Nvidia GTX 1080 Press Slides and Full Specifications leak

[Deleted User]

The user and all related content has been deleted.

ceratop001

Interesting comparisons thanks blue:) Think I will be keeping my 980ti for the time being...

Hellscream

if you have a high end 980ti the 1080 is not worth getting i would wait a while longer to see what else is in store for GPU cards

SEANMCAD

Hellscream said:

if you have a high end 980ti the 1080 is not worth getting i would wait a while longer to see what else is in store for GPU cards

true.

the general rule that has worked out well for me is 18 months on video cards. meaning if you bought your current card less than 18 months ago its likely to not worth it to upgrade.

Quizzical

18 months is an awfully short life cycle.  My general rule is wait until you can afford to double your performance.  These days, you'd typically expect that to take about four years if your budget doesn't expand.

SEANMCAD

Quizzical said:

18 months is an awfully short life cycle.  My general rule is wait until you can afford to double your performance.  These days, you'd typically expect that to take about four years if your budget doesn't expand.

used to be the performance doubled every 18 months on video cards and about 3 years for a PC. 
I say double but not completely sure.

interesting side note they found that the average lifespan of a PC (for all users) is 3 years before buying a new one. Now this was years ago so i dont know if its true anymore

Quizzical

Moore's Law says that you double your number of transistors about every two years.  It used to be that that basically meant you could double your performance, though at the expense of using about 40% more power.  When that was the difference between 5 W and 7 W, no one cared much about the power.  When it's the difference between 250 W and 350 W, 40% more power is a problem.  So now, you can get about 40% more performance in the same power envelope as before, which is the real limiting factor these days.  And that means it takes four years to double your performance.

SEANMCAD

Quizzical said:

Moore's Law says that you double your number of transistors about every two years.  I

its been more like every 18 months:

'The period is often quoted as 18 months because of Intel executive David House, who predicted that chip performance would double every 18 months (being a combination of the effect of more transistors and the transistors being faster)'

https://en.wikipedia.org/wiki/Moore's_law

Quizzical

SEANMCAD said:

Quizzical said:

Moore's Law says that you double your number of transistors about every two years.  I

its been more like every 18 months:

'The period is often quoted as 18 months because of Intel executive David House, who predicted that chip performance would double every 18 months (being a combination of the effect of more transistors and the transistors being faster)'

https://en.wikipedia.org/wiki/Moore's_law

An early quote said every 18-24 months.  If you plot data from the last 50 years and fit a line to it, it comes out to about every two years.

Malabooga

More detailed analysis:

Ridelynn

Quizzical said:

SEANMCAD said:

Quizzical said:

Moore's Law says that you double your number of transistors about every two years.  I

its been more like every 18 months:

'The period is often quoted as 18 months because of Intel executive David House, who predicted that chip performance would double every 18 months (being a combination of the effect of more transistors and the transistors being faster)'

https://en.wikipedia.org/wiki/Moore's_law

An early quote said every 18-24 months.  If you plot data from the last 50 years and fit a line to it, it comes out to about every two years.

It's like the difference in saying back to school is summer or fall... not really consequential to the discussion, and more a matter of semantics.

SEANMCAD

Quizzical said:

SEANMCAD said:

Quizzical said:

Moore's Law says that you double your number of transistors about every two years.  I

its been more like every 18 months:

'The period is often quoted as 18 months because of Intel executive David House, who predicted that chip performance would double every 18 months (being a combination of the effect of more transistors and the transistors being faster)'

https://en.wikipedia.org/wiki/Moore's_law

An early quote said every 18-24 months.  If you plot data from the last 50 years and fit a line to it, it comes out to about every two years.

''The period is often quoted as 18 months because of Intel executive David House, who predicted that chip performance would double every 18 months (being a combination of the effect of more transistors and the transistors being faster)'

filmoret

So, is it better then the proposed new AMD chips or wut?

Malabooga

Performance is as Nvidia said, 20-25% over Titan X/980ti, still bad in Dx12 and no Async from what has been seen.

Quizzical

filmoret said:

So, is it better then the proposed new AMD chips or wut?

