When it comes to chipsets for Broadwell, Intel is going to take a page out of AMD’s book for Trinity and Piledriver. This is the long way of saying the Broadwell generation isn’t going to have any new chipsets.
Technically speaking, the reasoning for doing this is quite sound. With added integration, the functions that a chipset provides are essentially limited to high voltage services that are pad bound and not very performance sensitive. That in turn generally means pad bound and eminently doable on a -1 or -2 process. For generations, Intel used it’s recently idled tools from the last shrink to make chipsets, and even now that makes sense.
If you look at the services that a chipset provides in any modern iteration of the breed, you see a bunch of truly commoditized functional blocks. Sound device, USBx, older PCI/PCIe lanes, a boot ROM, and drive controllers. Here and there you may find something interesting like a glorified video out splitter, but that is the exception to the rule. There really isn’t anything in a modern chipset that you can’t easily license on the open market from dozens of IP vendors.
With Haswell, Intel is putting in all the interfaces that are coming in the foreseeable future. In the Broadwell time frame what changes? Main memory type? PCIe? SATA? USB? All are fixed. How does the CPU talk to the chipset? Glorified PCIe3 overlays that Intel makes up curious marketing names for before they lock out competition with dubious patent methods. Intel is more than smart enough to realize that changing this for no reason is both counterproductive and costly. This is why (insert marketing name of the moment here) is pretty static across generations. The only reason it would change is when there is an update to the base PCIe spec, and that is not happening for quite a while.
So in the end, Broadwell isn’t going to get any new chipsets. Like AMD has shown, if there is no reason to update the damn thing, why waste the effort? Please note that this does not preclude making SKU changes like blowing a different set of fuses to enable/disable functionality that is there or upping clock by some insignificant number. Those are a time honored and sacrosanct way to milk the consumer, and profit is the name of this game, right? If you want more on how this will play out, look to AMD’s A75/A85 chipset ‘update’. Likewise with Broadwell, all you need to do is kick back and place your bet on what Intel will call it, not on what it will do.S|A
Latest posts by Charlie Demerjian (see all)
- More on Intel’s 10nm process problems - Sep 17, 2018
- Intel puts out another 14nm 2020 server platform - Sep 11, 2018
- Why Can’t Intel Supply Enough 14nm Xeons? - Sep 10, 2018
- Intel can’t supply 14nm Xeons, HPE directly recommends AMD Epyc - Sep 7, 2018
- AMD reintroduces the Athlon name with two CPUs - Sep 6, 2018