G80 pics - Talk about it here.

[sillyeagle]

3DM05 / 3DM06
8800GTX 16909 / 11843
8800GTS 15233 / 10071
7950GX2 12931 / 8781
1950XTX 11866 / 7007
1950PRO 10044 / 5600
(nana2)

When they get these scores what resolution/AA are they using? Is there a standard for this test?

[jAkUp]

Standard resolution is 1280x1024 for 3dmark06. 05 is 1024x768

[Fotis]

Sheesh, some of you ATi boys are so sensitive. Just for the record, I owned an 850XT before my 7900GTX, so you can call the guard dogs off now.

I'm not splitting hairs w/ either one of you regarding temps on past generation cards. I thought that both the 1800XT (because of its higher clockspeeds) and 1900XT ran significantly hotter than their nvidia counterparts on the high-end. If you don't, that's fine. Maybe I should preclude all my statements with "In my opinion" from now on out so you two won't get your panties in a wad.

And in regards to "features implemented," hasn't it always been a party-line of the ATi camp that their IQ is so much better than nvidia's? Doesn't that supposed increased IQ come at a cost? As in extra features, possibly? As in extra-generated heat, probably?
You are way to new on this site to call me an "Ati boy".I've used every nvidia card since the TNT age and many ATI cards too but I'm no companys fanboy.

That aside, where do you get that I said the R5xx series is cooler than G7x cause I wrote the complete opposite.I guess you don't take the time to read my posts.

You talked about unimplemented features and now you turn it all the way round and say that extra features that are used come at an extra transistor cost and probably heat cost.

I really can't follow your logic so I'll let this go.

[shadow001]

Thanks for the info , I also had some part of it in mind that the die does get bigger on the same process scale but I am also kind of speculating/predicting i-e Nvidia wont be a fool to introduce a mainstream card which wont be affordable by people.


Thing is,this isn't a card for the mainstream user,for that,you'll have to wait at least a few months more,and see the midrange and low end cards using based on this new architecture(the 150 to 350$ range)...


This if for those enthusiats who have to have the absolute latest,and as such,have to pay the price for it infortunately....No free rides for cards like this.


Nvidia would not go to that kind of transistor budget while still at 90nm,cutting the number of dies in half,even if the yields are perfect(wich they're never 100%),unless ATI has a similar monster waiting to be reveiled once it's ready...

[Redeemed]

You are way to new on this site to call me an "Ati boy".I've used every nvidia card since the TNT age and many ATI cards too but I'm no companys fanboy.

That aside, where do you get that I said the R5xx series is cooler than G7x cause I wrote the complete opposite.I guess you don't take the time to read my posts.

You talked about unimplemented features and now you turn it all the way round and say that extra features that are used come at an extra transistor cost and probably heat cost.

I really can't follow your logic so I'll let this go.

Ok,

Fotis you replied to Xion's statement that "recent trend" has had nVidia's cards running cooler than ATi's with a comment implying that G70 was the only line up of cards to run cooler than ATi's. Here is a quote where you said that:

You said "Nvidia's cards tend to run cooler than ATi's" and thats far from saying you were talking about the latest cards.Maybe you meant to say it.
Only G70 was cooler for the few last generations and thats because it has 100million trans less than Ati R5xx line.I don't think recent memory only covers the G7X gen and up.

Can you please name a few features that are unused in the R420-R5xx lines?

^^^Right there you were implying that G70 was the only GPU from nVidia to run cooler than ATI's cards. When infact the NV40 generation also ran cooler than ATi's R400 cards. And, upon release, R520 did run hotter than G70- and it performed far worse as well. That is why ATi was pretty quick to get R580 (the x1900 series) out to replace the R520.

Now, maybe you meant to include that the GeForce 6 and 7 series were cooler than ATi's past two generation of cards, but you didn't include such a comment thus your statement could have only been taken one way.

So, to reword Xion's comment (atleast I believe this is what he's getting at) the recent trend has been that nVidia's GPUs tend to run cooler than ATi's- as "recent trend" would be referring to the past two generations since everybody knows that ATi pwned the NV3x generation. ;)

As to features that are unused on ATi's card... that is insignificant. I have always wondered what the big deal was with ATi's "fully programmable memory controller"... what is the point of having a fully programmable memory controller? And no, I'm not implying that it is a useless and unused feature... I'm just questioning its necessity. ;) Ultimately, that argument of useless features on ATI's cards... if it is a true argument then I bet only a few "features" go unused- something you could probably just as easily nail against nVidia's GPUs.

[Redeemed]

Thing is,this isn't a card for the mainstream user,for that,you'll have to wait at least a few months more,and see the midrange and low end cards using based on this new architecture(the 150 to 350$ range)...


