533Mhz FSB or 800Mhz

[elektrotek1]

How big of a difference would a 800mhz FSB in a P4 processor be compared to a 533Mhz FSB P4 processsor that will be used for number crunching? It operation will take months, so would the 800mhz make a huge difference compared to the 533mhz FSB?

 

[4W4K3]

How big of a difference would a 800mhz FSB in a P4 processor be compared to a 533Mhz FSB P4 processsor that will be used for number crunching? It operation will take months, so would the 800mhz make a huge difference compared to the 533mhz FSB?

i believe faster FSB reduces multiplier...so you would have greater performance and more head room when number crunching. as intel's are locked multipliers higher FSB is always the way to go...the higher the better, more speed, more performance. but if you cannot afford a 800FSB CPU you can always overclock the 533 as high as it will go and hope you close the gap. but thats what i think...i could be wrong...doubt it though.

 

[Praetor]

The #1 bottleneck with computing performance is not the CPU speed but rather the memory speed. With the Canterwood setup, your FSB is running at a "clean" multiple of the (ideal) memory (PC3200)... if you run a 533Mhz FSB then the memory doesnt quite get a "clean" synchronization. If you are looking for performance, take a look at the Canterwood/Springdale platform and dual-channel PC3200 memory.

 

[4W4K3]

The #1 bottleneck with computing performance is not the CPU speed but rather the memory speed. With the Canterwood setup, your FSB is running at a "clean" multiple of the (ideal) memory (PC3200)... if you run a 533Mhz FSB then the memory doesnt quite get a "clean" synchronization. If you are looking for performance, take a look at the Canterwood/Springdale platform and dual-channel PC3200 memory.

thats my only beef with intel. if you want to run 1:1 you gotta have $$$$ to get high performance ram. but then even that expensive ram has really bad timings. im used to 2-2-2-5 timings at least...so if i went intel i'd be stuck with PC4200 or sumthin that runs like 3-2-3-8 (if i could even afford it)

 

[Praetor]

so if I went intel i'd be stuck with PC4200 or sumthin that runs like 3-2-3-8 (if I could even afford it)

Aggressive timings have their benifits and high-clock speeds have their benifits. I cant remember which is which but one favors burst mode and the other favors accesstime (i think clock speed favors burst mode however) so really it doesnt matter all that much however for people with specialized requirements (i.e., video editing where massive chunks of data are being transmitted so a higher clock on the memory would be ideal). For people dealing with lots of simultaneous accesses in memory, aggressive timings are the route to go.... of course if you push for either aggressive timings or high clock speed then "avg gamer Joe" should be fine.

 

[4W4K3]

Aggressive timings have their benifits and high-clock speeds have their benifits. I cant remember which is which but one favors burst mode and the other favors accesstime (i think clock speed favors burst mode however) so really it doesnt matter all that much however for people with specialized requirements (i.e., video editing where massive chunks of data are being transmitted so a higher clock on the memory would be ideal). For people dealing with lots of simultaneous accesses in memory, aggressive timings are the route to go.... of course if you push for either aggressive timings or high clock speed then "avg gamer Joe" should be fine.

AMD has kind of raised my expectations too high lol. im looking at ram and seeing really loose timings and thinking "thats crap" when really it probably performs better than my ram. once i break that mold of thinking though i'll realize i was such an idiot. i get too focused on those litttle numbers and forget about actual real life performance...its all about numbers lol. its probably only a fraqction of a second difference with some of the timings...something i cant even see...but i'll be stupid enough to pay an extra $30 for it lol.

 

[Praetor]

its probably only a fraqction of a second difference with some of the timings...something I cant even see...

Yes but those fractional-second differences happen trillions of times over and over and over and thus they add up reall fast. The timings are important and so is the ultra high clock speed. WHich is why you probably wont see PC4700 memory or something like that running with 2-2-2-4 becuase you cant really have both

 

[Sophocles]

The #1 bottleneck with computing performance is not the CPU speed but rather the memory speed. I'm thinking #2 right after the much slower hard disk.

 

[Praetor]

The #1 bottleneck with computing performance is not the CPU speed but rather the memory speed. I'm thinking #2 right after the much slower hard disk.

Bingo. Modern CPUs are more than fast enough for virtually every task we can conjure up... they are horribly crippled by the lack of memory bandiwdth (which was why DDR was such a big thing when it came around). As for 2nd place... not just harddrives but the entire IDE-spec

 

[4W4K3]

Bingo. Modern CPUs are more than fast enough for virtually every task we can conjure up... they are horribly crippled by the lack of memory bandiwdth (which was why DDR was such a big thing when it came around). As for 2nd place... not just harddrives but the entire IDE-spec

RAID seems to have wanted to fix that...but its too picky and $$$$. DDR2 isnt too far off as well..very nice bandwidth there. OCZ EB is the best solution as for now i think.

 

[Praetor]

Even with Striped RAID configuration, harddrives are several orders of magnitude slower than memory (i.e., memory is measured in nanoseconds, harddisks, in milliseconds); solid-state is the only thing that is comparable however it's bloody expensive.

If I recall correctly, as of now DDR2 doesnt offer much more than 10-15% realworld improvement in performance over dual-channel DDR; we'll have to wait and see though.

 

[4W4K3]

If I recall correctly, as of now DDR2 doesnt offer much more than 10-15% realworld improvement in performance over dual-channel DDR; we'll have to wait and see though.

wow..i thought it was going to hit BIG...guess not. like u said we'll have to wait and see lol.