Sign in with
Sign up | Sign in
Your question

Overclocking Guide Part 3: How To Gain 81% For $27

Last response: in Overclocking
Share
January 18, 2007 10:02:12 AM

How does an 81% clock increase from a $27 upgrade sound? Walk with us as we put our wealth of recommendations to the test and come out victorious!
January 18, 2007 1:31:56 PM

What are the CPU temps idle and under a load with that overclock?
January 18, 2007 1:37:09 PM

I continue to be confounded by RAM prices. I know if you want to play you have to pay, but these type of "for only $27" articles don't ever tell the price of the premium RAM that has to be used. The OCZ you used goes for around $270, but you can find 2x1gb sets for $190.

So I'd rather see the article title say $107, as it is the difference between a non-OC system price with RAM + HSF.

edit:grammar
Related resources
Can't find your answer ? Ask !
January 18, 2007 1:57:23 PM

Question: Which version of this board was used? 1.0, or 2.0
January 18, 2007 3:42:28 PM

what is the power consumptions?, compare to non-OC system.
these level of OC of coz need powerful and lilbit-pricey PSU.
January 18, 2007 4:34:41 PM

I'm pleased to see Tom's is starting to overclock with the right board, the right cpu, the right bios, and the right memory to get a respectable overclock. I'd been disappointed on those points over the past 6 months -- seeing articles reporting a low maximum overclock on a certain board or certain chip when they hadn't tested using the right components.
January 18, 2007 5:20:24 PM

Quote:
I continue to be confounded by RAM prices. I know if you want to play you have to pay, but these type of "for only $27" articles don't ever tell the price of the premium RAM that has to be used. The OCZ you used goes for around $270, but you can find 2x1gb sets for $190.

So I'd rather see the article title say $107, as it is the difference between a non-OC system price with RAM + HSF.

edit:grammar


The title is very misleading, i agree. As you already mentioned the whole hardware was choosen with overclocking in mind so the difference between what you need to overclock and what you need to get a complete system would be the final cost - not the 27$ of some HSF. The choosen title is quite a cheap shot to get attention, similar to the headings of heavily opinionated magazines like the "sun" or, and god forgive me for comparing THG to it, "the inquirer".
Apart from the title i liked the article. Everything that is needed to know is explained and written quite comprehensive. Even though i can´t really tell how comprehensive the article will be for someone with no technical expertise at all. But then again, such a person shouldn´t be overclocking at all. :) 
a b K Overclocking
January 18, 2007 6:21:07 PM

Overclocking with the "wrong" components can be good for comparing motherboard stability, but doesn't tell you what the highest bus speed the board can achieve is. Still, at least you get something out of it.
January 18, 2007 6:53:03 PM

Looking at this series as a whole I was expecting more of a generic "guide" for overclocking. However, the result is a specific guide on the specific set of hardware that was OC'd. I would think that the article was written for those with no overclocking experience. However the series arrives at conclusions without explaining how to get there.

Example: the recommended, maximum low-risk CPU voltage of 1.45v
The specs from Intel's site offer a range for the E6300 of 0.8v-1.35v
While the recommendation may be fine for the E6300 and similar CPUs it would have been better to give the reasoning behind the recommendation. Maybe something like, "Best low-risk practice is to increase the CPU's voltage by 10% max."


An improvement to Part 3 would have been more consistent listing of the starting values.
Example: it is explained that the memory voltage was raised from stock values to 2.2v. What is this 'stock' value?
OCZ specs list 1.9v-2.1v. DDR2 standard is 1.8v.


"...some speeds between 380 MHz and 400 MHz were unstable, while other higher speeds were perfectly fine. Increasing FSB voltage by 0.20 V got us past 420 MHz, but again, with some settings in the middle not working correctly. Choosing the +0.30 V setting eliminated all FSB stability issues and allowed us to shoot for the moon!"

What prompted further testing past the "settings in the middle"?
I have no overclocking experience. From all that I have read when reaching an unstable point you either up the voltage and try again, or decide the limit has been reached (you can also try reducing temperatures if possible, and test again). What would be the indicator that perhaps a higher speed would be more stable at current voltage settings? At what point does further speed-setting-only experimentation involve too great a risk of permanent damage to the CPU?

In the end section "Backing down" how were those values chosen?

What I would like to see would be the guidelines on how to set reasonable boundaries on the various parameters a noobi OC'er should experiment with. This guide would be geared toward OC'ing a system for continuous, long term use. Not for simply achieving the highest OC possible that will run just long enough to pass an arbitrary stability test. What percentage of the default/as-shipped settings is a low-risk margin to experiment with that will result in a system that will run at those speeds for 1-2 years?
a b K Overclocking
January 18, 2007 11:23:45 PM

Quote:
Looking at this series as a whole I was expecting more of a generic "guide" for overclocking. However, the result is a specific guide on the specific set of hardware that was OC'd.


You must not have read the conclusion of Part 2

Quote:
I would think that the article was written for those with no overclocking experience. However the series arrives at conclusions without explaining how to get there.


You must not have read the recommendations and reasons in Part 2.

Quote:
Example: the recommended, maximum low-risk CPU voltage of 1.45v
The specs from Intel's site offer a range for the E6300 of 0.8v-1.35v
While the recommendation may be fine for the E6300 and similar CPUs it would have been better to give the reasoning behind the recommendation. Maybe something like, "Best low-risk practice is to increase the CPU's voltage by 10% max."


The reason is covered in the E6300 recommendation, part 2. The old 10% rule of thumb is just that, a rule of thumb, and doesn't apply to all processors. Good voltage recommedations can only be made from experience, the experience was carried over in Part 2 recommendations.

Quote:
An improvement to Part 3 would have been more consistent listing of the starting values.
Example: it is explained that the memory voltage was raised from stock values to 2.2v. What is this 'stock' value?
OCZ specs list 1.9v-2.1v. DDR2 standard is 1.8v.
Stock value for DDR2-800 is 1.80V. OCZ specs 1.80V for this particular RAM. Most of OCZ's other RAM is specified at above stock values.

Quote:
"...some speeds between 380 MHz and 400 MHz were unstable, while other higher speeds were perfectly fine. Increasing FSB voltage by 0.20 V got us past 420 MHz, but again, with some settings in the middle not working correctly. Choosing the +0.30 V setting eliminated all FSB stability issues and allowed us to shoot for the moon!"

What prompted further testing past the "settings in the middle"?
I have no overclocking experience. From all that I have read when reaching an unstable point you either up the voltage and try again, or decide the limit has been reached (you can also try reducing temperatures if possible, and test again). What would be the indicator that perhaps a higher speed would be more stable at current voltage settings? At what point does further speed-setting-only experimentation involve too great a risk of permanent damage to the CPU?


Guesswork, the important part being the result, raising FSB voltage cured the problem.

Quote:

In the end section "Backing down" how were those values chosen?


CPU FSB was backed down by one setting level, the CPU was dropped to a FSB that could be run at that reduced voltage level, etc.

Quote:
What I would like to see would be the guidelines on how to set reasonable boundaries on the various parameters a noobi OC'er should experiment with. This guide would be geared toward OC'ing a system for continuous, long term use. Not for simply achieving the highest OC possible that will run just long enough to pass an arbitrary stability test. What percentage of the default/as-shipped settings is a low-risk margin to experiment with that will result in a system that will run at those speeds for 1-2 years?


Results always varry, in the end it's all about guesswork and experience. Part 2 included a wide variety of experience that came from other people's guesswork, Part 3 put some of it to use.
January 19, 2007 2:10:56 AM

Quote:
What percentage of the default/as-shipped settings is a low-risk margin to experiment with that will result in a system that will run at those speeds for 1-2 years?


I think the answer to that question is any setting that doesn't require an increase in voltage and runs within your RAM's memory specs. For example I OC my e6300 with DDR2-667 to a FSB of 333 MHz at stock voltage. It's a safe and stable OC, IMO.
January 19, 2007 2:35:55 AM

I have overclocked before and have tried to get more out of an older system in order to play games etc.

Some of the gains from OC this cpu are amazing.........but

If your already getting 115-130 frame in fear I would rather just keep it running stock. It performs great at stock speeds IMO for gaming. Now as newer games come out and you need more performance overclock the hell out of it.

I guess im a chicken I have never put a new system together and started off with an overclock. I usually end up doing it later on when it is needed.
January 19, 2007 6:45:36 AM

Hi.
I was reading OC guides for some time, tried with my new PC, but last week It sudenly died. :)  Some would say, ofcourse, you overclocked it.
But, here is the thing:

Asus P5W HD Deluxe
E6600 @ 3.2GHz (8x400) 1.35V
RAM @ 800Mhz 4,4,4,12 (which are declared specs for my Corsair) 2.1V
PCI @ 33.3MHz, PCIe @ 110MHz
CPU temp 44C under the torture, MB temp also 44C

And what is strange for me is that perfectly stable system with moderate voltages and temps, dies like that. Actualy the proc had died. I managed to change it under the warenty. They also changed the MB just in case.

Now I'm afraid to try the same OC-ing although I'm sure I didn't fry the proc. :) 

Any thoughts or suggestions on that?

Thanks!
B.
a b K Overclocking
January 19, 2007 6:50:01 AM

Some people have bad luck. Some parts die early, even without being overclocked.
January 19, 2007 8:00:50 AM

I think this OC system would not be good idea in extended periods of time. I think it will be stable in default, but there will surely be things that will crash system, maybe i`m wrong, but OCed system will never be as stable as it gets in stock, or at least in these high settings.
a b K Overclocking
January 19, 2007 8:25:06 AM

I've been running a 2GHz Athlon 64 at 2.4GHz for six months and not had a crash. That's as agressive an overclock as an E6300 at 3.0GHz.
January 19, 2007 8:42:18 AM

Yea and what do you do on your pc? Office work? Then i`d believe.
January 19, 2007 10:21:43 AM

hello

this is off topic but i need an advice...

i just have a question: I have bought one month ago a Pentium D 820 (at around 100$) , a Thermaltake Golden Orb 2 cooler and a Gigabyte 945PL S3 v.2.0 motherboard. I also bought a ATI R x1650XT from Sapphire + 2X512 DDR 533 from Corsair hoping that a low cost system like this one could give me a good gameplay experience (even if not at highest settings).

In the past few months I read de OC articles and I went home decided to try an oc on my system just to see how far it goes...

but then there is this problem: in my motherboard's BIOS I have no memory timing settings and jus CPU voltage and frequency modifiers. I said to myself that I had to get another BIOS version but Gigabyte doesn't offer any update...

My question is : have I chosen the wrong components for OC? I'm asking because on the MB box it is written that it was built for high end users and maybe OC...

thanks:) 
January 19, 2007 11:18:27 AM

Definitely total lack of any capability to read before you buy anything. If MB does not have any OC friendly notices, then it is no go. Normaly you can find out what mobos in bios provide. Box says a lot, my asus box have so many OC stuff written on box i can hardly notice anything not concerning OC friendly info, damn good options but as unstable as nforce can get. Damn next time AMD is no go for me...
January 19, 2007 11:54:02 AM

how about a overclocking A E6600? that should be interesting.
and why dont is chosed as mobo the DS4? or the P5B deluxe?
January 19, 2007 6:49:21 PM

I think because they were trying to keep a midline price range.

It was nice to see the components used so closely match my own. Makes me feel better about my choices for sure. :)  I have the S3 instead of the DS3, but same thing. Have the exact vid card too. G. Skill ram instead of OCZ, but same timings, etc.

Did I read the article right where they didn't manually set pci-e to 100? So the OC on the FSB also OCd the vid card? If so, how dangerous is that for the vid card? I'm all for getting more bang for my buck
a b K Overclocking
January 19, 2007 7:17:09 PM

Quote:

Did I read the article right where they didn't manually set pci-e to 100? So the OC on the FSB also OCd the vid card? If so, how dangerous is that for the vid card? I'm all for getting more bang for my buck


Great catch there on any article omission. Fortunately, it's in the pictures from Page 4:



Set to 100MHz PCIe
January 19, 2007 7:38:45 PM

Hmmm, so then I'm confused on how they got the overclocked speeds on the GPU. It did mentioned ATI Tools and the like, but said they weren't up to date (or some such) for the 1950Pro. I'll read it again and see if I missed something.
a b K Overclocking
January 19, 2007 8:06:34 PM

ATI Tray Tool was the one that supported the card they used. Others were not up to date on X1950 Pro support. Asus also included a utility, but it had a much more limitted range.
January 19, 2007 8:11:33 PM

Reread it and caught that. Was just about to edit my post. Thanks anyway for the heads up.
January 19, 2007 8:50:15 PM

Kick-ass, I love it, THG.

I only have one little spur. What is with people and the DS3? I mean, it is a great board, but haven't you been riding the DS3 bandwagon for a bit too long now? It was great earlier, but now there are better boards for less.

Take my prime example, the ASUS P5N-E SLI. You get a LOT of things you don't get with the DS3:

1. Firewire. The DS3 doesn't have Firewire on the back I/O panel, not even a header on the motherboard, while the P5N-E has both. (Cured in the 2.0 Revision of the DS3 board.)
2. SLI/More PCI-e slots. The Gigabyte doesn't even have a second x16 slot. No CF or SLI.
3. 2 IDE Connectors. Come on, this is a rarity in new boards today! A total of four IDE devices!
4. eSATA. Not that I would use it, but happy to know that it is there.
5. 1333FSB Support. This is a near deal-breaker for me. The P5N-E SLi has support for the upcoming Bearlake processors and others.
6. Superior Overclocking. This thing should hit at least to what the 965P/ICH8 combination can do, most likely more, possibly 500FSB! A few mods and you might get close to 600MHz (from what I've heard).

And it is $10 cheaper.

~Ibrahim~
a b K Overclocking
January 19, 2007 9:31:29 PM

OMFG the article was planned in December, following the DS3's HORRIBLE FAILURE to achieve top overclocking scores in its previous review.

Everyone else jumped some other bandwagon, this article contained all original content.
January 19, 2007 9:55:52 PM

I know this is a planned part from the previous ones, hence the title "Part 3", I'm just mentioning for those who are CURRENTLY making builds that there are better options!

~Ibrahim~
January 20, 2007 5:02:33 AM

Quote:
Kick-ass, I love it, THG.

I only have one little spur. What is with people and the DS3? I mean, it is a great board, but haven't you been riding the DS3 bandwagon for a bit too long now? It was great earlier, but now there are better boards for less.

Take my prime example, the ASUS P5N-E SLI. You get a LOT of things you don't get with the DS3:

1. Firewire. The DS3 doesn't have Firewire on the back I/O panel, not even a header on the motherboard, while the P5N-E has both. (Cured in the 2.0 Revision of the DS3 board.)
2. SLI/More PCI-e slots. The Gigabyte doesn't even have a second x16 slot. No CF or SLI.
3. 2 IDE Connectors. Come on, this is a rarity in new boards today! A total of four IDE devices!
4. eSATA. Not that I would use it, but happy to know that it is there.
5. 1333FSB Support. This is a near deal-breaker for me. The P5N-E SLi has support for the upcoming Bearlake processors and others.
6. Superior Overclocking. This thing should hit at least to what the 965P/ICH8 combination can do, most likely more, possibly 500FSB! A few mods and you might get close to 600MHz (from what I've heard).

And it is $10 cheaper.

~Ibrahim~


What do you think of the P5B Deluxe WiFi ?
January 20, 2007 12:36:05 PM

Quote:
What do you think of the P5B Deluxe WiFi ?


Too expensive. the eVGA 680i board isn't too far away.

~Ibrahim~
January 21, 2007 8:20:19 AM

Quote:
"...some speeds between 380 MHz and 400 MHz were unstable, while other higher speeds were perfectly fine. Increasing FSB voltage by 0.20 V got us past 420 MHz, but again, with some settings in the middle not working correctly. Choosing the +0.30 V setting eliminated all FSB stability issues and allowed us to shoot for the moon!"

What prompted further testing past the "settings in the middle"?
I have no overclocking experience. From all that I have read when reaching an unstable point you either up the voltage and try again, or decide the limit has been reached (you can also try reducing temperatures if possible, and test again). What would be the indicator that perhaps a higher speed would be more stable at current voltage settings? At what point does further speed-setting-only experimentation involve too great a risk of permanent damage to the CPU?


Guesswork, the important part being the result, raising FSB voltage cured the problem.

What do you mean, guesswork?

This is a well-known issue with the P965 chipset, that it may be unstable in the 370-400 FSB range, but perfectly stable above that. For example, people have reported 400 FSB being unstable on the DS3 motherboard, but 401 FSB being perfectly stable.

The reason is that the motherboard switches the northbridge from the 1066 strap to the 1333 strap somewhere in the 370-400 FSB range (the exact number depends on the motherboard). The northbridge is like another little CPU, and it has its own internal multiplier that relates its internal clockspeed to the FSB. While it's on the 1066 strap it has a certain multiplier, and the higher you raise the FSB the faster the northbridge chip runs. At above 370 MHz FSB with that multiplier, the northbridge can become unstable. But at around 400 FSB (on the DS3 board) it switches to the 1333 strap, with a lower internal mulitplier and looser memory timings. Then the northbridge's internal speed isn't as fast and it's stable up to 500 FSB and beyond (depending on the motherboard, of course). In other words, the northbridge chip keeps running faster and faster as you raise the FSB from 266 to 400, but then jumps down to a slower speed at 401 FSB and then gets faster and faster as you raise the FSB from there.

That's why you may get lower memory performance at 401 FSB than you had at 370 FSB... the memory bandwidth only gets up to its previous level at about 420 FSB.

Impkin, people know this from experience overclocking the 965 chipset. The solution is not necessarily to raise the FSB voltage as Crashman says, but to just skip from 370 FSB to 401 FSB.
January 21, 2007 6:18:09 PM

Hey, magreen. This problem, then, is restricted to the P965 boards?

~Ibrahim~
January 22, 2007 1:00:11 PM

I'd imagine every chipset has to switch from the 1066 strap to the 1333 strap at some point. I'm not sure where it happens on other chipsets.
January 22, 2007 3:15:34 PM

Thank you for the answers to my questions. Every bit helps the learning process.

I hope the guide's author and any forum readers did not take my post wrong. I did get useful information from all three parts.

I gather from the responses that the areas I had concerns about are difficult to cover in any general guide. The variations in hardware require any guides to be hardware specific. So if one doesn't have experience in OC'ing they should do extensive searching for hardware specific OC'ing info. This latter is assuming the lack of a budget to support experimentation and the cost of hardware replacement learning the boundaries. If a person can afford to discover the component life, stability, and performance boundaries on their own, go for it! And then please share your results. :D 
January 22, 2007 8:25:56 PM

Quote:
I'd imagine every chipset has to switch from the 1066 strap to the 1333 strap at some point. I'm not sure where it happens on other chipsets.


Got it. Wouldn't this, however, only be a problem on boards supporting 1333FSB or do they all already have an inherent design to allow 1333FSB?

~Ibrahim~
January 22, 2007 8:51:33 PM

Im trying to OC my 4200+ (939) and cant seem to get numbers even close to ones posted around on the net.
I'm using the following cooler Zalman CNPS7700 and 2 chase fans. This is how the setup looks

x. Big Ass Case
x. 480W (true) PSU)
x. Asus 7600GS PCI-E
x. 1x WD 250GB
x 2GB Dual Channel kit (Silver) OCZ
x M-Audio Audiophile 192
x Asus A8N-SLI premium board
Thats it. Nothing else in the box.

I can clock the btich up to 2.75Ghz (250x11) and can make it into windows, temperature is below 50degC. However the system is unstable, i cannot successfully complete a 3dMark 06 test without instant reboot.

The most stable clock i can reach is 2588.001Mhz (235x11)
the
HT multiplier @ 5x
FSB @ 235Mhz
Core Multiplier @ 11 (fixed)
RAM Bus @ 333Mhz
PCI bus @ 33Mhz
PCI-Express @ 100Mhz
CPU voltage @ 1.45v
HT @ 1.5v
RAM @ 2.8v

Why does my machine become unstable at 2.75Ghz even though the cpu is not over heating? Note: At 3Ghz the machine idles at 42degC and a 2.58 idles at 40degC (max out at 50degC)
a b K Overclocking
January 22, 2007 9:27:18 PM

Droppping your HT multiplier to 4x should get you to around 250, dropping it to 3x should take you to the limit of the CPU.
January 22, 2007 9:34:28 PM

Quote:
Droppping your HT multiplier to 4x should get you to around 250, dropping it to 3x should take you to the limit of the CPU.


This looks like the most obvious solution and it probably is. What memory is it, like speed and timings? 2.8V looks enough, either way.

~Ibrahim~
January 23, 2007 6:54:17 AM

Quote:
I'd imagine every chipset has to switch from the 1066 strap to the 1333 strap at some point. I'm not sure where it happens on other chipsets.


Got it. Wouldn't this, however, only be a problem on boards supporting 1333FSB or do they all already have an inherent design to allow 1333FSB?

~Ibrahim~I know the 975 boards that didn't switch to the 1333 strap couldn't OC past 370 FSB. I'm not sure if later they revised them to switch to the 1333 strap.

Basically, you'd have to have a mother of a northbridge chip to handle 400+ FSB on the 1066 strap... it'd be overclocked to kingdom come. It's like driving 75 mph in 2nd gear.

So I think all boards that OC well switch to the 1333 strap (though I don't have info on the nforce chipsets).
a b K Overclocking
January 23, 2007 3:40:15 PM

Quote:
I know the 975 boards that didn't switch to the 1333 strap couldn't OC past 370 FSB. I'm not sure if later they revised them to switch to the 1333 strap.

Basically, you'd have to have a mother of a northbridge chip to handle 400+ FSB on the 1066 strap... it'd be overclocked to kingdom come. It's like driving 75 mph in 2nd gear.

So I think all boards that OC well switch to the 1333 strap (though I don't have info on the nforce chipsets).


I'm fairly certain the 975X doesn't support 1333 strap. Intel will be replacing the 975X in a few months.
January 23, 2007 3:51:13 PM

Quote:
Droppping your HT multiplier to 4x should get you to around 250, dropping it to 3x should take you to the limit of the CPU.


Even the current state (posted previously) crashed the system last night. It was frozen solid when i woke up today.

The 2.75ghz was reached with the HT set at 4x. But i could not get it stable

I have not set any special timing for the memory, all default clock cycles.

What does changing the HT multiplier do?
January 23, 2007 10:57:50 PM

Quote:
Droppping your HT multiplier to 4x should get you to around 250, dropping it to 3x should take you to the limit of the CPU.


Even the current state (posted previously) crashed the system last night. It was frozen solid when i woke up today.

The 2.75ghz was reached with the HT set at 4x. But i could not get it stable

I have not set any special timing for the memory, all default clock cycles.

What does changing the HT multiplier do?

Got it, thanks so much for the info, guys!

Ah, what divider are you running, 1:1? That could be it. Your memory might not be able to run at 250MHz. You should change that. Some enthusiasts are of the party that removing your 1:1 ratio wrecks performance, so you should then loosen the timings to accommodate the frequency change. You could also up the memory voltage, but it is already a tad high. Any cooling on the sticks?

HT Multiplier: FSB times HT Multi < 1175. It is the rate in which your HyperTransport bus runs at, but other than that, I have no idea what it does, lol.

~Ibrahim~

~Ibrahim~
January 23, 2007 11:41:07 PM

Quote:

Ah, what divider are you running, 1:1? That could be it. Your memory might not be able to run at 250MHz. You should change that. Some enthusiasts are of the party that removing your 1:1 ratio wrecks performance, so you should then loosen the timings to accommodate the frequency change. You could also up the memory voltage, but it is already a tad high. Any cooling on the sticks?

I have an AMI bios , Pheonix version 4 something. Do you know where this divider is?
I understand what you are saying, but the ram clock is set at 400Mhz. It does not change according to FSB so what does this divider do?

What is the point of increasing memory clock speed, and then at the same time increasing read/write timings. That will just cancel each other out.

Here is a Link to the manual of my main board. It has a complete description (with screenshots) of all the bios features. I would be greatful if someone can help me reach a stable 3ghz.
a b K Overclocking
January 24, 2007 12:15:21 AM

Quote:
Some enthusiasts are of the party that removing your 1:1 ratio wrecks performance, so you should then loosen the timings to accommodate the frequency change.


