Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
Heat and temperature are different things. A 65W Haswell is pretty much 65W no matter what tim you have on it(temp can effect efficiency but not hugely at this level). A chip that they hoped/targeted to run at 65W running hot in temp is no big deal, but running 88W is a huge deal, that means they missed the power target for the specific clocks they aimed at for the 65W chip by over 30%.
Anyone should have been able to see proper higher end desktop Broadwells weren't just going to be all rosey, realistically they should have launched end of last year and were supposed to, with Skylake Q3/q4 this year. It was already delayed a year and they only released very low end chips which seemingly all need more voltage at the same speed the previous gen was offering and slim performance advantages as a result.
Intel's 14nm isn't doing great although it could be that they made some issues with the design of Broadwell and Skylake will fix a lot of them, think 480gtx vs 580gtx kind of situation. It was looking a long while ago that Broadwell was having serious issues and how useful it would be on desktop was extremely questionable. If the 95W version is scrapped and the 65W version has moved up to the ~90W range, then it means even less useful. Hopefully they didn't balls up Skylake as well and that is relatively on time and it works well at 14nm.
The only problem to a degree is, unlike TSMC and say Nvidia getting the 480gtx wrong, Intel produce their own chips, it's significantly less likely for them to mess up their own chip on their own process, which would suggest a bigger chance it was more their 14nm process that just isn't working well... which might not bode well for Skylake.
By June/July if we start hearing of Skylake delays as well then I think it will be a process problem and one they aren't close to getting under control.
Anyone should have been able to see proper higher end desktop Broadwells weren't just going to be all rosey, realistically they should have launched end of last year and were supposed to, with Skylake Q3/q4 this year. It was already delayed a year and they only released very low end chips which seemingly all need more voltage at the same speed the previous gen was offering and slim performance advantages as a result.
Intel's 14nm isn't doing great although it could be that they made some issues with the design of Broadwell and Skylake will fix a lot of them, think 480gtx vs 580gtx kind of situation. It was looking a long while ago that Broadwell was having serious issues and how useful it would be on desktop was extremely questionable. If the 95W version is scrapped and the 65W version has moved up to the ~90W range, then it means even less useful. Hopefully they didn't balls up Skylake as well and that is relatively on time and it works well at 14nm.
The only problem to a degree is, unlike TSMC and say Nvidia getting the 480gtx wrong, Intel produce their own chips, it's significantly less likely for them to mess up their own chip on their own process, which would suggest a bigger chance it was more their 14nm process that just isn't working well... which might not bode well for Skylake.
By June/July if we start hearing of Skylake delays as well then I think it will be a process problem and one they aren't close to getting under control.