Some of us are fine with 4.4-4.6 GHz quad core oc, but the stupid paste is sky rocketing temps. We just want the soldered cap back, that's all.
Well, you can't have it. If you use the soldered heatspreader on a die as small as the mainstream parts, the edges of the die crack under the repeated thermal shocks of CPU usage and you no longer have a working CPU at all after a while. No, Intel isn't dicking you over by putting in paste - they're just doing what they absolutely have to in order to ship these thing in actual quantities. It's particularly bad with long, narrow cores, which is what the mainstream chips are, unlike the HEDT parts which have been a fair bit more square in comparison.
If you want, I'll send you the published papers I've read the subject, but it's just a lot of boring, dreary reading on pretty deep material science analysis and statistics...
Good to see the HEDT variant coming out and I hope that it leads to better gaming performance.
I don't like the way the KL version is being deliberately hamstrung compared to Skylake, leading to a dilemma. This seems like a cynical move to sell the older platform, which should be obsolete by now and especially a year down the line.
I believe that's more a factor of incomplete validation rather than anything else. The bigger your core, with more features, the longer in takes to validate that everything works properly and all within spec etc. By the looks of it, KBL-X is validated up to basically mainstream desktop spec and they're willing to ship that out already with the rest of the die disabled. As to why they can't line up their validation to more or less end on the same date (because you're almost certainly wondering that), it has to do with the core computation/execution core/IP (pun not intended) being designed for all platforms (from Core M to 8+ socket Xeon E7s) at once, and starting validation essentially at the same time.
The other possibility is that they're re-packing the KBL-H (mainstream) parts into the new socket, and I suspect that that option won't take the soldered IHS route.
As everybody so far is only interested in number of cores and Hz I'll break the norm and say.
Give me a break, only 8 SATA ports? Already running 10 on X99 together with 20 SAS ports on RAID card+expander and they're ready to chop 2 ports with new platform. WTH? Expected improvement (ditching single SATA ports for SFF-8643 connectors) not moving backwards.
Performance gain negligible, flexibility minimal. Like I've expected, X99 is probably my last PC platform. Zen has to deliver big, massive, freaking time to even tempt me.
That's just the HEDT/2P platform split hapenning. Socket R is no longer the mainstream dual-CPU socket - that goes to the brand new Socket P 2016 (LGA3647) socket that's already been shipping with the Knights Landing Xeon Phi platform; and will extend to the E5-2xxx/4xxx and E7 series Xeons when the Purley platform launches sometime next year.
You want server-grade IO (like Intel's high-end C606 and C608's 8 SAS via the "SCU" + 6 SATA)? Then pay up for server-grade IO, cause last I checked machines with lots of disks are incredibly rare in desktop systems.
PS: don't expect Zen to be much better, current talk is 8 SATA (no SAS) per big CPU (SoC architecture) with no info on whether the SATA controller on the second CPU in a 2P system can be enabled and used.
