An extra 100MHz. This is progress, Intel style. I speak of the expected refresh of Intel's Haswell-vintage CPUs, due in a month or so. It's a PR upgrade to what was already an underwhelming family of desktop processors and yet another example of some pretty specular foot-dragging from Intel in recent years. Will Intel's next properly new family of chips, known as Broadwell, be any better? If not, we should at least be able to look forward to a big step up in SSD performance fairly soon in part enabled by Intel's upcoming 9 Series chipsets. Well, it's something to look forward to...
Officially, this falls under the rumour heading. But it all looks very plausible to me. Intel is prepping an interim refresh for its current Haswell Core iSomething-4xxx chips for the LGA1150 socket. And it looks like it boils down to a 100MHz speed bump.
The bump applies across a range of Core i3, i5 and i7 chips. For instance, the new Core i7-4790 is expected to clock in at 3.6GHz nominal, 4GHz Turbo, while the Core i5-4690 is a 3.5GHz/3.9GHz chip. Again that's a spectacular 100MHz increase on the existing 4771 and 4670 chips.
To be fair, marginal speed bumps are a long established tradition for interim refreshes intended to give a CPU family a temporary sheen of novelty before something genuinely new appears. Routine, fairly inoffensive capitalism at work.
The problem is that Intel's mainstream processors have been stagnating for so long now, every time the firm wheels out a new processor that underwhelms with incrementalism, like the drip-drip of water torture a minor blow lands with a painful thump. It's becoming tiresome.
Intel's cheapo Haswell packaging is likely to remain
You may also note that I haven't mentioned any new unlocked K-Series Haswell chips. None are expected. But that's arguably a moot point since these new models may well not even be new steppings and no improvements to Intel's increasingly shonky chip-packaging policies (Intel has been saving a few pennies by no longer soldering the heat spreader to the CPU die and electing to use cheap thermal paste inside the processor packaging) appear to be on the cards.
So, there's no reason for any new K-Series chip to overclock any better than existing models. As for pricing, it looks like the new models will slip in at around existing levels, so you may get that exciting 100MHz boost for free. Cue much rejoicing.
Of course, I've been arguing for a while that the notion of good-enough CPU performance has some validity. And yet I still yearn for something more from Intel. When you look at the wealth of its x86 server chips, which are now available with up to 15 CPU cores, the desktop offering looks feeble going on cynical. It's a familiar refrain, but this is what a lack of serious competition from AMD leads to, unfortunately.
If this Haswell refresh looks deathly dull, what hope for some excitement from Broadwell, the next major waypoint on Intel's CPU roadmap? First up, Broadwell and specifically its use of 14nm process tech seems to be suffering a painful birth.
Last October Intel conceded that yields from a test run of 14nm chips were poor enough to force a rethink on the roll out of Broadwell. Further delays have been rumoured lately and it's possible Broadwell won't be widely available until early next year, which would constitute a bit of a blow to Intel's increasingly precarious tick-tock strategy of launching new architectures and production processors in alternating years.
Anyway, from what I can tell, nothing much is expected from the CPU side of Broadwell. The cores will be pretty similar and once again we'll be stuck with four of them for the mainstream desktop socket. Power consumption will probably plummet and the integrated graphics performance will take another leap towards (but no doubt not actually achieve) true game-ability.
But I doubt there will be much to get excited about for us, the bedraggled gaming and desktop enthusiast community.
Die die, die. It's German for, 'The die, the'
That said, one Intel tech that does look promising for the desktop is the 9 Series chipset and its improved SSD support. OK, it will slightly piss me off that Broadwell reportedly won't be backwards compatible with existing 8-Series motherboards.
But the good news is that the spec list for the new Z97 chipset looks likely to include an M.2 SSD connector hooked up to a pair of PCI Express 2.0 lanes. Explicit chipset support isn't strictly necessary for M.2 support, of course. You can already get the odd Z87 board with M.2 and several laptops based on the 8 Series platform sport M.2.
But the idea is that M.2 will become a standard feature for Z97 mobos. And that means a boost in peak storage performance from today's 550MB/s courtesy of SATA 6Gbps to circa 1GB/s. A healthy speed bump by any metric, even if it's arguably IOPs and 4K random access SSD performance that really needs a bump, not peak sequential throughput. But I'm trying to find the positives. And the sequentials is all I got.