When you change the bus speed by overclocking it you also change the total CPU speed, so that comparison is flawed.

What Eug7 is wondering is if a 3GHz processor @ 800Mhz is much better than a 3GHz processor @ 533MHz.

What Eug7 is wondering is if a 3GHz processor @ 800Mhz is much better than a 3GHz processor @ 533MHz.

Soitanly.

http://www.dvinfo.net/conf/showthread.php?s=&threadid=18841
My own tests show that the difference is like 3%. However, I am only running single channel RAM (which no properly configured system should be using...) and my test may have other flaws. Running at a 3:2 divider to emulate lower FSB speed (533mhz instead of 800mhz) is slower than an actual 533mhz system (FSB and RAM have to wait around until they are in sync, which effectively worsens memory timings), but according to my own tests that would make no difference (memory timings make no damn difference).

Should you go for the 3Gz 800FSB Pentium instead of the 3.06ghz 533FSB? I don't know. The 3.06's 2% clock speed improvement helps lessen the gap a little, and the gap shouldn't be that big. Both have hyperthreading, so that's not an issue.

2- Why do you refer to yourself in the 3rd person?

>>>When you change the bus speed by overclocking it you also change the total CPU speed, so that comparison is flawed.

Oops that's right I forgot about that. By overclocking the bus speed the processor clock scales too. Okay forget about that! (that's why I like this forum) :D

I checked Tom's Hardware - for video encoding applications the 800MHz version of the P4 2.8GHz is between 5% and 15% faster than the 533 version (depending on the application).

http://www.tomshardware.com/cpu/20030521/800fsb-12.html

You can see that the 3.06 vs 3.0 are neck and neck but the 3.06 is a bit faster.

Increasing the BUS speed ALLOWS the CPU to access memory faster, actually MORE TIMES over a given (X) time span. That 's all you need to know. So yes, the higher the FSB, the faster the CPU can access memory otherwise lots of time in each clock cycle may be wasted with the CPU loafing kind of like traffic crawling along at rush hour on the "expressway".

Try to visualize it this way. A CPU clock cycle can be looked at a clock hand sweeping a clock one complete revolution. Of course its much faster happening many millions of times a second. In each cycle the CPU needs to do a bunch of housekeeping things like check if any of the IRQ's are requesting service, demands from the operating system, memory page swapping and so on.

For illustrative purposes assume those things happen between the second hand of a clock going from 0 seconds (12 o'clock) to 15 seconds past or (3 o'clock). The rest of the time the CPU is available to do what whatever running applications are asking of it unitil it gets back to 12 o'clock, then the cycle repeats.

So to over simplify things the CPU would be free to go grab memory pages between 16 seconds past and 59 seconds on our clock face example.

If you have a slow CPU and low speed FSB, over time X (lets say 10 seconds) it can only make Y number of trips to access memory. As you know a computer can't do anything unitl what it needs to be done is in memory.

Increasing the FSB speed (front side bus) is like making a real world clock run much faster.

While the CPU must still do its housekeeping in each cycle, because its now running faster it can access memory more times over time span X so the CPU won't be just sitting there doing nothing waiting for the next portion of the cycle when it can access memory again because the cycles now come much faster. The faster the CPU can get at what in needs and move it in and out of memory the faster you're applications will do whatever ithey are suppose to do.

The 800mhz and 533mhz of the P4 2.8ghz differ in that the 800mhz FSB one has hyperthreading.

As far as the 3.06 vs 3.0mhz CPUs go, they are pretty much neck to neck. I quickly ran over the benchmarks and render speeds scale nearly linearly with CPU clock speed. It seems like the FSB speed makes about a 2% difference in the 3.06B vs 3.0C. The CPU clock speed difference is 2%, while the difference in rendering speeds are 0.1% in the MPEG2 encode with the mainconcept codec [vegas uses a modification of the mainconcept codec I believe.

Combined with my results for Vegas rendering, it seems that FSB speed only makes a few percentage difference. I would have trouble telling the difference if Vegas didn't tell me how long renders took. The difference is small enough that you can mostly ignore it. Probably go for the cheaper processor. Consider the RAM too, you don't need as fast RAM with the 3.06B. Although getting slower RAM could hamper upgrading... but the Intel roadmap doesn't look so good (4ghz by the end of this year... only 33% faster AND you need a motherboard that supports Prescott). see anandtech.com

But really... I wouldn't worry over a few percentage difference.

Ok. I bought & installed the 2.8 800 FSB w/hyper onto an ASUS P4P800. My old system a P4 1.6, 100 MHz bus.

Old system renders AVI to MPEG2 (DVD) was about a 3x plus time frame. About 6.5 hours for a 2-hour project.

My test last night on a 15-minute project rendered in just under 15 minutes. Evidently, the upgrade will now yield a slightly better than real time result.

Yahooie!!!