How much does power delivery affect CPU performance?

[RenaKry]

One day, I decided to bench my i7-5820K clocked at 4.4 GHz on air using CPU-Z and I got some results, listed below. Obviously with a beast like the 5820K it should ideally be water cooled, but I got the thing on black friday and water coolers weren't on sale, so I was left with the Hyper 212 EVO. Cringe all you want, I know it is torture for the poor CPU, thermal throttling under max load, cooking itself to death, but worry not, I will eventually find adequate cooling for it; for now I lowered the thermal throttling temperature from 105 to 85 degrees Celsius, which is as low as the BIOS allows. But I digress; I had a billion PCs collecting dust in my room, so I decided to bench them using the same portable CPU-Z on a flash drive. 

 

Processor

Single Threaded Score / Multi Threaded Score (all threads)

 

Intel Core i7-5820K 4.4 GHz 1.3v on all 6 cores, 4.2 GHz 1.3v on Cache (I think that is also called Uncore, correct me if I'm wrong), with HyperThreading Enabled

1986 / 13246

 

AMD Phenom II X4 830 (non black edition) 3.36 GHz 1.25v (that is the readout CPU-Z gave me, if I had to guess this chip should run at least 1.45v stock) on all 4 cores, everything else is left on automatic (HyperTransport, etc.)

677 / 2032

 

AMD FX-6350 4.0 GHz 1.4v (CPU-Z readout, I'd lower this to 1.35v for a 100 MHz overclock) on all 3 modules, everything else left on automatic (voltage, etc.)

355 / 1462 (notice how shit these scores are)

 

and for shits and giggles, I benched a budget laptop sporting the AMD E2-3800 1.3 GHz (yes, you read that right) 1.075v, since it is a laptop, no overclocking was possible.

366 / 1033 (notice how this is better than the score above)

 

I left notes in parentheses where I want to focus you attention on. As you can see, the FX gets rekt by even the cheapo $300 laptop I bought a few years back, what's the big idea?

 

In one of my previous posts, I questioned why my FX-6350 was running so cold under 100% load, peaking at 45 degrees Celsius or so. The most common answer was that my motherboard for the FX-6350, the MSI 970A-G43, had weak power delivery. For comparison, my i7-5820K is installed on an ASRock Fatal1ty X99 Professional Gaming i7, which has "12+1 power phase" whereas I could only count up to 5 (or 4+1) on my MSI board. My Biostar TA790GX 128M (not the TA790GXM 128M) used for the Phenom seems to only have 4 (3+1) phases and still manages to destroy the FX-6350.

 

Basically what I am saying is that I am pretty confused as to why the FX chip performed the worst out of all. I took the single threaded scores and divided them by the clock speed (I am pretty sure that is NOT how you measure IPC but it is just for the sake of comparison) and got 451/GHz for the i7, 201/GHz for the Phenom, a staggering 88/GHz on the FX, and a whopping 281/GHz on the E2.

 

The question is why does my FX chip perform so poorly?

 

Edit: While I am at it, I looked at the references CPU-Z gave and the i7-5930K at stock speed performed just as well as my overclocked i7-5820K. Knowing that they are both the same chip but with different amount of lanes, why is it that my i7-5820K is performing worse than what essentially is itself? Bumping up the clock speed to 4.5 GHz makes the system unstable but it does push it over 2000 marks on single threads, so I know how it scales; basically my 5820K is performing worse than the 5930K by the GHz. Some explain?

Edited March 6, 2017 by nelsonpong

[Phentos]

A good motherboard should always deliver a steady flow of clean power to your CPU. To do this, the board needs high quality VRMs rated to steadily deliver their rated power output at 125C at least. It also needs high quality MOSFETs and capacitors to smooth out any remaining residual sine wave AC leaking through from the PSU, as the PSU is never perfect at eliminating AC sine wave current.

 

MSI's AM3+ motherboards are notorious for having poor quality power phases and merely average quality VRMs. Many boards either fry themselves, fry the processor, or just don't deliver anywhere near the required current for the processor. 

 

With that said, and taking from my personal experiences with MSI boards, I encourage others to avoid them like the plague, with the sole exception of their high end $200+ boards. Their high end boards are just as good as competing brands such as ASUS, Asrock, or Gigabyte.