Ordered the Asus X99 deluxe with the 5930k, and 32 GB RAM. Seems the faster RAM is still expensive and wasn't any available for shipping currently, so I got rather cheap 2133 mhz and will do with that for now and get 64 GB ~3Ghz once more vendors are available and prices drop. Already have 2 x 512GB Samsung 840 SSDs + a couple others.

Ordered a cheap PCI-express SSDs too.. that probably won't be of much use.. but gonna try and see if it can work with PCI passthrough in VirtualBox or VMWare to install a virtual machine on.

Yeah, i ordered the X99-Deluxe and 5930k as well. Got 4x8 GB of G.SKILL 2400 memory; I want 4 slots open for later (probably to drop in a 4x16 kit in 2016 for a total of 96). G.SKILL isn't my favorite, as I've had dead modules from them in the past (more than once). Their customer service has always been really good about getting it replaced though. Still too bad that Corsair isn't making 8 GB modules. Would've put a few extra bucks in for that.

The Asus was one of the few Thunderbolt ready boards, which surprised me. I thought Intel's new flagship platform would be all over the Thunderbolt thing, but apparently not? I don't get it. There was a big surge of Thunderbolt in Z87, and then it kinda died off for some reason. In any case I want it, so it was down to Asus and Gigabyte, and Gigabyte's board had a lot of gamer/overclocking garbage.

As far as the PCIe/M.2 SSDs... there aren't really a lot of options, and they're hideously expensive from what I've seen. I'm just going to get the 850 Pro for the main drive, and the array of WD Blacks or similar. The VM pass-through idea is cool though and I'd like to hear if that works (and if it's preferable to raw access to a SATA drive).

Choosing a case might be the hardest part. I went to the local Microcenter to check out some cases today evening and good god do most of them suck. Right now the Corsair C70 is the easy front runner.

As far as actual performance for dollar... Z97 and 4770K is almost certainly the better choice for nearly everybody. I had misgivings about doing the X99 thing several times during all this. Ultimately though, I'm doing enough production-type work, notably 4K video, that these more aggressive platforms are worth it. A traditional "recommended" system for that sort of thing is dual Xeons. Go price out the CPUs and motherboard for that. Six cores and a 3.5 clock is just about perfect for me. Plus the X series platforms are meant to last a good bit longer, and I'm anticipating being able to make effective, useful upgrades through 2018.

I'm always torn -- On one hand I would love to have something like that, so its one of those "when I'm wealthy" dreams; but on the other, it seems like a lot of the things I might care to do with 8 cores, if it really can be made parallel that well, might be equally well-served by a cluster of less expensive machines. If it were just the difference in processor cost, it'd probably be a win for a single, "bigger" system, but taking into account the new chipset and high cost of DDR4 today, becoming an early adopter is a hard sell.

I upgraded to a 4770k last year though, so I'm not hurting terribly for an upgrade. Maybe nextime they do a major refresh of the -E line, if I find myself without any more-pressing purchases.

You're limiting me to the consumer category?

If I'm going to dream, I'm going to dream big.

When compiling my biggest bottleneck is the pegged CPU and cache, not so much main memory transfer speeds.

Forget the "i" family of consumer hardware, go for the much more expensive "e" family designed for back-end computing. I'll go for a multi-CPU e7 rather than the i7, and all the memory the company is willing to pay for. I don't have time to wait for the CPU to compile code... The Xeon family of server-class processors have tons more on-die cache and are built for multiple cpus per box.

I mean really, why would you even?

[table][tr][td]Processor[/td][td]Cores[/td][td]Frequency[/td][td]L3 Cache[/td][td]Release[/td][td]Cost[/td][/tr][tr][td]Xeon E7-4890 v2[/td][td]15[/td][td]2.8 GHz[/td][td]37.5 MB[/td][td]February[/td][td]$6619[/td][/tr][tr][td]Core i7-5960X[/td][td]8[/td][td]3 GHz[/td][td]20 MB[/td][td]August[/td][td]$999[/td][/tr][/table]

Not listed are that several of CPU's internal structures, prediction tables, and other goodies are similarly bigger.

Right now today you can stick 8 of the Xeons in a box with Intel's high speed interconnect for your own personal 120-node computing cluster, for the low price of just $52,952 MRSP. (Motherboard and terabytes of RAM chips for your virtual drives are extra.)

Or going further, the soon-expected 18 core Xeon e7 which would be the paired equivalent of the announced e5. Based on the trends and the values from the announced eighteen core E5, it will probably be something like this:

[table][tr][td]Xeon e7-?? (49xx)[/td][td]18[/td][td]3.x[/td][td]50 MB[/td][td]Suspected Q4 2014 or Q1 2015[/td][td]$Lots[/td][/tr][/table]

Quite likely it will also use 4xDDR4-2133. With 8 in the box, that's 144 compute nodes with a combined 400MB L3 cache.

