Closing Thoughts


This is the first of many important, yet delicate discussions on the aspects of performance for the AMD FX processor and the Bulldozer architecture.  With the IPC decreases we saw in the the FX-8150 clocked at the same speed as the Phenom II X6 processor it makes sense that AMD needed to increase the clock rate of each core pretty dramatically in order to meet performance expectations.  It would appear that AMD wasn’t able to get frequencies as high as they wanted though and the performance of the FX-8150 in our series of benchmarks pretty clearly demonstrates that.  It seems doubtful that years ago when starting development the CPU team at AMD would have targeted the $245 price range for their highest end desktop processor.  

In applications that are very lightly threaded the FX-8150 does the poorest as you can see in our LAME MP3 encoding, Valve synthetic tests and more.  Even with a clock rate as high as 4.2 GHz in those cases, the FX-8150 was unable to to keep up with the likes of the Core i7-2600k and even the Core i5-2500k.  Even VirtualDub, used by many for video capture and transcoding, didn’t really see the benefits of 8-cores like we might have thought.

Some of our testing did show some potential for Bulldozer, in particular our highly threaded application workloads like CineBench, POV-Ray and Handbrake.  Programs like that can fully utilize the worlds first 8-core processor though even then the FX-8150 didn’t beat out the Core i7-2600k that is a quad-core Sandy Bridge processor with HyperTheading enabled.  Obviously the module design with 2 cores per module has helped AMD compress processing capability into a smaller die size but the truth is that 1 module does not TRULY equal 2 cores.  

AMD iterated over and over that many of the tests we are showing you today are "old benchmarks" and that instead we should focus on "new tests" like high-resolution gaming, media transcoding and even things like the new WinRAR 4.  The truth is that high-resolution gaming doesn’t see enough of a difference between platforms to really warrant it as a deciding point in my book and in several places of AMD’s "reviewer guide" they contradict themselves on that point.  Still, I feel that our collection of tests and analysis is fair enough to be a complete evaluation of CPU performance.  Just because it turns out to not be good for them all the time doesn’t mean it is wrong.  

Platform Considerations

Even though the processor might not have impressed us as we had hoped, there are some advantages the AMD platform has over Intel releases including a fairly static platform.  When the 990FX motherboards were first released in June we knew that most would support the upcoming Bulldozer processor even if we didn’t know exactly what performance would turn out to be.  AMD is hoping that many users bought compatible and FX-ready motherboards and will continue to take that path, buying up AMD FX parts to upgrade their rigs.

Also, because many of the new 990FX chipset motherboards support SLI as well as CrossFire options, they can be claimed as more than adequate gaming configurations for at least the next couple of years. 

Pricing and Availability

We mentioned and discussed on previous pages, but the cost of the AMD FX processor is another big selling point for them and was forced upon AMD by the performance numbers we reported today as well.

  • AMD FX-8150 – $245
  • AMD FX-8120 – $205
  • AMD FX-6100 – $165
  • AMD FX-4100 – $115

Compare that to the Sandy Bridge CPU prices on the Core i7-2600k ($314), Core i5-2500k ($219) and the Core i5-2400 ($189) and some interesting things take place.

In the best case, the FX-8150 is competitive with the i7-2600k and better than the i5-2500k, so the pricing is almost justified.  However, in many other cases, the FX-8150 has problems keeping up with the i5-2500k as well as the i5-2400.  In those instances, the price of the AMD FX part seems quite a bit higher than it has the right to be.

Just as we saw with the release of AMD’s Llano A-series APU, AMD is pricing its processors based on best case scenarios rather than the average or median.  The A8-3850 was priced in a way that expected consumers to put more value on the intergrated GPU than on the CPU performance.  The FX-8150 is priced in a way that expects consumers to put more value on highly threaded applications like Handbrake than on lightly threaded ones like iTunes or gaming.

How will consumers respond is the real question? 

Final Thoughts

The allure of having the "world’s first destop 8-core processor" is more than slightly muted by the performance results we saw in our review today.  Obviously the Bulldozer design team had to make some decisions years ago that couldn’t be easily rolled back but it appears obvious to me at this point that the "2 cores per module" design didn’t bring with it the benefits AMD expected.  And with the inability for the processors to scale to higher frequencies, the FX series from AMD is left holding promises that it couldn’t keep for consumers. 

The AMD FX processor release really comes down to the one thought: are you willing to give up performance on lightly-threaded everyday applications in hopes of better performance per dollar on highly threaded programs like Handbrake?  Even if the answer to that question is yes, Intel’s Core i7/i5 line of processors based on Sandy Bridge have competitive solutions that don’t require you to give up performance in either direction.  Will a system based on the new FX-8150 be competent and competitive while also making for a great gaming machine?  It definitely will but is that enough to pull consumers away from the Intel platforms that offer better performance in many areas for similar prices?  It is hard to see how it could be. 

Be sure to read Josh’s final thoughts on AMD’s latest release.

« PreviousNext »