changing of the guard from Northwood to 6xx?

lioninstreet · Post by **lioninstreet** » Thu Mar 10, 2005 11:57 pm

I just read the LEGIT article on the new 6xx series chips. An excellent, well thought out article that was quite useful. However there are some of us enthusiasts out here who, while impressed, still believe in the venerable (albiet mainstream) P4c Northwoods. It would be interesting to compare the old standard 3.2 or 3.4 ghz 512HT to the 640/650 from model to model, analizing operating tempratures and also true overclockability. While on the subject of OC, many seem to forget how easy it is to overlook the false sense of accomplishment some of the higher OC ghz's give. Ultimately, there are some great Ghz's to be seen by really cranking up the voltages, but the benchmarks usually start to suffer. And what good is a high Ghz if my computer is ultimately slower.... I have not seen many reviews that take this into consideration when overclocking, they always seem to only go for the big numbers...

infinitevalence · Post by **infinitevalence** » Fri Mar 11, 2005 1:37 am

Welcome to the forums, man i wish i was where you are right now. Most of my family is split between FL and VT so i figured i would split the diff and live in TN. Bad idea, cept for the women

... any way. Your right about the NW lots of people are still getting good overclocks out of that chip. Before i got Athlon64 tatoo'd across my chest i was thinking about going with the northwood. There are alot of people who have found that a NW at close to 4ghz still out performs a prescott at that speed and does so with less heat, as for seeing reducing returns on an overclock, well i think that may be a product of the archatecture. The p4 both northwood and prescott need lots and lots of fast memory bandwidth to keep the long piplines full. this is not a judgement but a statement. What can happen is as the speed and bandwidth increases you introduce more latancy from other async clocks on you system ie agp, pci, pata. so what your describing can happen where though the cpu and mem are maxed out the performance gain is lower than expected because of other latancy introduced into the system. Generaly the latancy is so small though that the improvement in performance from the cpu and memory overlcock out ways the disadvantages of runing your subsystems out of sync.

Now there is a good chance that everything i just said is total bull, this is my understanding of how things kind of work and i am by no means an expert just someone who reads alot of tech docs, so dont quote me as the end all be all there are people fare more knolageable than i.

teqguy · Post by **teqguy** » Fri Apr 08, 2005 12:19 am

I don't exactly understand how benchmarks could possibly be hindered from overclocking, unless of course you're talking about the throttling that occurs from the higher temperatures.

As transistor density increases, the need for sustaining tolerable temperatures increases.

It's an unavoidable concept that up until now has been plagued by attempted solutions that are far from ubiquitous.

LVCapo · Post by **LVCapo** » Fri Apr 08, 2005 1:24 am

Infinite isn't saying that O/Cing hinders benchmarking...but there isn't going to be a straight paralllel line of O/C to performance a you increase the O/C, the higher you go, other system components are going to have an affect on performance

teqguy · Post by **teqguy** » Fri Apr 08, 2005 2:36 am

I was referring to the thread starter's comment.

However, in response to infinite's post, the number one cause for concern right now is still hard drive bandwidth.

We have external hard drive interfaces capable of bandwidths as high as 384MB/s(3Gbit), while the internal interface has slowly progressed to an average of 60MB/s.

When you factor that in, memory bandwidths of up to 10.2GB/s seem futile.

LVCapo · Post by **LVCapo** » Fri Apr 08, 2005 2:41 am

HDD have been a systems #1 bottleneck for quite awhile....but there are other mitigating factors which limit O?Cing along the way......every system component has an affect at some point, one way or another

teqguy · Post by **teqguy** » Fri Apr 08, 2005 3:43 am

All overclocking aside, there are factors that are technologically irredeemable up to this point.

The point I'm trying to make is that until those factors are accounted for, no amount of overclocking will be satisfactory.

infinitevalence · Post by **infinitevalence** » Fri Apr 08, 2005 11:16 am

teqguy wrote:no amount of overclocking will be satisfactory.

i think that part of you statemnt made my day, thats exactly how i feel.