My new system (With a RAM oc ?)

[Fightnfire498]

Ok, so I told myself I wasn't just going to start posting "How do I overlock my my q6600 or my Ram?" I wanted to try and figure it out by myself. Thanks to all of the guides and threads I read here, it was actually rather easy, I just have a question or two about RAM.

I am currently, according to Prime95 100% (for as long as I've tested) stable at 3.2ghz (355x9) on my Q6600. I actually lowered, by quite a bit, the default voltage on the mobo to the chip I am now at 1.28, which put my temps according to core temps and everest, right at 47-50c at a full load for over 20 minutes. My longest test so far, was almost two hours ... it gave me the same temps and no errors.

Ram question: According to Vista's rating system my Ram is a 5.5, it actually varies from a 5.0 to a 5.5, I was disappointed with that because I thought the ballistix would stock easily get a 5.8 or higher.

Settings I've tried:
1066 Mhz 5-5-5-15 2.0v (2t) was stable
800 Mhz 4-4-4-12 2.0v (1t) was stable
711mhz 4-4-4-12 2.0v (1t) was stable, I underclocked this because I think it gives me that 1:1 ratio everyone says is optimal.

Which of these, since none change the vista rating, is the best performance?

Thanks!

EDIT: I put this in this forum because I am tying my RAM to the Intel CPU. If it should be somewhere else, lash me and let me know please and sorry.

 

[jason4207]

1:1 is best when determining your max CPU and FSB OC, so that your RAM doesn't get in the way. If your fine where your at then you could use a divider and run your RAM faster. Use benchmarks to decide what you like better.

If your set on running 3.2GHz, I'd suggest using the 8 multi, and running 8x400. Then you can run your RAM at 1:1 800MHz and go even lower on the timings, or try running the RAM w/ a divider at a faster speed. You'll just have to experiment to see what you like better. I like to use Everest to test bandwidth at each different timing/divider setting and see what gives the best scores. Some dividers work better than others.

 

[Fightnfire498]

Thank you, I'll try that, I didn't really think about changing the mulitplier. Makes good sense tho.

Also, what voltage are you running your Ballistix at? And, The timings you show after your memory, I haven't even thought about changing the second set of timings after the 4-4-4-12 and 1 or 2t. Is that something I should be looking at as well?

Thanks

 

[jason4207]

The sub-timings and 1T/2T won't give any real-world performance, but I think it's fun to tweak them. It does help in benchmarks. I use memset in Windows to test out different timings in real-time, and see what works (use P95 large fft to test for immediate stability & Everest to see if the changes help or hurt, then run memtest to ensure stability).

What you see in my sig is at 2.15v. I also ran 1191 5-5-4-10 w/ 2.15v, but w/ only 2 sticks.

 

[Evilsizer]

MS benchmark is a joke, use PCmark05/06.

the 2 ratios for best performance is 1:1 and 2:3.

 

[jason4207]

MS benchmark is a joke, use PCmark05/06.

the 2 ratios for best performance is 1:1 and 2:3.

I've heard good things about 4:5 as well, but my source may be unreliable.

I did get better bandwidth scores at 4:5 1125 5-5-4-10 than I did at 1:1 900 4-4-4-8.

With 4:5 I had my strap at 266 instead of 333 (1:1) which means the NB sub-timings were a bit tighter.

Best thing to do is experiment for your self.

 

[Evilsizer]

yea the NB sub timing affect your memory bandwidth and overall system performance more then you realize. just look at the numbers when people bench at 500+ fsb speeds.

yea the 2 ratios i talked about is from Gauntam's testing with superpi and some other benchmark program. i dont recall what it was/is been to long since reading it...

 

[Fightnfire498]

Thanks for all the help guys

I have tried several different combos. I lowered the multiplier to 8x400 but I like to keep really low temps at full load, with that setting I was pushing 53-55 because I needed a little more voltage than 9x355, which runs full load at 47-51.

With the Memory, I have gone back and forth from 711mhz 4-4-4-10 1t, which is one to one at 9x355, 800 (stock)4-4-4-12 1t and oc'd to 1066mhz at 5-5-5-15 2t. I seem to get the best performance out of the latter two. Keeping the CPU at 9x355 what do the other ratios work out to be on the memory?

I also had to RMA one of my 2gig kits, so I am only running at 2x1gig atm ... the other two are on their way. Maybe with all four in I will see different memory performance? Overall, I haven't been very pleased with the ballistix ...

 

[Evilsizer]

that would be your problem, cmd rate of 1t is promblematic on intel setups. that also has no RW gains for going to 1t, all it does is cause headaches.

 

[00MJR]

If you want to do ram benchmarks one program I love using is Everest Ultimate Edition. It allows you to run a Cache/Memory Benchmark. (under tools). Use this to see the differences between the cpu overclocks and ram overclocks. Just take some screenshots using the print screen key and use paint to save them and compare.

Also is you just want check ram stability I use Memtest. This gets a test in 30 minutes and lets you see if the mem is stable enough to run Prime95 v25.6.

Also, with that memory you might be able to get 1100Mhz stock voltages stable(2.2v). If you got any higher then 2.2v I would recommend using some kind of cooling on the ram.

my 2 cents

Matt R.