Sign in with
Sign up | Sign in

Power, Heat, And Efficiency

System Builder Marathon, March 2010: $3,000 Extreme PC
By

Efficiency is a comparison of work completed to energy used. Overclocking forces higher power consumption, but can still result in improved efficiency whenever performance is increased by a greater amount than power consumption. We start our analysis by checking the amount of power used at the source.

At 160W, idle power consumption was already high before we overclocked. Pushing our processor 61% above stock had only a moderate impact on idle power use, yet its full-load power went up by 64%.

A drastic increase in full-load CPU temperature could have a greater effect on power consumption than the moderate increase in voltage, since semiconductors become less resistive as they heat up. Once again, we remind readers that this temperature was reached by running eight threads of Prime95 simultaneously over the period of several hours and that operational temperatures were much cooler.

The liquid-cooled graphics card amazingly operates at only 25.6 degrees Celsius over ambient at loads and settings, a full 66% lower than temperatures typically achieved with air-cooled cards. The drastic difference in CPU and GPU temperatures is probably due to a flow restriction at the GPU water block, where the CPU block relies on high volume, while the GPU block relies on high pressure. Were we to add a second graphics card, we’d probably connect its cooling lines parallel to the first card for increased flow through the water block and radiator.

Because some of our games showed CPU limits at default settings, a gaming performance increase of 31% at a 31% GPU overclock appears to have occurred by happenstance. In other words, the same GPU clock probably would have provided far lower performance gains had it not been accompanied by a significant CPU overclock, though many hours of additional testing would be required to show an exact numeric correlation.

A 40% increase in performance corresponds to a 53% increase in power consumption, dropping the overclocked configuration’s efficiency by 8%. While certainly no “green” PC, we’re sure there are many users who would gladly pay the higher power bill for this performance improvement.

Ask a Category Expert

Create a new thread in the UK Article comments forum about this subject

Example: Notebook, Android, SSD hard drive

Display all 7 comments.
This thread is closed for comments
  • 0 Hide
    mi1ez , 16 March 2010 16:43
    RAID SSDs? No TRIM support there then!
  • 3 Hide
    ch1llboy , 17 March 2010 13:10
    920? really tomshardware! I do look to you for advice often and to see you recommend a chip that has been replaced by the 930 kind shocks and disappoints me. I have confidence that you will make good and update your article. It is 2% more expensive and 5% higher clocked at stock. It overclocks better because it has the 21x multiplier. It removes the cpu blk wall for those of us on air buying 300+ motherboards that can take the blk to 215. It is what I'm buying because 4.4 ghz is better than 4.1.
  • 2 Hide
    BlackKnight7891 , 19 March 2010 04:13
    It would be Nice to see some CPU\GPU temps through the testing especially when when testing the overclock against Crysis
  • 1 Hide
    LePhuronn , 19 March 2010 20:29
    Samsung SpinPoint F3s surely?
  • 0 Hide
    Anonymous , 13 July 2010 22:14
    hey all am looking for special desktop specs for animation design with high quality
  • 0 Hide
    mi1ez , 14 July 2010 00:02
    try the forums then rather than article comments...
  • 0 Hide
    LePhuronn , 14 July 2010 00:19
    Looking at it now, I'd also tweak things to run 2 5870s in Crossfire and drop the PSU to a 850W - games that don't benefit from Crossfire (i.e. GTA 4) will see better single-GPU performance from a full fat 5870 instead of the underclocked 5870 on the dual-GPU card, assuming of course that the 5970 is detected as an internal Crossfire setup by these games.