GTX 780 Performance ? Roughly ????

[brandon888]

Please don't start to bashing me im just wondering cause i saw many rumors that GK 110 beast with 2880 cuda and 384 will have 60-70% better performance then 680 ... so it matches 690 i guess ;/

im not trying to troll here ... i just notice that nvidia improves every generation with 10-15% and every new Architecture about 30-35%

so isn't it a bit strange to make Gk 110 with 60-70% gain ? and note that it's within same Architecture :/

no one want's Fermi 2.0 but 60-70% pefromance is a bit high... isn't it ? ;/

This thread is premature at best. It'll be a more meaningful discussion once GK110 is actually close to shipping and we know more about it
-ViRGE

 

[chimaxi83]

This thread will end well.

It will be 31.57% faster than GTX 680.

 

[imaheadcase]

Doubtful, more like 31.58%, plus you can press a button on back of card for power boost for extra .04% speed with 10c more heat.

 

[brandon888]

Doubtful, more like 31.58%, plus you can press a button on back of card for power boost for extra .04% speed with 10c more heat.

lol guys im serious as i remember there wasn't any leap about 60-70% ;/ biggest leap was from 200 series to 400 when 480 came out ...

 

[Arkadrel]

Since the node process is gonna stay the same....right? just like 480->580.
That "jump" was about ~15% in performance (when the 580 was released).

I dont think its unrealistic that the "jump" in performance might not be that big.
People that are expecting a 50% jump are probably gonna be really let down.

 

[Phynaz]

lol guys im serious as i remember there wasn't any leap about 60-70% ;/ biggest leap was from 200 series to 400 when 480 came out ...

Nobody knows. Does Nvidia even have silicon yet?

 

[brandon888]

Nobody knows. Does Nvidia even have silicon yet?

im sure that they even have maxwell but they won't relese it cause of marketing you know

 

[blackened23]

Please don't start to bashing me im just wondering cause i saw many rumors that GK 110 beast with 2880 cuda and 384 will have 60-70% better performance then 680 ... so it matches 690 i guess ;/

im not trying to troll here ... i just notice that nvidia improves every generation with 10-15% and every new Architecture about 30-35%

so isn't it a bit strange to make Gk 110 with 60-70% gain ? and note that it's within same Architecture :/

no one want's Fermi 2.0 but 60-70% pefromance is a bit high... isn't it ? ;/

Why the heck do you keep asking questions like this? Obviously the answer is 33.33333333333333% repeating.

 

[BenSkywalker]

If it is big K, around 690 performance is reasonable to expect. Thing is, we don't know if the 780 is going to be a big K or a little K respin.

 

[moonbogg]

Clearly, some people here are misinformed. Big-K will be 80% as fast as a GTX690.

 

[brandon888]

so as i understend big K will be for work and little K for gaming ^^ ? maybe they cut shaders to 2304 no ? then there will be no chance for 60-70% performance gain

 

[Granseth]

Clearly, some people here are misinformed. Big-K will be 80% as fast as a GTX690.

I am surprised anybody is informed. Maybe somebody should make a poll for everybody who want to guess.

Personally I would guess 35-50% increase (over GTX 680 ) in most games, and more in compute heavy games. But I'm just putting my finger up in the air and taking a wild guess

 

[VulgarDisplay]

The people claiming anything over 20% increase from the gtx680 have totally forgotten about things like power consumption, heat, noise, and manufacturing cost.

 

[njdevilsfan87]

My guess is 30-35% at 1440P with 4xAA or less, clock for clock vs a GTX 680. But where the card will really shine is with higher amounts of AA and higher resolution due a wider memory bus (and Skyrim modded since everyone feels the need to bring up the 2GB VRAM "issue" there), so there is where it might reach 50%-ish.

 

[AdamK47]

42

 

[BenSkywalker]

The people claiming anything over 20% increase from the gtx680 have totally forgotten about things like power consumption, heat, noise, and manufacturing cost.

The GTX 590 is ~double the power draw, more then double the die size of the 680. The 780 may well end up being 5% faster then the 680, but it won't because it was limited by power consumption, heat, noise and manufacturing cost. This is nVidia, monolithic die is taken to a whole nother level with them

Big-K will be 80% as fast as a GTX690.

I'd consider that around 690 performance. I'd say there is a rather large amount of wiggle room, particularly considering we don't know what BigK is exactly yet, but anywhere in the 75%-100% faster range wouldn't surprise me in the least.

