• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

9800 GX2 Pictures and Specs

9800 GT and GTX should have even shorter "life period" than GX2. Since those cards will also be replaced.
 
Wow I think my brain just melted from nVidia's stupidity :rolleyes: So this new VideoCard with a next generation name of "9800GTX" will not even be faster than last gen 8800Ultra, that is just craziness, why even make it we already have the 8800Ultra.

So from November 2006 to March 2008 the best nVidia can come up with for a new GTX is just 10% faster, and actually slower than their own May 2007 card, unbelievable...this card does NOT deserve the 9800 name it should be 8900GTX

...

And you know for a fact that the 9800GTX only has 512mb of memory on a 256-bit bus? Didn't think so.
 
A good guess would be arund the same time ATI RV770, rumor is around late Q2 2008 for RV770 since ATI is pushing it out faster. But no one really knows but i hope around mid 2008 for both this cards.
And that makes "9800" Gx2 a new shortlived 7950Gx2.


Oh well, then I am only going to upgrade my Motherboard, ram and cpu and wait for the G100 when it comes out thanks:)
 
We can all hope it will be a big performer, the 9800GTX or the G100 iteration. I know I'm going to build a new computer mid year and I want a monster video card. One that crushes the 8800GTX.
 
We can all hope it will be a big performer, the 9800GTX or the G100 iteration. I know I'm going to build a new computer mid year and I want a monster video card. One that crushes the 8800GTX.
G100 can be anything..haven't heard any specs. 9800 GTX is hold back because it uses G92 and so doing has 256-bit mem bandwith
 
Now G100 rumors are in full force??

I'll look forward to buying one, right after I get my H2 fuel-cell, + plug-in, & Lion battery equipped Chevy Volt :cool:
 
Now G100 rumors are in full force??

I'll look forward to buying one, right after I get my H2 fuel-cell, + plug-in, & Lion battery equipped Chevy Volt :cool:
Well G100 will come late Q2/2008 and will replace high end models from GF9000-series.

Basically nothing else is known from G100. Rumours has it that it will come with 55nm architecture. AMD's R700 will be dual RV770. AMD has already stated that their opinion is that future of high end cards are these multi-GPU solutions
 
=Niceone;1031959383 AMD's R700 will be dual RV770. AMD has already stated that their opinion is that future of high end cards are these multi-GPU solutions

That is a smart idea, I mean how many single core Intel cpu's do you see coming out now ?? They are all Dual and Quad core, so VideoCards should adopt that same thinking, we can't have 2ghz clocked cards, but 2-1ghz can be done.

I hope the ATI R700 kills the sleeping nVidia giant, bunch of douchebags for coming out with a card named "9800GTX" but no faster than the old 8800Ultra :rolleyes: That makes a lot of sense to me, and then come out with another faster newer card just a few months later to surpass it off the shelf for same price, the G100 ?? uh ok that makes even more sense :rolleyes:

nVidia "do the right thing" and name you new card it's proper name this new GTX should ONLY be named 8900GTX, it does not deserve the 9800 series name if only 10-20% faster than 8800GTX
 
That is a smart idea, I mean how many single core Intel cpu's do you see coming out now ?? They are all Dual and Quad core, so VideoCards should adopt that same thinking, we can't have 2ghz clocked cards, but 2-1ghz can be done.

I hope the ATI R700 kills the sleeping nVidia giant, bunch of douchebags for coming out with a card named "9800GTX" but no faster than the old 8800Ultra :rolleyes: That makes a lot of sense to me, and then come out with another faster newer card just a few months later to surpass it off the shelf for same price, the G100 ?? uh ok that makes even more sense :rolleyes:

nVidia "do the right thing" and name you new card it's proper name this new GTX should ONLY be named 8900GTX, it does not deserve the 9800 series name if only 10-20% faster than 8800GTX
Well to be accurate graphics cards have been multicore since 3DFX VoodooII era..for like 10 years now. These new 9800 GX2 and HD3870 X2 solutions are like having two quad core Xeons in one motherboard. For example Geforce 8800 ultra has like 128 "cores". 8800 Ultra manages 576 gigaflops when Intel Core 2 Quad's can manage like 40 gigaflops (though GPU's are limited on what they can calculate, but because of this there are now days supercomputers that uses multiple graphics cards for scientific calculation)
 
That is a smart idea, I mean how many single core Intel cpu's do you see coming out now ?? They are all Dual and Quad core, so VideoCards should adopt that same thinking, we can't have 2ghz clocked cards, but 2-1ghz can be done.