This if for those enthusiats who have to have the absolute latest,and as such,have to pay the price for it infortunately....No free rides for cards like this.


Nvidia would not go to that kind of transistor budget while still at 90nm,cutting the number of dies in half,even if the yields are perfect(wich they're never 100%),unless ATI has a similar monster waiting to be reveiled once it's ready...

Dude, so far all indications are that the GTS will cost around $449 USD, and the GTX will be costing around $650 USD. Even DT stated this when they released the specs of the G80.

I doubt the enthusiast segment has enough "enthusiast" in it to make a card that expensive sell. nVidia would only accomplish shrinking the enthusiast segment if they released the card at $800. They'd be shootin themselves in the foot.

[Daydre@m]

Dude, so far all indications are that the GTS will cost around $449 USD, and the GTX will be costing around $650 USD. Even DT stated this when they released the specs of the G80.

I doubt the enthusiast segment has enough "enthusiast" in it to make a card that expensive sell. nVidia would only accomplish shrinking the enthusiast segment if they released the card at $800. They'd be shootin themselves in the foot.
LOL you said exactly what I wanted to say. Even if they introduce the high end card at 800$ , the mainstream ( lets say half the performance ) would have to go around 400 or 300 in theory ... which I doubt. Then again its all speculation , let the product hit and let nvidia announce the final specs officially and you will know whats really going on .

[NoWayDude]

My understanding is that Nvidia does not sell cards to the public. They sell the chips to IHV's and they sell the cards
Also I was on the understanding that Nvidia puts a fixed price for the chips, but has no control on what the vendors charge afterwards . Am I wrong on this assumptions?

[shadow001]

Dude, so far all indications are that the GTS will cost around $449 USD, and the GTX will be costing around $650 USD. Even DT stated this when they released the specs of the G80.

I doubt the enthusiast segment has enough "enthusiast" in it to make a card that expensive sell. nVidia would only accomplish shrinking the enthusiast segment if they released the card at $800. They'd be shootin themselves in the foot.


Is that why just a few months ago,the fastest card was a 7800 GTX 512 meg,wich i remind you,was listed at 650$ a pop and routinely sold for 800$ each,and stayed that way for a couple of months at least,until ATI released the X1900 series...:p :D

Granted,those were cherry picked G70 cores that could hit 550 mhz stable and used the fastest memory available,but the specs of the current G80 suggest a huge die at 700 million transistors at 90 nm,wich means it'll be at least 2x times more expensive to make a G80 GPU,compared to a G71,since you only get less than half the GPU's per wafer,and that's assuming perfect yeilds...


Now add the 384 bit bus with the extra traces and layers needed for the PCB,and it will be more expensive to make than any previous card they've ever released....


Add the fact that the R600 is still 3 months away,and i still have no doubt it'll be a lot more expensive than 600$,at least until the R600 is released...


History will likely repeat itself,just like it did with the 7800 GTX 512 meg cards,you want the current fastest card around when there's no competition,be prepared to pay for it in a big way....

[Redeemed]

My understanding is that Nvidia does not sell cards to the public. They sell the chips to IHV's and they sell the cards
Also I was on the understanding that Nvidia puts a fixed price for the chips, but has no control on what the vendors charge afterwards . Am I wrong on this assumptions?

No, you are 100% correct. ;)

nVidia does not manufacture video cards, they only manufacture the GPUs. You are also correct that the price is up to the IHVs, not nVidia. nVidia can recommend a price, but the board manufacturers do not have to adhere to that price.

nVidia could say they feel the 8800GTX whould be sold for around $400USD, but if the IHV's decided it was worth $800 instead, they could freely charge that much for it. ;)

[Redeemed]

Is that why just a few months ago,the fastest card was a 7800 GTX 512 meg,wich i remind you,was listed at 650$ a pop and routinely sold for 800$ each,and stayed that way for a couple of months at least,until ATI released the X1900 series...:p :D

Granted,those were cherry picked G70 cores that could hit 550 mhz stable and used the fastest memory available,but the specs of the current G80 suggest a huge die at 700 million transistors at 90 nm,wich means it'll be at least 2x times more expensive to make a G80 GPU,compared to a G71,since you only get less than half the GPU's per wafer,and that's assuming perfect yeilds...


Now add the 384 bit bus with the extra traces and layers needed for the PCB,and it will be more expensive to make than any previous card they've ever released....


Add the fact that the R600 is still 3 months away,and i still have no doubt it'll be a lot more expensive than 600$,at least until the R600 is released...


History will likely repeat itself,just like it did with the 7800 GTX 512 meg cards,you want the current fastest card around when there's no competition,be prepared to pay for it in a big way....

This theory flawed.

And here is why:

How many gamers (i.e., "enthusiasts") have a 7800GTX 512? How many had one?

