Need someone to hold hand while OCing

[sellmesanity]

Yo everyone, I know how to overclock, except I am not too keen on the testing and the increments. My goal is to have a 1:1 overclock. I have ddr500 ram. Does this mean I can raise my htt to 250 without testing in increments? And if i do have to test in increments, how many mhz increments and how long would i run stress prime for during the increments? What are acceptable regular temps for an OC? I only have stock cooling, but all other components in my comp are ready to go. My goal is 2.4-2.6, it should be easy from a 2.2 3500+. HELP!!!

 

[Rik]

If you are not 100% sure of what you are doing then overclocking isn't a good idea. Especially if you only have stock cooling.

I got a water kit and 11 fans in my system before i even thought about overclocking.

 

[sellmesanity]

Thanks for not contributing anything to the thread. What about those stories of insane overclocks with stock cooling? I know how to overclock, I just want expert input. I have great airflow in my case. 1 Intake, 2 exhaust fans...and all my components are highly OCable.

Oh yeah and about your "you can never overcool a pc" actually yes you can. If an Athlon 64 goes below a certain temp it cannot run.

 

[kirock]

The only sure way to OC is go in increments. The fsb (HT) can be moved up in big chunks to start, then you see what the max clock freq the CPU can handle. Then you work on your RAM multi and resulting HTT.

So to start you set your RAM multi to X3 (dwn from X5) this ensures your HTT does not become the bottle neck. Now increase the FSB by 100M steps, 2.3, 2.4, 2.5 etc. Run a quick 5 min Prime95 test each time.
When you find the max cpu freq, back the fsb off from that a little. Run Prime95 for 10 min and check temps.
Now start increasing your RAM multi and play with your ratios to get the HTT as close to 1000MHz as possible. Run P95 for 5min each time.

If you get all fubared, then you have to reset the CMOS and start again.
Keep watching your CPU temps.
My 3500 runs at 30-35C at idle and 45-50 under heavy load. I would only let it go to 55-60C as a maximum temp ever. (65C is shut down). Mine is ever so slightly OCd to 2.3 b/c of the unstable PCI bus in my mobo after the fsb goes too high.

 

[sellmesanity]

So let me get this all straight.

The RAM multi is the LDT right? I increase the HTT (FSB) just enough so that it hits 2.3 then 2.4 then 2.5 and each time it goes up 100mhz I test with prime 95 for 5 minutes....gotcha. BUT in prime, do i do a blend test, cpu test, or ram test? And would it be logical to lower the CPU multiplier from 11 to 10 for cleaner numbers (in mhz)? How can I determine the max cpu frequency? At the end of the overclock, shouldn't I prime for a whole day or something?

 

[TCool]

Thanks for not contributing anything to the thread. What about those stories of insane overclocks with stock cooling? I know how to overclock, I just want expert input. I have great airflow in my case. 1 Intake, 2 exhaust fans...and all my components are highly OCable.

I've got stock hsf on cpu and 3 case fans and I overclocked my P4 prescott from 3.0 to 3.9 b4 I decided I didn't wanna bring up temps anymore. The temps I had were pretty high, but nothing compared to some of the temps I've read about from prescotts. I think that guy is just a lil too paranoid.

 

[sellmesanity]

Ok I will try this now, will keep posted. I don't plan to do insane ones like i said, as stability is INCREDIBLY important to me. 2.5 max. (250x10 on 1:1)

 

[sellmesanity]

Update:

My computer boots at 2.5 ghz, but fails Stress prime (CPU test) within 2 minutes. It doesn't fail it at 2.4. What does this mean?

Type: Small FFTs - stress CPU Min: 8 Max: 32 InPlace: Yes Mem: 8 Time: 15
CPU: 2497MHz FSB: 227MHz [227MHz x 11.0 est.]
8/21/2006 5:37 PM Beginning a continuous self-test to check your computer.
Press Stop to end this test.
Test 1, 800000 Lucas-Lehmer iterations of M172031 using 8K FFT length.
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.
Torture Test ran 1 minutes 33 seconds - 1 errors, 0 warnings.
Execution halted.