AMD hasn't officially announced specs yet, but there are just a lot of rumors.  My best guess is that the GeForce GTX 1070 will be a little faster than the top Polaris 10 GPU (e.g., 10% faster), but not a lot.  The GTX 1080 is substantially faster than both, of course, and further cut down Polaris 10 will be slower.  Meanwhile, Polaris 10 will be a much smaller die and use substantially less power.

Polaris 11 is a much smaller die yet, much lower power, and much lower performance.  I'm expecting Polaris 11 performance to be between a Radeon R7 370 and R9 380, for for Nvidia equivalents, between a GeForce GTX 950 and GTX 960.  I expect Polaris 11 to be the clear best chip for gaming laptops for some months, at least until Nvidia launches a small Pascal, but it's not that interesting of a desktop chip until 14 nm is more mature and AMD is more aggressive on pricing it.  Someday, Polaris 11 is likely to be the best $100 gaming desktop card on the market, but that day could easily be a year away.  Or two.  As a $150 desktop card, it's a lot less interesting.

Vega is the big next generation AMD chip, and that is due early next year.  I expect the GTX 1080 to be the top consumer graphics card on the market for a while, even if it's not widely available until Fall.  But I'd be extremely shocked if it doesn't eventually get beaten both by Vega and GP100, the big Pascal chip.

Quizzical

Malabooga said:

Performance is as Nvidia said, 20-25% over Titan X/980ti, still bad in Dx12 and no Async from what has been seen.

What matters is not whether Nvidia can do asynchronous compute, but how they perform in games that will use asynchronous compute if it's available.  These are related, of course, but a lack of asynchronous compute isn't directly a game-breaking problem in itself.

Ridelynn

The initial reviews today look promising. We won't be able to resolve the question about availability yet, but right now it's looking more plausible than it did a week ago.

Quizzical

Torval said:

Overall, it's overpriced, especially with the early adopter tax, but it looks like it will not disappoint. For 1080p gaming it's way overkill and way overpriced. I wonder where that will put AMD with Polaris.

Top end cards are nearly always a poor value if you just consider performance per dollar.  Among current generation cards, for example, a Fury X and a GTX 980 Ti are about twice as expensive as an R9 390 or a GTX 970, but nowhere near twice as fast.  So even at $700, the GTX 1080 isn't really out of line there.

mmoguy43

All I want to know is if the 1080 will be cost effective for a likely needed upgrade in the next year or two.  Hopefully it will be but Nvdia tends to keep price up.

Quizzical

mmoguy43 said:

All I want to know is if the 1080 will be cost effective for a likely needed upgrade in the next year or two.  Hopefully it will be but Nvdia tends to keep price up.

A year from now, competition from Vega should have pushed the price down to about $400 or $500.  But it will depend on exactly how good Vega and GP100 are, as well as if there are any other chips coming in that range that aren't yet announced.

Two years from now, we'll be close enough to Volta and Navi that you might just want to wait for that.

SEANMCAD

mmoguy43 said:

All I want to know is if the 1080 will be cost effective for a likely needed upgrade in the next year or two.  Hopefully it will be but Nvdia tends to keep price up.

personally I would look at the trend from the last generation of cards.

This generation will likely replicate that down to the same dollar and same month. ok maybe not that much but it will be very close to the same pattern

Loke666

ceratop001 said:

Interesting comparisons thanks blue:) Think I will be keeping my 980ti for the time being...

Lol, yeah. Not worth the upgrade for you. You should wait at least 1 more generation, here you wont notice much difference.

SpottyGekko

Looks like my 980ti will be fine for the next year or two.

I've not even seriously overclocked it, haven't even used it's water-cooling option, so there's some headroom yet.

[Deleted User]

The user and all related content has been deleted.

Quizzical

I've already got a Fury X, so I see no need to get a GTX 1080 already.  Nor a Vega or the big Pascal GPUs, either.  Maybe I'll upgrade when Volta and Navi come around, but there's a decent chance that I'll end up waiting for 7 nm.  It depends on how soon I run into situations where the card I have now just isn't good enough.

SEANMCAD

side note:

I just got an email that say Microcenter will have the 1080 GTX in stores by May 27th.

No price was posted $599 from my understanding