The answer is not many. Partly due to the price you mentioned, partly due to the fact it wasn't needed.

When G70 was released, it was no where near the $800 mark, and the 7800GT was no where near the $600 mark. If memory serves me correctly, the 7800GTX originally went for around $550 bucks, and the GT went for around $400 to $450.

As such, quite a few "enthusiasts" purchased the GTX, and quite a few more purchased the GT. When the 7800 GTX 512 was released, not many people ditched their current 7800GTXs for the 512- partly due to price and partly due to it not being logical with G71 just around the corner.

G71 is released and it is everything the 7800GTX 512 was and then some, for less might I add. ;)

As such, the enthusiast segment as it is right now is for people whom can afford and are willing to shell out up to $650 per card. That'd be $1300 for SLi.

If nVidia demanded these cards be sold for $800 USD, that would be $1600 USD for SLi.

Do you have any idea how dramatic the shrinkage would be to the enthusiast market? Do you have any idea how dramatic the shrinkage would be for nVidia's share in the market???

If the GTX goes for $800, then the GT will probably have to go for around $650.

Continuing with this logic, the 8600GT will have to cost around $400, and the very low end will have to cost around $150 to $250. Do you really feel that the 8200 will be worth up to $250? How many people would even pay $150 for the 8200?

Also, keep in mind companies like Dell, HP, Compaq, Alienware, Falcon Northwest, Gateway- they have to include the video cards in their computers. And if these cards cost as much as you are implying they will, then the computers they are in will have a hefty price jump as well.

For nVidia to charge so much for the G80... that would have a very dramatic effect on the entire computer market. And it'd probably give ATi the upper hand as they'd charge significantly less to be more competitive.

You cannot look entirely upon the transister count and die size to try and determine the cost. G70 was (if I'm remembering correctly) more costly to manufacture than NV40, yet it was right around the same price that NV40 was.

You cannot negate the ecomonic consequences of such a price jump, and neither can nVidia. nVidia wants video cards with the G80 to sell, not to barely sell. And at those prices, people would be better off getting two 7900GTOs and waiting till the R600 is released so that competition would drive prices down.

Heck, if the G80 costs even $700, I'm skipping it and going with two GTOs. I get roughly the same performance of one 8800GTS atleast while saving nearly $100. And I'm sure I wont be the only person doing such. Consequently, that'd be very bad for G80 sales.

No, I am quite confident that the GTX will go for about $650 and the GTS will go for around $450. This is what has been stated again and again even from websites with review samples. I think you are worrying for nothing, my friend. ;)

[shadow001]

This theory flawed.

And here is why:

How many gamers (i.e., "enthusiasts") have a 7800GTX 512? How many had one?

The answer is not many. Partly due to the price you mentioned, partly due to the fact it wasn't needed.

And the G80 is needed if you have a 7950gx2 like i have right,or any 7800 GT SLI,or 7900 GT SLi....Basically any pretty much any SLI setup purchased in the last 18 months or so...:)


When G70 was released, it was no where near the $800 mark, and the 7800GT was no where near the $600 mark. If memory serves me correctly, the 7800GTX originally went for around $550 bucks, and the GT went for around $400 to $450.

As such, quite a few "enthusiasts" purchased the GTX, and quite a few more purchased the GT. When the 7800 GTX 512 was released, not many people ditched their current 7800GTXs for the 512- partly due to price and partly due to it not being logical with G71 just around the corner.


True,but none of those cards were as expensive to build as this one will be,they used more convential 256 bit memory buses were pretty conservative transistor budgets for the fabrication process used,at least compared to the G80...


G71 is released and it is everything the 7800GTX 512 was and then some, for less might I add. ;)


True again,but you still had to wait about 4 months after the release of the 7800GTX 512 meg to see them,and if i'm not mistaken,jAKUp had a pair of those at one time....:p



As such, the enthusiast segment as it is right now is for people whom can afford and are willing to shell out up to $650 per card. That'd be $1300 for SLi.

If nVidia demanded these cards be sold for $800 USD, that would be $1600 USD for SLi.

That's what we're going to find out in 2 weeks at the public unveiling of the new cards,and i still hope i'm wrong here,but i don't think i will be...:D



Do you have any idea how dramatic the shrinkage would be to the enthusiast market? Do you have any idea how dramatic the shrinkage would be for nVidia's share in the market???

If the GTX goes for $800, then the GT will probably have to go for around $650.

Historically,the very high end of the market for those willing to shell out 650$ to 1300$ for video cards alone has always been tiny(maybe 2% of the overall market),and those cards have never been the biggest money maker for any company,since for every high end enthusiast market card they sell,they'll also sell 10~15 midrange cards for the casual gamer,and since there's likely won't be 200$ DX10 cards this xmas,you'll still see Nvidia and ATI dumping DX9 based cards into retail shelves by the tens of thousands,since it's those that keep both companies afloat...The midrange and low end of the market.


