Same Battle or a Different One?

[butchconner]

I just bought 4 sticks of G.Skill 2gig memory DDR2-8500. I have two ASUS MBs, one with a Phenom x4 9500 and one with a Phenom II 955. I have played with another brand of DDR2-8500 on the 9500 and got it work simply by changing the settings, speed and upping the voltage to 2.100.

I plan to try the same with my G.Skill Memory on the Phenom II machine. It is now at 400as many people are reporting. I will try 5-5-5-15-2t, etc. at 2.0 v and 2.1 v and see how it works.

Question 1: Should my F2-8500cl5d memory have any trouble with the above?

Question 2: A couple of times I have seem people recommend just standing at 800 and going to very fast timing like 4-4-4-12. How much real difference is there between running at 800 with fast timing vs. 1066 with regular timing?

Question 3: If I run at the 800 "fast" settings, can I run 4 sticks (8gigs total) with my 64-bit Windows?

Thanks for any info.

RevBC

[genetix]

Think I can give you the honest answer as I doubt none of manufacturer peons would not nor most of OC people neither.

#1:
You will have issues on 5-5-5-15-2T for sure. It will not POST even, if it does it will error. Controller will not stand all slots full under 1066Mhz on these latencies. (come to this topic later, if needed since this depends on clock speeds, DDR ratio, OC & voltages and I have seen some boards capable of doing clean 5-5-5-15, but not on stock conditions.).

#2:
People recommending 800Mhz with low latencies are correct. Memory probably will go even CL4-4-3-14 on 800Mhz and while you OC the core it will be around 931Mhz where you would have to even touch the latencies. Also even with the memory controller on board it will never go so high speeds even on phenom II that it would reach speeds of ~900Mhz in sync and anything out of sync is slower even while internal memory copying speed (inside the memory) would be faster it won't be faster in realistic use only on synthetic benchmarks.

However, considering AMD platform on this one. the 1066Mhz would probably still have lower latency even on something like CL6-5-5-16 than CL4 on 800Mhz which you can as well test yourself on something like 'everest', but this would probably be slower.

*Exception: Unless you clock the core high enough when 1066Mhz would become closer to the speeds.

3#:
Yes you can. it will (dunno what model you got), but it should run pretty damn fast too. (System won't be the issue there you should consider value called 'tRC' your issue there well what goes to performance.)

[butchconner]

Additional info: Caveat -- I do not claim to be an overclocker or an expert.

I set up a memtest86 disk and ran several tests on my F2-8500CL5D-4GBPK memory. I am running 2 sticks (4 gigs) in the first pair of slots on my ASUS M4A78 Pro MB.

1. I set up the 1067 test with 5-5-5-15-30-2T at 2.08v. No errors on memtest86, so I booted into Vista and CPU-Z showed the above configuration. In a couple of minutes of checking mail, etc. I did not notice any glitches.

2. I set up an 800 agressive test with 4-4-4-12-24-2T at 1.9v. Same results. No test errors and booted and ran for a short time with no difficulties.

3. I did not run the 800 8gig test mentioned in the original post. Going out of town this week so adding the last two sticks and testing that will have to wait a week or so.

Anyone have any observations or suggestions about the above?

RevBC

[Tradesman]

Think you got your eggs split pretty well, sort of 6 of one half dozen of the other, if your not really into OCing the 800 w/ tighter timings might run a bit cooler and you'll be closer to a 1:1. In all honesty though, not sure you'll actually 'see' much of a difference between the two scenarios you mentioned except in benchmark programs. I can't set up the exact system you have, but in some cases running at a slower clock with tighter timings will produce better performance. This is something Genetix and I have touched on w/ each other, and it's almost you actually have to see it happen to believe it.