• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Memory Timing Question? Latency?

mab45

n00b
Joined
Apr 20, 2004
Messages
9
I am relatively new to the overclocking scene (not new to the build-my-own-cpu scene). I'm now addicted to getting the most outta my rig. I've been able to find the answers to almost all of my questions on overclocking w/ the exception of Memory and timings/latencies.

I have 4GB (4x1GB) Ballistix DDR2 800 PC2-6400 EPP RAM, current settings (see system specs below as well):
Freq: 356mhz
FSB:DRAM 1:1
tCL - 5
tRCD - 5
tRP - 5
tRAS - 18
Command Rate - 2T

Is it worth messing with these settings for any performance gains? ie: cost/benefit. I do a lot of DVD encoding and image editting, but otherwise just normal day-to-day stuff.

Using CPU-Z, my "Rated FSB" is currently 1424 MHz, and I'm not sure exactly what this means in the whole equation... what's my upper limit?

All help/comments are appreciated in advance!
Mike

___________________________
My System: :D
Q6600 G0 stepping Overclocked to 3.2ghz (9x356)
Abit IP35 Pro Mobo
4GB (4x1GB) Ballistix DDR2 800 RAM
MSI 8600GTS OC 256MB PCI-E
Raptor 10k 120GB Hard Drive
Corsair 520-HX Power Supply
 
you could probably get 4-4-4-15 from those sticks....I don't really think you would get any real world benefit, but tighter timings do seem to help in some applications....benchmarks being one of them..:)

as far as your "rated" FSB..that number is just 4X your actual FSB.... yours is at 356...So 356 X 4 =1424..

If you were to set your FSB to 400...the rated would show 1600..and so on....
 
First, try CPU multiplier 8 x 400 FSB and see if it will run. This will give you the same CPU speed (so the CPU should do it) but it will be interesting to see if the memory subsystem with those 4 x 1GB sticks would be stable with your current timings and the higher FSB. Note that with a 1:1 (memory running at 2 x FSB = 800MHz) you memory is running right at its rated speed and that is faster than your current settings and would provide more memory bandwidth/transfer speed.

The rub is:
a) will the board be "happy" with a higher FSB of 400MHz (it should, thats not all that high)

b) will the board be "happy" with the 4 x 1GB at an FSB of 400. The four sticks makes things twitchy, some boards do it fine, some don't.

None of this costs a dime other than your time.

A bump in MCH voltage while keeping an eye on MCH temps (you finger is an excellent temp probe) may help stability.

This would be the best thing you can do as a starter. Then if the system is "happy" you can look at memory sub timings and regular timings for tiny increases.
 
>>>Note that with a 1:1 (memory running at 2 x FSB = 800MHz) you memory is running right at its rated speed and that is faster than your current settings and would provide more memory bandwidth/transfer speed.


Sweet. This makes sense to me and I just did it without a hitch. RAM transfering data at 1600Mhz should process things faster than those working at 1300Mhz. Maybe it's just placebo effect, but in the few minutes I've been jumping around, I think my cpu is a bit more responsive.... is this possible????

>>>b) will the board be "happy" with the 4 x 1GB at an FSB of 400. The four sticks makes things twitchy, some boards do it fine, some don't.


Small white lie from my first post - I'm currently running 2GB b/c my other two sticks show up tomorrow, so I guess I'll see if it makes a difference when I pop the other 2GB in... I'll let you guys know.


>>>A bump in MCH voltage while keeping an eye on MCH temps (you finger is an excellent temp probe) may help stability.


If I only bump the MCH up a little bit, does it matter to closely monitor it separate from core temps? I'm currently using CoreTemp and my Quad Q6600 runs all four cores at about 34/26/34/26 on a Thermaltake Ultra Tower. I guess, I'm asking, do I need to separately be concerned about my MCH temps????? and where would I place my finger?


AND FINALLY, should I attempt to bump my timings down to 4-4-4-12 or something close? would it perform any better for the potential cost???
 
Excellent !!! well we will find out about the 4 sticks, actually its good that you verified its good to go with 2 at those settings. Now I dont want to alarm you, and all may be fine with 4 sticks, I hope so, but kinda sorta expect trouble and then when it does work you will be happy. ;)

Maybe it's just placebo effect, but in the few minutes I've been jumping around, I think my cpu is a bit more responsive.... is this possible????

Sure, you got almost 50MHz more memory speed and probally that is what you are feeling, That 8 x 400 for 3.2 and 800MHz memory speed is called a "sweet spot" for for a reason. and now you can start upping the FSB in small increments while watching your temps and see where you can go and it will just get snappier. About 425 FSB your memory might start showing the strain of the higher than stock speed but determining where that limit is, is half the fun. Raise FSB 2 MHz, check temps, test if you want, use the machine a couple of days, no problems, give it another shot. If you raise it up and a problem develops back of 4MHz, use it for a while if all ok, leave it there if temps are good. If necessary back off a bit more. Cant hurt anything, both the cpu and the chipset chips have thermal protection and will just shut down (crash) to protect themselves and go back to normal when cooled off, since we are not talking about raising the cpu voltage (a later option if you really wish to see what the ulitmate performance is and safe if done with a little care) and the addition I am suggesting to the MCH/Northbridge voltage is so low, no worries.