You do know what a stream processor is, right?
 
yes yes Terran we all know that in fact GPU's are multi-core in terms of how many stream processors it has.

But you know as well as I that we want to see the concept of dual-core/quad-core CPU's being applied in the same fashion with GPU's. Stop being so damned technical. :p
 
Right, but I am making a point here- do we need gpu's to go the same route? Should they? We know that 2 cores does not equal 2x performance vs. one core. Should not then graphics cards focus on expanding the amount of stream processors a single core can contain (if you would rather expand objects as opposed to speeds)? By using a single core populated by stream processors, you only have the issue of "communication" between the stream processors themselves as opposed to "communication" between multiple sets of stream processors. Imo, the only way I could really see dual core gpu's being beneficial is if you could optimize the cores for different tasks- one for graphics, one for physics, etc... but that defeats the whole purpose of the unified architecture.
 
We know that 2 cores does not equal 2x performance vs. one core.
That sums up SLI and CF right there. But that isnt stopping NV and AMD is it?

Should not then graphics cards focus on expanding the amount of stream processors a single core can contain (if you would rather expand objects as opposed to speeds)? By using a single core populated by stream processors, you only have the issue of "communication" between the stream processors themselves as opposed to "communication" between multiple sets of stream processors. Imo, the only way I could really see dual core gpu's being beneficial is if you could optimize the cores for different tasks- one for graphics, one for physics, etc... but that defeats the whole purpose of the unified architecture.

Sooner or later, they arent going to be able to shrink transistors anymore. Moore's Law.
Not everyone has a case or PSU capable of powering these monstrosity's when they debut. I thought technology was supposed to shrink and be more energy-efficient? The X2's and GX2's contradict that.

As it stands, the GPU's in an SLI or CF configuration communicate with each other over an SLI/CF bridge. Nvidia makes that SLI connector permanent by pancaking 2 cards against each other (GX2). ATi has taken a step in the right direction with the 3870X2 by making each GPU communicate with each other on the same PCB.

Why not go one step further and make dual-GPU's communicate with each other across the same die?
 
G100 can be anything..haven't heard any specs. 9800 GTX is hold back because it uses G92 and so doing has 256-bit mem bandwith

Wrong. We only know that about the GX2, which will be basically two 8800 GTS 512.
 
That is a smart idea, I mean how many single core Intel cpu's do you see coming out now ?? They are all Dual and Quad core, so VideoCards should adopt that same thinking, we can't have 2ghz clocked cards, but 2-1ghz can be done.

I hope the ATI R700 kills the sleeping nVidia giant, bunch of douchebags for coming out with a card named "9800GTX" but no faster than the old 8800Ultra :rolleyes: That makes a lot of sense to me, and then come out with another faster newer card just a few months later to surpass it off the shelf for same price, the G100 ?? uh ok that makes even more sense :rolleyes:

nVidia "do the right thing" and name you new card it's proper name this new GTX should ONLY be named 8900GTX, it does not deserve the 9800 series name if only 10-20% faster than 8800GTX

lol wow. I'm speechless. You're ridiculous (and incredibly wrong in many ways), guy.
 
Sorry, but it's already known that Geforce 9800 GTX (D9E-20) uses G92's (as does 9800 GX2 (D9E-40)).

You don't have to be sorry, you just have to post the link that says that. Also, you keep insisting that a G92 based card, will have a 256 bit memory interface, which is not entirely accurate, given G80's architecture.

