Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Yeah ticks have been always been 12 months - if we consider Clovertown to Penryn and Nehalem to Westmere - but with the tock cycle getting longer and Intel's dominance I certainly won't be expecting Ivy Bridge EP within 12 months. 15-20 months before an update wouldn't be surprising in the least to me.

I'm missing the logic behind the tock cycle getting longer having impact outside of the tock cycle. If Haswell needs to ship later then it can just slide out. Sleeping on Ivy Bridge isn't going to shorten that cycle. The only thing that would 'slide out' Ivy Bridge is if they pulled some Haswell features forward and make the Ivy Bridge a hodge-podge tick+tock cycle. That's a dubious move.


It would surprise me (12 to 14 seems more likely). One of the primary reasons Intel has AMD so off balance is that 12 month regularity. If Intel throws that out and goes to extended development on both parts of the cycle, they are basically letting AMD get back into the game on AMD terms. I extremely doubt Intel is going to do that.

Right now what keeps happening is every couple of cycles AMD tries to 'swing for the fences" and hit a home run ball (usually a combined shrink+arch cycle). In part because there isn't enough time to 'catch' up.

I suspect Intel is going to have to let real, working engineering samples do the talking to the NDA partners at this point before putting voodoo dates down on paper. Ironically, Intel has joined AMD in the "yeah right, you are really going to hit those dates" club. That is not really a club they would want a long term membership in. Getting out is easy. Just stop rushing to write down dates before have enough info to write down a creditable date. If Intel posts a Q1 2013 roadmap date for Ivy Bridge EP in June the world won't end.



Your last sentence basically sums it all up sadly - Intel wouldn't have waited 2 years if Bulldozer Opterons were a good alternative.

Bulldozer would have been a much more creditable alternative if AMD had hit the targeted clock rate and got the OS scheduler updates out on launch. Intel pissing around for an extra 6 months would give them most of the time they need to get that corrected. The OS scheduler updates are already out and if give AMD 12 months to shrink and clock bump they probably will.
 
Anyone chip designer should be a bit apprehensive about more than two cores on per channel at these clock rates. It isn't going to scale very well to just keep piling up cores behind a memory choke point.

To keep the sockets smaller and the designs less complex the 4 core cap for mainstream CPU packages is probably going to be around for a extended period of time.
At the time Intel was looking at DDR3-1066 on the triple channel LGA 1366 and DDR3-1333 on the dual channel LGA 1156. They did mention the issue of over saturating the available memory bandwidth on their mainstream platform. We are looking DDR3-1866/2133 becoming standard soon enough.

Given the die size and lack of benefits for the price targets though, Intel probably will not make anything more than a quad core mainstream until 14nm.

Bulldozer would have been a much more creditable alternative if AMD had hit the targeted clock rate and got the OS scheduler updates out on launch. Intel pissing around for an extra 6 months would give them most of the time they need to get that corrected. The OS scheduler updates are already out and if give AMD 12 months to shrink and clock bump they probably will.
Bulldozer was already delayed in order to hit clock targets. AMD wanted something to show after ages on 45nm Phenom II and the 32nm Bulldozer die is a step down to 315 mm^2 down from 346 mm^2 on Thuban. Trinity utilizing a L3 cacheless Piledriver module is rumored to show remarked improvements over the Bulldozer core. Not that the available die shots comparing the modules look all that different.

The 2012 28nm low power APU appears to have been dropped either due to the new management team or to allocate more wafers to the much higher ASP HD 7000 line up. Atom has seen little change since its launch beyond lower power and lower prices. The performance is still based on the same ancient architecture from launch just on a new node and somewhat modified package.

That leaves 28nm CPUs and APUs for 2013 with Intel having about an 18 month process lead. AMD will eventually have all lines on 28nm but they can utilize their current 40nm bulk and 32nm lines competitively.
 

Attachments

  • 11.jpg
    11.jpg
    546.8 KB · Views: 80
  • 02.jpg
    02.jpg
    166.5 KB · Views: 5,015
My Mac Pro has never overheated and I don't have issues with my ATi video cards either. The article doesn't make any sense.

Not that it's best to go completely off of anecdotal evidence, but I had to replace my ATI 5870 a few months back--the fan would wind up like crazy making a terrible racket the whole time my computer was on. I also get strange beige screens at times when running windows--could be driver problems.

Anyways...this is my single experience, as is yours, so it's hard to know how many problems that apple has seen across the board with ATI cards.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.