Temperatures and Power Draw
Overall, we see very similar performance between the i5-2300 and the i5-2500, although the i5-2500 is a hair cooler and more power efficient. I didn't graph the results for our i5-2500K because it was negligibly different than our i5-2500, the difference being a few watts under load, and a watt or two at idle. Although they aren't remarkably different, those couple watts are what differentiate how the i5-2500 and the i5-2500K are binned, and that hint of added efficiency goes into the fully unlocked -K model chips. Similarly, the better-performing i5-2500 chips are binned because they're more efficient than the i5-2300s as they come off the manufacturing line, and we can clearly see that in our comparisons.
An interesting find is that we found a nearly identical "bottom" to our undervolting potential, at right around 1.00V. We were surprised to see that we were still able to obtain stable performance at 3.3GHz, the stock clock speed, and still push down to the 1.00V level. The allowed us to run the i5-2500 at lower power and thermal output than the stock i5-2300 at 2.8GHz, which is impressive.
The chips perform the same in terms of thermal output and power draw when the voltages and clocks are held the same, which is expected as their architectures are exactly the same. The Sandy Bridge family is wonderfully efficient, and at 3.8GHz, the i5-2500 only reaches just over 50 degrees Celsius, and a relatively low 84W of power. We do see the clear quadratic relationship of both temperature and power as you increase your clock speeds, and this is the reason chips hit a hard limit at around 4GHz; power and heat increase very quickly when the 4GHz threshold is reached. However, at only 50 degrees Celsius at 3.8GHz, it leaves alot of headroom and makes it obvious why we're able to push our i5-2500K to 4.2GHz with ease, and 4.4GHz isn't hard to do, either. However, as we've mentioned earlier, the non-overclocked i5-2500 is limited to 3.8GHz as its max overclocked frequency.