Well well, this weekend I gave it a try with the second socket.
All the time, since I wanted ECC for my 24/7 zpool machine, and finally got an affordable Xeon-EP board, that second socket was sitting there empty. And Haswell Xeons are now 10 years old and dead cheap. (The boards are not, they are still difficult to get - because they are still good.)
Observation now: the twin ops is inefficient regarding energy. The power consumption increases by some 25 Watt, just for idling. (The chip itself would then report 9 Watt.)
And there is the extra problem of getting all the energy, now in calories, off the board and out of the machine. I'm still using the high-tower case that came with a 80486 - because ATX form-factor hasn't changed - but fitting (and also cooling) dual Xeon and 20 disks is getting increasingly difficult.
I didn't get as far as to look into compute performance, because that second chip just died, too soon and for unknown reasons.
OT, you might already know this... but I hope you do not intend to put the (external) packet filter (often loosely called "firewall") onto the same physical machine than other services. Don't do that. It must be on it's own physical machine, solely for that purpose, and no other services on that host. In contrast, you can merge the internal PF onto the same machine as a DMZ host (with gateway services (proxy, load balancer, mail etc.) jailed or in VMs), but not the external one.
That issue doesn't apply to me, because I run all nodes internet-facing - just like in the good old times before IP4-exhaustion, when every coffee-maker on the campus was globally reachable. With IP6 we can do that again - and so each of my nodes has it's own firewall, and it's own DMZ.
The background is, simply, I am not afraid. (We all know from the stock exchange that greed and fear relate to each other and are the main motivation of business people. But I am not greedy, so I do not need to fear.)
This is actually a very important observation. Amateur computers, built using desk side cases, cheap fans (often only 1 or 2 fans), and whatever power supply is on sale at NewEgg this week, tend to be somewhat unreliable. Enterprise-class servers may have less CPU power or slower RAM, but they may up for it by having about 10 or 12 fans (each individually pretty small)
Yeah, I was looking for these fans.
Common understanding is, with a strong CPU (>100 Watt) you need kind of a monstrous heat dissipator (and the market is full with all kinds of these). But, 1U blade with two high-powered Xeon does not have or need any such thing, and they would not fit anyway.
So what's the secret? Must be these fans, they spin 14'000 rpm and cost about 50$ each.
Open question at this point: how noisy would that be? I might assume: noisy.
In contrast, a modern "super silent" fan does 1400 rpm, is super silent and moves not really much - practically nothing when run thru an air filter. I for my part tried to replace that with a modest 8 Watt, 3200 rpm model. That one was loud, so I added an inline thermosense regulator. Then it was wonderfully silent - because the 8 Watt had rightaway killed the regulator. Anyway, I am getting nearer to the optimal solution.
Considering: up to about Pentium-2 there was no real difference between server and desktop except the case - it was all just computer. Yes, there were multicore platforms, but these were very special and rare constructs for high-performance demands.
Since then, things have diverged. And the desktop market has become a cult of fetish believers.
Those monstrous heat dissipators are fetishes. (BTW: for what technical purpose does one need LED lights inside a CPU cooler? None? So then why are they reported first-place, instead of any useful info?)
Those fetishes are sold via the narrative of "technological advance" - which is bullshit because "technological advance" here does
not mean technological advance. 20 years ago, the performance of the machines would double every year - so there was indeed continuous technological advance. Now it doubles maybe every ten years, and the actual technological advance in the desktop field is negligible..
"Technological advance" in this sense is now a believe system: by constant repetition people are made to believe in it without thorough evidence, in order to make them continue to buy new stuff every year.
The personal computer market has become much similar to the cars market: there has to be a new model every year, but there might not be much of an improvement, and it might also just introduce new bugs. The main feature of the new model is that it is shiny and new.
But, with cars there is indeed an issue that they get more and more unreliable with increasing age, due to wear and tear. This is by no means true with computers (but still people are made to believe it would be).
All of this is then part of a larger trend in our society: we have done away with traditional religion, and, nevertheless depending on something to believe in, have made our flawed understanding of technology the new belief. And we have created new priests whom we worship: the big capitalist oligarchs (Amazon, Google, Intel, Facebook etc), who in fact rule this planet now, much like the churches of old.
Originally it was
us who assembled these servers. It was
us who made things work. And it was not considered a science, but a craft, similar to plumbing. Now the gadgets come ready-made from Google and Amazon, and they are not only built, but also managed by Google and Amazon, which is important because Google and Amazon as your new churches need to know every step you take.
