texas_techie

Distinguished
Oct 12, 2001
466
0
18,780
someone tell me some decent overclocking sites?

I know about hardocp... any other reputable ones?

Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.
 

girish

Distinguished
Dec 31, 2007
2,885
0
20,780
<A HREF="http://www.overclockers.com" target="_new">http://www.overclockers.com</A>
<A HREF="http://www.athlonoc.com" target="_new">http://www.athlonoc.com</A>
<A HREF="http://www.athlonxp.com" target="_new">http://www.athlonxp.com</A>
<A HREF="http://www.oczone.com" target="_new">http://www.oczone.com</A>
<A HREF="http://www.tomshardware.com" target="_new">http://www.tomshardware.com</A>
<A HREF="http://www.anandtech.com" target="_new">http://www.anandtech.com</A>
<A HREF="http://www.amdzone.com" target="_new">http://www.amdzone.com</A>
<A HREF="http://www.extremeoverclocking.com" target="_new">http://www.extremeoverclocking.com</A>

and a lot more, you can get those from a search engine like <A HREF="http://www.google.com/search?hl=en&q=overclocking" target="_new">this</A>

you can also post your queries on those forum

girish

<font color=red>Nothing is fool-proof. Fools are Ingenious!</font color=red>
 

texas_techie

Distinguished
Oct 12, 2001
466
0
18,780
thanks girish, im trying to discover the finer points of voltage, fsb, and asynchronos over-clocking.

Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.
 

girish

Distinguished
Dec 31, 2007
2,885
0
20,780
actually, I had discussed the theory behind overclocking, voltage in some thread in this same section. i will find it for you.

girish

<font color=red>Nothing is fool-proof. Fools are Ingenious!</font color=red>
 

girish

Distinguished
Dec 31, 2007
2,885
0
20,780
here is a compilation of some of my posts regarding overclocking. sorry if its too long.

----------------------------------
actually, the multiplier just affects the internal speed of the CPU, while the FSB affects both the CPU's internal speed and the other devices on the motherboard, the ones which are dependent on the FSB.

when you raise the multiplier, the rest of the devices on the motherboard namely the PCI and AGP bus (and the cards/chips connected to them) arent affected, just the processor runs faster. so its relatively easy to figure out what goes wrong when you just work with the multiplier.

on the other hand, raising the FSB might raise the CPU internal speed to a level that it wont boot, or make a PCI or AGP device work faster than than its capable of. the classic example of this is the IDE controller which fails to work or malfunctions at PCI speeds above 40 MHz, thats just 7 MHz above the nominal! many modern PCI-IDE controllers however do work at speeds higher than that.

now, one effect of overclocking the processor, either by raising the FSB or the multipler is that it gets overheated. the processor may not boot with higher core speeds because it needs more power which you supply it by increasing the core voltage. this in turn contributes to the power consumption and hence the heat generated, you must take care of the heat generated by the processor, else it will be fried.

---------------------
Its is recommended to lower the multiplier when you increase the FSB so that you have to work on one problem at a time. lowering the multiplier makes the CPU work within its rated limits while studying the effect of higher FSB on other parts of the board. once you get to the maximum FSB, now you can work with the multiplier. usually a processor running at a higher FSB but at the rated speed with a lower multiplier is better than the processor running at the same speed at the rated FSB!

Well, there are two methods to overclock the processor, eiter increase the FSB or increase the multiplier. To change the multiplier on AMD processors it needs to be unlocked. Intel processors have a permenantly locked multiplier so all we can do is to increase the FSB.

Now, assuming you have unlocked the multiplier many people will recommend you to lower the multiplier and go for higher FSB. Thats perfectly practical.

