Overclock.net banner

[Atomic MPC] HD7970 vs GTX 780: the future of graphics cards in 2012

10K views 72 replies 55 participants last post by  diabloSUCKS 
#1 ·
http://www.atomicmpc.com.au/Feature/288642,hd7970-vs-gtx-780-the-future-of-graphics-cards-in-2012.aspx

Quote:
With each new generation graphics are pushed further, delivering more power. We'll typically we see a ~30% increase per generation as well as new features (such as tessellation) thrown into the mix. Increased processing power in turn requires building new graphics architectures with more of a chip's basic constituent parts: transistors. Transistors form the building blocks of both Graphic Processing Units (GPUs) and CPUs alike; the long and short of it is that more transistors equals more 'oomph'.
Look like another excellent year for videocard enthusiasts!
 
#4 ·
That's crazy... I knew the defective chips were scattered through the trays but I didn't think there were THAT many in each tray. o.o

Guess we all know how the 5850 and GTX 470 really came into light now...
tongue.gif
 
#9 ·
Good read. Looks legit. Seems AMD can pump out 28nm so Nvidia should follow...hopefully that drastically larger die can produce little heat and clock WAY high! Looking forward to getting my hands on a 28nm chip once they're released by nvidia! Can't wait to crack that IHS if they have one and use my antec 620 to cool the puppy!

It's very surprising the performance increases they're talking about. I have read of 2x a performance increase...wow! Maybe it is worth the wait?
 
#11 ·
Quote:
Originally Posted by Cakewalk_S View Post

Good read. Looks legit. Seems AMD can pump out 28nm so Nvidia should follow...hopefully that drastically larger die can produce little heat and clock WAY high! Looking forward to getting my hands on a 28nm chip once they're released by nvidia! Can't wait to crack that IHS if they have one and use my antec 620 to cool the puppy!
It's very surprising the performance increases they're talking about. I have read of 2x a performance increase...wow! Maybe it is worth the wait?
Nvidia tends to be super ambitious with their goals, but usually don't achieve it in their first iterations of hardware. GF100 was supposed to be 2.5x faster than 5870, obviously that didn't happen. GK100 was supposed to be the same (compared to GF110) and was cancelled. Not a knock, just saying
smile.gif
It usually takes them 2 tries to get it right. Seems like Kepler will be the same way, GK110 will be kepler done right. Of course GK110 won't be here until later either.

This article has nothing of substance to it that I can see, and mentions GK100 which was cancelled. It even has fake slides which doesn't help matters.
 
#12 ·
Looking forward to what nvidia comes up with if the 28nm process is so solid.
 
#13 ·
Quote:
Originally Posted by Cakewalk_S View Post

Good read. Looks legit. Seems AMD can pump out 28nm so Nvidia should follow...hopefully that drastically larger die can produce little heat and clock WAY high!
Well that won't happen.
I think both Nvidia and AMD can create similar performing cards for the same die size. (same process and manufacturer)
Nvidia can only make a faster chip than the 7970 by making it larger and more power hungry. If the 7970 (4.3B) is a 250W part, then a 6 billion Nvidia card will be over 300W.

You know, power is expensive in Europe. Why can't they try to find performance the Intel way? Faster architecture and lower TDP.
 
#14 ·
Quote:
Originally Posted by toX0rz View Post

How exactly does it matter?
How does anything in the entire world actually matter? If you don't have an answer don't bother posting...

Quote:
Originally Posted by Nite View Post

the 600 series might be OEM only so more then likely 700 is what we will get, nvidia did the same thing with the 300 series as well in the past and skipped to 400
Still no official confirmation from Nvidia then yet?
 
#16 ·
Quote:
Originally Posted by djriful View Post

I want to clarify something. Is HD 7970 a Hybrid 28nm or a 28 nm High Power?

The way they stated it was; in order for AMD to get 7970 out of the door faster than Nvidia, it has adopted in the hybrid 28 nm process.
It is high power, I remember this being asked and mentioned at their financial analysts day. It is not hybrid.
Quote:
These massive speeds can in part be put down AMD's use of TSMC's most effectual 28 nanometre process, called 28nm HP (High-Power). Initially it was believed only Nvidia would be using this process, with AMD adopting a hybrid 28nm LP (Low Power) process in order to get its cards out of the gate more quickly; however AMD has surprised pundits by going the whole hog while releasing a good three months before Nvidia. With 2048 cores, 3GB of GDDR5 over a 384-bit bus and overclocking giving a guaranteed 50% speed boost over even the most overclocked HD 6970, the AMD HD 7970 is a beast - for now it's simply the fastest single card on the planet.
 
