How does a 680 perform vs. a 780?

ThumbtackJ

I've heard some people say that the 780 is a 'Titan for the masses'. So does it fall somewhere in between a 680 and a Titan? 

 

My plan is to wait for Star Citizen/Squadron 42 to release to see how my GTX680 performs, but by then I could probably just wait for the 800 series. Still, I'd like to know what some of the experts think.

 

Thanks,

TJ

iNeoki

the new 700 series architecture is a major leap over the past generation card improvements. Overall most reviews are stating 30-40% performance increases. Whether you should wait for SC to come out and then be possibly available to the 800 series, that's for you to decide.

Here's a link to a quick google source for good info.
680 vs 780 review / performance charts

Precusor

I'd wait for the 8 series.

Darkholme

I just bought my 680 GTX like 6 months ago, so I'm going to wait for the 8 series. It runs games perfectly fine right now, so unless I see a huge dip in performance when next gen titles start coming out (ala Watchdogs, etc) I'm just going to wait...

Sulaa

"for masses"  is greatly exagerrated - look at the price of GTX 780 and additionally it's not gonna get cheper at fast rate - considering it's size and production cost.

 

I would skip 700 series IMHO.

 

Next wortwhile GPU series will be Vulcanic Islands from AMD and then later on Maxwell from Nvidia maybe.

ThumbtackJ

Originally posted by NeokiNaomi

the new 700 series architecture is a major leap over the past generation card improvements. Overall most reviews are stating 30-40% performance increases. Whether you should wait for SC to come out and then be possibly available to the 800 series, that's for you to decide.

Here's a link to a quick google source for good info.
680 vs 780 review / performance charts

30%-40% is nice, but not for their current asking price, especially considering I haven't had the GTX680 for even a year yet. 

 

If I'm going to spend that much, I'll likely want at least a 50%+ increase in performance. 

Barbarbar

Originally posted by ThumbtackJ

Originally posted by NeokiNaomi

the new 700 series architecture is a major leap over the past generation card improvements. Overall most reviews are stating 30-40% performance increases. Whether you should wait for SC to come out and then be possibly available to the 800 series, that's for you to decide.

Here's a link to a quick google source for good info.
680 vs 780 review / performance charts

30%-40% is nice, but not for their current asking price, especially considering I haven't had the GTX680 for even a year yet. 

 

If I'm going to spend that much, I'll likely want at least a 50%+ increase in performance. 

My personal opinion is, that if your hardware upgrade isn't good for at least 24 months then you made a bad deal.

Quizzical

Originally posted by NeokiNaomi

the new 700 series architecture is a major leap over the past generation card improvements. Overall most reviews are stating 30-40% performance increases. Whether you should wait for SC to come out and then be possibly available to the 800 series, that's for you to decide.

Here's a link to a quick google source for good info.
680 vs 780 review / performance charts

That's complete nonsense.  For the most part, the 700 series is the 600 series.  The only exception is that the GK110 chip that was intended to launch as the top of the line 600 series part got delayed a year, renamed  from GK100 to GK110, and then released as the GeForce GTX Titan and GTX 780.

The rest of the 700 series is just rebrands of 600 series parts, possibly with clock speeds tweaked a bit.  In particular, the GTX 770 is exactly the same as the GTX 680, and you can convert the latter to the former by flashing the BIOS.

If you've already bought a Kepler card, then there's no real reason to buy another so soon as an upgrade unless either you're filthy rich or buying the previous one was a huge mistake and you want to get rid of it.  All that GK110 has to offer for gaming purposes is to take the same architecture on the same process node as the real 600 series cards that aren't Fermi rebrands and then scale up the die size and power consumption to get higher performance.

Now, that's not to say that the GTX 780 is a bad card.  Quite the opposite, it's a good card.  The "problem" is that it was delayed by about a year, but if you're looking to buy after it has launched, that's not your problem.

-----

Creating a GPU chip of a new architecture causes yield problems.  Creating a GPU chip on a new process node before it is well understood causes yield problems.  Creating a GPU chip with a huge die causes yield problems.  And if you want to do all three at once?  Well then, good luck with that.

That's actually why AMD has shied away from the huge dies to compete with Titan.  That contributed to the Radeon HD 5000 series crushing the GeForce 400 series, as AMD was able to get its cards out working properly on time and Nvidia wasn't.

GF100 was a train wreck, so Nvidia delayed it by several months before finally launching some disastrously awful cards on it.  Then Nvidia took a year to fix the chip and was able to launch the GeForce GTX 580.  For the 600 series, GK100 was likewise a train wreck, but this time, Nvidia decided not to bring some awful cards based on it to market, but just go with what was meant to be their upper mid range cards on down for a year until they had time to fix GK100 (and rename it as GK110) and launch it as the GeForce GTX Titan and GTX 780.

While Titan and the GTX 780 do crush AMD's top of the line, that's really just a matter of Nvidia having a properly working huge die on a good architecture.  It really shouldn't be any more surprising than the GeForce GTX 660 beating a Radeon HD 7770, or a Radeon HD 7870 beating a GeForce GTX 650.