G92 uses a 256 bit memory interface, because it's cheaper to produce that way and the costs are lower to the end consumers too. The 9800 GTX will be a "true" high-end product and thus, most certainly will not use a 256 bit memory interface. It will likely use the "old" 384 bit memory path, found in cards like the 8800 GTX and Ultra.
 
So 9800 GTX would use G90 or some other core that would have more than four 64-bit memorycontrollers like G92 has at best?
 
Sooner or later, they arent going to be able to shrink transistors anymore. Moore's Law.
I agree... but Moore's Law does not. Moore's Law is more of a 'general trend' that says that transistor density doubles every two years or so. It makes no account for physical or theoretical limitations whatsoever, rather it seems to disregard them entirely.

Right, but I am making a point here- do we need gpu's to go the same route? Should they? We know that 2 cores does not equal 2x performance vs. one core. Should not then graphics cards focus on expanding the amount of stream processors a single core can contain?
As we've so recently seen, smaller cores means cheaper cards. R600 was massive, and it was a flop. G92 nearly rivals G80 at a much lower cost, so much lower that it is possible to outperform a 8800GTX with two 8800GTs for less cost.

It may be such that a card composed of one single powerful core would be more expensive and ultimately less powerful than a card composed of 4 smaller cores, even if it isn't a full 4x advantage. We ought to be less concerned with exactly how well multi-gpu scales and more concerned with price/performance ratio.

How do I ignore someone? I'm worried I'll bruise myself from all the facepalming.

Sorry, but it's already known that Geforce 9800 GTX (D9E-20) uses G92's (as does 9800 GX2 (D9E-40)).
So 9800 GTX would use G90 or some other core that would have more than four 64-bit memorycontrollers like G92 has at best?
It's already known so well that you inquire about Silus's suggestion rather than disputing it?

You don't have to be sorry, you just have to post the link that says that.
Amen! Discussing unreleased and undisclosed tech stuff is full of rampant rumor and speculation. Niceone, you may have heard that the 9800GTX is G92-based from someone you trust, but where did they hear it themselves?

I would like to agree with Silus's prediction that 9800GTX won't be based on the regular 256-bit G92. Just because we haven't heard of a G90 doesn't mean it doesn't exist. They were able to keep G80 a fairly close guarded secret until very near its release. Or maybe that's just wishful thinking. The only real fact is that there aren't any facts regarding the 9800GTX yet, though I would be absolutely astonished if the 9800GTX was not faster than its predecessor.
 
So 9800 GTX would use G90 or some other core that would have more than four 64-bit memorycontrollers like G92 has at best?

1 x 64 = 64
2 x 64 = 128
3 x 64 = 192
4 x 64 = 256
5 x 64 = 320
6 x 64 = 384
7 x 64 = 448
8 x 64 = 512

Nvidia has done 1, 2, 3, 4, 5, 6 so far, so you can't claim that G92 is limited to 4 just because we haven't seen any higher, as its more expensive to produce and they don't really need the extra bandwidth.
 
Problem is that G92 only has 4 of these 64-bit controllers at best. If they build new chip..that would change things, but that chip wouldn't be G92.
 
Problem is that G92 only has 4 of these 64-bit controllers at best. If they build new chip..that would change things, but that chip wouldn't be G92.

How do we know they aren't disabled / not used? Eg: 6 on G80 for 8800GTX/Ultra vs 5 for 8800GTS.
 
How do we know they aren't disabled / not used? Eg: 6 on G80 for 8800GTX/Ultra vs 5 for 8800GTS.
well the G80 says six controllers in its specs and the G92 says four. the G80 8800gts has one disabled and the G92 8800gs has one of its controllers disabled. there are not more than six in the G80 or four in the G92...end of story
 
well the G80 says six controllers in its specs and the G92 says four. the G80 8800gts has one disabled and the G92 8800gs has one of its controllers disabled. there are not more than six in the G80 or four in the G92...end of story

Therefore I wouldn't be surprised if there was at least another 16 stream processors that are disabled on this particular product. When we asked whether the chip had clusters of stream processors disabled, the response we got back from Nvidia's Tony Tamasi was "the GeForce 8800 GT is a 112 stream processor product". There are potentially more ROPs and support for up to 384-bit memory bus widths too, but this takes us into the realms of speculation.

http://www.bit-tech.net/hardware/2007/11/02/nvidia_geforce_8800_gt/2

Guess we will see :)
 
That was Tim's own speculation..

