This is from my own personal experience and not "Synthetic Benchmarks" which do not account for real world performance.

Here is what you do for setting up the system.

1) Make sure you have a BIOS that can allow CPU overclocking and 1 : 1 Memory Synchronization.

2) You will be underclocking and overclocking and testing in with SMP set to 0 and SMP set to 1.

You will be using Quake 3 TimeDemo and four.demo 

Current Architecture actually has the CPU decide the limit of the video card itself. Your CPU past a certain speed can tap 100% of the video card's abilities, however as you reduce your processing power through downclocking, you restrict your video card ability eventually below 100%.

Without FANCY Synthetic Benchmarks..The very first thing you will notice in the Quake 3 time Demo between SMP 0 and SMP 1 is that your framerate is a lot higher at SMP 1 than SMP 0. 

Once you reach 100% Video Card ability, through your processing speed being high enough to tap all its power...then the way you increase your framerate is by overclocking the memory speed of the video card and synchronizing it with the core speed. 

You will find that as you underclock your processor, eventually your framerate starts going down. Another thing to increase performance.......and framerate is a 64-bit OS...

Here is how....

The maximum limit of registers a 32-bit program can use is 8. At 32-bits, the processors in 32-bit mode have 32-bit wide registers. In 64-bit OSes, the C2Ds and C2Qs while still using 8 registers, the depth of those registers go from 32 to 64 bit. Two 32 bit EAX instructions get flagged as RAL (covering bits 0 - 31) and RAH (covering bits 32- 63) allowing double the data to be stored in registers to make efficient use of them....

C++ has a keyword called Register. It stores a variable in a processor register which is 10 - 100 faster than L2 cache. All other register variables are stored in L2 when the registers are full.

The reality of the issue is that 3DMark and other synthetic Benchmark Programs don't tell you anything about the real world.

If you take a game like crysis and start underclocking your CPU, your maximum framerate will drop, from your CPU preventing your video card from using 100% of its abilities.

Start taking screenshots at the SAME RESOLUTION you normally use for gaming. Underclock and overclock your CPU and watch what happens. 

Anyone who wants the screenshots of this can email me and I will be glad to send the screenshots.

Don't you have to be running ME2 to run Vantage? What about those of us who are not masochists?

Another point of intrest if you have Vantage, is NEW Vantage out last Week. its Vantage 1.01.

If you bought Vantage, needn't Fret, Just take your OLD Activation Number & go to Vantage website of 3dMark, hunt bit & exchange old Activation number for New Number & download Improved Vantage 1.01
drashek

Some scores nearly doubled, yet NO where near reported 45,000 of Vantage X2 ATI in Tri CrossfireX hybrid, several months ago. So Many Extreme reports go relatively unnoticed or ununderstood(even fictionalized) & rest of industry takes years to catch up. At least frame rate goes to 50 fps in one test.

However, how about CPU2, scores of 8. Guess second cpu didn't do much.
drashek

Speaking as an owner of a ASUS Striker II Extreme. You CAN hit the 2000Mhz mark with this board, you just have to do it manually. ASUS implementation of EPP 2.0 (SPD) is crap. It just flat out does not work. But if you spend the effort to do a proper overclock you can his the 2000Mhz mark.

BIOS is the failure of this board. ASUS support is a joke too. Were I to be able to stomach throwing a $499 motherboard in the trash, I would. The EVGA 790i Ultra motherboard seems to be the best of the 790i's.

On topic: I agree completely with your observations from the article. Sorry you had to listen to my rant on ASUS. They make good boards... the Striker II Extreme just isn't one of them.

I like the way nVidia is going with this, we all know their chips are more efficient nearly tenfold and if made capable of regular binary processing, they would prove vastly more powerful than our current cpu's.

However, it's things like this that show how great an idea AMD Fusion was, because this whole thing is gonna start confusing the hell out of people.