Benchmark Results: Audio/Video Benchmarks

Just a little lesson in history. People have been overclocking for a very long time. In fact, it became so common on the original IBM PC/AT, IBM put a timing loop in the ROM so that people couldn't boot the machine except at 6 MHz. And that was during a time when you actually had to have some technical skill to overclock (you'd have to remove the crystal from the motherboard, and resolder a faster one in - it wasn't as simple as changing settings in the BIOS). You'd also be overclocking the system bus, and the memory, and sometimes, even be forced to replace the memory. You could get to 8 MHz easily though, with no problems.

It's also always been, since the PC/AT, that current applications don't need anything faster. The original PC/AT was so much faster than the PC, it didn't seem real or possible. But, sure enough, software got slower and more bloated. None other than Thomas Pabst said that he couldn't understand why anyone would need a 1 GHz Pentium III or Athlon.

Also, heat dissipation keeps going up too. Need I remind you the original 8088 wasn't even hot to touch, and until the 486, you didn't even use a heat sink on the processor? Even the original Pentium, which was known for being ferociously hot in the original incarnation, did not use a fan.

It's only recently they've paid any attention to power use and decided that a little more performance for a lot more power isn't worth it.

I'm guessing anyone very interested in efficiency would probably be more curious about the Clarksdale.

Processors without Hyper-threading are inherently less efficient on heavily threaded tasks. I'd be curious if the i5 750 would be competitive with the Clarksdale on these same benchmarks, especially using the IGP. I'm guessing it wouldn't be, but I could be wrong.

Since this is an efficiency comparison, wouldn't undervolting on default clocks also be an option?

@ta152h

Speaking of history. Early PC's had an actual TURBO button on the case, so you could manually "OC" the comp, from something like 4 MHz to 6 MHz, if I remember correctly.
When I got my first PC the turbo was left on constantly witch really bothered my neighbor who was an programmer. "The 286 is powerful enough at 12 MHz. You will only fry up the chip at 16 MHz. No one needs 16 MHz anyway." In his words...

It wasn't overclocking at all. The "turbo" mode was actually the rated speed, the non-turbo mode was for compatibility with timing sensitive programs.

So, it wasn't overclocking to run in turbo mode, it was underclocking when you didn't. There was no danger of frying anything. But apps with timing loops would only run at the lower clock speed.

One caveat though, there was some companies that would overclock AMD's 16 MHz 286 to 20 MHz. There was also some companies that would scratch off Intel's markings, or put a sticker over them, and then run the processors over Intel's specs. It was deceitful overclocking, not like Dell grabbing a top end i7 and running it 4.0 GHz. They would take a 20 MHz 386 and say it was really a 25 MHz(i.e. a part that already existed), and save the money. Needless to say, this wasn't too common, and the companies that got caught ruined their reputation in short order.

The most efficient overclock I saw: my brother and I had both bought similar systems: the legendary Celeron 300A ('factory-overclockable' to 450 MHz) on the classic Abit BH6. Of course (as was most often the case with these chips), we both could boot it at 450 MHz right out of the box (mine actually booted at 450 MHz on the very first boot, and since I hadn't put a larger heat sink on it yet, I had to 'downclock' it). However, his was more successful than mine: while I could boot at 450 with the default voltage, I couldn't reduce it to lower than 1.92V (base is 2.0V), his worked reliably for years at 1.75V - and didn't need more than the original cooling block. A 50% overclock with 13% undervoltage is rather good, ain't it?

@ta152h

At the time this turbo button was explained as a way to reach EXTRA speed.
It was the same story in computer magazines, computer stores and computer "gurus" (like my neighbor), probably to help sell computers (just push the button and you get more bang for your buck for free). It did strike me as odd even then: if this was such a special feature, why is it on every single computer?
I put OC in quotes because as you said it was a standard and tinkering with hardware was not involved.

Way to go
i5 750

What cooler are you using in the review? I can't find it anywhere.

Also, would Ocing to a 150 MHz BCLK be ok on the stock intel cooler?

Christ, do we REALLY need yet another overclocking article for i5/7? Tom's bias is starting to become too much to bear.

Nice review.

You guys could do a "Power Efficiency Chart" including all processors that matter from a (price/perf)/watt point of view IHMO.

Price/perf is a good metric, but nowadays processors can use a lot of energy (and think about whole systems with big graphic cards), besides more heat usually means louder systems and I love a quiet and powerful (yet) PC. Keeping the electricity bill low and PC performance high is a nice thing to have, lol.

I loved the final graph, btw. Keep up the good work!

Cheers!

Apps are great and all, but how does the chart work out in games. Particularly, with multithreaded apps that minimize the mhz bump from turbo boost.

Very interesting article, answered quite a few questions for me. Any chance you could do the same thing for the i3-530? Now THAT would be a truly efficient machine.

I like to know that too for HTPC purpose.

Really useful information. Thanks for taking the time and effort to put it together.

It wouldn't be that interesting... Most modern games are barely optimized for 3 cores, and don't scale with the CPU in any meaningful way after 3GHz, unless you're running multiple GPUs or at a very low resolution. There's an effect on minimum FPS, but its not really anything to write home about. How interesting are plateaued curves to you?

Is it really that big a difference, with respect to the electrical bill, to run an i5 full bore at 4.2 GHz than in an efficient mode? I honestly have no idea, not trying to be a troll.

Efficiency tests are great to read and compare for that particular chip, but can be misleading to some. All CPUs are different and require difference voltages for difference frequencies, naturally. My i7 920 requires a high amount of voltage to reach a particular frequency, while other i7 920s don't require much at all. The efficiency/cost ratio on my CPU will obviously be higher than someone who has a CPU which requires very little voltage. The max I can reach at 1.3V is 3.6GHz with HT. Lots of CPUs I've seen have reached that with very little, if any BIOS adjustments for voltage.

According to this article, "http://www.tomshardware.com/reviews/overclock-core-i7,2268-4.html", A i7 920 required 1.5V to reach 4.0GHz. We have seen others reach that same frequency with a D0 stepping at 1.3V. Also, that article shows that 4.0 GHz is inefficient, but on closer examination, one can see that it is just with that particular CPU, as others have reached frequencies with much lower voltages. With that being said my CPU requires 1.3V for 3.6GHz stable after 22hrs, while others have reached 4.0 GHz at the voltages. While it's great to see efficiency explored, it may not relevant to everyone persons' own experience due the the variances in CPU manufacturing/binning.

I'm not sure what Tom means.

"Please note that we had to switch off the Turbo Boost feature above 3.2 GHz, as the resulting maximum clock speed would have exceeded the highest reliable speed of roughly 4.2 GHz."

And then in the table below they list:

160 MHz BCLK

3,200 MHz (20x multiplier)
3,360 MHz (21x multiplier)
3,840 MHz (24x multiplier)*

170 MHz BCLK

3,400 MHz (20x multiplier)
3,570 MHz (21x multiplier)*
4,080 MHz (24x multiplier)*