AMD FX-8150, The Battle Continues…

Consensus suggests that the release of AMD’s much anticipated Bulldozer CPU was a little underwhelming. Performance improvements were a mixed bag with gains in some areas and only just staying on par with older AMD CPU’s in other areas. While AMD’s older Athlon and Phenom processors were due for an overhaul, AMD still took the gutsy approach making such aggressive and forward looking changes. The Bulldozer architecture represents a total redesign of AMD’s processors.

The new top of the line AMD chip, the FX-8150 contains an astounding 2 billion transistors on 315 mm² of silicon. While the FX is unable to compete with Intel’s 980 Extreme Edition processors on raw performance AMD has priced the chip to compete with Intel’s Core i7-2600K, the current Value Performance King on the desktop. At $269 USD the AMD FX-8150 is avoiding the whole Extreme Edition comparisons altogether and taking the fight to Intel’s mid range processors.

Three chips in the Zambezi family where released by AMD, FX-8150, FX-8120 and the FX-6100. The FX-8150 is AMD’s new top of the line processor. Running at 3.6Ghz, 4.2Ghz Turbo Mode, 8 cores, 8MB of cache and dual channel memory the AMD has excellent credentials. The mid range FX-8120 is a slightly clocked down version of the 8150, running at 3.1Ghz base frequency and 3.4Ghz turbo. The last member of the new FX family of chips I the FX-6100, a 6 core version of the Zambezi chip the 6100 is cheap and efficient, providing incredible value for money at $189. Clocked at 3.6GHz Normal and 3.9 GHz Turbo mode.

The 3 initial FX series processors to make use of the new Bulldozer architecture are the Zambezi chips. This first generation of Bulldozer for the desktop combines 4 dual thread processing units, the top of the line FX-8150 makes 8 threads available – with a thread roughly equating to a task. Part of the Bulldozers strength is its ability to multi-task. Unlike Hyper threaded processors – Intel’s 2600F – when AMD chips says there are 8 threads available, they means 8 threads. And while the 2600 will show 8 threads available in Windows Task Manager, there are only 4 CPU cores to execute the 8 threads.

A large proportion of all modern CPU’s have an area of the chip dedicated to doing math functions. Floating Point Units – FPU – make an enormous difference to video processing, photo editing , music production and games. AMD also chose to redesign the FPU of the Bulldozers, the results aren’t enough to close the gap to Intel’s 2600K but it helps. Bulldozer is the first x86 chip to support FMAC multimedia instructions New commands are constantly being added to processors instruction sets. The FX processors are the first on the market to support the latest set of instruction called fused multiply-accumulate – FMAC – which allow a processor to do two instruction in one pass, an addition and multiply – e.g. d = a + b * c -. The new instructions are included in the new SSE 4.1 and 4.2 extensions that also include most of the industry standard Advanced Vector Extensions – AVX – , including instructions for accelerating encryption and AMD’s own XOP extensions. All of this adds up to software developers upgrading applications, windows updates being created and eventually faster multi-media applications, better games and more processor time for surfing the web.

Intel’s entry into the performance desktop processor market is the Core i7 2600F. Based on the latest Sandy Bridge architecture this is Intel’s best mainstream desktop chip at the moment. At around $300 USD the Intel chip is an excellent high performance quad core processor at a great price. The Core i7-2600K specifications; 4 cores, 8 threads, 3.4Ghz, 3.8Ghz Turbo, 8MB Cache RAM, Dual Channel memory controller. The 2600 wins most benchmarks against AMD’s latest FX processors, with Intel doing especially well on FPU performance. The AMD processor draws back some performance with it’s 8 core’s during heavy multi-tasking, but it falls behind again when and application can’t use multiple threads. AMD is generally cheaper to setup, with the price of the AMD 8150 already dipping below $260 USD along with cheap motherboards makes an AMD setup fast at a reasonable price.

The new Bulldozer architecture is replacing the current Athlon and Phenom chips which owe their design heritage to the original K7 Athlon and later the K8 Athlon64 architectures, Both the original K7 and K8 saw some of AMD’s greatest achievements in competing with Intel. The original Athlon K7 CPU was the first processor to break to 1Ghz barrier which annoyed Intel so much that they engineered the P4 to allow higher clock speeds at the cost of processing efficiency. The Athlon64 K8 processors were the first 64bit X86 processor. Intel was a stalwart on this one and held off the 64bit upgrade for three years. All desktop processors except Intel Atom and AMD Brazos are now fully 64bit compatible, Windows 7 64bit has cemented the 64bit revolution with it now outnumbers 32bit installs on desktop PC’s. The Athlon64 chips were also the first to move the memory controller onto the processor, allowing much lower memory latency and higher memory bandwidths. Intel now also uses integrated memory controllers on all of its mainstream chips. The introduction of both the K7 and K8 Athlon chips saw AMD leading the performance battle with Intel, even if only for a year or two they did get Intel’s attention. The Bulldozer architecture can’t claim such performance victories against Intel for it’s launch, it does bring AMD much closer to Intel and improvements compared to AMD’s aging Phenom X4’s and X6’s are substantial for many tests. Keeping AMD in the game is only part of the equation though, the changes to the architecture do give AMD room to grow and develop. The old chip had been tweaked and prodded to it’s limits, the new Zambezi chips bring AMD back to the start of the performance ladder once again.

