INTEL VS. AMD: CHIP WARS

Are CPUs stagnating? Or are we on the cusp of a major leap forward in processor power?

What’s the best processor for your budget right now?

REMEMBER WHEN processors were the single sexiest PC component? Whether it was the first chip to smash the 1GHz barrier, or the move to multicore, CPUs were not only the beating heart of your system, but also the biggest headline grabbers. They were the measure of the PC’s progress as a platform.

Today? Like they say on Facebook, it’s complicated. For pure CPU performance and innovation, you could argue there’s been some stagnation. Believe it or not, it was nearly a decade ago that Intel launched its first quad-core processor for desktop PCs, albeit one that had two dual-core chips wrapped in a single package.

Here we are in 2015 and you can make the case for relatively little progress. Intel’s mainstream PC processors still top out at four cores, and those cores are pretty closely related to the Core 2 quad from 2006. Intel’s clock speeds haven’t exactly exploded over that period, either.

As for AMD, well, the core count has leapt up to eight, but whether AMD’s eightcore chips are truly eight-core is a whole can of worms. AMD’s CPU performance gains have been pretty modest compared to the huge strides taken by, say, its own graphics chips, or solid-state storage.

However, that’s just half of the story. For better or worse, Intel has bifurcated its PC processor platform and now has a separate high-end offering that runs to eight undeniable cores. Not as many as we’d have expected back when the first quads came out and massively multicore seemed to be the future, perhaps. But still a whole lot of CPU power.

Likewise, while more mainstream processors haven’t exploded in terms of CPU core count or individual core power, the picture of progress is a lot more complex. For starters, Intel has been chipping away at its core design, building performance incrementally. That adds up over a decade and multiple generations.

But the real change has been the non-CPU features, so to speak. We’re talking memory controllers, PCI Express links, and, yes, the horror that is integrated graphics. In some ways, then, CPUs have become more significant, more critical to your PC’s feature set and performance.

What’s more, depending on how old your current CPU is, there could be dramatic gains to be had. If your CPU is more than a couple of years old, odds are a new one will deliver on every level, from ramping up your in-game frame rates, to simply making everything feel more responsive. So, read on to get your head around the current state of the CPU market, and why it could be the right time for an upgrade.

Let’s start with the elephant in the room, the Intel CPU architecture that dare not speak its name. Yup, it’s Broadwell. And it’s still not truly here, not in proper desktop format. Despite being due out last year.

What's going on? In truth, nobody outside Intel really knows. The company’s move from 22nm, with Ivy Bridge and Haswell, to 14nm, with the upcoming Broadwell architecture, is at the center of everything. Intel protests that 14nm is all good. But it’s perfectly obvious to everyone that bringing 14nm to market has been a major headache.

Not that we’re criticizing Intel. As ever, it lives at the cutting edge of process technology. So, while it may have fallen behind its self-imposed, ambitious target of delivering a process shrink every two years, it’s not like anybody else is doing better.

Intel’s main rival is actually doing worse. AMD’s highest-performing CPUs remain 32nm chips, albeit with manufacturing out of AMD’s control since it spun off its chip-production arm. What’s more, independent analysis of early mobile Broadwell parts by Chipworks has revealed Intel’s 14nm tech is distinctly impressive, thanks to 13 layers and features like super-fine transistor "fins." Put simply, nobody is doing it better than Intel right now. However, distinct messiness in Intel’s CPU roadmap remains.

By the time you read these words, Intel may have released a pair of Core i5 and i75000 series quad-core Broadwell chips for the LGA1150 socket and existing 9 Series chipset. Both sport a "C" suffix rather than the "K" indicator that Intel has been using in recent years, but are nevertheless "unlocked" chips supporting overclocking. But get this. Clock speeds are apparently lower than the top quad-core K Series Haswell models available today.

BROADWELL OR SKYLAKE?

Odd? Yup, but probably a result of the follow-up to Broadwell, known as Skylake and requiring yet another new socket, probably turning up within a few months.

But how should that influence your decision-making? The good news is that Broadwell’s compatibility with the LGA1150 socket and the 9 Series chipsets, such as Z97 and H97, ensures some longevity. The bad news is that new socket for Skylake.

The thing is, if you already have an LGA1150 mobo, it’s going to be fairly recent. It only appeared in mid-2013. What’s more, it looks highly unlikely that Broadwell is going to be a major performance leap on the CPU side over Haswell. After all, Broadwell is technically a die shrink, not an all-new architecture. And it’s not expected to do anything spectacular with regards to clock speeds. Quite the opposite.

So, the case for upgrading from Haswell to Broadwell will be very marginal for most. The exception is if you went low-end with Haswell, but now have the cash for an upgrade. In that scenario, would it be worth waiting for Broadwell? Maybe. Not because we think Broadwell will be hugely compelling. But it’s probably worth holding out a month or two just to be on the safe side. It’s very possible that existing Haswells, like some of the models in our supertest this month, will remain the best bet.

Things get more complicated when you factor in Skylake. That’s partly because Skylake’s CPU performance is even more of an unknown quantity. It’s also because Intel has revealed some interesting info regarding the Skylake platform. It will require the new LGA1151, which is a pain, but we’re convinced there are good technical reasons. For starters, Skylake will support DDR4 system memory, which means more memory bandwidth. That’s unlikely to dramatically boost performance, but it’s never bad to support the latest tech.

Of greater potential consequence is support for more PCIe lanes and the new 100 Series chipsets that will come with it. In the past, PCIe lane count hasn’t been hugely critical outside of systems running multiple graphics cards. But PC storage is in transition from the clunky SATA interface to quicker, solid-state optimized PCI Express interfaces including M.2. Haswell and Broadwell chips on the LGA1150 socket have only 16 lanes. On paper, a performance graphics card wants all 16. You can’t just pinch a single lane from the graphics—take even one for your SSD and the card will drop all the way to eight lanes.

Anyway, it’s possible that handing over lanes to an SSD could impact graphics performance. But Skylake and the 100 Series chipset deliver 20 lanes, giving you four spare lanes for fast storage. Hurrah. In truth, it’s not actually clear how critical all of this will be. We don’t think it will truly be a factor for several years, but it’s clear enough that if you want to be absolutely sure of avoiding performance compromises, you need to wait for Skylake.

Unless, of course, you’re willing to invest in Intel’s existing high-end platform and the LGA2011v3 socket. Even the entry-level Haswell-E processor for LGA2011, the Core i7-5820K, packs a healthy 28 lanes. As for the AMD options, well, there’s no denying AMD desktop platforms are falling behind in both features and performance. But they can still make sense as value propositions. To find out a little more, catch our guide to AMD’s upcoming CPUs, opposite, and then flick on to our reviews of six of the best available CPUs from both Intel and AMD, for a great overview of your current options.