Basically, with every MHz you increase on the FSB, the core speed of the processor increases by the (multiplier) number of MHz, i.e. for a Athlon 1 GHz with a multiplier of 10, every increment in the FSB increases the core speed by 10 MHz. While all this is fine, you need still finer settings when you approach the max speed. Most of the times, higher core speeds can render the system unstable, but at the same time the memory speeds have some potential to work overclocked. In such cases, even if you cannot get the processor work faster but the memory can work faster, you can lower the multiplier so that you get core speeds near about the rated but a faster memory subsystem. That helps improve the performance even while the processor is still running at its rated or slightly overclocked speed.

Thats the difference between a Athlon 1000(100x10) and Athlon 1000(133x7.5)! You have a play of half the FSB with a single multiplier step.

---------------------
yes, all Athlons, all Durons, and interestingly as some reports say - both the Celeron and P-III come from the same wafer at the same time.

So initially, every Athlon is 1.4 but screening and some tests detects what the die will run best at and then it is labelled so. Durons have a different production line.

With Intel processors, all of them start as P-III, but screening will make some as P-IIIs and some as Celrons. The speed and cache is tested, the processor has some kind of control matrix PROM in which is set the multiplier, FSB and cache and CPUID details. Thus these settings differentiate a P-III 1 GHz from a Celeron 667 MHz!

Amusing they both start as P-III 1 GHz!

This is a common way of reducing manufacturing costs and that is why some 1.2G Athlon sometimes runs well at 1.4 and sometimes even more

----------------------
Well, although you dont have any PCI cards, your do have some PCI devices that are present on your motheboard! They might make FSB increasing difficult.

Even the AGP frequency (which is 66 MHz nominal) is derived from the FSB, 2/3 for 100 MHz FSB and 1/2 for 133 MHz FSB. Your Radeaon will probabely take more than 89 MHz, when you crank up the FSB to 166 MHz! That shouldnt be a problem.

What the problem is, you still have the IDE controller sitting on the PCI bus that cannot tolerate frequencies above 37~40 MHz! With a PCI divider of 4, you can have the FSB at max 160 MHz with effort, not more than that.

You have pretty good cooling given that the system is kept in a underground closet! And the temperature is also quite okay, you can run the processor at a bit higher temperature, so you have some room for the temperature to rise

------------------------
larger die area will certainly make cooling easier, not that the chip will dissipate less power, but lesser power will be dissipated per sq.mm, and that would help cooling.

but rather than reducing the cost benefit of going for smaller 0.13u technology with more die per wafer, it will actually be costlier because you havent improved the yeild even with a costlier process!

among the other things that would be affected by this larger die size is the maximum speed. longer the circuit traces, higher will be the inductance and capacitance between adjacent tracks (the silicon oxide acts as a dielectric) limiting the maximum frequency the die could achieve. one more factor is the voltage required to operate a longer conductor, it will certainly be a bit higher which would actually contribute to the powre consumed by the die and in turn neccesitate better cooling!

whoever thought of it, might have got it wrong. its actually difficult to balance the proc and cons of such a trick, especially in the market conditions like these. probabely the advantages slightly or more thn outweigh the disadvantages of this technology.

remember the Quantum's Bigfoot technology? Introducing a 5.25" hard disk when 3/5" ones were becoming mainstream was a marketing hype, old wine in a new bottle.

Its certainly not with silicon though, the heat spreader and active cooling are much better choices than making a larger die.

---------------------------
Before trying anything of that sort, make sure you have 1. a very good cooler, like SwiftechMC462 or OCz Gladiator, 2. have a good accurate temperature monitor onboard, or if you dont have that try touching bodies at different known temperatures to get an idea how hot is xx°C! May not be too accurate, but you can then judge by touching the HSF how hot the processor is getting! Continuosly monitor the processor core temperature. Update the BIOS to the latest version if possible.

<b>1.</b> how do i know if i have to raise the cpu core voltage and to what voltage do i raise it to?
<b>Ans.</b> If you increase the frequency and learn that the system isnt quite stable anymore or isnt booting at all, you have to increase the core voltage a bit, until you get a stable system. But dont overdo it, almost all of the times, it will be just a 0.1~0.2V increment over the previous setting. If you cant get a stable system even after increasing the voltage significantly, say about 1~1.2V over nominal rated value, you probabely are at the end of it. your CPU will not overclock more than that. set the voltage to lowest value at which the system is stable at that speed.

