Where are the Xeon-D 1548's?

[sth]

looking to upgrade my C2758s sometime soon.

 

[Dennis.A]

They're all at Facebook.

 

[Patrick]

They're all at Facebook.

Facebook uses a different Xeon D that is not the same as what you get at retail.

 

[sth]

thx guys, good to know. Any clue as to a potential ETA for the public? this year?

 

[Patrick]

thx guys, good to know. Any clue as to a potential ETA for the public? this year?

It is less of a chip availability issue and more of an issue where motherboard/ system manufacturers see enough volume to drive using the SoC.

 

[sth]

so what would be the optimal choice for a replacement board right now, I would have thought the "network optimised" 1528 but it appears it doesn't have future proofing QAT built in and needs a Coleto board anyway in the future. I can't understand why Netgate sell the 1540 as their xeon-d variant over a network optimised version unless theres a reason Im missing. thanks in adv.

 

[Patrick]

Better single threaded performance. The Broadwell-DE core is a much faster core plus you have built-in 10GbE. If you look at the Atom chip designs they only have L1/ L2 cache not L3.

The new QAT driver for FreeBSD I will guess can help with QAT first. I do not think the QAT accelerator on the C2758 can handle 10Gbps. (In the airport and I do not remember the exact spec.) C2758 is killer for 1GbE, maybe 2GbE but not 10GbE.

Interestingly enough, I have both a D-1537 and D-1557 platform that just finished getting benchmarked. The D-1557 completes our Linux compile benchmark about as fast as dual X5650's.

TBH a D-1518 is going to be an upgrade for most applications over the C2758.

 

[mstone]

QAT isn't "future proofing", it's a bet that Intel will keep developing that generation of crypto accelerator. The history of crypto accelerators suggest that's a bet with long odds.

 

[Patrick]

@mstone - that is a good point. We noted in the recent QAT benchmarks that the Rangeley accelerator is QAT 1.5 while the Coleto Creek accelerators are QAT 1.6. Those versions are not completely compatible.

 

[sth]

@Patrick Next time you're chatting with the guys at Supermicro, please feel free to let them know I'd love to see some more boards with onboard SFP+ (currently limited to 1508, 1518, 1537 & 1587's)!

 

[mstone]

Or how about including both copper and sfp interfaces?

 

[Patrick]

@Patrick Next time you're chatting with the guys at Supermicro, please feel free to let them know I'd love to see some more boards with onboard SFP+ (currently limited to 1508, 1518, 1537 & 1587's)!

The mITX D-1557 we have in the lab has dual SFP+

 

[sth]

Good to know theres more coming then... thx for the info.