It looks like the server variants of Haswell will have the same number of pins as their *Bridge predecessors, 2011. One slight little problem though is that it is not going to be the same socket, physically or electrically.
SemiAccurate’s sources are saying that Haswell-E/EP should use a 2011 pin socket, but it isn’t that 2011 pin socket, that 2011 is *SO* last year now. The number, it may end up varying by one or two before all is said and done, may be the same, but electrically they are totally different.
This very likely means that the physical socket will at a minimum be keyed differently to prevent expensive ‘whoopsies’. Server folk don’t like them, especially the kind accompanied by even more expensive smoke, and Intel is aware of this, so new physical keys at a minimum.
Since pin counts are dictated by power/ground pins and memory channels, you can read three things in to the 2011 number. First is that the chip will likely stick with four DDR3 channels for the initial launch. Second is that power feeds won’t be radically different from where things are now, even with the on-package voltage regulators. Last is that the QPI and PCIe3 links will be very similar to what you have now, both in count and capabilities.
In the end, socket 2011 is dead, long live the new socket 2011. It is close to the same, but totally different. Parts made for one won’t work in the other no matter how hard you push the chip down, nor will thermal goo applied by the bucketload help much. You have been warned.S|A
Latest posts by Charlie Demerjian (see all)
- SemiAccurate has Skylake-SP die shots, sizes, and more - Jul 21, 2017
- Skylake-SP has a diverged core - Jul 19, 2017
- Intel’s Purley platform architecture is a step forward - Jul 13, 2017
- Intel launches Purley aka metal Xeons - Jul 11, 2017
- AMD’s Epyc has lots of connections - Jun 26, 2017