If I only bump the MCH up a little bit, does it matter to closely monitor it separate from core temps? I'm currently using CoreTemp and my Quad Q6600 runs all four cores at about 34/26/34/26 on a Thermaltake Ultra Tower. I guess, I'm asking, do I need to separately be concerned about my MCH temps????? and where would I place my finger?

Nope not critical, we are going to go easy, when you read this and hopefully before you pop in the extra memory, just place a finger on the side of the MCH heatsink, this is the heatsink closest to the middle of the board and get an idea of how hot it gets, it does get toasty. Best to check after running your favorite game for 1/2 hour or so or do whatever you normally do so as to get an idea of its temp under load, Adding the extra memory will raise the temp as the MCH has to drive 2x the memory chips as before but you will probally not be able to tell with your finger. Pop in the extra memory and see how it goes, If trouble, shutdown, and give it the first notch of available extra MCH voltage, this will be such a small amount as to be almost neglible, (I have had mine as much as +.3V over the chips rating without incident trying to get 4 x 1GB working on my older P965 board. Your MCH is even more advanced with better thermal protection. Try like the next 3 bumps in MCH voltage before you give up and yell for help. FYI the chips normal voltage is 1.25V and its absolute max voltage rating is 1.75V so if the increases in voltage that abit provides are as small as I remember 3 bumps will not even get you anywhere near 1.5V much less 1.75. If you get the 4 sticks running just do what you do, or run orthos etc. and after a bit of time to let it get up to full temp, check it with your finger again. I should be hotter, but if you can keep your finger on it for a slow 5 count it will be ok. A lot of us find a small fan and hotmelt glue it on if we feel that it is getting too hot. Not a bad idea anyway as heat is the enemy of electronics and the cooler it is the better.

AND FINALLY, should I attempt to bump my timings down to 4-4-4-12 or something close? would it perform any better for the potential cost???

In general faster FSB memory speed is more powerfull than the latency timings. 4 4 4 12 will be very tricy with 4 sticks. In general I would say no, and most definately not worth buying new memory if your current stuff is rated 5 5 5 18, Now if your current stuff is 4 4 4 12 and you are using the default 5 5 5 18 auto settings, after you get the 4 sticks running, hell yeah, its worth a try but expect trouble. At worst you clear the cmos and have to put your settings back in.

LOL lets get the 4 sticks up and running first. That abit has some very fancy advanced settings that could also help but they are pretty complicated and we want to save them for a last resort.
 
Awesome. I'm having fun already - I'll post back tonight after I start the adventure!
 
Just popped in my other 2 Crucial Ballistix DRR2 PC-3200 sticks of RAM --- now 4GB total, all currently running at 5-5-5-18 even though they're specs report they can be run up to 4-4-4-12.

Last night, I had already changed my overclocked settings to 8 x 400Mhz = 3200Mhz.

Everything's working like a charm. I haven't pushed the system super hard yet, but I personally wouldn't expect any reason for it to run differently... I'm gonna make the jump and try to decrease the latency/timings to 4-4-4-12 and see what happens!
 
Didn't work out well... I was moving along at 4-4-4-12 for about 5 minutes and then I started multi-tasking. My cpu blue-screened and started over, and then wouldn't start up. I had to clear my CMOS and put them back to 5-5-5-18 defaults.

Interesting though because apparantly my RAM has specs of 4-4-4-12 at DDR2-400MHz, so where did it go wrong??? Was this because I have 4GB?

PS-I even had the MCH voltage kicked up one notch.
 
go back to 5 5 5 15 or 5 5 5 18 for a few days to make sure that is stable.

In a word, yes, its normal, a dirty little secret is tha a LOT of boards will not run 4x1GB at high performance settings or OCed.

If the above proves stable try 5 4 4 12 and see what happens, give it another bump of MCH voltage.

Dont worry too much, you have the buss speed correct now to get the data transfers huming along, those timings are mainly set up and addressing latencies and once a row and column are addressed the data is burst mode (more than one memory location can be read and sent without having to do the whole addressing deal again). What I am saying is that there is not really any big performance difference between the 5 5 5 15 and the 4 4 4 12 for the things we typically do. Yes you can measure the difference with benchmarks etc, but otherwise /shrug. I warned ya, I could not get 4 x 1GB to even work at completely stock settings and crazy voltages but I do have an older board.

All that said I am not an expert in Crucial memory, and do not have that board. Higher MCH voltages, tweaking the voltage given to the memory chips themselves etc. might help but as I have no experience with Crucial I hesitate to say more. And as I mentioned there are some very advanced setting we have not even discussed. If an Abit expert does not wander by, I would consider finding a forum devoted to Abit or perhaps post in the motherboard section as now your issue is mainly with the motherboard.
 
I'm tempted to just leave things alone now b/c everything's stable at 5-5-5-18, especially because I don't have to play with any of the voltages.

Would it have any perceivable benefit to lower the 18 to 15 ???

Is Vista planning an upgrade at any point to allow utilization of more than 3.5 GB of RAM?

Thanks guys!
 
Back
Top