The very highest end if for visibility and PR value more than huge sales,and to develop the underlying technology that later on,will make it into the lower price brackets next year....It's only then that the really huge money comes in from DX10 card sales,not before.


So the price they'll ask for this first high end generation,are for the hardcore gaming freaks out there,that have plenty of disposable income to shell out the kind of money that these days,the average user will spend on a new system,not just for video cards



Continuing with this logic, the 8600GT will have to cost around $400, and the very low end will have to cost around $150 to $250. Do you really feel that the 8200 will be worth up to $250? How many people would even pay $150 for the 8200?

You're assuming the GPU transistor count will be anywhere near the same as with the 8800's and that they will use a 384 or even a 320 bit memory bus....0dds are pretty good that when those do show up,they'll be much cheaper to build by far,using more conventional 256 bit memory busses and transistor budgets not much above the current cards,but they will be DX10 cards none the less...



Also, keep in mind companies like Dell, HP, Compaq, Alienware, Falcon Northwest, Gateway- they have to include the video cards in their computers. And if these cards cost as much as you are implying they will, then the computers they are in will have a hefty price jump as well.


Considering that Dell sold out the renegade series at 10 000$ a pop in record time,wich at the time had the first Quad SLI setups ever released, and that companies like alienware,falcon northwest and voodoo PC,sell so called boutique PC's that are custom made to the clients specfications(including custom painjobs) and often exceed 5000$ or more,i don't really think they or their prospective clients,really care if they'll pay an extra 300~400$ for a pair of G80's in SLI....:p



For nVidia to charge so much for the G80... that would have a very dramatic effect on the entire computer market. And it'd probably give ATi the upper hand as they'd charge significantly less to be more competitive.

Not on the entire computer market,not even close...Only the the hardware freaks out there willing to shell out whatever it takes to get the best,made worse when it's direct competitor(R600),won't be out for another 3 months at least....Means having the high end market to themselves for at least that long...


You cannot look entirely upon the transister count and die size to try and determine the cost. G70 was (if I'm remembering correctly) more costly to manufacture than NV40, yet it was right around the same price that NV40 was.

Actually,about the same,since the first NV40 chips were built at 110 nm for 226 million transistors,the G70 pushed that upwards to 301 million at 110 nm,wich did make it more expensive,but then re-released what is essentially the same GPU at 90nm(G71),still keeping the same transistor budget,and about the same size as the original NV40 chip at 226 million...


G80 is a 700 million transistor behemoth at 90nm,so way bigger than either the original NV40,G70 or G71 by far....Not even close.



You cannot negate the ecomonic consequences of such a price jump, and neither can nVidia. nVidia wants video cards with the G80 to sell, not to barely sell. And at those prices, people would be better off getting two 7900GTOs and waiting till the R600 is released so that competition would drive prices down.

Heck, if the G80 costs even $700, I'm skipping it and going with two GTOs. I get roughly the same performance of one 8800GTS atleast while saving nearly $100. And I'm sure I wont be the only person doing such. Consequently,that'd be very bad for G80 sales.


I will without question,since it's not like we have 20~30 DX10 titles on store shelves right now to even justify a DX10 card,where the transistor count spent make making it DX10 certified,will be just using up a lot of power and not really doing anything in the way of graphics work,for a long while still...


Not to mention that i don't feel like being a Vista betatester since DX10 is restricted to Vista,and go thru the next 6 months after it's release,with still raw drivers that will take a while to sort out in terms of performance and stability,so that they'll be at least on par with the current state of Winxp drivers....


No, I am quite confident that the GTX will go for about $650 and the GTS will go for around $450. This is what has been stated again and again even from websites with review samples. I think you are worrying for nothing, my friend. ;)

I hope so,i really do,but like i said before,i have serious doubts on that...

[Xion X2]

You are way to new on this site to call me an "Ati boy".I've used every nvidia card since the TNT age and many ATI cards too but I'm no companys fanboy.

*snort* Yeah, whatever. The problem is, when you debate me on the point of ATi's last two generations of cards running hotter than nVidia's last two generations of cards, you begin to look like what you're trying to deny. There really is no debate on this subject, and you shouldn't have had a problem with my comment in the first place.

Redeemed--you hit the nail on the head, bro.

[Redeemed]

And the G80 is needed if you have a 7950gx2 like i have right,or any 7800 GT SLI,or 7900 GT SLi....Basically any pretty much any SLI setup purchased in the last 18 months or so...:)




True,but none of those cards were as expensive to build as this one will be,they used more convential 256 bit memory buses were pretty conservative transistor budgets for the fabrication process used,at least compared to the G80...




