Overclocking ram : cpu ratio

Boyd117

Member
Nov 14, 2007
101
0
0
Hey I plan to overclock for my first time as soon as I choose an air cooled HS for my Q6600 and wanted to know what the difference of running the ram as the same speed of the cpu's FSB or running the ram overclocked as high as it will go.In my case I have 2x2gb of patriot pc2-6400 ram that I want to overclock, but wanted to know if I will loose or gain performance when overclocking my cpu? Trying to understand the different ratios (ex 1:1, 5:6 ....)

more specs below
 

n7

Elite Member
Jan 4, 2004
21,281
4
81
With a P35 chipset, generally the higher you can get the RAM, the better performance will be.

IOW, here's our hypothetical situation:

You get to 9x389 (3501 MHz).

You can run 1:1 DDR2-778 4-4-4.

Or you can run 4:5 DDR2-973 5-5-5.

In pretty much everything, the higher speed with looser timings will be faster.

But here's the thing, you'll never notice it other than benchmarks.

I encourage all those who ask these question to bench for themselves & see.

Try out different configs & see.

But as i've said a zillion times over, in general, higher bandwidth > timings on everything but nForce chipsets where you can 1T.

Again, in reality, you'll never see the difference though.

 

cozumel

Senior member
Nov 29, 2007
337
0
0
I'm not in full agreement with n7 as I think it depends on many factors such as the individual cpu/cores, mobo, chipset, ram ICs and whether they happen to work well together or not. In some instances I've got better performance with a slower frequency but tightener timings. I haven't yet finished my overclocking with this rig, still trying various combinations and am yet to lap the CPU. But I reckon that if I had my Q6600 at 400x9 (3.6GHz) and the 2x1GB Ballistix PC2-8500 (D9 ICs) running at 1:1 so they're underclocked at 800MHz I might be able to get something like 3-3-3-8 that would outperform me doing 1066 or 1100+.

But the point is Boyd is that there are no definites. You need to experiment with different configurations and see what works best for your set-up with benchmarking. You need every fps you can get when you are playing Crysis, Gears of War etc etc

Whether it is a difference that you will notice is another matter...
 

Boyd117

Member
Nov 14, 2007
101
0
0
Will having the ratio 1:1 be more effective in the cpu overclocking stage? I cant really test things yet because I'm still deciding on buying the Thermalright-ultra 120 extreme or wait on AC new heatsink to be released, so at this point I'm trying to get as much info as I can so I will have some understanding of what should be done when overclocking.
 

cozumel

Senior member
Nov 29, 2007
337
0
0
When you are overclocking the CPU you should always have the memory running at the same speed or slower than the CPU so that when you run stability tests you can be sure the only potential problem would be with the cpu overclock. So, for example, if you wanted a 3.0GHz Q6600 you could have 333x9 or 375x8 and use a 1:1 ratio which would give a DDR667 or DDR750 and would loosen timings to say 6-6-6-24 during stability stress test.

Use prime 25.5.1.0 to stress test as this version starts 4 threads that will check all four cores at the same time. Your temps should be ok with the stock cooler up to around this level of overclock as you probably will not need to raise the vcore voltage but I would advise using CoreTemp 0.95.4 to monitor temperatures to be on the safe side. With a G0 Q6600 you should stay below 70c (71c is Intel thermal spec) but I recommend under 65c.

Once you have had the stability test running at least 8-10 hours (although preferably 24 hours) you could tighten the timings on the memory. If wanted to see what you could then play around with overclocking the meory which is basically the same principal as the CPU but in reverse. Remember to keep the cpu speed as low as possible while checking for stability with the overclocked so you know if there is any problem it is due to the memory. Use memtest+ 1.70 to check stability on test #5 which you only need to run error free for around ten minutes. Once you reach a memory speed that you are happy with metest overnight for 8+ hours or prefereably overnight. Again, remember that memory can overheat and I tend to use the 'touch test' with my finger on the heatspreader and if it feels hot to touch then I would be concerned. I have a fan on an arm about an inch above the RAM blowing accross the modules and onto the northbridge to keep everything cool.