But what IS a 1:1 ratio? You're not going to clock your RAM to 2000MHz, and probably not even DDR2-1000 in dual channel mode. Let's look a little deeper!

I'm fairly certain the HT link is only 16-bits wide. That makes 5x200MHz x16bits roughly equivalent to 5x50MHz x64bits, DDR-250 Single Channel! Of course you want your memory to be twice as fast since HT is full-duplex but RAM is only half-duplex, and that's still only DDR-500!

And that's not even factoring in dual-channel mode! So do you really need anything more than the lowest speed setting for good performance?
January 24, 2007 12:45:45 AM

How can the HT link be 16bits wide?
The AMD X2 is a 64bit cpu, doesnt it require a minimum of 64bit wide memory address space? The registers are 64bit wide too right?

Look at this link
3Ghz on AMD 4200+

I cant even get over 2.6ghz without becoming unstable

The A8N-Sli Premium does not have a memory "divider", i can sett mem clock manually though.

Here is the current setup...
Mem 400Mhz
HTT 235Mhz
CPU Multi 11x (stock)
HT Multi 5x
Mem voltage 2.8v
Cpu core 1.45v

I managed to get 250x11 with 4x HT multiplier, but i could not get passed even 1 3dMark test without a immediate reboot.

How do the guys in the link i posted manage to hit those speeds and keep it stable? Im getting frustrated. The only difference on their setup is they are using a AM2 and i have a 939, is there that much of a difference?
a b K Overclocking
January 24, 2007 1:23:04 AM

You can put 64-bit instructions through a 4-bit pipe, just look at Serial ATA.
a b K Overclocking
January 24, 2007 1:24:49 AM

Oh, I'd drop the memory to the lowest speed possible while searching for the highest stable CPU speed, then raise it later to find the highest stable memory speed.
January 24, 2007 2:38:47 AM

Quote:
Oh, I'd drop the memory to the lowest speed possible while searching for the highest stable CPU speed, then raise it later to find the highest stable memory speed.


yah ur right.
I seemed to have figured it out too. my memory is not able to keep up with cpu clock at high speeds. i reached 250x11 and managed to run 3dmark06 (for stress test) . and it ran well. but windows began to complain, at this point my mem was set to 266Mhz, which is probably not worth it, there is no point on having a cpu soo fast, that the memory cannot keep up. so i just said screw it and settled for a mere 350Mhz increment of the stock clock, which seems to be rock solid. and stability is more important to me than speed. Got my cheap thrills, was fun, now back to normal.

i feel the multiplier is set far to low (11x), if i could have it at about 20x, (and if HTT was 110Mhz at stock) i could keep my memory satisfied at 200x2 and push my fsb right through the roof. But who cares. im thinking of chucking this box and going for a Sparc workstation all ways wanted to get one of those, so might as well.

thanks guys for the help...
January 25, 2007 12:12:11 AM

Quote:
Some enthusiasts are of the party that removing your 1:1 ratio wrecks performance, so you should then loosen the timings to accommodate the frequency change.


But what IS a 1:1 ratio? You're not going to clock your RAM to 2000MHz, and probably not even DDR2-1000 in dual channel mode. Let's look a little deeper!

I'm fairly certain the HT link is only 16-bits wide. That makes 5x200MHz x16bits roughly equivalent to 5x50MHz x64bits, DDR-250 Single Channel! Of course you want your memory to be twice as fast since HT is full-duplex but RAM is only half-duplex, and that's still only DDR-500!

And that's not even factoring in dual-channel mode! So do you really need anything more than the lowest speed setting for good performance?

I'm part of the other party, one who does not buy uber-fast memory. I don't believe a word of it (fast memory for performance), I thought it just might be nice to mention it, you know, both sides of the story.

Nice info, though.

Which makes DDR2 near useless, lol...Minus power efficiency.

~Ibrahim~
January 29, 2007 3:46:35 PM

Hello. I enjoyed the article but I am frustrated that I can't find the memory mentioned in the article. This isn't the first time.

Try finding the memory mentioned at www.newegg.com or www.zipzoomfly.com.

You can't...

The only place I've found it online is through someplace in Canada.

Any idea where to go or to use instead?

Thanks.
!