It looks like you're new here. If you want to get involved, click one of these buttons!
There is, in American politics and probably elsewhere, a time-honored tradition of the Friday afternoon news dump. Basically, if you have bad news that isn't publicly known yet but you know is going to come out eventually, you release the news on a Friday afternoon to minimize the publicity it gets. The theory is that you release it too late for Friday news, people don't check the news much on Saturday, and by next week, you can dismiss it as old news. In an Internet era, I don't think this works as well as it used to, and big enough scandals can drag on for weeks or years, but that doesn't stop people from trying.
There were rumors that Intel was going to launch Haswell-E today, on the Friday before Labor Day. If you wanted to time a launch for minimal publicity, this is basically when you'd do it. I started to write a post a couple of weeks ago asking why they'd do such a thing and what must be wrong with the chip. But then I thought, nah, they wouldn't dare, and canceled the post. Well, they just did: and on a Friday afternoon, no less.
I can only recall one significant computer part in the last several years to launch on a Friday before today: the GeForce GTX 480. And that was on a Friday evening, no less. Granted, Haswell-E isn't as bad as that awful chip, but then, "not as bad as the worst GPU chip of the last seven years" isn't exactly high praise.
So let's look at the damage. With enough cores, the clock speeds are limited by how fast the chip can go inside a given thermal envelope, not just how fast the cores can go, period. That has long been true of the higher end Xeon E5 and E7 chips, but this is arguably the first time that we're seeing markedly reduced clock speeds in desktop chips. Granted, this is also the first time we're seeing an eight-core Intel chip, but even the six core chips have reduced clock speeds, with turbo topping out at 3.6 and 3.7 GHz, as compared to 3.8 GHz or more for all of Intel's previous Core i7 desktop chips of the last four years--including the six core variants of Sandy Bridge-E and Ivy Bridge-E.
It's not that they're trying to burn less power, either. The TDP goes up: 140 W, as compared to 130 W for Ivy Bridge-E. In spite of that, the clock speeds are down. At stock speeds with six cores, Haswell-E is mostly slower than Ivy Bridge-E. Overclocking can close that gap, but if you compare overclock to overclock, Haswell-E still isn't any better than Sandy Bridge-E, and you're probably using more power than Ivy Bridge-E.
But Haswell-E does offer DDR4, at least, right? Well yes, but it's not at all clear that that's a good thing. Even if you decide that you need 2133 MHz DDR3 to match the clock speed of DDR4, it's a lot cheaper to buy a given quantity of DDR3 than DDR4. DDR4 will use less power than DDR4, but that is likely to get overwhelmed by Haswell-E using more power for the CPU than Ivy Bridge-E.
So does that make Haswell-E completely useless? Well, no. If you want eight cores in a desktop with a desktop feature set (overclocking allowed, non-ECC memory, etc.), Haswell-E is the first Intel CPU to over that. It's over $1000, but still, if you need it, it's there.
But there's one other thing: you can get a six core Haswell-E CPU for $390. Ivy Bridge-E started at $560 for six cores. Needing DDR4 memory and a more expensive motherboard will eat up much of that price different today. But DDR4 will probably drop in price as time passes, as will X99 motherboards once they're out. And cheaper is certainly good from a consumer perspective.
But this should also remind us of just how good Sandy Bridge really was. If you bought a Core i7-2600 three and a half ago and are willing to overclock, you still have a legitimate high end CPU today. It will be a year or so before anything with even a remote chance of changing that launches. It used to be that you were doing well if a four year old CPU was not completely awful.