 

[kirock]

Well it can mean you didn't turn your memory mulit down from 5 to 3! Or the CPU is reaching it's limit of stability. If memory is eliminated by going to X3, then try adjusting up from 2.4GHz in smaller increments. Finding the limit of the CPU is the whole point in this first step. (maybe you found it, i.e. 2.48 or 2.46, etc. and at 2.5 you are already past it).

 

[KingCody]

it can mean you didn't turn your memory mulit down from 5 to 3!

I think you mean the HTT multi...

 

[sellmesanity]

The whole idea for the HTT multi (or memory multi) is that it doesn't go over 1000. The max HT I used was 250. 250x4 is 1000. I wanted this 1:1 ratio. I used no dividers. At stock, my ram is DDR500. Just by lowering the HT by 5mhz to 245 HT, it became stable. Is this a sign of a CPU wall? 2.45 max? It has got to be voltages. Some forum I read specifically said not to push the voltages too far past spec if you have stock cooling, so I guess 2.45 is my wall.

I have a new problem now though, I am not sure it was because of the overclock. It shouldn't be, unless it caused damage. I will start a new thread to describe the BSOD that happened even after I stopped overclocking.

Edit:

Oh yeah my voltages were: 1.7 chipset (from 1.5), 2.9 Vdimm (from 2.7) 1.475 Vcore (from 1.4)

 

[i_am_a_newbie]

The whole idea for the HTT multi (or memory multi) is that it doesn't go over 1000. The max HT I used was 250. 250x4 is 1000. I wanted this 1:1 ratio. I used no dividers. At stock, my ram is DDR500. Just by lowering the HT by 5mhz to 245 HT, it became stable. Is this a sign of a CPU wall? 2.45 max? It has got to be voltages. Some forum I read specifically said not to push the voltages too far past spec if you have stock cooling, so I guess 2.45 is my wall.

I have a new problem now though, I am not sure it was because of the overclock. It shouldn't be, unless it caused damage. I will start a new thread to describe the BSOD that happened even after I stopped overclocking.

Edit:

Oh yeah my voltages were: 1.7 chipset (from 1.5), 2.9 Vdimm (from 2.7) 1.475 Vcore (from 1.4)

Was it necessary to up your chipset voltage? I was able to bump my 3200+ from 2.0 to 2.5 on stock cooling(upgraded to a zalman now) and stock voltages. Even so, I don't think(unless you OC'd your NB to SB or vice versa) it's required to up your chipset's voltage.

I was given advice to leave the chipset freq's/voltages alone.

 

[sellmesanity]

I will try again except I wont change my chipset voltage, after I fix this BSOD problem. BSODs really irk me. They make me feel like something is wrong with my 1000+ dollar investment here...

I would think the opposite about chipset voltages because when your HT is upped, your whole system bus speeds up, this you needing the motherboard to be able to withstand the new speed.

 

[sellmesanity]

Update: The problem is not because of the OC so i started trying again.

Turns out that when i upped the Vcore to 1.52v it primed stable for 19 hours (CPU test) and I am currently testing ram at 2.8v. It is at 2.5 ghz just like how I wanted. I can push higher but I do not have adequate cooling solutions.


My original plan was 250 x 10 = 2.5 --That way I could use my DDR500 to it's full potential. This posed a problem because it caused my computer to produce excess heat because of the RAM, despite the fact that it was rated to run at that speed.

I though to myself, why am I not taking advantage of my 11x multiplier? I lowered the RAM down to 227

227x11=2.5ghz STILL with less stress on my ram, letting me lower volts and having slightly cooler temps and a more stable system.

I am pretty sure I will pass the rest of my RAM stress test and I will succeed at my overclock! DO NOT underestimate first time OCers!

 

[sellmesanity]

Update part 2:

It's been almost a week and my computer is still stable at 2.5ghz. I passed stress prime for ram at 23 hours, and cpu for 19. Only con is that I get 44c idle temps on the cpu and max temp of like 51c on load. Wish they were lower, but I do not want to buy extra cooling. I DO however want to buy an air conditioner so i think this will help immensely. My room is very hot.