Which is the same we are doing here. Neither you or I, know what chip will power the new 9800 GTX. We do know that the GX2 will use two G92 chips, essentially the same as the chip used in the new 8800 GTS 512.

Niceone said:
Problem is that G92 only has 4 of these 64-bit controllers at best. If they build new chip..that would change things, but that chip wouldn't be G92.

At last! Seems you got it, though I think you are still confused about a given architecture and its derivatives. G92 is a derivative of G80, which architecturally allows more than what you see in the existing versions of G92. Existing versions of G92 use a 256 bit memory path, but that does not mean it's not expansible to the full blown 384 bit in the original G80. Architecture is the same and what the original one allows, its derivatives will to.

Will they use the same codename ? I doubt it, but why not ? The changes they did from G80 to G92, will remain, so the chip will essentially be the same and just have its memory path extended. As for Stream Processors, if the past speculation of those 32 Stream Processors disabled, since the time of the original 8800 GTX is true, that means a total of 160 Stream Processors. But I'm still betting on 192 SPs. NVIDIA had plenty of time to prepare G92 (or whatever its codename will be). Couple that with the old 384 bit memory interface and 1 GB of fast VRAM and we have ourselves a 60-70% faster card than the 8800 GTX. With 160 SPs, I guess performance would not go beyond 30-40%, when compared to the 8800 GTX, but that's still very good IMO.
 
Which is the same we are doing here. Neither you or I, know what chip will power the new 9800 GTX. We do know that the GX2 will use two G92 chips, essentially the same as the chip used in the new 8800 GTS 512.



At last! Seems you got it, though I think you are still confused about a given architecture and its derivatives. G92 is a derivative of G80, which architecturally allows more than what you see in the existing versions of G92. Existing versions of G92 use a 256 bit memory path, but that does not mean it's not expansible to the full blown 384 bit in the original G80. Architecture is the same and what the original one allows, its derivatives will to.

Will they use the same codename ? I doubt it, but why not ? The changes they did from G80 to G92, will remain, so the chip will essentially be the same and just have its memory path extended. As for Stream Processors, if the past speculation of those 32 Stream Processors disabled, since the time of the original 8800 GTX is true, that means a total of 160 Stream Processors. But I'm still betting on 192 SPs. NVIDIA had plenty of time to prepare G92 (or whatever its codename will be). Couple that with the old 384 bit memory interface and 1 GB of fast VRAM and we have ourselves a 60-70% faster card than the 8800 GTX. With 160 SPs, I guess performance would not go beyond 30-40%, when compared to the 8800 GTX, but that's still very good IMO.
IMO none of what you said is likely to happen with the G92. also there is no way a 384bit part would have 1 gig of ram. it would be 768mb or 1.5 gig.
 
Wow, so the GPUs are facing each other. Interesting. I'm not sure I'm sold on the cooling solution. I'm a big fan of the standard dual-slot exiting out the back of the case coolers.

Woah I didn't even realize that when I first looked at them, I was just thinking it was pictures of one side :p

Will be strange to see how that works out for temps and expelling heat...
 
I noticed on the 9800GX2 it has 1x 6pin and 1x 8pin connector and thats a problem for me because I have the silverstone ST85ZF with 4 x 6pin connectors.:(
 
I noticed on the 9800GX2 it has 1x 6pin and 1x 8pin connector and thats a problem for me because I have the silverstone ST85ZF with 4 x 6pin connectors.:(

8pin connector is for PCIe 2 requirements. You can still plug in 2 six pins.
 
Woah I didn't even realize that when I first looked at them, I was just thinking it was pictures of one side :p

Will be strange to see how that works out for temps and expelling heat...

I don't generally do this, but I'm going to put some faith in nVidia that the new cooling set-up will be better because otherwise I don't see why they would change what they were doing previously.
 
Back
Top