HELP NEEDED: A Stable Overclock Technique

[Caribouuu]

Hi, I have a Athlon64 3500+ (New Castle core),

My main question is, when I set the cpu values at 95% of my maximum speed to attempt 12hour prime stability, do I have to keep the vcore voltage as high as it was when I achieved the maximum speed, or can I lower it to all that was needed to run Prime95 for +/-10 minutes at the speed I'm setting it at for stability?

Someone please let me know this so I can continue with my overclocking project.

In other words, my maximum speed is 2538mhz at 1.65v, do I set it to 221x11 (2431mhz) but keep the voltage at 1.65v, or can the voltage be lower? I originally found 221x11 stable for at least 10 minutes at 1.5v, and that would be good because I might want to still run AMD Cool'n'Quiet; which automatically sets voltage to 1.5v.

Question #2: I haven't tried using lower multipliers yet, should I? What are the advantages, other than more variability in cpu speed?

Here are the notes I took.

max 1.5v:221x11=2431 (10 minutes); 222x11 (2 minutes)
max 1.55v: 225x11:2475mhz (9 min stability) / 226x11:2486mhz (no stability)
41c
max 1.6v:227x11:2498mhz (stable 6 minutes) / 228x11:2508mhz (no stability) 42c
max 1.625v:228x11:2508mhz (9 minute stability) 52c


@1.55v=270x9=2430mhz; 275x9=2475 (44c) (5 minutes barely);
@1.6v=276x9=2484; 277x9=2493(7 minutes); 278x9=2502 (7 minutes); 279x9=2511 (10 minutes+?) (46.5c); 280x9=2520 (9 minutes\.);
@1.625v=281x9=2529 (2.5 minutes)
@1.65v=281x9=2530 (8.7 minutes); 282x9=2538 (2 minutes)

maximum CPU speed 2538mhz. (2538x0.95=2411.1)

The Athlon64 3500+ New Castle has a 130mm core which runs at 1.5v factory default. Word on guides and forums is max voltage for overclocking 130mm chips safely is 1.6-1.65v.

Rightmark CPU Info:
minimal: 4x, 0.8v
maximal: 11x, 1.55v

This third and last question is partly related to Question 2:

FOLLOWING THIS RIGHT-HERE OVERCLOCKING GUIDE, the highest CPU speed I've booted up into was 282x9 = 2538mhz (Prime95 ran for 2 minutes before blue-screen) at 1.65v. With the multiplier set to it's maximum, I could get only 228x11=2508mhz, (Prime ran for 9+ minutes before I stopped it, or it blue-screened, I can't remember, but it probably blue-screened) at 1.625v. 229x11=2519mhz, significantly lower than 282x9=2538mhz, so why doesn't it work?

So Question 3 is can anyone give me any clues on why 229x11 won't let Windows boot up and when, if at that point, attempting to go to 1.65v actually sets the vcore on the A8V Deluxe Rev.2 to like 1.77v (according to the bios hardware monitor, which scared the crap out of me and I immediately turned the HTT/FSB back to 228 and the vcore down to 1.5v and restarted). Switching to 1.65v before did not cause this glitch, when I was successfully able to OC to 282x9 = 2538mhz at 1.65v. Furthermore, I've heard of processor/motherboard combos like mine doing significantly better on air cooling with a 11x multiplier, around the 270mark. It is just that my cooling isn't good enough? The cpu temp gets to about max 47*c on full load at the top speeds.


Oh one last thing I just discovered: The memory timings on OCZ's website specify 2-2-2-5, but that crashes Memtest86 on test 2, no overclock. I have to run 2-3-2-5 instead. Motherboard vltage maxed at 2.8v, A8V Deluxe Rev.2 and OCZ Platinum PC3200 4x1024mb. Any ideas on that?


Thank you in advance for any clues you can supply.