True again,but you still had to wait about 4 months after the release of the 7800GTX 512 meg to see them,and if i'm not mistaken,jAKUp had a pair of those at one time....:p




That's what we're going to find out in 2 weeks at the public unveiling of the new cards,and i still hope i'm wrong here,but i don't think i will be...:D




Historically,the very high end of the market for those willing to shell out 650$ to 1300$ for video cards alone has always been tiny(maybe 2% of the overall market),and those cards have never been the biggest money maker for any company,since for every high end enthusiast market card they sell,they'll also sell 10~15 midrange cards for the casual gamer,and since there's likely won't be 200$ DX10 cards this xmas,you'll still see Nvidia and ATI dumping DX9 based cards into retail shelves by the tens of thousands,since it's those that keep both companies afloat...The midrange and low end of the market.


The very highest end if for visibility and PR value more than huge sales,and to develop the underlying technology that later on,will make it into the lower price brackets next year....It's only then that the really huge money comes in from DX10 card sales,not before.


So the price they'll ask for this first high end generation,are for the hardcore gaming freaks out there,that have plenty of disposable income to shell out the kind of money that these days,the average user will spend on a new system,not just for video cards




You're assuming the GPU transistor count will be anywhere near the same as with the 8800's and that they will use a 384 or even a 320 bit memory bus....0dds are pretty good that when those do show up,they'll be much cheaper to build by far,using more conventional 256 bit memory busses and transistor budgets not much above the current cards,but they will be DX10 cards none the less...





Considering that Dell sold out the renegade series at 10 000$ a pop in record time,wich at the time had the first Quad SLI setups ever released, and that companies like alienware,falcon northwest and voodoo PC,sell so called boutique PC's that are custom made to the clients specfications(including custom painjobs) and often exceed 5000$ or more,i don't really think they or their prospective clients,really care if they'll pay an extra 300~400$ for a pair of G80's in SLI....:p




Not on the entire computer market,not even close...Only the the hardware freaks out there willing to shell out whatever it takes to get the best,made worse when it's direct competitor(R600),won't be out for another 3 months at least....Means having the high end market to themselves for at least that long...



Actually,about the same,since the first NV40 chips were built at 110 nm for 226 million transistors,the G70 pushed that upwards to 301 million at 110 nm,wich did make it more expensive,but then re-released what is essentially the same GPU at 90nm(G71),still keeping the same transistor budget,and about the same size as the original NV40 chip at 226 million...


G80 is a 700 million transistor behemoth at 90nm,so way bigger than either the original NV40,G70 or G71 by far....Not even close.





I will without question,since it's not like we have 20~30 DX10 titles on store shelves right now to even justify a DX10 card,where the transistor count spent make making it DX10 certified,will be just using up a lot of power and not really doing anything in the way of graphics work,for a long while still...


Not to mention that i don't feel like being a Vista betatester since DX10 is restricted to Vista,and go thru the next 6 months after it's release,with still raw drivers that will take a while to sort out in terms of performance and stability,so that they'll be at least on par with the current state of Winxp drivers....



I hope so,i really do,but like i said before,i have serious doubts on that...


Dude, you brought up some really great points that I actually hadn't considered...

Regardless, I believe you are not quite grasping what I'm saying.

First off, G80 is not the GPU, it is the architecture that the GPU is built upon.

As such, any midrange and low end card will sport most features found in the G80 architecture.

The memory bus will probably be atleast that of the 8800GTS. The pipes (including the geometry pipes) will probably be scaled down, and I'm sure the low end wont sport any sort of Quantum Physics Processor that'll be found in the high end- heck, the midrange might not even see that.

I guess, I could see the low end having a 256 bit memory bus, that'd be a quantum leap above today's low end. And it'd be about time as well. ;)

http://www.nvidia.com/object/7_series_techspecs.html

^^^ At that page is a break down of nVidia's GeForce 7 series lineup, from low end to top dog.

Look closely. You'll notice nVidia already has a GPU with a 512bit memory bus. Now, for the GX2 I don't believe it actualy uses a 512 bit memory bus- I'm sure they are referring to 256 bits per card on the GX2. But the 7950GT and the 7950GTX both have a 512bit memory bus.

Oh, and the 7950GT is going for under $400.

Why am I pointing this out to you? Simple. You seem bent on the 384 bit memory bus causing the price of G80 to sky rocket.

Hmm... the 7950GT has a 512bit memory bus, while the 8800GTX only has a 384 bit memory bus. The 7950GT only costs, what, about $350. And the 8800GTX, in your opinion, will cost somewhere around $800 per card?