A bit of theory behind this: Increasing the core speed affects two important things. The signal integrity on the silicon and power consumption of the chip.
<b>a.</b> Any digital signal does not change instantaneously, it takes some time to reach the desired value. This is due to the capacitive and inductive effects on silicon. When you increase the core speed of a processor, you reduce the period of the clock pulse, allowing lesser time for this signal to reach to a voltage level that can be treated as a valid signal. It is in fact, the ratio dv/dt (rate of change of voltage) that has to be increased so that the change in voltage is faster and it reaches the desired level in shorter duration. And this is acheived by increasing the core voltage. This way, the signal reaches the level which can be asserted valid by the destination circuit faster, in lesser amount of time!
<b>b.</b> All processors are manufactured by MOS (Metal-Oxide Semiconductor) technology or its advanced versions. all MOS devices consume more power when they are operated a higher frquencies. Eventually you increase the speeds to such a level that the power you supply isnt adequate anymore for it to function properly. You increase the core voltage so that mor epower can be supplied to the core and iit again starts functioning properly, until you make it work still faster when even this extra power becomes inadequate!

<b>2.</b> Raising the FSB is a way to overclock, but what effects does it have on the other components (memory, AGP, PCI, etc..)? how fast is too fast when increasing the FSB?
<b>Ans.</b> Raising the FSB can acheive limited results, because <b>a.</b> raising it too high will make your other devices that are dependent on the FSB run faster than they are rated to, like the PCI and AGP bus. and <b>b.</b> The increment in FSB is multiplied by the multiplier to get the core frequency. That means, with every MHz you increase the FSB, your processor will run say 10 MHz faster with 10x multipler. This increment may be too high for the processor although your memory and other parts might run quite well at still faster speeds. This is a condition when you might need to lower the multiplier in favour of faster memory subsystem!

PCI bus runs at a fraction of the FSB, FSB/3 for 100 MHz FSB and FSB/4 for 133 MHz FSB. Its nominal speed is 33 MHz, but it can work at a bit higer speed. If you are working with FSBs higher than 133 MHz, check if your motherboard supports FSB/5 divider. As far as possible, try to keep the PCI clock to 37~40 MHz, your PCI-IDE (hard disk controller) might fail with >40 MHz FSBs, although many modern controllers can work well at even 42~45 MHz.

The AGP bus works at 2/3 or 1/2 of FSB for 100 MHz and 133 MHz respectively. It should be run at a nominal value of 66 MHz. If your motherboard supports manually setting 2/3 ratio its good, else it will run at speeds about 89 MHz which most new cards will take, but almost all older cards will fail.

Some chipsets supports asynchronous memory, operating at a different speeds than the FSB. FSB=memory speed is desirable for best performance, so run the memory as fast as the FSB. Even a slow bad noname memory cannot work at higher speeds, so make sure you have good branded memory preferably rated at 133 MHz or better (for all types of memory, RDRAM for P4 is poor at overclocking)

If you are using AMD processor, unlock it and lower the multiplier to some lower value so that you can get the memory to work at its best, while the processor runs nearabout its rated speed. And then try cranking up the processor speed. You can get the unlocking info from a number of sites like Anandtech, THG, athlonoc etc.

Hope its enough. Keep posted on your results.

---------------------------
no overclocking would reduce the performance of a system except that of a P4! usually overclocking results in higher power consumption of the processor, and the P4 slows itself down to save its skin, to as much as 12.5% of its rated speed! That means a 2 GHz P4 can run itself down to a 250 MHz processor if you try get too much out of it!