Did I make you proud, rik? hahahaha

Can anyone give me their Athlon 64 OC results and their divider?

 

[KingCody]

Can anyone give me their Athlon 64 OC results and their divider?

sure...

2518MHz (+2.5GHz) was the highest I was able to achieve on my 3000+. FSB=279 (anything higher and it won't boot), CPU MULTI=9 (9 is the max for a 3000+), RAM DIVIDER=14 (my mobo does not have a RAM divider of 13 when using a CPU multi of 9 ) CPU VCORE=1.47

Since I can get over 2.5 with a 3000+, I would imagine that you could easily achieve 2.7GHz since you have access to a CPU multi of 10 (of course each CPU has a different overclock ceiling).

good luck

 

[sellmesanity]

Ah KingCody, that is cool that you got 2.5 with only 1.47 volts. I had to do 1.5 otherwise prime would error within 10 minutes. My multiplier is actually 11...11x227. I have no divider. I could definitely go higher. I can do 11x250 on a 1:1 ratio and get like 2.750, but I am held back by my stock cooling and CPU vcore.

 

[KingCody]

you should be able to hit 2.5GHz without having to raise the VCORE all the way up to 1.5v. it's only a 300MHz overclock. the extra voltage creates the extra heat. what are your idle and load temps now?

your RAM may be holding your overclock back. I know you have DDR500, but that doesn't mean that it can run that fast with the default settings that your mobo has it running at. you should take it out of the equation in order to find your max CPU overclock. try these settings:

a. BUS(FSB) to 250MHz
b. CPU multiplier to 10
c. HTT multiplier to 4
d. RAM divider to 13 or 14

if it boots up and runs stable, then you know that your RAM settings were holding you back, not your CPU. you can then raise the BUS even higher or raise the CPU multi to 11 or maybe even both.

I wouldn't worry about running your RAM at a 1:1 ratio. AMD64 systems are not as bandwidth dependant as intel systems are, so you benefit more from tighter timings than you do with faster speeds. it's best to have both high speed and tight timings... but if you have to choose between the two then tighter timings will usually yield better performance in AMD64 systems.

that being said... you have high latency RAM. it's not bad RAM, but it's designed more for Intel based systems. even though it can run at 250MHz, it suffers from high latencies at that speed.

:wave:

 

[sellmesanity]

Ah this pleases me to know that there is improvement to be done in my overclock, and a cool guy to help me out.

My idle temp is about 44c. What I don't understand is why ram would hold my OC back when in Stress Prime i test only the CPU and not the ram and it errors below 1.5Vcore, but if I up just the CPU vcore then itll work at 2.5ghz without erroring for 19 hours.

Ah and one more thing...what speed would I need to set the ram at to get a 13 or 14 divider? Im still new believe it or not! hahaha

 

[sellmesanity]

Nvm i figured out the divider. My computer boots up. When I lower the vcore to the same as yours, my Stress Prime errors within 5 minutes. I think some processors just need more juice (mine 1.5 haha). Maybe it's my motherboard or something. Anyway, I'm gonna try my same settings except i'm going to see what a divider can do so that I have cas 2 timings. Suggestions?

 

[KingCody]

yeah your cooling is definitely going to hold you back, that's too bad.

when I refer to your "max overclock", I'm not talking about stress testing. I'm simply referring to the fastest speed you could achieve and still have the PC boot up. to achieve your optimal system overclock (not just the CPU) you want to isolate components to see what each one can handle. then balance them out.

personally I could care less about stress testing. I really don't care if it passes or fails a stress test. as long as it doesn't crash while playing halo/fear/cod2 etc then I'm satisfied with my overclock. if it crashes then I back the overclock down a bit, that's all.

I googled your RAM a little bit and from what I read it can operate at 2-3-3-6 @DDR400 @2.8v so you could try that.

:wave:

 

[TimeParadoX]

I dont want to say anything because I might break your computer =P

 

[sellmesanity]

ya i run it now at 2-3-3-6 but at 2.7 . my new divider is 12 apparently. I just don't understand why my cpu needs more vcore than someone else's....