I hope you see where I'm going. I'm not sure why nVidia isn't giving the 8800GTX a 512bit bus, obviously they could have. My theory is this:

At 90 nanometre and with the current transistor count of G71- they were allowed to give the 7950GT a 512bit memory bus without it breaking the bank.

G80 on the other hand is a whole other animal from G70 or even G71. Sure, it shares the same manufacturing process and it shares the interface- but that is all it shares with G70 and G71. Everything about G80 is a completely new deal, designed from the ground up and having nothing relating it to any previous generation.

You are getting on about the transistor count of G80. Well, you must remember that both the GTS and GTX will have a crap load of unified pipes- the GTX having 128 and the GTS having 96. Not to mention the geometry pipes and the Quantum Physics Processor. Then there is GPU clock which, supposedly, is designed to scale up to 1.5GHz.

Now, from what information about G80 is available, my opinion is that those transistors are mostly due to the extremely high level of pipes (geometry and unified combined) and the added Quantum Physics Processor. I am quite confident that high transistor count has very little to do with the memory bus.

Otherwise, if the memory bus was such a costy feat, the 7950GT would probably be going for around $600 a pop.

Here is what I'm thinking:

nVidia can't charge $800. That would alienate too many potential high-end customers.

Look at it this way:

nVidia can charge $800 per card and only have a handfull of "customers" (including the ones that buy two for SLi). Or, nVidia could charge only $650 per card and have around two handfulls of customers (including the ones purchasing two cards for SLi). If nVidia would make far greater profit by charging around $650 bucks for the GTX than they would if they charged $800.

http://www.ati.com/products/RadeonX1950/specs.html

^^^There we see ATI's greatest has 384 million transistors. This is on 90 nanometer, with only a 256 bit external memory bus. It is selling for only $400.

http://www.newegg.com/Product/ProductList.asp?Category=38&N=2010380048+1305520549+106790809+1067923257&Submit=ENE&SubCategory=48

My whole point is you seem to believe that due to the 384 bit memory bus, combined with the 700 million transistors, the 8800GTX will sell for $800.

I'm hoping I've successfully shown that the memory bus isn't going to cause the card to cost $800.

Now, the only obstacle left would be the 700 million transistors. Hmm...

Okay, from an economical standpoint, this is more than likely what would happen if the 8800GTX went for $800:

Right now, the super high end is actually mixed in regards to individual income. I'm quite confident Jakup doesn't own a mansion, and doesn't have and underground garage full of exotic sports cars. Yet he has one of the fastest gaming computers possible.

Jakup prioritizes the hardware in his computer enough to where he'll do what he has to inorder for it to be one of the fastest.

Okay, looking at the cost of his computer, the graphics cards combined might have cost him about $1300 (I'm forgeting what the GX2 went for when it was first released). I don't think he sold both his GTXs as I believe NaluGrl is using one of them in her rig right now. So he forked out the $1300 (or less) bucks for Quad-SLi goodness. He could have sold just one GTX, thus bringing the price down to maybe the price of a single GX2. Regardless, in the worst case scenerio he probably only spent $1300.

Now, Jakup is in the "high end" market segment right now. There are many other enthusiasts like him that don't have tons of cash but have decent top-dollar rigs. That is because due to the price of the high end cards, they can make it happen.

There are enthusiasts that can afford to spend up to $1k on their graphics cards. As such, they'd be the ones that purchase two 7800GTs or 7900GTs. That is still considered high end. If the GTX were priced at $800, the GTS (which is replacing the "GT" moniker) would have to go for about $600 to $650. That is more expensive than today's fastest GPU. As such, those customers that in previous generations bought the "GT" branded GPUs, wont be able to. That is about $500 more for an SLi setup. And over $200 more for just a single card. Thus, those customers get bumped down into the midrange catagory- the 8600 range.

Now, those customers, such as Jakup- the ones whom bought two 7900GTXs and then later two 7950GX2s when the cards were released- those "enthusiast" customers, will now be brought down to the "GTS" level- still spending a total of about $1300 for a SLi setup.

Now, the ones whom dump down the case for the 8800GTX, will be those with quite the large income- and those people probably comprise about half of the "enthusiast" segment. Thus, the sales for that segment will take a negative hit.

As I demonstrated earlier, it'd be more profitable for nVidia to charge around $650 for the GTX, and around $450 for the GTS. So far, every website with people "in the know" are claiming these prices. The only thing we don't have is nVidia themselves stating those will be the prices.

[Xion X2]

The 7950GT has a 256-bit memory bus/interface:

http://www.newegg.com/Product/ProductList.asp?DEPA=0&type=&Description=7950GT&Submit=ENE&N=0&Ntk=all

I think the link you're referring to is talking about the GX2 only, and it's kind of misleading because it looks as if they're doing the same thing that they do with the video memory claims. They're adding the two PCB's together (256 busGPU1 + 256 busGPU2 = 512).

