running a low FSB affects performance, RAM in particular?

Jotosuds · Sep 14, 2008

it seems from reading around that the e5200 maxes out at ~320 fsb. i'm trying to figure out how to run the RAM because of this.

assuming 3.5ghz (280 x 12.5) the RAM (DDR2-800) speed at 4:5 would be 700 Mhz and at 1:1 would be 560 Mhz correct?

i'm guessing you want to run your RAM as close to its rated speed of 800 Mhz for best performance so you'd want to use the 4:5 divider?

how about using a higher multiplier like 2:3 and having the memory clock be 840 Mhz? (280x2)/(2/3)

would this be faster than running the RAM 1:1 and lowering the timings?

ImSpartacus · Sep 15, 2008

Jotosuds said:
it seems from reading around that the e5200 maxes out at ~320 fsb. i'm trying to figure out how to run the RAM because of this.

assuming 3.5ghz (280 x 12.5) the RAM (DDR2-800) speed at 4:5 would be 700 Mhz and at 1:1 would be 560 Mhz correct?

i'm guessing you want to run your RAM as close to its rated speed of 800 Mhz for best performance so you'd want to use the 4:5 divider?

how about using a higher multiplier like 2:3 and having the memory clock be 840 Mhz? (280x2)/(2/3)

would this be faster than running the RAM 1:1 and lowering the timings?

Keep your RAM at 1:1 if it's that close.

Also keep in mind you don't have to try for 12.5x280. Why not try for 9x333.

If you really want an acceptable overclock, just snag an E7200 and run 8x400 or 9x400 to make use of that RAM.

th3 · Sep 15, 2008

Jotosuds said:
how about using a higher multiplier like 2:3 and having the memory clock be 840 Mhz? (280x2)/(2/3)

would this be faster than running the RAM 1:1 and lowering the timings?

Yes thats faster, speed is more important than tight timings (just look at DDR3 performance on Intel chipsets for proof). The FSB has twice the bandwidth of RAM at same frequency, theoretically you need 1:2 to max it out.

Evilsizer · Sep 15, 2008

i have something different for you to do.... in the bios lower your multi to 9 then set fsb to 400mhz. this with the 1:1 ratio will put your ram in spec this will how ever put the cpu at 3.6ghz instead of 3.5ghz. though the cpu should have no issue at this speed...

It is better to run a higher fsb as it yeilder a more EFF cpu per clock and has overall better system performance. as well in terms of latency when looking at ddr speeds to timings, DDR2-800 cas4 = DDR2-1000 cas5.
http://www.thetechrepository.com/showthread.php?t=160

th3 · Sep 15, 2008

Good idea, im sure someone else would have thought of that also if E5200 could do FSB 400. Which they cant, looks like some cant even do 333. Mine maxes out at 336-337. Anyway low FSB doesnt affect real world performance, it barely affect any benchmarks also, if RAM is set to a somewhat equal level in both scenarios.

Bottomsup · Sep 15, 2008

th3 said:
Good idea, im sure someone else would have thought of that also if E5200 could do FSB 400. Which they cant, looks like some cant even do 333. Mine maxes out at 336-337. Anyway low FSB doesnt affect real world performance, it barely affect any benchmarks also, if RAM is set to a somewhat equal level in both scenarios.

I have no proof of this, but I'd imagine it does show a noticeable difference in games or media encoding. I'm glad I read this thread because I have G.Skill HZ memory, but am only running my Q6600 at 333 mhz. I may try to lower my multi to make better use of my RAM.

th3 · Sep 15, 2008

Experimenting is good =) It has been done before and FSB speed is really not important and 1:1 is not better than any other divider, what matters is the RAM speed and even that has only small impact in benchmarks (except WinRAR and a few synthetic ones)

Bottomsup · Sep 15, 2008

th3 said:
Experimenting is good =) It has been done before and FSB speed is really not important and 1:1 is not better than any other divider, what matters is the RAM speed and even that has only small impact in benchmarks (except WinRAR and a few synthetic ones)

So in that case I'm better off leaving it running at 333x9 1:1 and tightening up my timings?

th3 · Sep 15, 2008

No, loosen the timings and increase the RAM speed. As mentioned, look at the DDR3 performance, its all speed and loose timings.

I like to remind about 2 years ago when the "high FSB era" started with the introduction of Intel P965. One of the really popular boards was the Asus P5B-Deluxe, nicely performing board but performance dropped noticeably at FSB 400. Some boards still do that, they set the strap automatically based on the FSB you selected, in the case of P5B the 400 strap loosened northbridge timings badly. So if your board will not let you override the strap, or if you are not sure of what you are doing, higher FSB can hurt performance.