 

[chimaxi83]

The people claiming anything over 20% increase from the gtx680 have totally forgotten about things like power consumption, heat, noise, and manufacturing cost.

This. I would like a cupful of green tears of fanboys who will be defending the increased power consumption of more shaders and a wider bus, and what will most likely be around 25-30% performance increase. You know, since Tahiti gets blasted for it day in, day out, which is funny coming from former Fermi users (I owned two, highly clocked, and I didn't pretend to care about watts).

Anyway, near 690 performance, possibly less than a year after they released the card? Really?

 

[exar333]

Faster than 680 but slower than the 880?

 

[Crap Daddy]

The people claiming anything over 20% increase from the gtx680 have totally forgotten about things like power consumption, heat, noise, and manufacturing cost.

GK104 is almost 2 times faster than GF114. I'm sure the design for GK110 was trying to follow the same pattern in comparison with GF110 (remember this monster chip has 7 billion transistors) 70% over the GTX580 is not so unrealistic and will give us roughly 35% over the GTX680 (GK104). Of course if power consumption, heat, noise, and manufacturing cost will permit this.

 

[Keysplayr]

This. I would like a cupful of green tears of fanboys who will be defending the increased power consumption of more shaders and a wider bus, and what will most likely be around 25-30% performance increase. You know, since Tahiti gets blasted for it day in, day out, which is funny coming from former Fermi users (I owned two, highly clocked, and I didn't pretend to care about watts).

Anyway, near 690 performance, possibly less than a year after they released the card? Really?

Take your trolling line out of the water. Nobody is biting today. Right guys?

ON topic. So lets see. Nvidia went with a compute light (at least double precision light) arch with the 6 series. It appears that it is possible Nvidia is segregating it's desktop products from pro products not only with the drivers and software, but hardware as well.
They might keep the lean green fighting machine approach in the gamers desktop, but go with holy heavy tonnage compute for the pro sector. BigK or even BigM at this point and little M will be for the desktop.
This is pure speculation, but who knows.

 

[toyota]

GK104 is almost 2 times faster than GF114. I'm sure the design for GK110 was trying to follow the same pattern in comparison with GF110 (remember this monster chip has 7 billion transistors) 70% over the GTX580 is not so unrealistic and will give us roughly 35% over the GTX680 (GK104). Of course if power consumption, heat, noise, and manufacturing cost will permit this.

according to the TPU performance summary the gtx680 is right at 70% faster than the gtx560 ti. and it took really high clocks just for gk104 to do that to gf114.

 

[brandon888]

This. I would like a cupful of green tears of fanboys who will be defending the increased power consumption of more shaders and a wider bus, and what will most likely be around 25-30% performance increase. You know, since Tahiti gets blasted for it day in, day out, which is funny coming from former Fermi users (I owned two, highly clocked, and I didn't pretend to care about watts).

Anyway, near 690 performance, possibly less than a year after they released the card? Really?

same here .... they are not idiots .... i don't mean that nvidia is so weak that can't make gk 110 more powerful then +25-30% over 680 ... no ! but they such politics will cost them money ....


isn't 780 a bit overrated and too "wow" ? same feeling was about 680 ... but what we got ? ... i Love Nvidia ....but some people avoid buying 600 series cause they expect such a bit leap ;/ and maybe help them to net waste time ? maybe it's not my business but

 

[Silverforce11]

Depends on whether you think gtx780 is a huge ~600mm2 die or just a respin gk104. The difference is massive.

 

[chimaxi83]

Take your trolling line out of the water. Nobody is biting today. Right guys?

ON topic. So lets see. Nvidia went with a compute light (at least double precision light) arch with the 6 series. It appears that it is possible Nvidia is segregating it's desktop products from pro products not only with the drivers and software, but hardware as well.
They might keep the lean green fighting machine approach in the gamers desktop, but go with holy heavy tonnage compute for the pro sector. BigK or even BigM at this point and little M will be for the desktop.
This is pure speculation, but who knows.

Meh, I brought up a valid point, but I'm glad to see you're on duty

Power consumption is brought up in every other post now, bashing Tahiti (even though it's faster) and praising Kepler. Let's just wait and see what their next card does, and see who reverses their positions.

 

[Keysplayr]

Meh, I brought up a valid point, but I'm glad to see you're on duty

Power consumption is brought up in every other post now, bashing Tahiti (even though it's faster) and praising Kepler. Let's just wait and see what their next card does, and see who reverses their positions.

So, you weren't trolling then? Because it sure as hell looked that way to me bud.