GeForce Titan coming end of February

[SirPauly]

Not sure if it's legit, but a pretty monstrous score if it is.

Indeed! Interesting though!

 

[Ferzerp]

Only way it's real is if all 15 smx's are enabled and boosting to or above 900mhz.

For that to be true, I think we can only suppose that yields are far better than projected for it, or the demand in the HPC space is less than they projected, and they need to unload some chips.

Otherwise, why sell these as consumer cards when they can be marked up as Teslas?

Or it was planned all along and their manufacturing contracts were sized for it. Who knows.

 

[tviceman]

For that to be true, I think we can only suppose that yields are far better than projected for it, or the demand in the HPC space is less than they projected, and they need to unload some chips.

Otherwise, why sell these as consumer cards when they can be marked up as Teslas?

Or it was planned all along and their manufacturing contracts were sized for it. Who knows.

They sold GT200, GT200B, GF100, and GF110 cards in consumer spaces when they were also marked up as quadro and tesla cards, how is this any different?

There is a significant amount of transistors dedicated to GPGPU on GK110 that would be fused off for the Geforce part anyway, so yielding to get Geforce chips out of GK110 has a higher chance of success assuming defects are randomly spaced out.

 

[Ferzerp]

They sold GT200, GT200B, GF100, and GF110 cards in consumer spaces when they were also marked up as quadro and tesla cards, how is this any different?

There is a significant amount of transistors dedicated to GPGPU on GK110 that would be fused off for the Geforce part anyway, so yielding to get Geforce chips out of GK110 has a higher chance of success assuming defects are randomly spaced out.

That falls in to the "planned all along" option.

It just seems a strange way and time to release a new product for nvidia is all, in my view.

 

[f1sherman]

Only thing that's strange, is NV adding another product class.
Ultra-high-end or whatever market experts agree to call it.


GT200 @ 65nm = 576 mm2

 

[96Firebird]

Maybe they finally got some extra chips after they shipped all the GPUs for the Titan Supercomputer?

Heh, just realized they share the Titan name.

 

[Ferzerp]

576 mm^2 @ 65 nm = 107 mm^2 @ 28 nm ... That's less effective area than the 640. I don't get your point.

(hint: die area comparisons on different nodes are worthless)

 

[f1sherman]

No particular point. Other than another big die has been mentioned, so I noted it's size.

I don;t get the point of your ideal shrink

 

[RussianSensation]

NV should just go all out = 15 SMX clusters, 1Ghz clocks and 300W power consumption and cool it with an AIO watercooling system. A single rad on the Ares II is able to keep the card at just 53*C despite it using 431W of power.

If this score is just for a 732mhz 2688 GK110 part, then I don't see how that's possible unless there are some 3dMark11 quirks/inconsistencies (such as when comparing the scores of GTX580 to GTX680 on Extreme setting which make no sense). That score is impressive for sure if legitimate. Can't wait for gaming benches.

 

[f1sherman]

At least Titan is still on schedule


The next generation GeForce and Radeon postponed to Q4 2013

 

[blackened23]

At least Titan is still on schedule


The next generation GeForce and Radeon postponed to Q4 2013

It's kinda strange that Titan is being released and the GK114 (kepler refresh) is reported to be 15-20% faster than the 680? Product timing seems very, very strange. Unless the GK114 is much cheaper, which i'm sure it would be as a 680 replacement.

Anyway I hope that rumor isn't true, at least coincide the 8970 / GK114 release with haswell; a lot of people like doing spring PC refreshes once a year.

 

[Silverforce11]

I seriously hope they go all out and hit the 300W barrier. A single gpu thats awesome at gaming and HPC, huge vram, no sli/cf hassles.. gtx690 perf, i'd buy it.

 

[MisterMac]

So lets get this straight.

Titan is coming as a Super High end Single GPU something inbetween 690\680.

and a GK114\Kepler refresh is coming Q3?



That's some weird product line ups nvidia has going

 

[VulgarDisplay]

It's kinda strange that Titan is being released and the GK114 (kepler refresh) is reported to be 15-20% faster than the 680? Product timing seems very, very strange. Unless the GK114 is much cheaper, which i'm sure it would be as a 680 replacement.

Anyway I hope that rumor isn't true, at least coincide the 8970 / GK114 release with haswell; a lot of people like doing spring PC refreshes once a year.

I don't mind because I want a CPU upgrade next anyways. At least I hope that Haswell offers enough benefit over an i5 2500k to warrant the upgrade.

 

[tviceman]

I seriously hope they go all out and hit the 300W barrier. A single gpu thats awesome at gaming and HPC, huge vram, no sli/cf hassles.. gtx690 perf, i'd buy it.

You'd spend $900 on a video card?

 

[Ferzerp]

You'd spend $900 on a video card?

I certainly would for the right level of performance.

 

[VulgarDisplay]

I certainly would for the right level of performance.

I would consider it if it was as fast or faster than current flagship dual GPU setup. In this instance however I highly doubt Titan will be that fast.

 

[2is]

If it performed twice as fast as my 680, I certainly would.

 

[VulgarDisplay]

The ridiculous thing is that I know some of you crazies will have quad sli on these things.

 

[Ferzerp]

The ridiculous thing is that I know some of you crazies will have quad sli on these things.

Not me. Quad SLI doesn't appear to scale all that well from what I've seen.

 

[Ajay]

It's kinda strange that Titan is being released and the GK114 (kepler refresh) is reported to be 15-20% faster than the 680? Product timing seems very, very strange. Unless the GK114 is much cheaper, which i'm sure it would be as a 680 replacement.

Anyway I hope that rumor isn't true, at least coincide the 8970 / GK114 release with haswell; a lot of people like doing spring PC refreshes once a year.

This doesn't make any sense, unless it because Apple bought out most of TSMC's 28nm wafers for 1h13!

 

[2is]

This doesn't make any sense, unless it because Apple bought out most of TSMC's 28nm wafers for 1h13!

Well that'll be a temporary supply constraint unless Apple starts innovating again. Several years of incremental updates to their product lines are starting to show with lower than expected iPhone sales and they're also starting to lose market share to Android in the tablet space.

 

[Ferzerp]

We can only hope we are at/have passed peak Apple.

 

[Ajay]

Well that'll be a temporary supply constraint unless Apple starts innovating again. Several years of incremental updates to their product lines are starting to show with lower than expected iPhone sales and they're also starting to lose market share to Android in the tablet space.

Yeah, they are in a pickle alright. Two of my nephews in HS (and their friends) consider the iPhone to be an uncool device. Something that should only be used by *gasp* old folks like me!

 

[RussianSensation]

At least Titan is still on schedule


The next generation GeForce and Radeon postponed to Q4 2013

I am not worried. My early sample Titan with an overclock reaches nearly X9000. I promise I have the card with me but I am not allowed to post pictures.

On a serious note, if HD8000/GTX700 are postponed to Q4 2013, what's the point? In that case most of us will just wait 6 months for 20nm Maxwell/Volcanic Islands. Aren't 20nm GPUs slated for 2H of 2014? Seems wasteful to upgrade to 28nm parts in Q4 2013 if we might see 20nm parts by late Q2 / early Q3 2014. Why would both NV and AMD mysteriously decide to postpone their next gen parts to Q4 2013? Did AMD and NV sit down at a round-table meeting this week and mutually agree to delay their launches?

So many of these rumors do not add up.