Sign in with
Sign up | Sign in
Your question

FSB to DRAM frequency ratio really important?

Last response: in Overclocking
Share
May 26, 2008 3:28:30 PM

my computer is already not a 1:1 even when not overclocked. 333mhz fsb: 400mhz DDR2
is the ration really important?
right now my computer is overclocked to 1500mhz rated FSB (375MHZ) and my memory is at 500 MHZ.

specs

MSI p7n 750i platinum

Xeon e3110 (8400) overclocked to 3375 MHZ on default voltages (i can get up to i believe a 1530 ish rated fsb before my computer starts to crash because i am still on stock voltage)

2x1gb corsair xms2 ddr800 overclocked to 1000mhz 5-5-5-15 on default voltage (not sure how much higher i want to go but i know i can reach 1066mhz at this timings and it is stable - havent checked for errors for a long time but a 10 min test in prime95 showed none so far)
http://www.newegg.com/Product/Product.aspx?Item=N82E168...

psu: pc power and cooling 610w (49A 12v)

xfx geforce 8800gts 320mb (core/shader/memory) 580/1500/900 (i believe xfx had this card overclocked so i left it as is)

case antec lanboy (old model but has 2x 120mm fans -1 front 1 back so decent airflow... overall temps decrease at most 1C when side door is off, load and idle.)

cpu cooler is artic freezer 7 pro. its running about at 60% fan speed and wont speed up untill cpu reaches 40C which only happens after running small ffts in prime95 for a long time.



my question is how important is a 1:1 fsb dram ratio? i have better performance when my ram is overclocked to 500mhz (1000mhz) 5-5-5-15 than when it is 409 mhz at a reduced timing (4-4-4-12) where it is closer to a 1:1 fsb dram ratio.
also how do you know when you should stop overclocking RAM. there are no temp readings from it that i know of. it seems to be stable at 1000mhz (200mhz overclocked) at 2.1v (stock voltage)
May 26, 2008 7:01:34 PM

What I have discovered is that FSB to RAM ratios dont really mean much. As in, FSB 333 and ram ran higher.

If you wanna find out for your self, simply use the CPU multi to adjust the speed so its near the same and up the FSB from 333 to 400 and leave all else the same and then test! (Make sure you test at 333 first!)

You'll easily see that a higher FSB effects your RAM.

I only wish I could run mine faster and maintain my timings!!

365 FSB in sync at 4-4-3-15 1 Tick command rate! Need more power!!

--Lupi
May 26, 2008 7:06:09 PM

yeah i could try that. ill have to up the voltage on the fsb though as it wont go higher than a 1520-1530 rated fsb on stock voltage.
Related resources
Can't find your answer ? Ask !
May 26, 2008 7:21:39 PM

Yup, 1.45! I did the super pi 32 million test, because its 15 mins long, and from a simple change of FSB, I lost 30 seconds.

That may not be alot to me and you! But if you notice it after 15 mins, its fast to the computer, because RAM was made for burst mode!

Just be sure to adjust your multi to compensate for the FSB so the CPU speed remains the same!

--Lupi
a b K Overclocking
May 26, 2008 7:45:26 PM

@OP: I recommend you run Memtest86+ since you are OCing the RAM.
May 26, 2008 8:08:58 PM

yeah i did run memtest for 2 complete runs when the memory was at 1000mhz, no errors.
now im kinda having a problem figuring a stable overclock. i have it at fsb 1600 (400) and ram is back at 800. my cpu is 3.6ghz which is nice but im having ahard time trying to find a stable voltage.
right now my cpu voltage is between 1.248-1.27 which is +0.1375 from default. and fsb is at 1.3v if i increase fsb more it becomes red in the bios (meaning its unsafe according to the bios)
my cpu core temps are running a bit hotter under full stress from prime 95. 70C for core#0 and 63C for core#1, on stock its around 64C/58C respective. granted this is the individual core speed which is within the limits according to tjmax on the temp calibrating guide here. nonetheless im wondering should i be increasing the vcore more if this current voltage proves unstable? or increase fsb voltage even though the next step up says its unsafe and could cuase instability.

edit: crashed during prime95 after couple mins in the middle of editing this. this time around i increased fsb voltage to 1.325 (one step up). still wondering though if this could be causeing more harm than good by increasing voltage this much to get 3.6ghz stable.
May 26, 2008 8:18:37 PM

I believe the Max is 1.45 on that chip? So stick with 1.4 or lower if you can. Some boards think they are cool by showing green, yellow and red values. Dont worry about it to much. Just be sure to look up the max voltage of your chip.

What FSB voltage? You mean NB voltage? Leave FSB termination Voltage alone! (VTT!) Use the North Bridge voltage to support the higher FSB!!

--lupi
May 26, 2008 8:19:42 PM

Ahh, yes. VTT would be red at 1.3000 for a 45 NM chip. Leave that on auto.
May 26, 2008 8:23:53 PM

Well... VTT is a strange device. You never want it over VCore unless its at its lowest value. (1.1 I think for 45 NM chips?)

VTT is the voltage termination point! Your chip will always receive this voltage on a low level logic transition to high level. So if its 1.3000 your chip always sees 1.3000 when going from a zero to a one!

--Lupi
May 26, 2008 8:24:50 PM

really? ive heard people do well overclocking fsb and just increasing fsb voltage on my motherboard: msi p7n platinum. also it seems more stable now. yeah its the VTT FSB voltage. so i shouldnt mess with that one even though it makes my system a bit more stable?

edit:
voltage of the e3110 is 1.225V-0.956V according to the specifications on newegg. although people reviewing did say they were running between 1.25-1.4 vcore. im just wondering though should i decrease the vtt fsb voltage and increase vcore even though it will increase heat and possibly wont be as stable?
May 26, 2008 8:30:48 PM

alright my system was still not stable at the above voltages. so i went and reduced vtt back to default 1.2v and increased vcore to about 1.3v. however windwos crashed shortly after loading. i then increased northbridge voltage by 2 steps up to 1.32v from 1.26 i believe. ill try stress testing this again.
whats safe for northbridge voltage?
May 26, 2008 8:45:45 PM

pretty high, 1.55-1.60 But 400 FSB should only need 1.45 or so, you may only have 1.55 on NVidia chipset. Whats it called, MCH? MCP? Hmmm.. MCP is NB, and the other is SB, I think! media and communications processor?

I hate all the different names!

It may simply be VCore, because NVidia is famous for their Max intel spec VDroop and drop.

As in, the voltage gets lower when booting into windows, and then really low when you run the chip at maximum frequency as well as maximum power consumption. That = BSOD.

--Lupi
May 26, 2008 8:46:57 PM

A 1:1 ratio is not much gain with an Intel CPU...but when used with AMD and FSB/HT is boosted it will make the memory bandwith skyrocket.

Even my older Optrons with G.Skill PC-4000 HZ RAM OC'd my bandwith is 1K points in Everest over the fastest DDR2 ram (stock) sold today by OCZ/Corsair.
All my OC's are on stock volts (CPU/MB/RAM) with an average 1Ghz on the slower CPU's.
May 26, 2008 8:59:01 PM

yeah at idle my core voltage is 1.312v but when i was stress testing it it dropped to 1.28v although it seems increasing the northbridge voltage has made my computer stable. im just wondering now to try and lower core voltage to try and bring my temps down. was reading 72C and 66C on core temp but was 6-7C lower on realtemp. ill jsut do some calibrating on realtemp to try and get this more accurate.
May 26, 2008 9:20:56 PM

Thats a smart plan! Its always wise to find the lowest stable VCore voltage for a given speed. Glad its gaining in stability!

--Lupi
May 26, 2008 9:35:07 PM

wish me luck , stress testing at 1.32v NB (+2 increments) and 1.168 V core (+4 increments/ +0.05V) during load, vcore is a bit higher on idle. 3.6ghz @ 400x9.

+3 increments on vcore resulted in an error in prime 95 within 2 mins. so far no errors in 5 mins.

thanks for all the help so far guys, deffinitly saved time for me.
May 26, 2008 10:18:10 PM

:) 
May 26, 2008 10:31:37 PM

meh had to increase voltage several times on the vcore. im thinking that from reviews of this cpu that my mobo has a lower voltage default. im at 1.216v under load 1.23 under idle atm. from what i hear from other reviews of the e3110 they had their mobo vcore default at 1.2v and were able to obtain 3.6ghz overclock.

once i get this stable im wondering if i can then enable intel eist, lowers multiplier to save energy when idle. im not sure but i think it also lowers voltage and ide hate for it to crash everytime it went into idle. you guys know anything about that?
May 26, 2008 10:37:30 PM

Not sure about that, and I guess it all depends on your initial VID as far as how much VCore ya need for a speed.

Run Core Temp and look for the field, VID and then list that here.

--Lupi
May 27, 2008 12:03:48 AM

VID = 1.1500

not sure what VID is.

computer locked after about 2-3 hours of prime95 at the previous 1.216vcore under load. upped vcore again its now 1.248 under idle. wierd thing is it goes up to 1.3v under load but thats probably because i have EIST enabled. im wondering if instead of upping the vcore i should have just upped northbridge or even the vtt fsb 1 notch. now my temps on my cores are about 3-5C higher just 1 jump in voltage vs marginal before in my increments.
May 27, 2008 12:24:34 AM

VID is your starting factory assigned Voltage! (Thats what it is when on auto at stock speeds.)

It is best that you disable EIST and speed step so it doesnt confuse whats going on!

Leave VTT alone for now, its only useful for high OCs or low under clocking. North bridge should be 1.45 for 400FSB and there abouts.

--Lupi
May 27, 2008 12:25:33 AM

1.45 NB? mines only at 1.32 which is already 2 steps above default. i guess i could go more there.

edit: as of right now though im not seeing much of a performance gain from a 1500fsb with 1000mhz dram.

fsb 1500mhz ddr1000mhz 5-5-5-15
11137 3DMarks
4814
4827
3114

fsb 1600 ddr800 4-4-4-12
11339
4831
4844
3350

ill try it out with 1000 mhz ram again but i dont think thats gonna make a big difference. 1500fsb was all on stock voltages.

edit:
fsb 1600 ddr1066 5-5-5-15
11389 3dmarks

so really i think having a 1500 fsb with my ram overclocked to 1066 all on stock voltages provides the optimal performance and power consumption wiht lowest risk dmg to parts...

ive also compared fps in cod4 on all occasions. and the variance is marginal. 1-2 fps gain/lost etc tops. granted there are other benchmarks i could run that would specifically target the cpu. however i mainly use my computer for games and video playback.

thanks for all your help. it was nice to see that i could indeed reach 3.6ghz even though the benefit of that is not great enough to warrant a permanent overclock, and from what i hear the diference between 3.6ghz and 4ghz is even less than 3ghz - 3.6ghz i think ill stop here for now and just revert back to 3.375ghz and 1066 mhz dram - a small overclock but is stable under stock voltages.
!