always keep the goodies for the future.
Releasing it now, can kill the prices & sales of the G92 and especially the 9800GX2 & the G92b.



Check my posts... people & rumors on the web talk about 1bil, but I'm talking about 1.3b [ info from my source ]



That's my performance estimate, if the specs I've posted are true ;)

---

looks like we're on the same page for most things. Taped out doesn't necessarily mean anything at this point since nvidia has time to kill. ATI pushed the rv670 launch extremely hard since they were losing tons of market share and forced nvidia to rush their g92 launch to combat it (imo, I don't think g92 was ever supposed to become part of geforce 8, the g80 design was doing so well I actually think nvidia was planning on releasing g92 as the geforce 9 high end and performance, but was worried about the rv670 head stomping their performance segment). Just look at my last post, you could make rational arguments about whether it was taped out or not very easily

But I don't think g92b will ever be a high end segment. The only reason I can see g92 being shrunken down to 55nm is to improve yields to combat the rv770 in the performance section. That and I think the gt200 dies will be so large there won't be any cut downs for the performance section, all of them will have to be put towards high end for nvidia to make money off it. So if that's true, gt200 could never hurt g92b's sales since there's going to be such a large price difference between performance and high end. Think about it, I mean its not like the 9800gx2 is killing the 8800gts 512mb sales (its taking some, but a very small percentage)


As for 1.3 billion trannies, honestly I couldn't care if it had 5 trillian trannies so long it performed well in games. The r600 proved that numbers on paper mean nothing until they are proven to do so. While g80 is a lot more efficient in current games than the r600, who knows what either the r700 or gt200 will do (especially if the new architecture rumor is true, gt200 doesn't necessarily have to perform like a 240 shader version of g80 as the pipeline can be significantly different; not to mention a lot of the r600's major bottlenecks are finally being removed which can make all the difference).


As for your performance estimate, I have to agree. Without a doubt, the sli gts 512mb will be nothing in comparison to gt200 if gt scales like g80 because we're talking a single die here vs sli's poor scaling (though it has greatly improved, xfire still scales better imo and even that can't do 100% scaling in 10% of the time)

@AliG

If the specs are true, they can easily disable some clusters, and get the GT & GTS out.
Even with 6 clusters you have a beast card to show.
144 SPs
72 TFUs
72 TMUs

Quote:

---

Originally Posted by AliG

The r600 proved that numbers on paper mean nothing until they are proven to do so

---

Well... the numbers were clear for people to figure that it won't be any good in gaming ( especially with AA & AF enabled ) since... it only had 16 TFUs, and lacked hardware AA resolve.

It's like designing a RWD Mitsubishi Lancer for off-road races against a 4WD Impreza :D

true, but still, I'm talking about die costs due to the high size not about the performance. Unless if you can just chop off parts of the die, something tells me none of the gt200 dies are going to the performance section. Perhaps a gts to go with the gtx (or whatever the new naming scheme will be), but nothing else

Quote:

---

Originally Posted by AliG

true, but still, I'm talking about die costs due to the high size not about the performance. Unless if you can just chop off parts of the die, something tells me none of the gt200 dies are going to the performance section. Perhaps a gts to go with the gtx (or whatever the new naming scheme will be), but nothing else

---

A GTS is planned.
GT ? If ATi comes up with something worth the effort :p:

I realize a gts is planned as they'll have to do something with the chips that don't meet the bin, but unless if g92b will be a mainstream product, I'm almost certain it'll be the gt or whatever the performance models are (who knows, maybe we'll see a x700 from now on).

As for ati coming up with something that compete, I'm sure of that, the biggest concern I have for ati is that unless games come out designed specifically for multi gpu setups (which to be honest would be quite nice to have a few of those every year as it would mean we'd have a few new games like crysis every year as the coders would put in a lot more detail than the regular games), the r700 will be a decent high end, but won't be powerful enough to justify getting it over the rv770 due to budget constraints or gt200 for pure performance

I'm happy with my G92 GTS 512MB, it plays all my games at a good framerate and Image Quality on my 20" NEC LCD.
Well...all except for Crysis, but I'm of the opinion that Crysis is a bit like Farcry in the sense it was built to last the test of time longer than other games released at the same time.
I might upgrade to the 9900 series IF they let me crank up the details @ 1680*1050, otherwise I shall wait until the next big thing. (ATi or nVidia).
John

Quote:

---

Originally Posted by BenchZowner

june = G92b ( 9900XXXs )

---

Hi BenchZowner...

Just wondering performance difference in the 2 eVGA 9800GTX SSC cards I just bought from the egg? Then the 9900XXX???? 5% 10%?? Thanks... :up:

Quote:

---

Originally Posted by Lu(ky

Hi BenchZowner...

Just wondering performance difference in the 2 eVGA 9800GTX SSC cards I just bought from the egg? Then the 9900XXX???? 5&#37; 10%?? Thanks... :up:

---

G92b will more than likely be a intact G92 @ 55nm, with GDDR5 memory.
The GTX will probably be clocked at 800-850MHz GPU, GDDR5-3000 - GDDR5-3600, 1998 - 2052MHz SPs.
256BIT BUS Width.

A single 9900GTX will be slower than your SLI setup there.

Wait....so when the GT200 comes out will we finally have something that can play Crysis at very high settings at 60 frames or better? :D

Quote:

---

Originally Posted by EternityZX9

Wait....so when the GT200 comes out will we finally have something that can play Crysis at very high settings at 60 frames or better? :D

---

Yeah, it should run as sweet as butter with the GT200 High End part.

gt200 by christmas. but that might be optimistic.

seeing as how there are so many cf and sli options around right now:lol:, but crippled by vram, bus and shaders.

Please come out within 90 days of April 15 :rofl:
(I get my video card on the 15th)

My fund is already being saved for GT200 ;). One shall be mine :D! Hope for sooner release...

Quote:

---

Originally Posted by GoldenTiger

My fund is already being saved for GT200 ;). One shall be mine :D! Hope for sooner release...

---

November, at least ;)