If anything, the real aberration was the GeForce GTX 580.  While it was a properly working huge die, it was based on a mediocre architecture.  Not only did the GeForce GTX 570 fail to handily crush the Radeon HD 6970, but it actually tended to lose more than it won.

But as cards become increasingly limited by power rather than transistor count, huge dies may become a thing of the past.  We can already see this in laptops, where neither GK110 (GeForce GTX Titan) nor Tahiti (Radeon HD 7970) made it into laptops.  Likewise, neither GF110 (GeForce GTX 580) nor Cayman (Radeon HD 6970) were used in laptops.  For that matter, the next chips down in the current generation, GK104 (GeForce GTX 680) and Pitcairn (Radeon HD 7870) are only used in a handful of very expensive specialty laptops because even they are a major pain to cool.

If at 20 nm, a huge die gets you awesome performance if you're willing to burn 400 W, will anyone be willing to make such a card?  If you're trying to stay inside of, say, a 250 W TDP, a 550 mm^2 die size doesn't offer you that much advantage over a 350 mm^2 die size.

ichihaifu

Meanwhile my 5850 still rocks every game on the market.

grndzro

AMD  has all next gen consoles so that's it. no point in getting NV.

Dihoru

Originally posted by Quizzical

Originally posted by NeokiNaomi

the new 700 series architecture is a major leap over the past generation card improvements. Overall most reviews are stating 30-40% performance increases. Whether you should wait for SC to come out and then be possibly available to the 800 series, that's for you to decide.

Here's a link to a quick google source for good info.
680 vs 780 review / performance charts

That's complete nonsense.  For the most part, the 700 series is the 600 series.  The only exception is that the GK110 chip that was intended to launch as the top of the line 600 series part got delayed a year, renamed  from GK100 to GK110, and then released as the GeForce GTX Titan and GTX 780.

The rest of the 700 series is just rebrands of 600 series parts, possibly with clock speeds tweaked a bit.  In particular, the GTX 770 is exactly the same as the GTX 680, and you can convert the latter to the former by flashing the BIOS.

If you've already bought a Kepler card, then there's no real reason to buy another so soon as an upgrade unless either you're filthy rich or buying the previous one was a huge mistake and you want to get rid of it.  All that GK110 has to offer for gaming purposes is to take the same architecture on the same process node as the real 600 series cards that aren't Fermi rebrands and then scale up the die size and power consumption to get higher performance.

Now, that's not to say that the GTX 780 is a bad card.  Quite the opposite, it's a good card.  The "problem" is that it was delayed by about a year, but if you're looking to buy after it has launched, that's not your problem.

-----

Creating a GPU chip of a new architecture causes yield problems.  Creating a GPU chip on a new process node before it is well understood causes yield problems.  Creating a GPU chip with a huge die causes yield problems.  And if you want to do all three at once?  Well then, good luck with that.

That's actually why AMD has shied away from the huge dies to compete with Titan.  That contributed to the Radeon HD 5000 series crushing the GeForce 400 series, as AMD was able to get its cards out working properly on time and Nvidia wasn't.

GF100 was a train wreck, so Nvidia delayed it by several months before finally launching some disastrously awful cards on it.  Then Nvidia took a year to fix the chip and was able to launch the GeForce GTX 580.  For the 600 series, GK100 was likewise a train wreck, but this time, Nvidia decided not to bring some awful cards based on it to market, but just go with what was meant to be their upper mid range cards on down for a year until they had time to fix GK100 (and rename it as GK110) and launch it as the GeForce GTX Titan and GTX 780.

While Titan and the GTX 780 do crush AMD's top of the line, that's really just a matter of Nvidia having a properly working huge die on a good architecture.  It really shouldn't be any more surprising than the GeForce GTX 660 beating a Radeon HD 7770, or a Radeon HD 7870 beating a GeForce GTX 650.

If anything, the real aberration was the GeForce GTX 580.  While it was a properly working huge die, it was based on a mediocre architecture.  Not only did the GeForce GTX 570 fail to handily crush the Radeon HD 6970, but it actually tended to lose more than it won.

But as cards become increasingly limited by power rather than transistor count, huge dies may become a thing of the past.  We can already see this in laptops, where neither GK110 (GeForce GTX Titan) nor Tahiti (Radeon HD 7970) made it into laptops.  Likewise, neither GF110 (GeForce GTX 580) nor Cayman (Radeon HD 6970) were used in laptops.  For that matter, the next chips down in the current generation, GK104 (GeForce GTX 680) and Pitcairn (Radeon HD 7870) are only used in a handful of very expensive specialty laptops because even they are a major pain to cool.

If at 20 nm, a huge die gets you awesome performance if you're willing to burn 400 W, will anyone be willing to make such a card?  If you're trying to stay inside of, say, a 250 W TDP, a 550 mm^2 die size doesn't offer you that much advantage over a 350 mm^2 die size.

Didn't understand the last bit but still credit where credit is due: