Same for me. I can see me hanging onto this 3930K for a couple of years unless Maxwell does cause a bottleneck with a pair of them.
-
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
- Thread starter Kaapstad
- Start date
Same for me. I can see me hanging onto this 3930K for a couple of years unless Maxwell does cause a bottleneck with a pair of them.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
There won't be a low end Maxwell on 20nm anyway, and like hell will anything translate to the finfet designs at 16nm anyway so why bother.
Nvidia barely does low end even when low end was probably 2-3 times the current volume(which is also shrinking by the quarter). Due to costs there is basically no financial incentive to make the low end uber small. Say a 20nm wafer costs double compared to a 28nm wafer, with 30% lower yields, and the die shrink only means you get say 50% more chips per wafer. A chip that cost $15 per core to make, ends up costing around $27-28 or something instead. There just isn't any worth to doing that, which is why normally low end lags significantly behind.
Older process, particularly as people migrate to the new ones, have so much availability the cost goes down, yields are up. You can't make a theoretical GM110(titan successor) on 28nm, you can at 20nm... so where do your dedicate your early small number of wafers with lower yields. Highest profit cores that aren't possible on the older process.
Low end is the worst chip to do on a new process, and due to the lack of performance push, on a design aimed at say 1Ghz clocks, doing 600-700Mhz at lower voltage reduces leakage drastically. Biggest gains on newer processors are at the high end where leakage is worst.
The real question is will AMD/Nvidia bother with 20nm, for a year, when they can ignore it and work straight towards 16nm. Mobile guys, qualcomm/samsung/apple, they sell in a quantity and profit margins that make it a no brainer to do 20nm. Huge graphics parts.... when it's the 16/14nm drop which is bringing most of the power reduction.
Answer is they probably will, but both companies should be focused on 16/14nm and with the power drop being the biggest feature of 20>16nm, I wouldn't be surprised if a 7970/680gtx level card come at 20nm, but the 290/titan sized cores only arrive with 16nm.
My guess would be that with 20nm delays they don't want to leave the R2xx series unanswered (dropping your last gen card prices and releasing a roided up Titan isn't really a proper answer) that long and so are rushing out 28nm cards as an interim measurement.
Associate
- Joined
- 28 Feb 2013
- Posts
- 2,468
- Location
- Birmingham
Lol
Anyway, this isn't that surprising really - when 20nm arrives we'll have plenty of upgrades available on the market hopefully. For now, I'll just get another 780 to last me until 20nm Maxwell hits (or maybe I'll even wait for Volta).