Overclocked to 4Ghz the chip is very competitive, if it was released at full speed the game may have been different. In fact the 8150 should be a hit with the overclocker’s. FX branded CPU’s have always been known for having unlocked frequency multipliers. The frequency multipliers set the final processor speed by applying the multiplier factor to the CPU bus speed. In the 8150’s case standard out of the box it runs at 3.6Ghz, which uses a 18X multiplier on a 2000Mhz bus to provide the 3.6HZ processor clock speed. Most CPU’s on the market keep this multiplier locked once they leave the factory, allowing chip makers to easily predict the lifespan of the processor. The FX CPU’s on the other hand allow the multiplier to be changed at will, letting you choose the speed of your processor. Using the AMD sealed unit water coolers 4Ghz clock speeds are easily achievable. Be sure to get an overclocked friendly motherboard, a button on the motherboard to reset the BIOS is an overclocker’s best friend. You will find once you push a CPU too far the motherboard and CPU will freeze, requiring everything to be reset back to factory default.

Bulldozer for the Servers include the Opteron 4200 series code named Valencia, with up to 8 cores and Opteron 6200 series code named Interlagos, with up to 16 cores. The 16 core server processor is one of the most powerful and best value server chips in that segment of the market, big tin. Both HP and Dell have dual CPU servers based around these chips. The fully tricked out HP server with two CPU’s has 32 threads, up to 64 GB of RAM and Raid based hard drive storage controllers all for under 10k. That’s enough grunt to run a medium sized enterprise in one small 4U rack, or in Dells case one Tower case. These chis are also being used by Cray in the latest Supercomputers.

Listings of a mysterious FX-8170 are already appearing. At 3.9Ghz the 8170 should offer a little more competition to Intel. According to a recently leaked company roadmap, the start of 2012 will see AMD release two new FX-Series CPU’s, the FX-8170 and the FX-6120.

AMD may not have blown away the competition with the release of the latest FX range of CPU’s but it has stayed in the game. While the Core i7 2600K is faster the AMD CPU still manages to come away with a few wins and stays within 10 to 20% on the tests its looses, it’s in the right ballpark. Obviously the new architecture is going to need a few tweaks. Once AMD can hit the 4Ghz mark with this CPU, especially if they can do it relatively quickly they may still be able to take the game to Intel. The FX-8150’s performance at 3.6Ghz is good but overclocking tests are showing very good scaling up over 4Ghz, with performance numbers that draw away from the 2600K.

If you have an existing AMD AM3+ motherboard this is a good cheap upgrade but be sure you have an AM3+ motherboard before buying an FX chip. If you wait a month or two it will drop well below $250. Intel won’t sit buy and let AMD even think it has a chance, they will drop the price of the 2600F well below $300 without even blinking an eye. Intel’s chips are getting high yields on a mature process, they are churning out chips like a Doritos factory and people are eating them up.

So there you have it, a little healthy competition is good for everyone. Not to say the initial performance numbers for the FX-3850 aren’t a little disappointing, but not totally surprising. The Bulldozer architecture that AMD has launched here is a necessary re-alignment of the chips internals. The weaknesses of this new design seem to be single-threaded application speed. Applications that are optimized for multitasking and using multi-core chips perform very well, especially games that are optimized for multi-core x86-64 chips.

AMD FX-8150 AMD64 Specifications

  • L1 Cache: 128KB (64KB + 64KB)
  • L2 Cache: 8MB
  • L3 Cache: 8MB (shared L3)
  • Hyper Transport™ Technology up to 4000MT/s full duplex, or up to 16.0GB/s I/O Bandwidth
  • Integrated DDR3 Memory Controller
  • Memory Controller Width: 128-bit
  • Type of Memory Supported: Up to DDR3 1866
  • Memory Bandwidth: Up to 21 GB/s dual channel memory bandwidth
  • Total Processor-to-system Bandwidth – Hyper Transport plus memory bandwidth -: Up to 37 GB/s
  • Process Technology: 32 nanometre SOI – silicon-on-insulator – Technology
  • Packaging: AM3+
  • Thermal Design Power: 125W, & 95W