So as long as I'm going into fantasy land, I'll pick that little supercomputer and use it as a Christmas present to myself.

Hook me up with 144 real processors, a high speed interconnect, a few terabytes of DDR4-2133 memory for a virtual disk faster than any SSD, enough L3 cache to keep the entire dependency source tree and the compiler and tools in CPU cache. Incredibuild and other distributed compiler systems can suck it; I'll crank up the parallel builds option and build locally.

Also, hook me up with some serious cooling, I'm guessing it will make more heat than the PlayStation devkit space heaters.

I guess I'm just anchored to some semblance of personal practicality -- these midwestern roots of mine run deep.

Actually, though, I'm kind of interested to see what will come out of the Xeon-socketted Knights Landing -- 72 silvermont cores running at (reportedly) ~2.4Ghz, and with 4-8 GB of fast RAM on-package, each with 512-bit-wide vector instructions. Given the combined floating-point throughput (2Tflops single-precision, 1 TFlop double-precision, per socket) and x86 compatibility, I've kind of wondered if they won't be ideal for hosting a game's build system -- they'd handle both compiling and media transformations really well.

I guess I'm just anchored to some semblance of personal practicality -- these midwestern roots of mine run deep.

It's also the difference between 'work' and 'play' machines. These hardware costs are trivial when balanced against how much of my time is spent waiting for builds, for example. I mean, I'm waiting for a build as I write this. It's also an experiment for me, finding out just how much hardware I can throw at the problem of day-to-day downtimes. The other half of the experiment is finding out whether going for a higher class platform is worth it. Which is to say, X99 doesn't really make sense above Z97 for normal use except for a few rather specific tasks (which I do happen to need). But does it lead to a longer viability for the computer and even out cost-wise?

Still don't know the answers, but it's something I'm eager to find out. If it were just a home/gaming machine, Z97 would've made far more sense. But if the gigantic CPU and storage cut a good chunk out of my compile times, that along will be enough to make me quite happy.

I'm busy finalizing some misc specs on the machine. Looks like it'll be Samsung 850 Pro boot drive (512 GB), Crucial MX100 512 GB x2 (RAID 0) data drive, and WD Black 4TB x3 or x4 RAID 0 large files drive. VS had damn well better be blazing with the ridiculous amounts of processing power I am throwing at it.

Oh, no -- totally get it. I was mostly responding to Frob. If you can use the extra 2 cores, that hex-core Haswell-E is totally worthwhile, IMO -- actually, I might even suggest you just go whole-hog an get the 8-core, simply because once you've chosen the Haswell-E, you're already fronting all the platform expense. Or, maybe you're planning to upgrade when they refresh (Haswell-E platforms, like Xeon, aim for twice the socket longevity of the consumer platforms) down the line.

If was building a serious workstation I might be tempted to swing the an -E platform -- I considered it last year when I build my last desktop PC even priced a system out, but it wasn't worthwhile, given that Haswell-E was barely in sight (and now we see it, nearly a year later) and the old platform was on its last legs. So I built a nice, small ITX gaming rig instead -- I do most of my serious work on my laptop anyhow (no match for a high-end desktop, but quite tricked out for a laptop).

Extremly dissapointed with the new Enthusiast Platform. Only six core CPUs for reasonable prices, the only octa core is the extreme edition.

Now I have my ageing i7 970 that is still doing great today, already has six cores and only was 500$ three years ago. I am keeping my cash until I get at least 100% more multithreaded performance on a reasonably priced (less than 600$) CPU, especially as I will have to practically get a new PC by that time (DDR4 Memory instead of DDR3 to get the most memory performance, new mainboard needed, my GTX 580 graphics card is also ageing (and I am running out of VRAM for certain 3D applications)).

Granted, paying 500$ more for the CPU is not THAT bad when compared to the total cost of such a workstation. Still, 3 years in and the Enthusiast platform has not moved much, Intel is really keeping the nice CPUs back for the Xeons because AMD is just dropping the ball constantly when it comes to powerful CPUs.... I will not bend over and thank Intel for basically ripping me off at the same time, not as long as my 970 is still going strong, and I don't really HAVE to upgrade.

Maybe next year Broadwell-E will be more interesting. Maybe we have to wait until AMD is finally lifting the cover off the mysterious new CPU architecture they claim to be working on for 2016, and hope they don't botch it up again and really give Intel a good run for their money this time (though I will not hold my breath... Please, no more crappy module architecture, AMD! )