Kepler Nvidia GeForce GTX 780

[NapalmV5]

since you guys mentioned 1080p/2gb/skyrim

heres skyrim/latest update/official hd pack/misc pack/step v201a pack/2x ultra settings/1080p/3x gtx580 3gb

i can only show @ 4096 shadowmaprez because @ 8192.. 3060mb is reached very fast and i have to esc the game before it crashes

iShadowMapResolution=4096
fShadowDistance=2000.0000

high rez short distance shadows


no shadows at long distance


iShadowMapResolution=4096
fShadowDistance=32000.0000

low rez short distance shadows


shadows at long distance

[BababooeyHTJ]

Fallout 3 even maxes out the vram on my GTX480s with the mods that I'm using at 1920x1080. I'm sure that higher resolutions could easily top 2Gb.

[UrbanSmooth]

But IF, (everything based on rumors)

1) GK104 has almost the same die size as 7970
2) GK104 has almost the same price as 7970
3) GK104 will be called GTX680,

can´t we consider it high end?

... Will we see a GTX 685 as a high end card?

... GTX 690 won´t be two GPUs on the same PCB?

I think there´s been a long time since AMD and Nvidia are not sharing the same vision on what a high end part is.

My thoughts exactly. If Nvidia is calling this so called "mid-range" card the GTX 680, what will the so called "high-end" card be called? GTX 680 Ti? GTX 680 Ultra? GTX 685? GTX 780?

Oh, Nvidia, you so craaazay!

[bill_d]

But IF, (everything based on rumors)

1) GK104 has almost the same die size as 7970
2) GK104 has almost the same price as 7970
3) GK104 will be called GTX680,

can´t we consider it high end?

... Will we see a GTX 685 as a high end card?

... GTX 690 won´t be two GPUs on the same PCB?

I think there´s been a long time since AMD and Nvidia are not sharing the same vision on what a high end part is.

the 7970 is not their high end the 7990 is, and the duel card has been AMD's high end for the past few years

maybe nv is doing the same thing now

[solofly]

Food for thought: maybe NVIDIA realized that they don't NEED what would be deemed a "high end Kepler" to trounce AMD this round. Maybe their mid range core is more than enough.

/runs back out of thread

All the rumors definitely point to that possibility and I'm willing to bet that is the case. Midrange card with a high price tag and it's all thanks to AMD/ATI...lol

[snoid_zero]

$550 launch prize
$339 in 2-3 months after launch...

[The Jesus]

A mid-range card sold for $550 would be taking the piss. Smells like price fixing.
With a small die like that and just 2GB of VRAM it can't possibly cost more than $200 to manufacture (probably WAY less).

This seems to be a common misconception from lots of people. CPUs and lots of other hardware cost very little to manufacture (imagine the SB-E chips are probably $50 a piece, if that). However, the reason they cost so much is because of R&D. It costs hundreds of millions in research for each chipset and years of work. I once spoke to a man who tried to start manufacturing his own CPU and by the time he got done he was in about $50 million and the chips were $10 each to make.

[bamtan2]

If Nvidia is calling this so called "mid-range" card the GTX 680, what will the so called "high-end" card be called?

you've missed the plot! they skipped the big chip. gk104 (which is late) IS the high end.

the next big chip will be so late it need not be 600-anything. expect gtx 780 for christmas...

[STEvil]

are we there yet?

[kuroikenshi]

^^ LOL I wanna put down bets that we'll hit 75 pages... any takers?

oh by the way Stevil
http://www.xtremesystems.org/forums/...=1#post5067626

[Cold Fussion]

if i had to guess i would say that the shadow res is stretched across the entire shadow field of view rather than implemented on a per shadow basis as reducing the shadow view distance also sharpens the shadows

Ah ok.

[minpayne]

GK104=GTX680, NDA ends on March 22nd

Pictured at chiphell:





As of now, the author says he still haven't got the driver yet (so no benchmark possible).

Link: http://www.chiphell.com/thread-367824-1-1.html

[aussie-revhead]

If chiphell can't benchmark it because they don't have drivers then are all other leaked benchmarks fakes?

