Sod it. I've bought one and will plug it in tomorrow.
I've had no issues at all with it to this point but what's the harm?
On a slightly unleated note, I'm excited to start looking out for the next upgrade. I change the CPU when it halves my render times. Since I've been rendering, I've gone i7 950, 6700k, 1800x and now 3950x.
I've stayed away from HEDT but Threadripper does look appealing. Fun times for this hobby.
I wonder if AMD will move AM4 class up a slot with Zen3 or not? Would 4 chiplets and 32C/64T be too much to make sense for thermal density and memory bandwidth on 7nm+? Will 16C be the floating edge for consumer socket for a while to come?
One of the issues facing these extremely dense processes is hotspotting and the difficulty in cooling such tiny little red hot surface areas when you want to ramp clocks up through the 4Ghz range towards 5Ghz. Efficiency was already on an observably downward spiral as higher clocks were chased in various other process tech nodes, but with 7nm it appears like an even more complicated issue.
Maybe hybrid clocks could be a solution? We already have separate cores to peak at higher rates, but what about :
2 chiplets of 'efficient' 7nm+ 8C/16T type, capped at ~4Ghz, weighted towards one half of the package.
1 intentionally more 'diffuse' (less dense/prone to hotspotting) chiplet of 4C/4T in which can ramp up to perhaps 5Ghz or beyond for turbo. Perhaps have the packaging itself focused heavily on horizontal heat dissipation, copper shimmed or however they can help with areal heat distribution. This 'max IPC' side could occupy the opposite half of the package vs the more standard type described above.
Sort of thinking out loud there. I'm not sure what to expect of Zen3. There is no immense leap in process tech waiting for the release, so if it's just something like a couple more cores per SKU, 100-200Mhz increase in clock potential for extremely short bursts, etc, or perhaps even less notable than these possibilities, then we may know we're in for a fresh round of stagnation. I still haven't gotten rid of my 2920 Threadripper, but in regular desktop use, more cores aren't really a worthwhile endeavor for your regular Joe. My 3700X unit feels as fast in most things, and faster in some (games/etc). This isn't a knock on TR, it's fantastic for people that CAN and DO need and use the extra cores for specific tasks. General use though, it's approaching irrelevance. 3600 is as good as 3950X for running Windows, Office, and anything short of bleeding edge GPU gaming.
A 5Ghz+ all-core Zen3 8C SKU though? If IPC moved even a hair up, the combo of clock and IPC would make it pretty epic and enough to speed up performance across the board, not just the most highly distributable/threaded load examples like rendering, folding, etc.
Opportunity cost is interesting to me. Someone who renders something once a week or once a month may see very little benefits from going from even moderate performance to double it. Say they're waiting 5 minutes to render a handful of times a month at most, for their occasional social media project for their kids or work etc. Getting a 50% increase would cut that 5 minute wait down to something around 3 minutes and 30 seconds. Still enough time they might be getting up to stretch their legs, grab a cup of coffee, take a leak, whatever. And not common enough to make a difference to their day/week/year.
Take a different person who is doing this kind of thing 40-60 hours a week, and suddenly they can either make a LOT more projects fit into their schedule, or have more time to polish/post-produce the work, or simply accomplish the same amount of work and have more time for themselves, their family, their health, etc. Instead of basically zero gains in practical terms, you're talking ENORMOUS gains, all from the same generational step in tech.
All about perspective.
Oh nice thread loading btw on post 14%. Very nice to see it hard at work!
Facebook
Twitter