Sunday, November 6th 2011
New 28 nm Graphics Cards To Be 45 Percent Faster And Overclock Like Never Before?
The next generation NVIDIA and AMD GPU's are going to be built on Taiwan Semiconductor Manufacturing Company's (TSMC) brand new 28 nm process, which may offer a 45% speed boost to these new products. Smaller geometries usually mean that a processor will use less power and can be made smaller, faster and more cheaply than previously. It's good news then that TSMC reports that the new process is ready for mass production and is running very well indeed. So well in fact, that unnamed sources within TSMC claim the new 28 nm process allows for a 45% increase in clock speed over the current 40 nm process. Put this together with improved GPU architectures and the next generation of graphics cards could be wickedly fast, something every enthusiast likes to see. However, it appears that NVIDIA and AMD may go for a blend of performance and power usage, rather than outright performance, since power use of modern graphics cards is already hitting limits of acceptability. So, does this mean that these new cards will overclock like never before? We will have to wait and see. AMD's cards should be out around the new year and NVIDIA's a couple of months later.
Head on over to DailyTech for a lot more detail on this.
Source:
DailyTech
Head on over to DailyTech for a lot more detail on this.
64 Comments on New 28 nm Graphics Cards To Be 45 Percent Faster And Overclock Like Never Before?
I'm looking forward to AMD's Graphics Core Next, seems a big jump for them :rockout:
45% higher clock rates DOES NOT EQUAL 45% higher performance, there are tons of other factors that go into the final performance of a chip.
That being said this is interesting news, coupled with what I hear about AMD using XDR 2 memory this is shaping up to be an interesting graphics card generation.
Architecture is most important.:confused:
For example if I overclock my 6870 to 1ghz core I get a %10 performance increase ( 11% overclock)
And running it at 1100 core nets me a 20% improvement ( unfortunately it's only noticable in source games but heh!)
Not quite perfect scaling but close it it.
(i hope its true)
and its like asus mars just with 6 pins power
They would of done very very well if they had!
Silly AMD.
Making claims liek that need to be verified with an offical statement, not by an anonymous source. If there is no officail source, then it is jsut rumour, and shoud be ignored. I dunno why I still gotta say that...
I am certain at least with my GTX460 and GTX570 voer the last 12 months, that clocked up 15% across the core and memory, they both perform damn close to 15% faster overall, it is generally within the margin of error unless the game engine is outdated or say has a CPU limitation for example.Very very much so but what they are aiming to hit here, as indeed with many past generations, is improve the arch greatly while also increasing clock speeds across the board. lets have a look at this done historically.
8800GTX 575 mhz 128 sp's
GTX280 607 mhz 240 sp's
GTX480 702 mhz 480 sp's
HD4870 750 mhz 800 sp's
HD5870 850 mhz 1600 sp's
HD6970 880 nhz 1536 sp's (VLIW4)
naturally there are examples inbetween too but for instance GTX280 to 285, all the die shrink accomplished was higher clocks/lower power, but having said that they did that same thing with the 480 to 580 without even jumping to a different node.
in short from 40 to 28nm is a big jump alright, and both camps are launching new arch on it, so in all likelyhood the GPU's themselves will be considerably more powerfull, but the other side of the same coin is that we could/should still see higher clockspeeds emerge across the board of low to high end, AMD or Nvidia.
it's not as if they're haven't been any Awful games/games with bad coding that were aimed at the PC