For people that are shopping for CPUs, notebookcheck's processor list at http://www.notebookcheck.net/Mobile-Processors-Benchmarklist.2436.0.html (which includes both mobile and many desktop processors) is invaluable. I took the data and put it into a visualization tool (Many Eyes at IBM Alphaworks) - this requires Java:
http://manyeyes.alphaworks.ibm.com/manyeyes/visualizations/comprehensive-overview-of-notebook
Interesting observations:
1. TDP scales very roughly linearly with overall performance.
2. The multithreaded-ness of various benchmarks can be determined. Plotting Mhz vs 3d marks clusters processors into 2 neat groups, with quad and dual cores forming almost parallel lines. Plotting the same thing but with SuperPi results in almost no clustering.
3. SuperPi vs 3dmark 06 forms a nice inverse relationship except for some odd outliers (i.e. SL9400)
4. (with an external tool) The relation 0.407598*Mhz*Cores is an excellent predictor of 3dmark scores.
Anything interesting that other people discovered?
http://manyeyes.alphaworks.ibm.com/manyeyes/visualizations/comprehensive-overview-of-notebook
Interesting observations:
1. TDP scales very roughly linearly with overall performance.
2. The multithreaded-ness of various benchmarks can be determined. Plotting Mhz vs 3d marks clusters processors into 2 neat groups, with quad and dual cores forming almost parallel lines. Plotting the same thing but with SuperPi results in almost no clustering.
3. SuperPi vs 3dmark 06 forms a nice inverse relationship except for some odd outliers (i.e. SL9400)
4. (with an external tool) The relation 0.407598*Mhz*Cores is an excellent predictor of 3dmark scores.
Anything interesting that other people discovered?
Last edited:
