Does overclocking work

DanGerMus

Does over clocking a CPU lets say an e6600 to the same clock as an e6700 or ex6800 actually make your system perform as good as the higher CPU's in the real world. Because as far as a know the clock speed is really only half the battle. I only ask cos when ever anyone says i'm gonna buy an e6600 somone always pipes up and says dont bother just buy an e6300 and overclock it and get the same speed. It seems to imply that all the CPU's are the same and manufacturers like Intel and AMD just under clock them for the laugh and then charge you more for the higher clock.

DirkVoodoo

They are all the same.

At least architecturally. All kentsfields will be the same, conroes and Allendales. Due to manufacturing of the silicon wafers, different batches will perform differently (Went on a tour of Intel there, can't remember specifics, google) and as such, Intel takes samples from each batch and uses a process of "speed-binning" to see how fast they can go. Intel also locks the multis, except on extremes, to stop crafty OEMS from buying 6300 and clocking them to be resold as more expensive 6600s. Of course, this means for you to overclock, u have to increase you overall system speed to increase your CPU speed, but a 6300 will achieve the performance of its more expensive counterparts, provided u get a good batch, of which nearly all core2duo seem to be, hence the reason for them being a popular overclockers choice, much like the old K8 AMDs.

EDIT: A more eloquent and detailed explanation here http://pcpitstop.com/news/maxpc/overclock.asp

B00MSTICK

Simple answer: yes

papu

a 165 euro e4300
http://www.komplett.ie/k/ki.asp?sku=330167
can clock to 3.8 with alotta Luck! it gets to 3.2 no problem thats better than an X6800
http://www.komplett.ie/k/ki.asp?sku=322640
which costs nearly 950 euro or more!!, bang for buck is the best motto ,

Anti

In simple terms yes it is, if you are willing to take the risk.

For example i gave a friend of mine my old 3500+ for free so he can build a new(ish) pc for free.

To start off i got him to run super pi to calc 1M. Over 2 tests he got 40 and 41 seconds. Straight after i got him to go into the bios and increase the fsb from 200 to 200, and change nothing else at all.

Straight away he scored 38 and 37 seconds. Now thats pretty impressive for a 10% Cpu overclock. B the end of the week ill have him running at 2.8ghz and doing a 1M calc in 30 seconds or under. So you get a 3500+ ( 40€ ) and get to to run at fx 5x speeeds ( 120-150€ ) for a few mins work.

Gurgle

happygandalf wrote:

It seems to imply that all the CPU's are the same and manufacturers like Intel and AMD just under clock them for the laugh and then charge you more for the higher clock.

Sort of.

The manufacturers produce a batch of CPUs and test them to determine the top speed at which it will perform with full load and standard cooling.

Real world use is less demanding on the CPU than full load test conditions, and better cooling can be installed.

So overclocking will give you better performance from the CPU you have, but buying a higher rated CPU will get you a better CPU.

In terms of performance for your money - You'll always do better buying a cheaper CPU as they massively overprice the best ones compared to the performance difference.

L31mr0d

I asked myself this same question when I ran a benchmark and it compared my old 3700+ as better than an FX55 because i'd overclocked it.

The first part of understanding this is that you can't view a processor like other normal devices, in that each different model has different features. The names and clocks that are given to processors are purely indications of the PHYSICAL properties of the material they are made with. For examble a wafer of silicon could go through the whole process of being turned into a CPU, but in the end product when the wafer is cut up into individual CPUs and tested there could be a few X6800 on that wafer, along with some E6600, maybe a E6300 as well.

Sometimes the physical properties of the silicon, or maybe just inconsistancies in the manufacturing process cause some chips to come out weaker than others, so are clocked lower and sold cheaper. What the Overclocker does is cash in on the paranoia of the chip manufacturer by realising that the ACTUAL limit of the chips clocking ability is usually a lot higher than what the manufacturer tests it to be.

But its all about getting the best bang for buck in the end. It would be cheaper for you to buy an E6300 and an excellent air cooler, than to buy a E6600 and use the stock cooler.

But in saying that, you are right in that clocks aren't the only consideration when comparing ALL cpus (i.e. Athlons at lower clocks performed better than P4's, and a Pentium D @ 4Ghz would perform worse than a C2D @ 3Ghz)

Within a single architecture though, clocks to performance are proportional.

d_advocatus

Gurgle wrote:

So overclocking will give you better performance from the CPU you have, but buying a higher rated CPU will get you a better CPU.

I'm sorry but that's just wrong. I have an E6300 that beats the X6800 in benchmarks.

It used to be that the chips with defects are sold as lower end chips but it's not that way anymore with Intel. They have refined their manufacturing process so well that nearly all their chips could be X6800's. They are all made with 4MB Cache also. The difference comes where Intel physically lock the multiplier and disable half the cache for the lower end chips. Also, if you do even a little research from google then you will see that the extra 2MB cache makes almost no difference in power.

DirkVoodoo

True.

It depends on demand also. If a large portion of chips are eligible to be X6800s, say 700 out of every 1000 batch, intel can't sell them all as x6800s because there isn't that much demand for high end components.

Instead they will lock them at a lower multi and sell them as 6300s. It costs intel the same amount of money to produce these chips in a mature production line.

EDIT: the extra cache is useful in gaming though.

po0k

and HPC

Gurgle

d_advocatus wrote:

I'm sorry but that's just wrong. I have an E6300 that beats the X6800 in benchmarks.

Comparing like with like:

Take an E6300 with 1.86GHz rated clock speed. This is the speed at which the manufacturers determined that it could run at full load without the current it drawn exceeding the specified limits.

Overclock it by 15% and it might work just fine. It might draw more current, get a bit hotter when you're gaming (or as most overclockers seem to spend all their time, benchmarking), but it probably won't get to a temperature that will damage the chip.

What have you got now?
You've got an overclocked E6300.

Your overclocked benchmark might beat a stock E6400 (2.13 GHz), but you've still got an overclocked E6300.

There is a physical characteristic with your CPU that meant when it was graded by Intel it didn't qualify to be rated as an E6400.

And btw, if your overclocked E6300 beats an X6800 on benchmark, that tells me more about the benchmark software than about either CPU.

DirkVoodoo

Gurgle, read above.

As d-advocatus mentioned, intel is producing a lot of top quality chips at the moment. Chips dont just get speed binned, a lot of X6800 capable chips are marked down to keep the production numbers of each chip in line with demand. At no cost to intel.

What may look like an overclocked 6300 by label may in fact be of the same quality as a 6800, in effect, an overclocked underclocked 6800.

Anti

overclocked underclocked.... LOL

Gurgle

DirkVoodoo wrote:

As d-advocatus mentioned, intel is producing a lot of top quality chips at the moment. Chips dont just get speed binned, a lot of X6800 capable chips are marked down to keep the production numbers of each chip in line with demand. At no cost to intel.

That may be true to an extent, when speed grading produces higher numbers of better grades than expected and orders for the lower grades are higher than the number of those grades produced.

I would expect though that in the mature stage of the production cycle for a particular architecture, orders are filled from warehouse stock not from the line.

DirkVoodoo wrote:

What may look like an overclocked 6300 by label may in fact be of the same quality as a 6800, in effect, an overclocked underclocked 6800.

Or it may be a CPU that just barely passed the speed grading as a 6300. You don't know until you put it in your system and try to overclock it.

Anti

Lets take amd opterons(2way) for a example, I have a opty 170 which is clocked at 2ghz at factory speeds, This chip is a rebadged fx-62. It has exactally the same architecture as the opty170. And when you overclock the opty to 2.8ghz(on air) it will return better performance then the fx-62 as you have a much higher memory bandwidth speed. as the fsb is usually around 280-300 where the fx62 is 200.

DirkVoodoo

Gurgle wrote:

You don't know until you put it in your system and try to overclock it.

True. And therein lies the fun

I have an L638A, which on OCUK a prominent member has denounced as a poor clocker. I got it up to 3GHz on stock cooling and stock volts, just to see would it go that far. Once i get my ultra 120 i can go further, but if I only got 3GHz I wouldn't be too upset. Another guy hit 3.6GHz on a 38A, so its all a bit of pot-luck and good overclocking practice.

However, my AM2 3800+ (Orleans) refuses to go above 220 HTT :mad:

SouperComputer

It depends on demand also. If a large portion of chips are eligible to be X6800s

QFT.

I think this is the main reason that some chips are so overclockable. I bought a P4 1.8Ghz a couple of years ago for 20 bills, it ran at around 3hz (without raising voltage!) for two years without issues. I have since moved to an 805D @ 3.8Ghz.

Anti

IIRC the northwood and pentium 4C were hugly overclockable for no apparent reason at all. Same with the pentium D's

philologos

Speaking of overclockiing does anyone know how to get more juice out of a AMD 64 3700+ on a ASRock 939NF4G-SATA2?

L31mr0d

erm... overclock it? how n00b are we talking here? what have you tried to get more "juice" out of it?

Anti

Jakkass wrote:

Speaking of overclockiing does anyone know how to get more juice out of a AMD 64 3700+ on a ASRock 939NF4G-SATA2?

There have been many threads on similar amd cpus, do a search and you should find a few.

hopeful

Jakkass wrote:

Speaking of overclockiing does anyone know how to get more juice out of a AMD 64 3700+ on a ASRock 939NF4G-SATA2?

Have a look HERE...ASROCK specific forum section...very handy.

DirkVoodoo

yeah, enthusiast forums are generally the best place to go (not that anyone here isn't an enthusiast!). I have found the DS-3 thread on OCUK to be most helpful as it deals with specific settings/issues with the board rather than generic overclocking (which you should be familiar with before attempting any overclocks).

Capt'n Midnight

Not sure how true it is but, it used to be said that if Intel had an order for 1000 CPU's at the highest speed, they would test all the CPU's that came off the line at that speed, but once they got the 1000 they would only test the rest at the next speed to match the next order and so on..[see next post ] [/edit]

If a CPU is rated at a speed then with good cooling and clean power it should run happily at that speed for years. If you overclock you might get lucky or you might not depending on the individual chip. Also the type of chip matters - some are more overclockable than others , an outstanding example being the celerons PIII's where the cache ran at half the speed as it would on the full PIII's, turned out the cache was the limiting factor on overclocking, and so the celerons could ramped up nearly twice as much as the ordinary ones ( Ok not exactly but you get the idea )


Overclocking does wear out CPU's faster - then again you won't be keeping it for years. Also makes the system less resistant to things like voltage spikes and hot rooms. Maybe a cheap air con unit that keeps the rest of the PC and the air going into it just above the dew point can't hurt.

mathias

if Intel had an order for 1000 CPU's at the highest speed, they would test all the CPU's that came off the line at that speed, but once they got the 1000 they would only test the rest at the next speed to match the next order and so on..

Not true captain , not true ,

From the horses mouth here , I worked in Sort for 8 years ( IBM and Intel )

Speed sort is applied to all chips that come off the line , they are graded according to their performance on the test machine ,( which is a high speed tester and handler/Prober ) the machines ( and test program ) are not flexible enough to keep stopping at certain counts like you suggest , they are all run on the same program and all sorted according to the results automatically. ( The test program is not modifiable / scalable on the fly , that would be a nightmare to manage , the program is fixed and only modifiable after a 6 week review process )

What needs to be kept in mind , and what most overclockers do not realise , is that for any stepping , say core2duo , all the chips coming off the line are identical and vary only in tolerance allowed in the fabrication process.
All chips are marked according to the highest speed test they passed and here is the crunch point , any speed higher than the chip is marked to run at , it has already failed this speed in an operational test , so an overclocked chip is running at a speed it is known to fail at. ( and a rigorous test it is too)

In very rare cases , and I mean very rare , some higher speed chips are marked down to fill an order from a big OEM customer. These are always big OEM customers , never retail box sets , just big OEMS.

So overclockers beware , overclocking the chip considerably reduces its life span , but i Guess that you guys never hold on to the chip long enough to find that out !

L31mr0d

mathias wrote:

So overclockers beware , overclocking the chip considerably reduces its life span , but i Guess that you guys never hold on to the chip long enough to find that out !

I think this is a given with overclockers, you know you aren't going to have a chip for more than 2 years (lol, it seems to be around 6 months nowadays) so you want to use up the usuable life of that chip as fast as possible.

Can I ask mathias, is it only clocks that determine the failure of a stress test, or do they stop testing once the chip reachs the average Tcase max, because I always assumed they clocked chips until they reached their thermal limit under the heatsink they would be provided with, and therefore if an overclocker bought a more efficient cooler it would allow for a higher clock before reaching the chips thermal limit.

Chosen

Overclocking doesn't necessarily wear down components. I have been using my procs at the maximum stable frequency that can be achieved with default voltage (that is 3.33GHz for my QX6700 and 3.45GHz for the E6700 that I used for 10 days). Ofcourse I found the limits of the chips with up to 1.5V, but that was for a few benchmark runs.
My sister is still using my old Celeron 533A @800MHz, since mid-2000 and she never complained about restarts or BSODs.
It's the prolonged heat and overvoltage that can reduce the lifespan of a processor, from a theoritical 10 years to as little as 2 years. But nowadays, a 2-year old processor usually costs around 30-40€, and people don't really care about overclocking it, especially since the benefits of an overclocked system can be seen *today*

I already feel bad for my L628A that I got as a temporary proc, until my QX arrives, and it's been sitting in its box since January...

mathias

Can I ask mathias, is it only clocks that determine the failure of a stress test, or do they stop testing once the chip reachs the average Tcase max

There are a number of parameters that determine whether a chip can run reliably at a certain clock speed , not just temp , in fact , the chips are tested under the best thermal conditions that the module is ever likely to see in its lifetime , using ATC test heads ( active thermal conditioning and liquid nitrogen ) because you dont want to kill or damage the the product to determine its limits.

Its performance characteristics , like speed of response , propagation delay , etc are the main factors , but a lot of other tests are thrown in there too. Its the performance as a whole at certain clock speeds that are taken into account, so temp for testing at sort is really just one many parameters.

Fanny Cradock

Anti wrote:

To start off i got him to run super pi to calc 1M. Over 2 tests he got 40 and 41 seconds. Straight after i got him to go into the bios and increase the fsb from 200 to 200, and change nothing else at all.

I was considering OCing my 3700+, but just as a simple test. Could you elaborate on the numbers in bold.

So far I get 46/48 sec, which considering your friends results seems very high for a 3700+

SouperComputer

Mathias wrote:

In very rare cases , and I mean very rare , some higher speed chips are marked down to fill an order from a big OEM customer. These are always big OEM customers , never retail box sets , just big OEMS.

I was wondering that. The 1.8 Northwood I was talking about, I actually bought two that both came from Dell machines and overclocked like crazy with no extra voltage and no major increase in temp. The one I kept fopr myself ran on stock voltage and a stock cooler for two years with a 50% overclock.

It still ran fine the day I stopped using it, I just moved to a prescott at that stage (YUK!, but hey it was free!) it was a power hungry chip but gave me another 600Mhz to help with some of the video projects I was working on. It lives in the DVR backend now. If I can get a decent board, I may run the old 1.8 in a DVR frontend.

DanGerMus

Well guys thanks for the huge response. All very helpful indeed. Points that stick out in my mind are that overclocking seriously reduces the life of your CPU. And to counter, that most cpus (for, well people like me) will never realise there full life span anyway. And i think that if you shorten your CPU's life to say 2 years then to replace it with the same model would probly cost you about a fiver by that time. So it seems that a low-medium overclock would deffinatly be worth it. Now i'm a bit lazy so what i'm looking for is someone with a similar build to mine that can recommend settings for a long term overclock. gigabyte DS3 (rev-1) bios f5, e6700 c2d, 2gb geil pc-6400

DirkVoodoo

OCUK have a detailed Gigabyte DS-3 thread, also Hardforum i think has a good guide. The DS3 is a good board, make sure you get BIOS F10. From what I have heard/read, Geil is good RAM for overclocking, don't know much about the voltages for it though.

Sorry that I don't have those links, I have them bookmarked on my home PC.

Chosen

Have fun reading

DanGerMus

well guys i've officially done my first overclock. i used these setting from the OCUK guy
1. Make sure that you have the latest chipset drivers and flash the board to the most recent bios version
2. On the main bios screen press Ctrl+F1 to open the advanced options within the menus
3. In the Advanced Bios Features Menu, Disable all of the following: CPU Enhanced Halt (C1E), CPU Thermal Monitor 2(TM2), CPU EIST Function
4. In the PC Health Status menu, set the FAN Speed Control Method to "Legacy" & FAN Speed Control Mode to "PWM" to maintain highest speed
5. Go into the MB Intelligent Tweaker (MIT) menu at the bottom of the page, set the System voltage control to "Manual"
6. These boards have a default VDIMM of 1.8v, set the VDIMM to co-ordinate with your ram. The Geil I have is specced as 2.1v so I have mine set to +0.3v
7. Set the MCH Voltage to +0.1v
8. Set the FSB Voltage to +0.1v
9. Set the VCORE to 1.45v - Please be aware that this will increase temps!
10. Set the CPU Host Clock Control to "Enable"
11. Set the System Memory Multiplier to "2x"
12. You're set to go, just increase the System bus Frequency to Overclock
13. I would reccommend that you use Intels TAT to monitor temps

Handy that i have the same RAM as him. )
i left cpu multi at x10
i have currently clocked at 3.65 but under load test in TAT system locks.
I tried 3.7 but got bsod and 3.8 wouldn't boot properly and 4 would boot at all. lol. learning curve i guess. my only benchmark is super pi for now so went from 46 at norm 2.66ghz to 35 at 3.65
my temps at 3.6 are core0,1 - 46,45idle to 66,63 load

some screenies attached
So now that i've tried it i'm left asking my self one thing..... 4????!?!?!?!

awhir

whats is it for a e6600 thats nice man

DirkVoodoo

Yup, all part of the learning process, gentle tweaks here and there, just like tuning an engine. There is no quick fix, click button here, to overclock (properly).

Is that on stock cooling? I was getting the same idle temps on stock, didn't test it though because it was way hotter than i liked. Heat is the enemy comrades!

Don't run your system OC'd like that without better cooling, or expect more crashes and shorter CPU life.

EDIT: also, consider getting a little fan (50mm) for the northbridge. I think they used to supply one, not with mine anyway. Anyhow, use speedfan as well, and keep an eye on your northbridge temps. They can get quite hot on the DS3. Also, I've read that a dab of superglue is good for fixing it in place.

Cooling Northbridge

DanGerMus

it's a e6700. with and arctic freezer 7 pro. and these readings are with all fans in my case blowing full blast. Antec p180 with its three regular fans plus a papst blowing air in the middle accross th gfx. have a zalman... something.. on my x1950 pro, the copper expensive one lol. think i might invest in a north bridge cooler they seem fairly cheap. i also didnt recieve a fan with my ds3. wasnt expecting one though lol. the greatest temp jump seem to come with the initial voltage increase 10 degrees. after that it's fairly gradual with the FSB increases a degree or so per 20mhz. So do i now have to increase voltages to get stability at higher clock? whats my safe thermal limit here?

DirkVoodoo

Keep it below 70C, but I would keep it below 60/65C.

Nearly had a heart attack the other day, intel fan control wasn't kicking in when I played oblivion, CPU was goin into 80s (briefly, before I turned speedfan to software controlled).

Case fans are good for air flow in the case, but they won't cool a cpu. Invest in a good cooler, its money well spent.

EDIT: Run orthos as well, see what you load temps are. If they go over 65C, shut it down and reconsider your overclock.

DanGerMus

right then i have water coolin gear. double 120 rad and a pump and hose. just no blocks. what would you recommend. OCUK have only got the dd tdx or zalman wb4 in stock. which is better? or other blocks from somewhere else?

DirkVoodoo

ooh, watercooling is not my specialty. Good route to go tho, plenty of experts here.