Updated: We were wrong about the memory, it is Unbuffered ECC.
Yes, there are 18 full machines here, sharing nothing other than power supplies, and those have 2 + 1 redundancy. Each node has a ‘next-generation Xeon’, so called because Intel’s naming scheme is so broken that even it’s execs can’t use it with a straight face, 32GB of unbuffered memory, 2 GbE ports, and 2 2.5″ HDs. The Ethernet switches are on top of the ports per module. It all looks like this.
Lots and lots of modules in a box
While this may sound good, there is one major problem, the chipset. Intel’s ‘next-generation Xeon’ is nothing more than a desktop Sandy Bridge chip with the price tag multiplied by an AMLOCT (Arbitrary Monopoly Lack Of Competition Tax). It isn’t really a server chip, it doesn’t even support ECC, (Update: It does support ECC, just not buffered ECC) but it hits several marketing checkboxes, so that wins. This may be fine for some customers, but I can’t see any serious deployments until Intel provides an actual server part in this space. That will be Q3.
Once that happens, Tyan looks to have a killer chassis on their hands, I can’t think of anything like it on the market that doesn’t have the ‘blade tax’. The FM65-B5511 is going to start a trend, by this time next year everyone will have something similar. For those that don’t need ECC, this is a great start, and that gap will be filled in soon enough.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