There are some kernels of truth to this IVR scare, but basing investment advice on such incomplete information is downright criminal. So here's a little inside info about this integrated VR stuff. I'm not involved with IVR technology, but I do have some insight into the Haswell project.
First of all, the primary benefit of an integrated VR is cost savings. It's integrating yet more things that used to be on a motherboard into the chip itself, and that means a lower BOM for OEMs. Not that there aren't other benefits as well, but those other benefits by themselves probably wouldn't be enough to take on the technical challenges (and they are numerous).
Secondly, there were indeed some problems identified with Haswell's integrated VRs when silicon came back over a year ago. You know who found those problems? The team in Israel. I've been told that the folks on that team do not like some aspects of the IVRs on HSW (I don't know if it's the implementation, or the whole concept, or whether they doubt the ROI, or what). One of the benefits of having two world class design teams at your disposal is that they are always pushing each other, always questioning each other, not to mention the friendly competition that always goes on.
Well, the folks in IDC wrote a funky power virus test that showed some holes in the IVR circuit performance on Haswell silicon. This did indeed cause an all-hands-on-deck response, and various circuit improvements that were already done on the Broadwell IVRs were pulled back into the Haswell design, along with some other potential fixes that were identified. I won't go into details on the problem, the impact, or the solution, but continuous improvements were put into each successive stepping, and my third-hand knowledge is that the problem is 99% contained. Meanwhile, Haswell's PRQ schedule marches on unabated.
It's worth noting that in the grand scheme of processor design, there really wasn't anything new here. This was by no means the first all-hands-on-deck emergency on the road to releasing a brand new processor, and it won't be the last. And it's also not the first time that a major problem was 99% solved instead of 100%. I think the reason this one got into the public domain is because it's likely that the next tock project that will be designed by IDC, Skylake, will probably not have IVRs (I don't actually know that, but it seems like a safe bet to me if they don't believe in the technology for whatever reason). Since this is a feature that directly impacts board and system manufacturers, I'm sure that many folks outside Intel asked "Why?". And if the person answering that question has some bias against the technology, there could be some bad-mouthing or doom-saying that slipped into the answer (or maybe that needs to be part of the answer to cushion the blow of removing a feature that is helpful to system makers). If you worry that maybe things like cost or power could get out of control with such a decision, remember that those two vectors are still priority #1 with Intel's processors, and you can bet they'll be pulling out all the stops to put out a killer product. Those folks over in IDC are no dummies, and if their ROI analysis tells them a project doesn't need IVRs to be successfull, I trust them. Again, I have no idea what is or isn't being included in Skylake and other future products, this is just my logical conjecture.
So that's what I know, along with my conjecture. It really is amazing how we've all become so jaded to the enormous complexities of building better and better microprocessors, as if new designs and products come out every year all on their own, with no hefty engineering challenges cropping up along the way. Most on this board know better, and know that it doesn't happen that way at all. The sausage-making can get pretty ugly sometimes, and nothing is ever easy. Intel is where it's at today because it responds to those challenges better than anybody, IMO. Have a little faith, folks...
AI