#17 ·
Quote:
Originally Posted by John-117 View Post

Well that won't happen.
I think both Nvidia and AMD can create similar performing cards for the same die size. (same process and manufacturer)
Nvidia can only make a faster chip than the 7970 by making it larger and more power hungry. If the 7970 (4.3B) is a 250W part, then a 6 billion Nvidia card will be over 300W.
You know, power is expensive in Europe. Why can't they try to find performance the Intel way? Faster architecture and lower TDP.
You almost make it sound easy. The problem is that demands for GPU's are no longer a single endeavor. GPU's are being used increasingly for GPGPU tasks, and that requires extra die space for the hardware. Each time a feature is added onto the graphics card, an area of the die has to be allocated for the task. Unlike CPU's, making the architecture more efficient is very difficult when the current process is already very VERY efficient (much more efficient than a CPU in most tasks.) If you have some insight into this, I'd love to hear it.

AMD has the more power efficient architecture, while nvidia has a more feature-rich architecture. There's a trade-off of power consumption in this case.
 
#18 ·
It think it was the LP of 28nm and the next 8xxx would be the HP of 28nm from AMD.

Quote:
Originally Posted by djriful View Post

I want to clarify something. Is HD 7970 a Hybrid 28nm or a 28 nm High Power?

The way they stated it was; in order for AMD to get 7970 out of the door faster than Nvidia, it has adopted in the hybrid 28 nm process.
 
#20 ·
Currently waiting for Kepler benchmarks (official, not rumor) to decide on my next GPU. It'll probably be an Nvidia even if it matches up with a 7970 just because I use my GPUs for folding, and Stanford cannot see to properly utilize OpenCL, so AMD cards are terrible for folding, with my GTX 275 outpacing the 7970 in folding.
 
#21 ·
Quote:
Originally Posted by John-117 View Post

Well that won't happen.
I think both Nvidia and AMD can create similar performing cards for the same die size. (same process and manufacturer)
Nvidia can only make a faster chip than the 7970 by making it larger and more power hungry. If the 7970 (4.3B) is a 250W part, then a 6 billion Nvidia card will be over 300W.
You know, power is expensive in Europe. Why can't they try to find performance the Intel way? Faster architecture and lower TDP.
lol my mind read OVER 9000!!!
 
#22 ·
Quote:
Originally Posted by xoleras View Post

Nvidia tends to be super ambitious with their goals, but usually don't achieve it in their first iterations of hardware. GF100 was supposed to be 2.5x faster than 5870, obviously that didn't happen. GK100 was supposed to be the same (compared to GF110) and was cancelled. Not a knock, just saying
smile.gif
It usually takes them 2 tries to get it right. Seems like Kepler will be the same way, GK110 will be kepler done right. Of course GK110 won't be here until later either.
This article has nothing of substance to it that I can see, and mentions GK100 which was cancelled. It even has fake slides which doesn't help matters.
Lol, did nVidia really say that?
A GF100 single GPU that was supposed to be 2.5 times faster than a HD 5870? Not even the GTX 590 can do that and its based on later gen GF110 and is even a dual-gpu.
I cant imagine nVidia was so insane to actually claim that.

EDIT: Actually, I think you got something mixed up there. What you mean is the tesselation performance. The GTX 480 was rumoured to have 2.5 times the tessellation power of a HD 5870 and it turned out to be true aswell...
Quote:
Originally Posted by Phoenixlight View Post

How does anything in the entire world actually matter? If you don't have an answer don't bother posting...
Still no official confirmation from Nvidia then yet?
Easy there, it just bugs me that people act like a GPU generation would be skipped while it really is just a name that has no impact on the card.
If the 5xx cards were named 6xx they would still be the same thing just with a different branding... no technology or generation would be "skipped" here.. and the same applies for Kepler.. 680 or 780, who cares?
 
#25 ·
I will probably need a new video card by summer...once Arma 3 is out lol
 
#26 ·
Two things I am waiting for.

1: Nvidia releasing new beastly cards (hopefully)

2: Intel releasing 8-core socket 2011 thats not XEON.

Why oh why, won't you share some info to me Nvidia?! I want benchmarks...
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top