Reducing the CAS latency is a trick to make te memory run faster, so usually it either slows it down slightly or compensates with faster speed, depending on the increase in memory speed. AGP and PCI busses would run faster if all the devices connected to them run at that speed. one failing device can make the system either unstable or it just wount boot.

so in a sense, overclocking (any other processor than P4) either improves its performance, or brings it down to zero!

------------------------------
when you are overclocking, you at times need to increase the core voltage of the processor. faster a processor runs, more power it consumes. at the rated voltage, the power delivered migt not be enough to run it well, the system then either becomes unstable or just doesntboot.

by increasing this core voltage, you make sure the procesor is supplied with enough power to run at faster speed. of course, a side-effect of this is that it starts giving off more heat, which has to be taken off as fast as possible. that is why overclocking also demands use of better and better heatsink/fans.

---------------------------------

hope this helps.

girish

<font color=red>Nothing is fool-proof. Fools are Ingenious!</font color=red>
 

texas_techie

Distinguished
Oct 12, 2001
466
0
18,780
Thank you girish, that is precisely the information I needed. I also want to thank you for taking the time to paste all that info, im sure it took some work.
Knowing how FSB affects all the components, and how dividers work was the best. I wasnt very informed on dividers, so that info was very welcome.
Im not real interested in overclocking itself. But the info that comes along with it helps me understand PCs better.
I recently came to the realization that I understand the basics of everything, but the details of very lilttle. So Im on a mission to get the details. Mission one accomplished, thanks to you :)

Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.
 

girish

Distinguished
Dec 31, 2007
2,885
0
20,780
Glad it was useful.

I did have to dig into my own posts, but I wanted do that for a long time, just to collect this info for later reference as well as use it in future for somebody who needs it!

BTW just a addendum to the above info, some board like the Abit TH7-II now decouple the PCI and AGP frequencies from the FSB so that they stay at 33 and 66 MHz no matter what FSB you set. Very useful feature when you plan to ramp up the FSB to extra high speeds, now dont have to worry about overclocking your PCI or AGP cards! Hope more and more boards have this feature in future.

So whats your next mission?

girish

<font color=red>Nothing is fool-proof. Fools are Ingenious!</font color=red>
 

texas_techie

Distinguished
Oct 12, 2001
466
0
18,780
The decoupling sounds very good. I hope that trend carries over into the next generation of Mobos (v-link, serilal ata, hypertansport, usb2 etc etc)
Here are some of my missions:

1. Understand how latency affects performance. I know that the lower the latency, the better/quicker the system. But I wanna know how chipset makers reduce latency. The nuts and bolts of it all..
2. Understand how the OS interfaces with the hardware. Things like IRQs, the Kernal, registers, memory allocation etc.. I know next to nothing about Operating systems. Sure I can do a fresh install, tweak, edit the registry, move stuff around, defrag etc etc.. But I wanna know the inner workings, especially where hardware meets software.
3. Understand internet protocols. Im not really interested in java, html, active X and the like. I mean the actual hardware behind the internet. How signals are sent, routed, interpreted, TCP/IP, packets, lag.... all of it.

4. And this is the biggy! Understand graphics cards and game engines.
Engines: I know what they are, I know what they do. But I wanna understand them to get a better idea how to map, and asses the quality of a game.
Graphics cards: With the geforce4 out. I REALLY REALLY wanna be able to map for games that are optimized for the Geforce 3/4 series. That involves some high level programming for the GPU. SO i need to understand programmable shaders, particle effects, blinn mapping, environmental bump mapping etc etc.. if im gonna map for them. I know thats one hell of a challenge. But from mapping with older engines, i know that the real magic of a map comes from the programmable stuff like shaders.

Mission 5 - (optional) - If i havnt killed myself by this point, i want to learn the subtle aspects of video compression and video/image enhancement. Dont necessarily want to do digital video editing. Just understand whats some good formats, good ways to convert between them and other basic quality issues.

Thats it for now... im sure thats enough to keep anyone's plate full.

Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.
 

TRENDING THREADS