• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Maxwell is a 28nm chip

Kaapstad

Kaapstad

Kaapstad

Man of Honour
Joined
21 May 2012
Posts
31,953
Location
Dalek flagship
GM107 that is

Geforce Maxwell GM107, Nvidia’s upcoming entry level chip, is a 28nm product. This was not hard to guess but now we can confirm it. There will be a 20nm Maxwell line, but this will naturally come at later date but we don’t have any timeframe for these chips.

The GM 107 will be the first of many Maxwell chips and we have been hearing that Maxwell has a much better performance per watt ratio compared to the current Kepler architecture. You can imagine that it is destined to do well in notebooks.

The follow up chip, the one that addresses the mainstream Geforce 660, 760 market, or its successor will be the one to watch, as this architecture aims to produce a worthy successor to the current line-up.

Nvidia operates at its best when it is winning the market battle, and its leadership was shaken by the immense performance and aggressive prices offered by AMD’s Hawaii generation launched last year. Nvidia is still doing quite well in the GPU market and it has a huge market share.
Click to expand...

http://www.fudzilla.com/home/item/33657-maxwell-is-a-28nm-chip

These are not the GPU chips you want.

Nothing interesting here.

Move along now.
 
Nothing_5d865f_2837850_zps971f28eb.jpg
 
******* 28nm :mad:

Oh well I can at least look at it as money saved until the 20nm chips finally, preferably sometime this decade, arrive.
 
The GM 107 will be the first of many Maxwell chips and we have been hearing that Maxwell has a much better performance per watt ratio compared to the current Kepler architecture. You can imagine that it is destined to do well in notebooks.
Click to expand...

Doesn't really make sense as a good part of the performance per watt gain of Maxwell comes from the 20nm process over using the 28nm process.
 
I thought the first gen of Maxwell was on 28nm and then the next gen of cards was on 20nm? I'm sure I read this some where?

NVM just read the post properly, yea it's supposed to be at the end of the year and will replace cards such as the 780 and 780Ti
 
pgi947 said:
******* 28nm :mad:

Oh well I can at least look at it as money saved until the 20nm chips finally, preferably sometime this decade, arrive.
Click to expand...

I think my next upgrade will be Haswell-E and some 20nm GPUs this time next year after the dust has settled.
 
Rroff said:
Doesn't really make sense as a good part of the performance per watt gain of Maxwell comes from the 20nm process over using the 28nm process.
Click to expand...

Not especially, I mean to some degree, sure, but there is plenty you can do with the same design. These aren't performance chips, if they used the same process and exact design of GK110, it wouldn't be very power efficient, but if they use a different process, lower target clocks, use different transistor design aimed at lower power rather than clock speed you can absolutely create something noticeably more power efficient than what is available. Doesn't mean they'll reduce it loads, but some is more than possible. I think Tegra K1 will not quite meet Nvidia's promises(like all their other Tegra's) but simply using a newer tweaked A15 core, all their previous 28nm experience and using different design goals they will have improved performance/w fairly significantly over their other 28nm chips.

Case in point, the 290x uses what, 15-20% more power than a 7970 but is around 40% faster. Some of that architectural, some will be down to tweaking for the process based on their previous experience.

Fact is, they aren't sticking a GM107 in tablets, and not mostly not desktops, but in 35-70W laptops. It's not a gaming part and it's not a mobile part, it's a there to fill up the numbers part. At this stage even Nvidia have said the low end barely matters to them, because APU's are making it go away. There is little profit in brilliant low end parts, certainly wouldn't want to take up precious wafer starts on 20nm to be wasted on parts that aren't seeking ultra low power usage, nor performance.
 
WalderX said:
I'm not surprised at this due to all the issues they are having getting 20nm fabrication working in a cost effective way.

This article makes for a very interesting read and explains some of the problems with 20nm:

http://www.pcper.com/reviews/Editor...ation-20-nm-and-Beyond/20-nm-and-Below?page=1
Click to expand...

That is also why I find it strange that Maxwell would appear on 28nm - it was designed from the ground up for 20nm with TSMC making the design kits available early on and as the article illustrates the lithography and layout is very different to the same design implemented on 28nm meaning they'd have to go to a whole load of extra work implementing and debugging for 2 seperate processes with no direct shrink path available though I guess it might be possible to port a low spec design directly back in an inefficient manner.
 
Can't see myself upgrading the CPU in a hurry, I don't even really need the OC I have on it atm though it does add a little tiny bit to the fps in some games.
 
You must log in or register to reply here.
Back
Top Bottom