G80 will be the first video card to have a bus higher than 256. It's one of the reasons everybody's so psyched about it. It will significantly increase memory bandwidth.

[CaptNKILL]

Wow... I post long winded messages but you guys are putting me to shame. :p

[Redeemed]

The 7950GT has a 256-bit memory bus/interface:

http://www.newegg.com/Product/ProductList.asp?DEPA=0&type=&Description=7950GT&Submit=ENE&N=0&Ntk=all

I think the link you're referring to is talking about the GX2 only, and it's kind of misleading because it looks as if they're doing the same thing that they do with the video memory claims. They're adding the two PCB's together (256 busGPU1 + 256 busGPU2 = 512).

G80 will be the first video card to have a bus higher than 256. It's one of the reasons everybody's so psyched about it. It will significantly increase memory bandwidth.

Yeah, you're right. I looked up the specs for the 7950GT and the 7950GTX, both only have a 256 bit memory bus. And I already stated that I knew that GX2 didn't have a "true" 512bit memory bus. So, I was wrong- GET OVER IT!!! :p ;) :D :D ;)

[Redeemed]

Wow... I post long winded messages but you guys are putting me to shame. :p

Are you trying to imply something extremely derogeratory about me or the extremely intellectual and mind altering posts of mine?

Or are you just jeolous? :p ;)

[Xion X2]

Yeah, you're right. I looked up the specs for the 7950GT and the 7950GTX, both only have a 256 bit memory bus. And I already stated that I knew that GX2 didn't have a "true" 512bit memory bus. So, I was wrong- GET OVER IT!!! :p ;) :D :D ;)

S'all good. Just didn't want the wrong information to get out there. http://xtremesystems.org/forums/images/smilies/cool.gif

[Xion X2]

Are you trying to imply something extremely derogeratory(sic) about me or the extremely intellectual and mind altering posts of mine?

Or are you just jeolous(sic)? :p ;)

LOL. Never, ever refer to yourself as an "intellectual" and misspell 2 words in the same breath. :D

[Daydre@m]

Some more CPU numbers from the ol-trusty inquirer

Link : http://www.theinquirer.net/default.aspx?article=35258

Sorry if someone already posted

[shadow001]

Regardless, I believe you are not quite grasping what I'm saying.

First off, G80 is not the GPU, it is the architecture that the GPU is built upon.

As such, any midrange and low end card will sport most features found in the G80 architecture.


I know that too,since you don't need 700 million transistors to make a DX10 card at all....you just need that many to hit a certain performance level when doing DX10 graphics operations....

By comparison,Intel will undoubtedly have a DX10 certified graphics chip built into their northbridge,on the new motherboards they intend to release in about 9 months from now,but they're going to suck in performance as usual,DX10 or not....:p



The memory bus will probably be atleast that of the 8800GTS. The pipes (including the geometry pipes) will probably be scaled down, and I'm sure the low end wont sport any sort of Quantum Physics Processor that'll be found in the high end- heck, the midrange might not even see that.

I guess, I could see the low end having a 256 bit memory bus, that'd be a quantum leap above today's low end. And it'd be about time as well. ;)

I'd be pretty impressed if a possible 8600 series packs all the speed of a 7900 GTX right now,but also adds DX10 features and gets released for 250$.


Those however won't even use the same GPU as the 8800 series or their 320~384 bit bus at all...It'll likely be based on the G80 architecture,but will be a seperate die in all respects...



Why am I pointing this out to you? Simple. You seem bent on the 384 bit memory bus causing the price of G80 to sky rocket.

The bus and the transistor budget on the GPU are closely tied together,since the actual GPU packaging will have a substancial increase in overall pin count because of that 384 bit bus,and the GPU die itself needs to be bigger,since you have to have enough room to route those extra connections from those extra pins to the GPU die itself....


Currently,all GPU's that use a 256 bit memory bus,use a crossbar memory architecture,wich splits that 256 bit memory bus into 4 seperate 64 bit memory channels,and wich have 2 memory modules connected to it's own 64 bit channel....It's the main reason why you have 8 memory modules on your average card these days.


With the G80 however,now you have a 384 bit memory bus,so it'll be 6x 64bit memory channels with 2 memory modules per channel,and the reason why you have 12 memory modules instead of the usual 8, and the odd sized total memory amount(768 megs).


Those 2 extra 64 bit channels means more traces on the video card PCB,wich may require extra layers to route them from the memory modules to the GPU packaging,wich like i said before will increase the number of pins on the GPU packaging as a consequence,and indirectly,also affect the transistor budget of the die to an extent,since those 2 extra channels need to be connected from that GPU package to the die itself....

