Sign in with
Sign up | Sign in
Your question

Memory/FSB: Correct Effective Speed for the Rated FSB

  • Memory
  • DDR2
  • Product
Last response: in Memory
October 11, 2007 10:03:18 PM

I'd like to find out whether or not I've got a good take on this, once and for all.

For Intel systems the rated FSB is quad pumped and therefore is divided by 4 to get the real FSB.
ex: 800mhz FSB divided by 4 = 200mhz FSB

Now the ram is double data rate and therefore the effective (advertised) speed is divided by 2 to get the real speed.
ex: PC2-3200 (DDR2-400) 400mhz divided by 2 = 200mhz

Technically speaking and using DDR2 just as an example, the requirements for a 1:1 ratio of the memory bus and FSB are as follows:

FSB@800mhz = PC2-3200 (DDR2-400mhz)
FSB@1066mhz = PC2-4200 (DDR2-533mhz)
FSB@1333mhz = PC2-5400 (DDR2-667mhz)
FSB@1600mhz = PC2-6400 (DDR2-800mhz)

Now, what I would like to know with regards to the Memory Bus and FSB is this: When you are NOT over clocking, what benefit is there in running a higher ratio than 1:1 in your system? I didn't think it made much of a difference whatsoever.

When I see MB specs advertising stuff like this I often wonder if I've missed something somewhere:

4 x DIMM, max. 8GB, DDR3 1800(O.C.)/1600(O.C.)/1333/1066/800 MHz, non-ECC, un-buffered memory

It would seem they are advertising that DDR3-1800/1600 are Over Clocked but 1333/1066/800 are not, yet you'd have to OC the FSB to get the full use of any of those speeds. I also wonder who has a computer that could even benefit from DDR3-1800mhz, that would mean you would need to have a real FSB speed of 900mhz (3600mhz advertised) to take full advantage??? That's Insane! I thought I used to know this stuff good enough, but anymore I'm not sure. I'm either getting duped by marketing hype or I've missed something along the way. I've looked at benchmarks before and the increase in bandwith never amounted to much from what I remember, as the latencies always had the greater impact. Again the question was about the ratio 1:1 and what benefit there is when running over that.

Thoughts/Comments appreciated.

More about : memory fsb correct effective speed rated fsb

October 11, 2007 10:57:34 PM

This is what I've been wondering as well. I've been having problems with my system lately and I think its RAM related.
October 11, 2007 11:51:09 PM

Remy said:

FSB@800mhz = PC2-3200 (DDR2-400mhz)
FSB@1066mhz = PC2-4200 (DDR2-533mhz)
FSB@1333mhz = PC2-5400 (DDR2-667mhz)
FSB@1600mhz = PC2-6400 (DDR2-800mhz)

It *looks* right to me if you take into account dual channel doubling your memory bus width. Of course I can't shake the feeling that something *feels* wrong about it.Because of the price there really isn't much reason to get anything slower than PC2-6400 these days. A decent number of people probably match 1066MHz memory to the 1066MHz bus.

Related resources
October 12, 2007 12:24:44 AM

I completely agree about the price thing. As far as the memory it's almost like you said, I keep seeing 1066mhz memory and 1066 fsb and I see the mb manufacturers marketing for the new processors, rated 1333mhz fsb with native support for DDR3 1333 and then "O.C DDR3-1600" and I start to second guess myself, it borderlines false advertising. Price you pay for memorizing how something works instead of understanding how it works.
October 12, 2007 4:15:27 AM

In the computer world there seems to be ALOT of this... examples...

Harddrives cannot transfer data as fast as what SATA150 can handle yet we have SATA300

The PCI-e1.0 bandwidth doesnt get fully utilised by an 8800GTX yet we have PCI-e2.0 with double bandwidth we cannot utilise yet

Optical drives have no hope in hell of transferring enough data off the disk to utilise ATA-133 let alone SATA bandwidth

When 4 core CPUs first came out no program could utilise them all

Most of it is all marketing hype to get people to buy the more expensive stuff. Im not saying that im all knowing, or even know enough to confirm without a doubt that a Q6600 at stock will not utilise 1066mhz ram. With the knowledge i do have though, there are alot of things that dont add up in the computing world and the only explaination is companies wanting more money for there "high-end" products that dont nessecarily make any difference in the real world.