AMD introduces first "Bulldozer" FX processors

Today AMD launched Bulldozer micro-architecture, and first four FX desktop microprocessors. Bulldozer microarchitecture takes radically different approach than previous generation of K10 chips, and, instead of separate cores, integrates from 2 to 4 modules. Each module contains two CPU cores, that share some circuitry, such as instruction level 1 cache, Floating-Point Unit and level 2 cache. According to AMD, shared cores are supposed to offer the same core efficiency as Phenom II, while offering extra performance and better power management. Higher performance claims were confirmed by many official reviews, that were published today:

The biggest let down of new processors, and micro-architecture, is their poor efficiency. According to benchmarks, FX-Series CPUs perform slower than older Phenom IIs in single- and lightly threaded applications. Another disappointment with the new chips is their high power consumption. At stock frequencies (not overclocked), the FX-8150 draws almost as much power as Phenom II X6 1100T, and much more than competing microprocessors from Intel. The last disappointment, or more like an annoyance, is marketing of the FX products. That includes misleading phrases like "epic performance" and "unmatched multitasking and pure core performance", unfair comparisons with Intel counterparts, limiting launch reviews only to FX-8150 model, and re-using "FX" brand name for chips, that could be better called Phenom III.

It's not all that bad though, and there are quite a few good things about FX CPUs. The chips scale better with frequency, and can be clocked considerably higher. Overall performance of FX-Series processors in heavy multi-threaded applications is better than on 45nm K10 chips, in particularly due to higher frequency, and two extra cores on FX-8000 series parts. Thanks to added support for SSE4, AES and AVX instructions, the performance will only get better in the future, when more and more applications are optimized to use AES and the latest SIMD instructions. Prices of new models are not sky high, and they are in line with processors' multi-threaded performance. All in all, FX-Series products have their own niche, and they will be beneficial to overclockers, or users, who can put to use all eight cores.

As we already mentioned, all reviews, linked above, are only for flagship FX-8150. Three other parts, released today, are FX-8120, FX-6100 and FX-4100.

AMD FX-8150 operates at 3.6 GHz, and, when necessary, can "overclock" all cores as high as 3.9 GHz, or up to 4.2 GHz when at least half of the cores are idle. The CPU has 8 MB level 2 and 8 MB level 3 caches, and 125 Watt TDP. Official price of the FX-8150 is $245, which is slightly higher than Intel Core i7-2500K, and lower than Intel Core i7-2600K.

Like the 8150, the FX-8120 has eight cores, but its clock speeds are lower, from 3.1 GHz stock frequency and up to 4 GHz maximum Turbo Core frequency. This processor is priced at $205, i.e. identical to Phenom II X6 1100T.

AMD FX-6100 comes with 6 CPU cores, 6 MB L2 cache and 8 MB L3 cache. This model features 95 Watt TDP, and it operates at 3.3 GHz. When Turbo Core is enabled, the FX-6100 can increase its frequency up to 3.9 GHz. Official price of this chip is $165.

FX-4100 is a quad-core processor with 4 MB of level 2 cache, and 8 MB of last level cache. The CPU supports Turbo Core feature, although it offers only 200 MHz boost over stock 3.6 GHz frequency. The FX-4100 fits into 95 Watt thermal envelope, and is priced at $115, which is on a par with Phenom II X4 955.

Specifications and prices of 4 new chips are provided in the table below:

Model	Cores	Frequency	Turbo Frequency	L3 cache	TDP	Price
AMD FX-4100	4	3.6 GHz	3.8 GHz	8 MB	95 Watt	$115
AMD FX-6100	6	3.3 GHz	3.9 GHz	8 MB	95 Watt	$165
AMD FX-8120	8	3.1 GHz	4 GHz	8 MB	125 Watt	$205
AMD FX-8150	8	3.6 GHz	4.2 GHz	8 MB	125 Watt	$245