If i'm not mistaken,the current GPU's packages with the G71 and R580, already have over 1000 pins,split between power and ground pins,the memory bus pins,and the ones leading from the PCI-e connector to the GPU,the SLI connector on the top part or the card,and of course,the 2 DVI connectors as well,and all those pins need to be connected to the GPU die somehow...Yes,it's a tight fit:D


Now, from what information about G80 is available, my opinion is that those transistors are mostly due to the extremely high level of pipes (geometry and unified combined) and the added Quantum Physics Processor. I am quite confident that high transistor count has very little to do with the memory bus.

I've already answered why the transistor count and the bus are related to eachother...





^^^There we see ATI's greatest has 384 million transistors. This is on 90 nanometer, with only a 256 bit external memory bus. It is selling for only $400.

True,but then again,this is now,not when it was introduced and Nvidia has a card than can compete with it,so it forces both companies to make prices more reasonable...


The GPU die of the G80 will be roughly 2x bigger than the 384 million present in the R580 since both use the same fab process,and more importantly,Nvidia has the High end market to itself for the next 3 month after it's release,since it's true competitor(R600) won't be out till the end of january,so there's no pressure on Nvidia to keep the prices reasonable...


In short....No competition,the chances increase quite dramatically that you'll pay thru the nose for the absolute best out there,it's really that simple.





Now, the ones whom dump down the case for the 8800GTX, will be those with quite the large income- and those people probably comprise about half of the "enthusiast" segment. Thus, the sales for that segment will take a negative hit.

And like i said before,those willing to dump as much money for 2 video cards as most people do in their entire systems,are a really small minority,not even amounting to 2% of the market,and the extreme high end market like this is for visibility,and PR value as well as developing the basic tech that Nvidia will use for the next 2+ years in it's product introductions,until they release a brand new architecture yet again...


The refresh for the G80 will be automatically much cheaper for the simple fact that once that one is released,sometime next year,it'll likely be built at 65nm,either from TSMC or UMC,thus making the overall die size the same as current chips like the G71 or R580 GPU's are at 90nm...So no extra die cost since you can crank out the same volume of G80 chips at 700 million transistors for every wafer as you do with current GPU's at 380 million at 90nm

[CaptNKILL]

Some more CPU numbers from the ol-trusty inquirer

Link : http://www.theinquirer.net/default.aspx?article=35258

Sorry if someone already posted
10,000 on an FX-62??? Come on. :|

This thing better have some amazing performance in games... my slightly overclocked 7950 pulls 8,100 in 3dmark 06 on a slower CPU. I cant BELIEVE this "uber" card is only going to get 1,900 more points.

If this is true, I'm very disapointed... but I'll hold out for the results from real games.

I'm probably still going to sell this 7950 and "trade up" anyway, but this is sounding less and less amazing every day...

[Daydre@m]

10,000 on an FX-62??? Come on. :|

This thing better have some amazing performance in games... my slightly overclocked 7950 pulls 8,100 in 3dmark 06 on a slower CPU. I cant BELIEVE this "uber" card is only going to get 1,900 more points.

If this is true, I'm very disapointed... but I'll hold out for the results from real games.

I'm probably still going to sell this 7950 and "trade up" anyway, but this is sounding less and less amazing every day...

Nah I think its another one of those inquirer's crap to grasp attension...

Btw I found this too , lol this says that Nvidia to ban overclocking on G80 chips .... and its again from inquirer :captnkill: :o

nVidia told all of its partners that it simply won't allow the overclocking of G80 cards. It doesn't want another Geforce 7900 GT overclocking fiasco that caused a too high return rate for the cards.


Remember, Nvidia didn't allow anyone to overclock Geforce 7950 GX2 and they all worked at the same reference 500MHz clocks as this was what Nvidia guaranteed for these cards.

We still believe that there is a chance that some Geforce 8800 GTX and GTS might end up a little bit faster, as some of the vendors depends on that. Maybe you can overclock the water-cooled card if anyone makes one.

News source: The Inquirer

Link : http://www.flexbeta.net/main/comments.php?catid=8&shownews=21896


Oh Boy if its true then nvidia will be more like ATI.....

PS : Sorry if it has been posted before

[CaptNKILL]

Nah I think its another one of those inquirer's crap to grasp attension...

Btw I found this too , lol this says that Nvidia to ban overclocking on G80 chips .... and its again from inquirer :captnkill: :o



Link : http://www.flexbeta.net/main/comments.php?catid=8&shownews=21896


Oh Boy if its true then nvidia will be more like ATI.....

PS : Sorry if it has been posted before
Well, I hope you're right. I've actually found the inquirer to be reasonably accurate in the past, but just because I want this card to kick ass, I'll wait till we have more solid numbers.