The CPU industry has always been a game of oneupmanship. Whether between the two main competitors in the consumer market, Intel and AMD, or between generations of chips by the same firm.
They've always followed what Intel calls a tick-tock cycle, where you get a brand new architecture or design, followed by a refinement of it. Then the cycle repeats, things improve by a certain jump, and then by a little more.
AMD's brow has been decidedly lacking in jewellery in the past few years, as Intel's Core 2 Duo chips stole the show in 2006, finally toppling AMD's reign with the Ahtlon 64 CPUs that had been the darling of over-clockers and gamers for a couple of years. And that's the way it's been since, with Intel dominating through the Core 2 years, into the new “tick” of the i series without much of a hiccough. Sure AMD held sway in the budget game somewhat, remaining relevant by having a good bang for buck in the cheaper systems, but if you wanted performance, it was all about the Intel.
Things were supposed to turn around for AMD this month, with the impending release of its next generation of chips, the Bulldozer line. This is the first redesign since the 2007's Phenom blueprint and was supposed to be giving the firm back the edge that they'd lost in the past few years.
Unfortunately, as the first reviews have trickled in, things have not gone as expected. Multi-threaded performance is indeed quite strong as AMD had hoped, with the new core design being based around this principle. The problem however, is that single-core scores are atrocious; worse than the last generation of Phenom chips in some cases.
While we are certainly moving in a multi-threaded direction when it comes to computing in general, games and applications, we're still not quite there yet. Many games still utilise only a single core (especially those that are more GPU bottlenecked) and there are plenty of applications that are designed similarly. Try convincing an enthusiast or gamer to buy something that's going to be great in a couple of years, but for now it's just about average across the board. It's hardly a strong selling point.
The potential problem with this, is that once again, Intel has no competition at the top. The outfit doesn't need to do much in order to flog the highest performing part in their chosen field; so why should they try? Competition and stress has driven development throughout this planet's history and it's the same for the PC hardware market. When AMD became king of the hill with its Athlon 64s, Intel had to step up its game from the rather poor Pentium 4s, delivering what was one of the biggest performance gains in generational jumps with the C2D line.
But at least AMD can hang on to the value side of their business right? Perhaps with a low-cost-but-still-strong-in-multithreaded-applications rival, the contender could cause Intel to keep a keen eye on its pricing structure. At least we'd be getting cheaper chips out of it, even if they don't perform as well as they could if there was more push and pull.
The price of Intel's arguably highest value and best bang-for-buck chip, the i5 2500k: £170. The price of AMD's poorer performing, nine-months younger, flagship Bulldozer chip, the Fx-8150: £205.
So now we have AMD committed to a line of CPUs that for now at least, doesn't keep up with Intel's performance, in fact scores lower in many cases than their own previous generation of chips and still costs more. It's hard to see what motivation, Intel would have, if any, to pursue anything bigger and better at this point. All it needs is to do a little better than before with the Sandybridge E chips, due out in early November, and it will not only retain the performance crown, but they could potentially steal the budget spot too, by slashing the prices on their left over generation-one Sandybridge chips.
This leaves AMD in quite a predicament. It has got other markets to further explore for sure, it doesn't need to focus on the desktop CPU game. The firm has the netbook market with its APUs doing quite well, in-fact the Llano is performing better than the mobile Intel core chips and has improved features like DirectX11 support. As consumers in the desktop game though, a refocusing of AMD's efforts would be a blow to us, as it would be the equivalent of a white flag. Intel would have won and without anyone else to challenge it Why try? These chip makers don't owe us anything.
The real interesting thing about the whole Bulldozer debacle is that something went fundamentally wrong. It's not like AMD released these chips not knowing what they would perform like compared to Sandybridge. Internal testing would havetaken place . So why the fanfare during development? AMD touted that these new CPUs were game changers. They were going to really shake things up. So what happened?
The design is also a strange one. The FX-8150 was supposed to be eight cores, but it's more like four cores - only it's four modules with two cores a piece. Everything is shared between the cores on each module and that's where the problem comes in architecturally. Per core, performance can actually end up being lower than previous AMD chips that didn't employ this sharing as each core had its own fetch and decode hardware. How did AMD not spot this problem in the design process?
The outfit promised 10-15 per cent performance gains with a few upcoming tock cycles throughout 2012-2014, but that's hardly going to be enough to keep up with Intel's next tick - Ivy Bridge; especially since it's not too far away. That one drops things down to 22nm too, bringing with it all the benefits of a die shrink.
Going forward, how does this look for AMD? Well, it could play the price game, figure out some method of making itself cost effective and therefore try and hold on to the budget multi-core field it has been the heart of for the past few years. It could also keep its fingers crossed that over-clockers fall in love with the new chips. If water cooling can do 5GHZ or more with the right motherboard and a bit of tweaking, the processors could find a niche among enthusiasts. There is some potential here as we've already seen AMD take the overclocking crown with an impressive 8.429GHZ. But that was with liquid helium - and really is that enough to secure main stream sales?
Unless this is some ugly early batch and something went horrifically wrong during the manufacturing process, it seems nigh on impossible that AMD can claw back a shot at the performance crown, so whatever it does, Intel is sitting pretty at the top for at least another year, probably longer. Hopefully something comes about that keeps them on their toes, but for now at least, they have a nice coasting period to enjoy.