1st O/c

[p3rfect]

What the.. how are you getting a dram frequency of 400mhz with that FSB.. :S

 

 

1:1 = DDR480

5:4 = DDR384

3:2 = DDR320

 

I can only choose 266, 300 and 400 in the BIOS, so you tell me.

[road-runner]

Lock it on 300 and see if it will go higher, it should if you have not reached the limits of your cpu or some other hardware, then try the 266 and see what it does. Nuclear always said use 5:4 that would be the 266 also set the ram voltage to high which on my board is 2.85 which is not enough, if your cpu does hold you back at around 3.8ghz is when I had to add the vdimm mod to get more volts to the RAM to get a higher clock speed. You do realize that overclocking can burn things up? I am not responsible I am just telling you what I had to do with my system to get over 4ghz out of it, and all cpus are different so you can only get what that chip will give you, do you have the box the cpu came in? if so see what fpo/batch# it is and post it back here or PM me.

Edited May 10, 2004 by road-runner

[p3rfect]

Thanks road-runner. My chip is a SL6WK, but I don't have the box so I can't find out the other info.

 

Isn't it faster/better to run at 1:1?

[Vampire]

Isn't it faster/better to run at 1:1?

I'd much rather be running 1:1 on my rig but my ram is only PC3500. Usually yes, it is better.

[JackRussell]

The recommendation I've always come across is 5:4 rather than 1:1 provided your ram can do tight timings.

 

Seems counter intuitive, but my theory (not fact) for why that is, is this: 5:4 actually reflects most closely the rate at which the cpu actually accesses ram on average. So 1:1 is overkill... at 5:4 and tight timings, from the CPU's perspective the ram is actually faster than at 1:1 and loose timings, because @1:1 ram is available for access more often than it needs to be and when it is accessed it's running loose timings and hence slower access.

 

I don't know if that's right. Just my theory as to why 5:4 seems to be recommended as prefered for higher performance.

 

In short, underclocking the RAM enables stable RAM at tighter, faster timings... Am I even close? Correct me if I'm wrong... I'd like to understand this once and for all.

[r_target]

It's strange,but it seems like you can make up for the loose timings by raw fsb speed to the point where it makes little to no difference.For an experiment I ran the Sandra memory test at my usual settings (267FSB,5:4,2-2-2-6) and with loose timings/higher FSB (275FSB,5:4,3-4-4-8) and the results were almost identical.

[JackRussell]

For an experiment I ran the Sandra memory test at my usual settings (267FSB,5:4,2-2-2-6) and with loose timings/higher FSB (275FSB,5:4,3-4-4-8) and the results were almost identical

I was thinking along the lines of comparison like the following:

 

ram in both cases pc4000 for instance

 

1) 250 fsb, underclocked ram 5:4, tight

2) 250 fsb, ram @speed 1:1, loose in order to be able to be @speed

 

1st would score better, becuz in 2nd you sacrifice timings to run ram "faster". cpu prefers underclocked ram that's accessable-more-efficiently as apposed to ram that's accessable-more-often

 

again I don't actually know... is that how it works and why people recommend 5:4 when possible with quality ram?

 

I've stayed with 1:1 and loose just coz it's simpler less hassle.

 

EDIT: If this qualifies as "hijacking the thread", Sorry. Just thinking this is relevant to the direction this has taken. (apart from wanting to know myself)

Edited May 11, 2004 by JackRussell

[Nuclear]

I'm flattered to see my name in all of these posts...kinda makes me feel special

 

I have said time and time again...no chip, stepping, or batch will guarantee a good OC. If it were possible, I wouldn't buy 3-5 chips at a time, I'd buy 1 and save some money. Unfortunately it doesn't work like that, so I buy multiple chips and test them, you guys only hear about the good ones.

 

I got extremely lucky, a 3.0c that does 4GHz on air is rare. I had to go thru 7 chips before I found it, and 4 other chips when I found my 4.4GHz chip. So, out of 11 chips, I have 2 that are definite 4GHz chips. Yeah, I could throw all but about 2 in the MachII and push them extremely hard, but that's too easy.

 

Now...I have never noticed a time when 5:4 ratio gave me better scores or performance than 1:1, but the gains are only noticeable in gaming or benching.

 

- At 250FSB 5:4 2-2-2-5 timings, I score around 5600-5700 in SiSandra MemTests

 

- At 250FSB 1:1 2-2-2-5, I usually score around 6500-6600 easily in the same tests

 

So, I pick up around 1K points by going 1:1. It's a lot harder, because I'm on PC3500. Timings and dividers all depend on the user and his level, everyone will have different ways.

[JackRussell]

Now...I have never noticed a time when 5:4 ratio gave me better scores or performance than 1:1, but the gains are only noticeable in gaming or benching.

Thanks Nuclear. Didn't mean to waste space... just that I've even seen OC'ing guides (in addition to the advice) with this 5:4 claim. Been trying to make sense of it, hence my wacky theory

[FanATIc205]

guess i lucked out then nuclear since my dinky slk800 and smartfan 2 with a fan placed blowing inside the case resulted in3.0 sl6wk@ 3.9 prime stable 12 hours. with out the added fan blowing in the case i could not reach this. it seems as soon as the temps hit 56+c the cpu acts up. cant wait for water cooling as i think 4ghz+ is attainable maybe more seeing as 1.7 volts was stable at 3.9

 

 

btw whats the packdate on the one you have at 4ghz ?

[Nuclear]

You too have a really good chip, and I've come to the conclusion that really good chips don't require an awesome spectacle of copper to get high speeds. You may do a bit better with a different HSF, but H20 will really unleash it.

 

My chip has never gotten over 50C, but I've read of some becoming very unstable when OC'ed and around 55-60C. I think the temp increase is a direct result of the added core layers and over twice the amount of working caps. I could be way off, but all of the hotter running chips have been newer D1 chips.

 

My 4GHz SL6WK was packed on 3-2-2004 or 3-4-2004, I'll look when I get home, but I'm pretty sure it was 3-2-2004.

[FanATIc205]

lol we might have the same pack dates mine was 3-2-04