Interesting... Fermi vs Kepler

[kipper]

Imagine what Nvidia could have done if they die shrunk Fermi and bumped it to 1ghz...

 

This is a GTX 480 at 865mhz.

 

Both tests are maxed out at 1080p using 4xMSAA.

 

 

 

 

 

 

 

[ZetZet]

480 is powerful, and hot. 

[AlwaysFSX]

The problem with Fermi is that it was more of a brute force design instead of more intelligent load processing.

[i_build_nanosuits]

Imagine what Nvidia could have done if they die shrunk Fermi and bumped it to 1ghz...

they did, it's called the GTX 580.

[kipper]

they did, it's called the GTX 580.

No, 580 is just a few more cores, exact same process and it shipped at 775mhz.

[i_build_nanosuits]

No, 580 is just a few more cores, exact same process and it shipped at 775mhz.

still...your point was ''imagine what they could have done with fermi'' well..that's what they did...they did the GTX 580 which is the unleached version of fermi...that's how good they could made it.

[kipper]

still...your point was ''imagine what they could have done with fermi'' well..that's what they did...they did the GTX 580 which is the unleached version of fermi...that's how good they could made it.

Not really no, the reason they did not push it to 28nm was the power consumption and thermals.

 

Due to AMD and other press giving them a hard time they decided to strip their GPU's of any compute function, which is why Fermi is so damn strong in Hitman, Kepler falls to its knee's in comparison.

[i_build_nanosuits]

Not really no, the reason they did not push it to 28nm was the power consumption and thermals.

 

Due to AMD and other press giving them a hard time they decided to strip their GPU's of any compute function, which is why Fermi is so damn strong in Hitman, Kepler falls to its knee's in comparison.

Kepler is VASTLY superior to fermi...and your example of a kepler card is the GTX680...which is the small kepler GK104 chip operating at low clockspeeds...compare the GTX 580 to a 780ti or titan black which are full fledged GK110 GPU which is MUCH more powerful..even in compute.

GTX580(best fermi) vs GTX 780ti (best kepler)

http://www.anandtech.com/bench/product/1072?vs=1350

As you can see, kepler is a MUCH stronger GPU architecture...hands down...even pumping 1500mhz on the core if it was possible the 580 would still be way slower.

[kipper]

Kepler is VASTLY superior to fermi...and your example of a kepler card is the GTX680...which is the small kepler GK104 chip operating at low clockspeeds...compare the GTX 580 to a 780ti or titan black which are full fledged GK110 GPU which is MUCH more powerful..even in compute.

GTX580(best fermi) vs GTX 780ti (best kepler)

http://www.anandtech.com/bench/product/1072?vs=1350

As you can see, kepler is a MUCH stronger GPU architecture...hands down.

You missed what i was talking about by a land slide, Good Job.

[i_build_nanosuits]

You missed what i was talking about by a land slide, Good Job.

you said a 1000mhz fermi would be a beast...i'm prooving you: IT WOULD NOT.

[kipper]

you said a 1000mhz fermi would be a beast...i'm prooving you: IT WOULD NOT.

You do realize the 700 series is 3 newer gens vs 400 series?

[rjfaber91]

Here's your answer:

 

560: TDP = 150W

660: TDP = 140W

760: TDP = 130W

960: TDP = 120W

 

Consider how much faster these cards have become, while consistently using 10W less power for each generation, and it becomes obvious why Kepler and Maxwell have a lot more potential than Fermi. Hell, I'm saving 25W when I, hopefully next month, switch out my 560Ti for a 970, and those two cards aren't even in the same performance range.

[kipper]

Here's your answer:

 

560: TDP = 150W

660: TDP = 140W

760: TDP = 130W

960: TDP = 120W

 

Consider how much faster these cards have become, while consistently using 10W less power for each generation, and it becomes obvious why Kepler and Maxwell have a lot more potential than Fermi.

 

Not really no, the reason they did not push it to 28nm was the power consumption and thermals.

 

Due to AMD and other press giving them a hard time they decided to strip their GPU's of any compute function, which is why Fermi is so damn strong in Hitman, Kepler falls to its knee's in comparison.

[rjfaber91]

[some quote]

 

So you say, but is that true? Nobody has been pushing the GPGPU concept like Nvidia has, so to say they stripped their GPUs of computing functions seems a bit untrue at best...

[i_build_nanosuits]

You do realize the 700 series is 3 newer gens vs 400 series?

YES!! that's the point: FERMI is an old, slow and inneficient achitecture i don't see why you keep beating a dead horse like that?! you're right...i really don't get your point.

[kipper]

I will just give up as it seems people here are on a totally different wave length.

[Manny Calavera]

Here's your answer:

 

560: TDP = 150W

660: TDP = 140W

760: TDP = 130W

960: TDP = 120W

 

Consider how much faster these cards have become, while consistently using 10W less power for each generation, and it becomes obvious why Kepler and Maxwell have a lot more potential than Fermi. Hell, I'm saving 25W when I, hopefully next month, switch out my 560Ti for a 970, and those two cards aren't even in the same performance range.

and those two are both 104/204 chips

[i_build_nanosuits]

I will just give up as it seems people here are on a totally different wave length.

of course they are, nvidia has more than doubled the performance of the cuda cores with kepler...and added another 40% on top with maxwell which can reach the performance of kepler by using 40% less cores...fermi is nowhere near the performance of any of these newer cards...they moved on for good reasons and we all appreciate the much powerful gpu's available these days, and honestly: you should do the same

[kipper]

of course they are, nvidia has more than doubled the performance of the cuda cores with kepler...and added another 40% on top with maxwell which can reach the performance of kepler by using 40% less cores...fermi is nowhere near the performance of any of these newer cards...they moved on for good reasons and we all appreciate the much powerful gpu's available these days, and honestly: you should do the same

I don't use Fermi.

[i_build_nanosuits]

I don't use Fermi.

good

there's reasons why nvidia moved on from fermi, and we can fully grasp the full extent of why this happened today...and pascal will AGAIN be much stronger than any previous architecture, yes they could have cramped 4000 kepler cuda cores on a smaller die and reach the performance of a 980ti with kepler...but they didnt because nvidia is not AMD, they improve things, not just make the bigger.

[rjfaber91]

and those two are both 104/204 chips

 

The 560Ti is 114 according to Wikip. Not that I have the foggiest idea what the individual numbers in Nvidia's code names stand for...

[Manny Calavera]

The 560Ti is 114 according to Wikip. Not that I have the foggiest idea what the individual numbers in Nvidia's code names stand for...

they are moving smaller chips to have bigger number

 

if we compare 580 to 680 its unfair, as 780ti is kepler fully unlocked

 

so 970 is even lower end chip than it should be (just under 980 if this was pre-kepler chip naming) while being still extremely powerful

[kipper]

My gtx 480 can clock 900 core 2100 ram, pretty sure there few out there but very very rare that can clock 1000

 

If not upgraded before cos i did't need to, current games are boring only thing i look forward to atm is the division, so probably upgrade around the end of this year.

http://forum.overclock3d.net/showthread.php?t=45446

 

 

http://www.techpowerup.com/gpuz/details.php?id=v8m2x

[Alurit]

I'm happy with nvidia choosing to do kepler instead of making fermi even more powerful, and then doing maxwell, even with the r9 300 competitive pricing I don't like recycling old architectures, cause look at amd, if you're buying low end you had to make sure that card is dx12 compatible cause it's so fucked up right now

[Qyoon]

Imagine what Nvidia could have done if they die shrunk Fermi and bumped it to 1ghz...

 

 

 

I do imagine.