High end GK110 will be a GTX780 ..... Unless AMD can pull something magical out of their hat/bag/ass lolol

Nvidia should call this a 670 and leave the 680 name in reserve, otherwise we have the weird situation where next release will have no new tech/new architecture etc. I want the 110 now, even of its $1000 per vid card, not like we haven't done that before ( maybe you guys haven't paid us$1000 but we have paid over au$1000, that's for sure)

[Olivon]

294mm2
3.54 billion transistors
195W

http://www.chiphell.com/forum.php?mo...9&pid=11424899

[zalbard]

This seems to be a common misconception from lots of people. CPUs and lots of other hardware cost very little to manufacture (imagine the SB-E chips are probably $50 a piece, if that). However, the reason they cost so much is because of R&D. It costs hundreds of millions in research for each chipset and years of work.

Oh, I am fully aware of this. However, it makes the calculation far more complex, since you'd have to account for money gains from future products as well (R&D doesn't have to pay off from a single card model, there will be a whole new line-up using the same tech, quite similar professional cards, and future refreshes may be largely based on it as well).

My point simply is, while there may be lot of R&D involved, there is no doubt that calculated profits from selling these cards (at rumoured $549 MSRP) are going to be FAR greater (per card) than what you'd typically expect from such products, thus implying that they are priced too high and Nvidia is hardly trying to undercut competition and bring more reasonable pricing to the table.

To put it short, this is a rip-off. These cost way too much (due to poor competition, price fixing or just sheer greed). If all of this is true, I sure hope that these sell in low quantities, so while profits per card are going to be WAY up, the overall profits from selling these are lower than expected, teaching Nvidia a lesson.

And talking about performance... So, let's say these are 5% faster in games than 7970 (which is 25% faster than GTX580), making them ~30% faster than GTX580. Is this impressive? Not really. A simple die shrink of GTX580 (which has a hot and large die) would easily allow for a 30% overclock, so we'd end up with a cheaper (smaller die), 30%-higher-clocked-than-GTX580 card with the same 250W TDP. So, according to the rumours, we are going to pay good $250 extra for Nvidia lowering the TDP by 50W (and some GPGPU performance improvements most people don't care about). Doesn't really strike me as an awesome investment.

We need GK100 (GK110?).

[Lanek]

Food for thought: maybe NVIDIA realized that they don't NEED what would be deemed a "high end Kepler" to trounce AMD this round. Maybe their mid range core is more than enough.

/runs back out of thread

Something is sure with 2x SLI connectors on it, i doubt they will realise anything faster before a while.. Or they have drastically change their politic about it. ( allow for tri and 4 quad SLI with midrange.. huum. )..

IF there's a higher end parts who should drop, i hope we will get will get real infos. Not just marketing and rumors for 4-5 months more.

[kgk]

Food for thought: maybe NVIDIA realized that they don't NEED what would be deemed a "high end Kepler" to trounce AMD this round. Maybe their mid range core is more than enough.

/runs back out of thread

If that's true then Nvidia is adorable. I have no problem hanging onto a GTX580 3GB until they release GK110 or GK112 (whichever model number is the high-end single GPU). If the 7970 were a better value (or anything remotely approaching value) I'd have jumped on it already.

Unless GK104 gives significantly better performance than an overclocked GTX580 3GB @ 2560x1600, Nvidia can bugger off. And if they're serious with 2GB VRAM then that'll mean a few games will either be pushing or maxing out its VRAM already, let alone any games that haven't even been released yet. I mean 2GB VRAM on a "top end" card.........in 2012? I sure hope not.

And talking about performance... So, let's say these are 5% faster in games than 7970 (which is 25% faster than GTX580), making them ~30% faster than GTX580. Is this impressive? Not really. A simple die shrink of GTX580 (which has a hot and large die) would easily allow for a 30% overclock, so we'd end up with a cheaper (smaller die), 30%-higher-clocked-than-GTX580 card with the same 250W TDP. So, according to the rumours, we are going to pay good $250 extra for Nvidia lowering the TDP by 50W (and some GPGPU performance improvements most people don't care about). Doesn't really strike me as an awesome investment.

We need GK100 (GK110?).

Whoa, whoa, whoa wait a minute. a 7970 is 25% faster than a stock GTX580 1.5GB and that's in best case scenarios. A brand new 3GB GTX580 is still 40-50 Euros cheaper than the 7970.

Minimum frame rate is equally (or more) important than the average. A 45FPS average is worthless if you're dipping down into the teens. At 1920x1200 and below the stock GTX580 1.5GB actually beat the 7970 in minimum frame rate in Dirt 3 (coincidentally the same title that lots of 7970 manufacturers seem to bundle with their cards) http://www.anandtech.com/show/5261/a...7970-review/18

40-50 Euros higher for a whopping 6.7 fps increase in minimum frame rate @ 2560x1600 and potentially equal or lower minimums at lower resolutions? I don't see the incentive. It's still overpriced. If you ignore brand-specific features (eyefinity) I fail to see any justifiable reason for paying the 7970 MSRP.

