I got a Q. Which one is better and faster, running at 2.8GHz with RAM timing of 4-4-4-12 or 3.0GHz with RAM timing 5-5-5-15?

it's not juz to bring out the performance and for prime and bench...
it to supply the proc with enough power to maintain it's clock speed...

always up your cpu speed 1st

This post has been edited by dblooi: Jul 5 2007, 10:09 AM

i do OCing.. not for benchmarking... show off n what so ever...
i spot for perfect setting of my rig compartment.. for better usage n performance.. and not neglecting the power consumed..
for me, it give more fun.. more variables to play on..
there's no point pushing so high.. and resulting unstable system..
in the end.. reduce their life span..

No one can be sure, it depends on what program you're running. Programs which are ram bandwidth/latency sensitive will suffer from the 5-5-5-15 timing, while processor intensive applications will suffer from a lower clockspeed. Even games, some are known favour better ram performance while others benefit from more processor power. Its your call, up to you which is better for your use. Use benchmarks to help you make up your mind.

Misconception here-the proc does not need more power. The real reason is that the transistors can switch faster if they have higher voltage, particularly on the MOSFET gate. Transistors also switch faster at lower temperatures, though the temp difference has to be quite great (>10C) to have any substantial difference. You will know this very well if you read transistor datasheets, the different response times of a transistor will be given at several temperatures and gate voltages in the datasheet.

Danger is this, the tiny MOSFET gates are very vunerable to being burned up with higher voltages (currently they're only a few atoms thick). So, processors with smaller transistors are generally more vulnerable to complete burnout at high voltages-don't worry if it hasn't happened, if it happens the processor will immediately stop functioning; no slow death. Usually, electromigration is a more common problem for overclockers.
http://en.wikipedia.org/wiki/Electromigration

Another explaination on vcore concerning signal integrity, in addition to what I've mentioned already.
http://www.hardwareanalysis.com/content/article/1482/

This post has been edited by lohwenli: Jul 5 2007, 11:25 AM

Suddenly heard about burn-in... Wanna try to fry my sempron right now...
It's clocked at 2.2Ghz rite now... Unable to get more than that... Wat to do now? Put more Vcore and run prime for 24 hours? Then clock higher see if it's run again?

If you wanna permanently fry that chip, just set the highest voltage your board can support and hope for the best..too bad taking off the HSF will no longer help as it will shut down when too hot, before it can burn But if you're gonna burn the chip might as well test out if the burn-in 'myth' has any merit.

From what I've heard, overclock the chip to the max (with as much voltage as necessary), then push it a bit more until starts becoming unstable (can prime for a few minutes to an hour before it crashes), then bump up the voltage another notch. Leave it to prime a couple of hours to a day, restarting and continuing when it crashes.

But be warned, prepare to kiss your chip goodbye if anything goes wrong..which is quite likely..

Tried max 2240mhz wif 1.45v ... Prime a while but RAM cannot go through... Use divider cannot get over 2170mhz also... MAybe memory controller too bad liao...

I guess my ram fried already... Pc cannot start even i reset bios..

This post has been edited by Amedion: Jul 5 2007, 12:40 PM

1.45v is not that high, thats the voltage that the Athlon FX uses (it uses a similar core design). You're probably right, ram KO already..

If can, get some dirt cheap ram and try again, but this time don't OC the ram, only the processor. Heard ram can also benefit from burning in, but better if try out one thing at a time..less trouble finding out what got fried..

Respect you man

Problem is the memory controller... When i use HTT 275... The ram already 8xx MHz which cap ayam ram cannot do...
If use DDR533 or DDR400 divider.. cannot go over 270htt also..
Tried changing ram slots and sata cable slot.. no use..
maybe the processor really max out at 2.2ghz stable..

I see, thanks. Currently I'm running at 2.8GHz, RAM at 803 4-4-4-12. When I push to 2.9GHz, I need to change the RAM timing to 5-5-5-15, or else can't boot. Although fail in Orthos, but SuperPI calculation is better with 2.8GHz at tighter timings than 2.9GHz.

And agreed with 8tvt, as long it is great for my usage, I'm satisfied already OCing is fun, as long it is stable enough for personal usage.

I am using the same RAM like yours but one piece 1gb only...
This DDR667 value ram can do 890+ MHz only...

Oh yeah... Can anyone give me some good reason to buy 3600 brisbane or 3800 windsor... If let say i able to clock brisbane to 2.7ghz... It's the same speed with windsor 2.6ghz or lower, rite?

Overclocking higher but cannot obtain more performance is something that i dun want.. I want stable, 12hours per day usage, lifespan maintained, etc...

This post has been edited by Amedion: Jul 5 2007, 01:30 PM

2.7ghz windsor u need to supply 1.5v but brisbane 1.3v..
performance depends also the ram bandwidth and timing..

can't really recall about that though.... if yes, maybe i was saying something in a different context.

or most probably i was talking to a noob regarding the content of this section, which is about overclocking.

if u wanna oc very high, get brisbane as windsor cannot reach the clock as high as brisbane...
summore the vcore needed is higher than brisbane on same speed...
bt at the same speed, windsor wins...
so it depends on wad u wan la...

According to soulfly, the gap between Brisbane and Windsor gets bigger as clockspeed is increase..my guess is a 3.0GHz brisbane is equal to a 2.7GHz windsor

You overclock very much like me..I started overclocking because I couldn't wait for my video editing projects to take so long to render. And I didn't want to kill the chip and it also had to be rock stable because even overclocked it takes a few hours to render with the CPU at full load; if crash means start all over again and waste all the time. Basically, performance boost depends on how much is the CPU bottlenecking the program you're using-in my case video rendering is very CPU dependant, so a 30% overclock translates into 30% less time needed to render (eg 2 hours instead of 3 hours). Usually the performance boost is not noticable unless the overclock is 20% or more. Processors are designed to last 5-20 years under non-overclocked conditions, but because the circuit in the chip is identical for the same core and revision, the operating conditions of the highest end chip can also be used for the lowest end chip without significant reduction in lifespan. Example, an Athlon X2 3800+ can run at the spec of an Athlon FX-62 (2.8GHz@1.40v) without suffering a shorter lifespan because both are using the same circuit (Windsor F2) though its unlikely for a 3800+ to actually be stable at 2.8Ghz; the FX-62 can because its specially picked from the entire batch of chips.

2700 mhz winsors win almost same as brisbane 3000mhz ?
that's wrong and loose penang face only =X
see here
http://www.hothardware.com/printarticle.aspx?articleid=922
can say the cpu speed both running 2000mhz is almost same....
so brisbane 3000mhz guarantee faster than 2700mhz winsors
cheers guy

If based on cpu itself, I think Brisbane wins if having the same clock speed.

Windsor 3800+ 2.7GHz = 270x10
Brisbane 3600+ 2.7GHz = 285x9.5

Brisbane uses higher bus speed, therefore it performs better right? Correct me if I'm wrong. Still learning here

I said I'm guessing only..from your article it looks more like 3GHz Brisbane = 2.9GHz Windsor.


Nope, bus speed has very little effect in AMD K8 processors because the memory controller is on the processor, and memory access does not need to use the front side bus. In your comparision, Brisbane will surely lose, as the poorer cache latency means it can never outperform a windsor of the exact same clockspeed (you can refer to the article mentioned by cyberloner).