AMD’s new Bulldozer architecture is a major change from its previous Phenom and Phenom II processor technology, as it introduces us to an eight-core processor. That’s eight hardware cores, rather than four cores using Hyper-Threading – with the release of its first Bulldozer chip, the FX-8150, AMD genuinely is breaking new ground.
It may sound unlikely, but it’s actually AMD that has a track record of introducing innovative new features and technology, while giant rival Intel’s expertise lies in the fabrication process, production, marketing and sales.
AMD Opteron moved the memory controller from the chipset to the processor and also introduced native support for 64-bit x86 software. Phenom arrived in December 2007 and introduced a genuine quad-core processor rather than Intel’s two-dual-cores-on-the-same-socket-fudge, Pentium D.
With Bulldozer we have eight cores on the same AM3+ processor socket. This socket is backwards-compatible with Phenom II, and is combined with the AMD 9 series of chipsets that was introduced in June 2011. If you need a clue that Bulldozer arrived later than AMD intended, we’d point you to the fact that you could buy an AM3+ motherboard months before the new processor was available.
So what is Bulldozer?
There are three versions of Bulldozer. The first is code-named ‘Zambezi’, the desktop processor, which has a single HT (Hyper Transport) link. Then we have two server chips: ‘Valencia’ can be used in single- or dual-processor mode; while ‘Interlagos’ is essentially two Valencia cores on the same socket, and can be used with 1-4 socket server motherboards.
For the purposes of this review, we’ll be looking exclusively at the version of Bulldozer in the desktop Zambezi chips.
Bulldozer uses a 32nm fabrication process, and this appears to be the source of at least some of the delays to this new processor. The original plan was to debut Bulldozer using AMD’s tried-and-tested 45nm process; however, it seems that these processors – code-named ‘Shrike’ – would have been too large, hot, juicy and expensive, so they were scrapped and AMD made the leap to 32nm.
Bulldozer’s design is unusual, in that it uses a series of modules, each with two processor cores. Each module carries front-end components such as integer pipelines and L1 cache that are shared by the two local cores. The floating point pipelines and L2 cache are available to all the cores in the processor. There is 2MB of L2 cache for each module and 8MB of L3 for the processor as a whole.
The memory controller has been improved to support dual-channel DDR3-1866MHz.
AMD hasn’t included a graphics core in Bulldozer, it rationale being that the feature is reserved for the low-end Fusion APU, and has no part in a high-end performance part.
Turbo Core and Max Turbo speed boosts
AMD has taken a bold approach with regard to the TDP (Thermal Design Power), clock rating and power consumption of each model of Bulldozer. For starters, we have the base clock speed with a 300MHz Turbo Core boost when the processor is under load. If the cores remain fully loaded there is an extra spurt of speed called Max Turbo that adds another 300MHz, provided the processor remains within its thermal limits.
Buying a Bulldozer
At launch AMD delivered three models of Bulldozer to the buying public. The 8-core FX-8150 has a base clock speed of 3.6GHz, Turbos up to 3.9GHz and can run at 4.2GHz in Max Turbo mode, provided it stays within the 125W TDP. This model costs £199.99, which falls neatly between Intel Core i5-2500K at £169 and Core i7-2600K at £240.
To test its mettle, we lined it up against the 6-core Phenom II X6 1100T Black Edition, which currently costs £150.
The other two Bulldozer launch models are FX-8120 (8-core, 3.10GHz and £169) and FX-6100 (6-core, 3.30 GHz and £140).
We tested the FX-8150 using an Asus Crosshair Formula V motherboard, 4GB of Kingston KHX1600 RAM, a Radeon HD 5850 graphics card and Intel X25-M SSD.
At stock speed, Bulldozer performed well enough and was a match for the Phenom II X6. We were, however, a bit concerned at the loaded power draw of FX-8150, which was 15W higher than Phenom II – itself renowned for sucking too much power.
Overclocking the FX-8150 to 4.5GHz was simple enough using the traditional approach of feeding an extra 0.2V to the CPU core and chipset, and then cranking up the processor multiplier. We experienced no trouble in running the FX-8150 at 4.5GHz, but that’s hardly surprising given that AMD has overclocked the FX-8150 to 8.43GHz using liquid helium cooling.
AMD is also putting a lot of weight behind its plans to increase clock speeds with future models of Bulldozer. Unfortunately, the chip’s performance doesn’t appear to scale especially well as the clock speed increases – and the last thing we need on the desktop is an increase in power draw.
It’s a moot point whether Bulldozer is better than Phenom II, but on balance we like the new processor and give it a thumbs up.
The real problem for Bulldozer is that it gets kicked all round the park by Intel’s Core i5-2500K. Admittedly this quad-core processor is one of a select bunch of ‘K’ models produced by Intel that are left unlocked for overclocking, whereas every Bulldozer is unlocked. This takes us away from the business of Black Edition AMD processors of yore.
Core i5-2500K runs like a dream at standard speed and, like Bulldozer, benefits from a Turbo boost – but it also overclocks to 4.6GHz with no extra power draw whatsoever. At this speed, the Intel chip draws just 175W compared to the Bulldozer’s 260W – and still delivers stomping performance.
You can see the FX-81500 key parameters in the CPU-Z validation page below:
- Eight cores and loads of clock speed at a reasonable price.
- Performance is unimpressive, and the power draw is too high.
Bulldozer is an impressive piece of silicon, but it's not a winner. On balance, it's a better bet than its predecessor, Phenom II, but you'd need to have a very good reason to build an AMD AM3+ system when Intel LGA1155 does such a superb job.