AMD Bulldozer / FX-8150 Desktop Performance Review

Computer hardware enthusiasts have literally waited for years for AMD's Bulldozer architecture to come to market and we finally see this today in its desktop form, code named Zambezi, brand named AMD FX. In this article we share with you our analysis of Bulldozer's performance in synthetic benchmarks and desktop applications.


Video Encoding

Video encoding is a process that is very well know by many computer enthusiasts. The scenarios shown below are very much real world "benchmarks." Again, we are using a workload across all systems, and timing how long it takes to get it done in these first three examples. You might ask, "How do you get HandBrake to give you an encoding time?" Well if you know a way to trace it down to the second, let me know so I don't have to sit there with my stopwatch any longer.

This first example is a short movie that was ripped from a DVD....that I produced myself. wink We are using handbrake to take the DVD VOB file and encode it to a M4V file sized to fit an iPhone4 screen. You can set up handbrake to encode for whatever device you might need to, such as an EVO4G which I use every day, but we use the iPhone preset to make it simple.

Article Image

HandBrake does a very good job of taking advantage of the resources available to it. If you open up Resource Monitor you can watch how it will utilize almost all threads available to it at times through the encoding process.

First off let's say that we are happy to see in a real world test that we have a set of data that graphs such that we actually have to look at it to tell who "won."

Our primary comparison we are looking at is the FX-8120 and the 2500K, priced respectively at $205 and $215. At stock, the 2500K is 7 seconds faster than the 8120. At our 4.6GHz and 4.8GHz overclocks, the 8120 is 4 seconds faster than the 2500K. When we look at the 2600K, which is basically a 2500K with HyperThreading turned on, it opens up the gap a bit, but costs $100 more.

In this HandBrake scenario we are taking the Ironman movie trailer off the Transformer Blu-ray disk and encoding it with to an MKV format and reducing it in file size.

Article Image

Again we have some results that bear nothing in resemblance to our synthetic or IPC results.

Looking at the 8120 and 2500K again, we see the 8120 is 3 seconds faster at stock, and 11 seconds faster when overclocked. The Bulldozer is actually showing better IPC scaling than Sandy Bridge with HandBrake.

Videora is a free video encoding tool for iPod devices. It is VERY simple to use. It works well and is very thread aware of its resources. I am using Videora to encode the same DVD movie above to the iPhone4 preset inside the software.

Article Image

In this example, the stock 8120 is 19 seconds faster than a 2500K, and the overclocked 8120 is 42 seconds faster than the overclocked 2500K. Again we see better IPC scaling as we clock the Bulldozer. We do however again see the HyperThreading enabled 2600K outstretch the 8120.

This encoding benchmark shown below is NOT a real world experience but I wanted to add it in order to give you an easy to access point of reference on this since sharing huge VOB and Blu-ray test files is a bit prohibitive. You can download the benchmark here. This test explains that it takes a short DVD clip and encodes with the x264 video codec in two passes. Passes are shown below in frames encoded per second.

Article Image

Article Image

The two passes show tremendously different results. The workload per frame is much less on pass one and much higher on pass two. Intel excels across the board on the less stressful workloads (except for the $1000 990X), and AMD's Bulldozer when overclocked shows to excel at the more stressful workloads in pass two.

I am not sure how much credence to give this benchmark, but I thought the results were interesting enough that these should be included.