DT: Next-generation 28nm GPUs Could Be 45 Percent Faster

[SickBeast]

It's too bad that we're down to hoping for a 45% improvement year-over-year. Not only that, but I think most people would be very excited to see a GPU 45% more powerful than a GTX 580.

I remember when GPU power would sometimes triple year-over-year.

 

[Idontcare]

It's too bad that we're down to hoping for a 45% improvement year-over-year. Not only that, but I think most people would be very excited to see a GPU 45% more powerful than a GTX 580.

I remember when GPU power would sometimes triple year-over-year.

When was the last time we saw a new node give us 45% higher clocks?

 

[crisium]

45% higher clocks would give us around 1300MHz on the high end AMD cards. Do you think this is feasible?

 

[SickBeast]

When was the last time we saw a new node give us 45% higher clocks?

Off the top of my head, I don't know.

That 45% figure is potentially very misleading. Would the chip clock 45% faster if it were a simple die shrink? Perhaps according to this article. What about a far more complex GPU, which we are certain to see with the next generation? Then we would probably be looking at a 25% improvement which is in line with what we've seen in the past.

I'm very skeptical about leaked information prior to product launches now. I will believe all this when I see it.

Like I said, my prediction FWIW is that we will be lucky to see a single GPU that's 45% faster than a GTX 580. Your own data corroborates the fact that microprocessors are hitting the wall and we're down to diminishing returns.

All of that said, a die shrunk GTX 560 that clocks 45% faster and costs less would be nice. :thumbsup::thumbsup:

 

[LiuKangBakinPie]

Easy to see how much more beef it got.
Shader count x shader frequency x3 should give you the single precision G/flops. Then compare it to the current batch

 

[Skurge]

Off the top of my head, I don't know.

That 45% figure is potentially very misleading. Would the chip clock 45% faster if it were a simple die shrink? Perhaps according to this article. What about a far more complex GPU, which we are certain to see with the next generation? Then we would probably be looking at a 25% improvement which is in line with what we've seen in the past.

I'm very skeptical about leaked information prior to product launches now. I will believe all this when I see it.

Like I said, my prediction FWIW is that we will be lucky to see a single GPU that's 45% faster than a GTX 580. Your own data corroborates the fact that microprocessors are hitting the wall and we're down to diminishing returns.

All of that said, a die shrunk GTX 560 that clocks 45% faster and costs less would be nice. :thumbsup::thumbsup:

Rumors point to 7870 being a die shrunk 6970. So that could be a 6970 at 1.2ghz + a few tweaks.

 

[LiuKangBakinPie]

this

 

[WMD]

It's too bad that we're down to hoping for a 45% improvement year-over-year. Not only that, but I think most people would be very excited to see a GPU 45% more powerful than a GTX 580.

I remember when GPU power would sometimes triple year-over-year.

Nowadays 30% gain in 15 month will be amazing. Just look at how successful the 6000 series are.

 

[toyota]

its been 3 years since I bought a gtx260 and its still not possible to get a card that can fully double its performance for the same price I paid back then.

 

[Concillian]

I remember when GPU power would sometimes triple year-over-year.

7800 GTX TDP: 84 Watts
Top of the line in that generation

84 watts isn't even a 5770.

Part of the increases in the past, a pretty large part, came from just increasing power usage. They're to the point where the highest end cards now are over 3 times the power of a 7800GTX and almost double a 9800 GTX.

MFRs can't keep increasing power, they've kinda reached the limit on that side of things.

 

[RichUK]

When are these new AMD/ATI 7 series cards coming out?

 

[Fanatical Meat]

more importantly how much will they cost?

 

[Soccerman06]

this

Weve been over this before, rambus wont be used in AMD cards.

 

[AtenRa]

^ surely the 1.75V is in error? Maybe 1.05V?



This is where you REALLY have to watch the EXACT words used in the marketing-speak. They cited "gate leakage"...neither the gate leakage (no mention of the leakage from all the other contributing sources) nor leakage in general (adding in all other contributors) say much of anything about the dynamic power consumption of an IC under load.

From here, dynamic power dominates leakage (static) power by nearly a ratio of 6:1.

Saying the leakage power is the same when the clockspeeds are 45% higher is literally speaking to power-consumption at idle clockspeeds when static leakage power dominates.

That 300W GTX580 is 300W because 225W is dynamic power and 75W is static power, raising clocks 45% just because the static power will remain 75W says nothing about what happens to the 225W of dynamic power when you raise those clocks.

IMO the entire basis of the OP is a red herring.

First of all a BIG thanks for the formula and the graphs.

The article most probably took the 45% figure from TSMC's PDF

http://www.tsmc.com/download/brochures/2011_28 Nanometer Process Technology.pdf

The 28nm high performance (HP) process is in risk production and will be the first process from TSMC to use high-k metal
gate technology. Featuring superior speed and performance, the 28HP process targets CPU, GPU, FPGA, PC, networking,
and consumer electronics applications. The 28HP process supports a 45 percent speed improvement over the 40G process
at the same leakage per gate.

Just a question, does the "same leakage per Gate" as written in the PDF translates in to Static leakage, dynamic leakage or leakage in general??