 

[cmdrdredd]

You need to prime test for the full 12+ hours (prefer 24) to be sure it's 100% stable at the voltage you set. If it's stable at 1.65v and 2.5Ghz and you go down to 2.4Ghz you can drop the voltage down too, but you have to test how low you can go.

Let me give you an example of my system. I set my Q9550 at 3.6Ghz and it takes 1.29-1.3v to remain stable (actual voltage under load). This requires me to set the voltage to 1.38 in BIOS because this particular board has serious vdroop issues. That is when your BIOS setting is not what you actually get because the voltage dips a bit when the CPU is under some load. For 3.8Ghz I need 1.32v actual which is 1.41v in BIOS. I tried lower but it errored Prime95. So you see you can lower voltage for lower speed. The lower the voltage the cooler the CPU runs and the safer it is. So test until you see it crash and then go up a bit until you find the lowest voltage you can run at a given speed. It is a lot of trial and error but that's how it's done. Remember that not ever CPU will do the same thing. Your CPU may just not be as good an overclocker. Go down a bit when it crashe

 

[Mothergoose729]

Overclocking is also about finding a good balance between speed and voltage. At 2.5ghz it may only take about 1.5v to get stable, while at 2.6ghz it may take 1.65v. Obviously 2.5ghz would be a better choice. In the end you have to decide how much heat you want and how much performance you need. I recently got a E2180 for one of my computers, and I found the best balance overclocked to 2.7ghz from 2.0, undervolted from 1.3v to 1.15. Setting the speed to just 2.8ghz made me have to increase my voltages by more then a tench of a volt, which is kind of a lot. My quad core 940BE can OC to 3.7ghz stable, but most of the time I keep it at 2.5ghz. The important thing, whatever speed and voltage you set, you need to test it. I run a custom test with the threads in place, with 2 minutes of each iteration, then if it passes that, and can go more then 6 hours stable in prime95 I consider that good. Everybody has their own standards. Just make sure to keep it at those settings, changing them takes away the whole point of testing to make sure everything is stable.

 

[Caribouuu]

well i guess nobody bothered to have a look at my notes sheet.

i'll try something tonight, maybe .. 219x11 @1.5v. I've heard it takes 12 hours on prime95, and that most systems appear stable until the last four hours- wish i could find the guide where i read that now for reference.

i don't expect 221x11 @1.5v will pass the 12 hour test, although it is roughly 95% of my cpu max, due to the fact that 222x11 didn't get past 3 minutes on prime without a voltage boost, so i won't try that until i see how 219 does. if someone would have a look at my post in detail and try to answer the questions and also give me an educated guess on which overclocks to try, i'll be very happy .. nearly delighted. goodnight.

 

[Mothergoose729]

well i guess nobody bothered to have a look at my notes sheet.

i'll try something tonight, maybe .. 219x11 @1.5v. I've heard it takes 12 hours on prime95, and that most systems appear stable until the last four hours- wish i could find the guide where i read that now for reference.

i don't expect 221x11 @1.5v will pass the 12 hour test, although it is roughly 95% of my cpu max, due to the fact that 222x11 didn't get past 3 minutes on prime without a voltage boost, so i won't try that until i see how 219 does. if someone would have a look at my post in detail and try to answer the questions and also give me an educated guess on which overclocks to try, i'll be very happy .. nearly delighted. goodnight.

The simple truth is overclocking is not very difficult, and there really isn't a magic button or a mysterious manipulation of the bios that can make an overclock that can't go 3 minutes pass 12 hours without a significant voltage increase. 221x11 looks like a good clock to me, and that is about as far as the athlon 64 usually get. Instead of monitoring clock speed I would take a look at temperature. If you decide 70c under load is about as far as you ever want to go, then I would test various overclocks to see which ones get the closes to the market, and stable.

Just a few things you can try though, set your memory to a lower strap, just to rule them out as a cause for failutre. Also, don't forgot to experiment with any sub voltages, like FSB or PLL or whatever your board has. Be sparing on those though, because they can kill a chip or a motherboard pretty fast if volted to high.