I have zero loyalty to Nvidia but let's try to avoid Apple-esque reality distortion.

I agree that we need GK110/112/100/whatever the hell the high-end single GPU is called.

[Tim]

Has anybody actually discussed how awesome it would be if 190w beats the current high end cards by a fair margin? I mean 190w!!! That's fantastic! (if true)

[DilTech]

you've missed the plot! they skipped the big chip. gk104 (which is late) IS the high end.

the next big chip will be so late it need not be 600-anything. expect gtx 780 for christmas...

A few months behind a competitor isn't late. 6+ months is late. We've discussed that fact earlier.

GK104 is only being looked at as high end because NVidia found themselves in a unique position where their midrange gpu was strong enough to beat their oppositions high-end. As such, they can sit and reap major profits and force AMD to show their next card, which will have to deal with the GK110.

[boxleitnerb]

Has anybody actually discussed how awesome it would be if 190w beats the current high end cards by a fair margin? I mean 190w!!! That's fantastic! (if true)

Fantastic? Hm, I would say it's normal

[Pantsu]

While competing with Tahiti using a smaller chip sounds like a great improvement for Nvidia, the same can't be said about GK110 launch schedule, so I guess it balances out. Most 580 owners probably won't "upgrade" for Tahiti level performance even if GK104 were a lot more efficient. It's not surprising though that Nvidia will follow AMD in slapping their performance GPU with an enthusiast price tag.

[kgk]

A few months behind a competitor isn't late. 6+ months is late. We've discussed that fact earlier.

GK104 is only being looked at as high end because NVidia found themselves in a unique position where their midrange gpu was strong enough to beat their oppositions high-end. As such, they can sit and reap major profits and force AMD to show their next card, which will have to deal with the GK110.

That's my reasoning as well. Until we see actual GK104 benchmarks though it's hard to be certain.

The problem this presents for some people is that if GK104 is only a smidgen ahead of the 7970 that we wind up with two overpriced and unimpressive cards because AMD doesn't have anything ready to regain the top spot (dual GPU monstrosities are irrelevant in this case). So Nvidia rides out GK104 until AMD can release an 8000-series card at which point they finally dump the flagship GK110/112/100/whatever it's called.

While competing with Tahiti using a smaller chip sounds like a great improvement for Nvidia, the same can't be said about GK110 launch schedule, so I guess it balances out. Most 580 owners probably won't "upgrade" for Tahiti level performance even if GK104 were a lot more efficient. It's not surprising though that Nvidia will follow AMD in slapping their performance GPU with an enthusiast price tag.

And that's what I don't understand. The GTX580 has been out since November of 2010. Plenty of 580 owners are ready to upgrade (myself included) but I'm not going to do it for a little gnat-fart GK104 when everyone knows the big leap will be the GK100/110. GK104 is seeming more and more like a stop-gap card to keep Nvidia from hemorrhaging any more customers to AMD.

[minpayne]

If that's true then Nvidia is adorable. I have no problem hanging onto a GTX580 3GB until they release GK110 or GK112 (whichever model number is the high-end single GPU). If the 7970 were a better value (or anything remotely approaching value) I'd have jumped on it already.

Unless GK104 gives significantly better performance than an overclocked GTX580 3GB @ 2560x1600, Nvidia can bugger off. And if they're serious with 2GB VRAM then that'll mean a few games will either be pushing or maxing out its VRAM already, let alone any games that haven't even been released yet. I mean 2GB VRAM on a "top end" card.........in 2012? I sure hope not.

I'm with you. Recall that the S3 ViRGE had 2MB VRAM about 15 years ago. Now we have 2GB VRAM on mainstream cards. It's a 1024-fold increase, which is strictly and strikingly consistent with Moore's Law (doubles every 18 months)!

For an enthusiast dumping serious money into high-end graphics cards (especially for SLI to max out graphics including AA), it doesn't make sense to make a quad-SLI with only 2GB VRAM in Year 2012, not even for 1920x1080.

[Handrox]

[spicypixel]

I guess 50% better than 580 in 3dmark11 isn't to be sniffed at.