Evilsizer · Sep 15, 2008

as tested by Gauntam here 1:1 and 2:3 ratio do show the best performance. some have shown the 4:5 ratio to be good as well.

as far as the fsb limitation, pad mod the cpu to 266 from 200 that should solve that issue. as it was the same thing with E2000's and E4000's. i also happens on celeron-l's but i got a lucky 420 that did 400mhz fsb no issues on p35. no some boards have been able to work at higher fsb's without padding modding. i think it was some gigabyte boards, been a while since i read exact models so i dont know atm. i do suggest tring the pad mod to fix the fsb walls being seen since like i said has worked in the past. now be mindfull that gigabyte boards will not take to pad modding, they ignore it like you just put in the cpu in.

as far as a higher fsb goes, its been proven that. the higher FSB does effect gaming and other areas. it also improves the cpus efficiency of doing work per clock, agian its been proven in past reviews on the net.

Jotosuds · Sep 15, 2008

Evilsizer said:
as tested by Gauntam here 1:1 and 2:3 ratio do show the best performance. some have shown the 4:5 ratio to be good as well.

2:3 would probably be best then as i could do (280x2)/(2/3) for 840Mhz memory clock? i don't see how 1:1 could work in my situation. this is all assuming there is a FSB wall and i can get ~3.5ghz oc w/safe volts

i read the OC sticky and it said that running a ratio can stress the NB or SB and you may have to up the voltage for them. i'm guessing it's worth the performance gain?

Evilsizer · Sep 15, 2008

well given your board is a P35 you will more then liky not need to. as the only case i noticed i needed to was on p965 at around 350ish, p35 doesnt need voltage like p965 does. the NB only gets stressed because it houses the memory controller so at higher speeds it will be needed at higher fsb speeds. you only going for DDR2-800 or a little higher so no you wont/dont need to increase the nb voltage. if your about gaming then yes its worth doing...

benbaked · Sep 15, 2008

Evilsizer said:
i do suggest tring the pad mod to fix the fsb walls being seen since like i said has worked in the past. now be mindfull that gigabyte boards will not take to pad modding, they ignore it like you just put in the cpu in.

Some (all?) Gigabyte boards are even worse then that, they flat out will not POST if you do a pad-mod.

Mr.Guvernment · Sep 15, 2008

havent past tests shown FSB vs Mutli debate is useless, or is that mainly only for the CPU? 3.2ghz is 3.2 is 3.2 kind of thing..

i guess when adding ram factor in it can make a diff, but then i guess the question is - how much?

Evilsizer · Sep 15, 2008

benbaked said:
Some (all?) Gigabyte boards are even worse then that, they flat out will not POST if you do a pad-mod.

i swore some stated it resulted in the same fsb as the cpu. though its been a while and im still "brain fried" from tring to find the right encoding program for my needs, things i do for a review. LOL

Mr.Guvernment said:
havent past tests shown FSB vs Mutli debate is useless, or is that mainly only for the CPU? 3.2ghz is 3.2 is 3.2 kind of thing..

i guess when adding ram factor in it can make a diff, but then i guess the question is - how much?

this my actually depend on the app more then anything... could also be kind of explained like this... if your trying to tow something would you rather have your engine at a lower rpm or higher rpm? even though your travling the same speed the higher rpm has more power behind it then the lower rpm does.

wingman99 · Sep 15, 2008

Evilsizer said:
this my actually depend on the app more then anything... could also be kind of explained like this... if your trying to tow something would you rather have your engine at a lower rpm or higher rpm? even though your travling the same speed the higher rpm has more power behind it then the lower rpm does.

Yes i agree with that but there is a point of saturation when more FSB cycles cant be used by the CPU, also i can't find any test past 1066 fsb to 1333 fsb in real world applications. :soda:

sonictron · Sep 19, 2008

Hey there, just dropping in to also say that its true, the E5200's are gimp. They are incapable of high FSB. Mine in fact will not POST over 310, on a 430+ capable board. Garbage chip, avoid at all costs, spend the extra 20-60 and get an e7200 or e8400.

running a low FSB affects performance, RAM in particular?

Jotosuds

Member

ImSpartacus

Member

th3

Member

Evilsizer

Senior Forum Spammer

th3

Member

Bottomsup

Member

th3

Member

Bottomsup

Member

th3

Member

Evilsizer

Senior Forum Spammer

Jotosuds

Member

Evilsizer

Senior Forum Spammer

benbaked

Folding/SETI/Rosetta Team Member

Mr.Guvernment

Member

Evilsizer

Senior Forum Spammer

wingman99

Member

sonictron

New Member

Similar threads