Multitasking

In order to use the performance of all four cores the QX6700 features, you can run more than one application simultaneously. Obviously, a human being cannot do two things on a computer at once - so, for example, playing two games at once on a dual-monitor system is out of the question (for the moment).

Let's take the following scenario: you have just been on a holiday to Hawaii with your family. Being the geek that you are, you captured thousands of photos and hours of video. Upon returning home, you know that you want to resize all your photos, encode your video into DivX, and play a game (because after a holiday, you have to be tired!). Now, this is exactly what the quad core CPU will allow you to do! Provided that other components can keep up with the speed.

This is where the main problem with quad-core (or x-core CPUs, for that matter) comes into play more than ever before. Feeding one core with data was no problem for our current generation of hardware. 2 cores, OK, a little bit of a struggle, but still doable. With 4 cores, I must say - either have a full blown top notch system, or forget it.

The slower your PC is in the first place, the less it will benefit from your quad core investment. This is a thing that struck me during testing, and wouldn't go away. Many editors from famous web sites around the Internet have tested and benchmarked Intel's Quad core CPUs, and they were taken aback by the performance. They surely used top of the line memory (preferably 2 gigs of it), a Raptor hardrive and the latest and greatest graphics card. Yes, the QX6700 does shine in this case.

Our test bed was far from ideal - 1 GB of DDR800 memory (with horrible CL-5 timings) was used, and a definitely out-of-date GeForce 6800. Here, instead of the graphics card being limited in games, it was the CPU - it just couldn't show its full potential.

I decided to run SuperPI 4x, in order to see how the whole system would do. Below is a screenshot of the run:

The 1M run was just as expected, slower than running SuperPI once. I assumed that the same scenario would occur if I ran SuperPI 4x with the 32M option. How terribly wrong I was. The system took off fine, but after a while, I could hear loud sounds of the hard-disk swapping - the system had ran out of memory, and the whole calculation was terribly slow.

After backing off to 16M, the run completed successfully:


Now, things like 2 GB and 4 GB memory kits start to make sense. Super power graphic cards too. However, there is a grave problem to the new path both manufacturers, AMD and Intel, have chosen to follow. Before, only the company could be blamed (or blame itself) for a slow processor. Now, all the components come into play - the memory, hard-disk, graphics card. I am afraid that one day, we may find ourselves at a dead end again - the CPU power will be here, but other components will prevent us from utilizing it all.

Intel has realized that this problem exists, and has presented its visions of tackling this problem at IDFs all around the world.

Multi-core scaling

How much performance does one gain by switching from one to two cores, two to three and finally three to four? We have investigated several multi-threaded apps, below are our results. First up, Quake 4:


From one to two cores, the FPS heightens by about 30 frames. Adding more cores, on the other hand, did not show such results - the performance levels off, and actually decreases from two to three cores.



Cinebench performance is much more pleasing when increasing the number of cores. The graph is almost linear, adding a core gives approximately a 400 point increase in the benchmark.



The 3DMark graph is most similar to that of Cinebench, but we can see indications of the performance slowly levelling off.



Lastly, we have PCMark performance. From one to two cores, there is a large performance increase, and from two to three and three to four, the line seems to follow a different equation, but remains linear.