nvidia "D8E" High End solution, what can we expect in 2008?

well nV did still have around a 20% profit margin on the GX2, Arun and I had a disagreement about this before, and I thought they didn't have much margin, but suprisingly they did, don't know if they sold alot of them, but profits where there on a per card basis.

well ... i won't say they don't have profit on GX2 style.but what about single card profit ?!
 
It will certainly be interesting to revisit this thread once the next "high-end" hardware is out.
Thats always the case when it comes to speculation threads. :D

My estimate for development costs (for a dual chip solution) would be low, seeing as how AIBs can afford to do it themselves even though the target market is much much smaller. The only major headache for this kind of a solution are the drivers ..
 
I hope the new top end is not a multi chip or SLI only based. I'd gladly pay to help.. :D
Although marketing can and will always argue that having high visibility for your company even if it's only one product can help sales for all of your products.
 
I hope that NVIDIA will prepare something more interesting (even if slower on average) than another GX2 card. Such a solution will falter more often than not whenever a new title will be coming to the market (due to SLI issues). Besides, even with two 8800GT chips, the card will not be that much faster than GTX or Ultra. Just take a look at www.bit-tech.net/hardware/2007/11/02/nvidia_geforce_8800_gt/1, where two GT based cards are tested in SLI.

If such a card is ever released, it will be obsoleted by a single chip high-end solution within one year.
 
If such a card is ever released, it will be obsoleted by a single chip high-end solution within one year.

Boy, there's a bold prediction. You mean unlike a new single chip high-end solution that still won't be obsolete two or three years down the road?
 
Well, good point. GPUs are getting old very fast, although 8800GTX has not been made obsolete yet :). But let me explain. I perceive GX2-like cards as a stop gap solution. Note that a new architecture from NVIDIA should be coming sometime next year (if everything goes as planned). In my opinion, if we see a GX2 card in January, this new architecture will debut sooner rather than later. I would say in the first half of 2008. Thus, GX2 cards will be made obsolete very quickly.
 
Well, good point. GPUs are getting old very fast, although 8800GTX has not been made obsolete yet :). But let me explain. I perceive GX2-like cards as a stop gap solution. Note that a new architecture from NVIDIA should be coming sometime next year (if everything goes as planned). In my opinion, if we see a GX2 card in January, this new architecture will debut sooner rather than later. I would say in the first half of 2008. Thus, GX2 cards will be made obsolete very quickly.

Aren't refreshes by definition a "stop gap" product to hold the competition at bay until a new part is released? :p

Regards,
SB
 
Well, you`re saying NVIDIAcould push GX2 like card (with two G92) as their high-end solution. OK but let me see what they said a few months ago?? As you remember M. Hara said there will be a new Highend GPU with performance close to 1Tflop in Q4. The second thing is that some guy from NVIDIA said their next-gen GPU will support double precision floating point and it will be launched by the end of year. ;)
So what does it mean?? Did they suddenly change their plans?? They said about 1Tflop beast but there is "only" 0,5Tflop GF8800GT which doesn`t support DP Floating Point.

So i wonder what could happen that they have changed their previous plans or maybe NVIDIA prepares something more powerful than 1Tflop GPU or maybe they have already had 1Tflop GPU but they are still waiting on AMD/ATI move??

But i`m a little dissapointed that there is only one card launched (GF8800GT). I belived we will see all new family based on G9x SKU - something like NV40-->G70. But i hope this year we will se something even more powerful than GF8800GT or new GF8800GTS from NVIDIA (i`m not thinking about GX2 solution). :)
 
as long as the next high end at the time of release, however its put together, is faster than GTX in SLI and costs less than SLI GTX, I think that would be good and I'd probably get one. i guess what im saying is i want a similar situation to when the 8800GTX came out and was faster than sli 7900 GTX in most cases.
 
As you remember M. Hara said there will be a new Highend GPU with performance close to 1Tflop in Q4. The second thing is that some guy from NVIDIA said their next-gen GPU will support double precision floating point and it will be launched by the end of year.
What makes you think G92 doesn't fit your description? 8 Clusters @ 2GHz+ fits the 'close to 1TFlop' requirement just fine, as long as you weren't expecting 999GFlops... And that's not contradictory with a GX2 either, as long as it's clocked lower.
 
Even though I'm a firm believer that we'll see a G92 based GX2, personally I prefer a single chip sku over the dual-chip solution. But this topic was already beaten to death in other threads so lets just leave it ..
 
What makes you think G92 doesn't fit your description? 8 Clusters @ 2GHz+ fits the 'close to 1TFlop' requirement just fine, as long as you weren't expecting 999GFlops... And that's not contradictory with a GX2 either, as long as it's clocked lower.

That's very generous. The clock has to be far north of 2GHz to come anywhere close to 1TFlop. 2.4GHz+ is more like it.
 
That's very generous. The clock has to be far north of 2GHz to come anywhere close to 1TFlop. 2.4GHz+ is more like it.
800-900GFlops would perfectly fit within the definition of 'close', IMO. If you were expecting 975GFlops or something stupid like that, why wouldn't they just be overvolting it by 0.001V and go above 1T for marketing.
 
Margin of error I suppose. Personally, I won't consider the promise fulfilled at anything < 900GFlops. Might be a moot point anyway since the rumour mill is pretty quiet on that front. At least for Q4.
 
Margin of error I suppose. Personally, I won't consider the promise fulfilled at anything < 900GFlops. Might be a moot point anyway since the rumour mill is pretty quiet on that front. At least for Q4.
Yeah, I think that's largely down to inventory management though.
 
800-900GFlops would perfectly fit within the definition of 'close', IMO. If you were expecting 975GFlops or something stupid like that, why wouldn't they just be overvolting it by 0.001V and go above 1T for marketing.

So by that definition G80 actually IS half a teraflop even if you don't count the MUL. Since it's close enough. ;)

IMO - Close is within 5% of a stated value, not 10-20% of a stated value. Although I'm quite sure the Nvidia marketing department absolutely love you right now. :)

Regards,
SB
 
Back
Top