Wednesday, September 4, 2013

Comparing Apples and Oranges: Primary vs. Secondary

A couple of weeks ago, I introduced my secondary PC on this blog; a machine consisting of spare parts and hand-me-downs. After upgrading the CPU cooler to combat the high temperatures I was experiencing, I was keen to benchmark the system so I could compare it with my primary gaming rig.

Clock Speed and Temperature

While installing all the tools I use for profiling a system, I noticed that the processor clock speed reported by CPU-Z was sat around 1600MHz, no matter the load. This was strange, as the CPU's stock clock speed is 2.13GHz, so I suspected something was awry with the BIOS settings. After much tweaking, I finally was able to get the CPU to run at the intended clock speed, but I had to disable two settings in the overclocking section:

  • CPU Internal Thermal Control - causes the CPU to throttle back when it's temperature reaches a certain threshold.
  • Enhanced Intel SpeedStep(tm) Tech - switches the CPU's clock speed and, more importantly, core voltage when it's full performance isn't required.
  • .
However, I wasn't particularly happy with this, especially the thermal control as this would prevent damage to the CPU should temperatures reach dangerous levels. I was even less happy when I booted up to find that forcing the CPU to run at this speed greatly increased temperatures; even just performing basic tasks such as watching video in a browser brought the core temperatures close to 60°C.

There was no way I would be able to benchmark the system with the system in this state, so I decided to take the side of the case off and set up a desk fan to point directly into it. After running Intel CPU Burn, I was satisfied that temperatures wouldn't reach dangerous levels and I was able to benchmark; clearly I need to improve case airflow for this system at some point!

The Tests

As I was just after a basic comparison between my two machines, I opted to only use synthetic tests and only run them on the secondary system; I would just compare the results to those I obtained recently during my overclocking/unlocking investigation. To keep things as fair as possible, I used the same versions of the software I used when I benchmarked my primary system, including the AMD drivers:

  • AMD Catalyst 3.1
  • Cinebench 11.529
  • POV-Ray 3.7 RC6
  • Unigine 3.0

I didn't really have any expectation other than my primary system being faster, but the results showed a significant difference between the two machines:

The Cinebench scores show that in single-threaded performance, my unlocked and overclocked Phenom II system is around twice as fast as the Core 2 Duo, and in the multi-threaded benchmark it's three times as fast.

Looking at the POV-Ray results, they're almost the same; with the Phenom II only just falling short of achieving the same lead as before.

The Heaven benchmark was interesting, as the Intel system produced more a more consistent frame rate; the difference between minimum and maximum FPS was much less than the AMD system. However, it's maximum frame rating only matched the average of the Phenom, while it's average was in line with the Phenom's minimum.


It's great to see how superior my primary system is to my secondary, but it's important to not draw any conclusions; the Intel CPU is two years older than the AMD, while the two graphics cards are a generation apart. After these synthetic tests, I'm keen to try some real-world comparisons, using some of the games I have benchmarked with in the past.