trials and tribulations my first oc'ing adventure

[BoNg]

hello again after about a grueling week of patiently waiting i have decided it is time for my post, and so it starts. 3.6 ghz was original goal....my system is as follows:

phenom II x4 955 BE ?? GHz

thermalright true extreme push/pull

msi 790fx-gd70 bios ver 1.4

8 gigs - OCZ platinum amd edition 1600 MHz @1.72 v and 7-7-7-24-28-1T - OCZ3P1600LVAM4GK

msi 4890 oc edition 955/1150

2x western digital caviar black 640GB

kingwin mach 1 1220 w psu

coolermaster cm690 nvidia edition (LOL)first post with new system

log bios -cpu-vdd voltage ambient temps were normally 70-71 F
v
first post with new system
nb and ht link oc

3.2ghz - x16 - 1.30 v - 45-46 C - stable
3.4ghz - x17 - 1.30 v - 45-46 C - 1 hour prime(i stopped) i know 3 hours...cont.

3.6ghz - x18 - 1.325 v - n/a - crash prime 95 start
3.6ghz - x18 - 1.3375 v - n/a - crash 2 min
3.6ghz - x18 - 1.35 v - 47-48 C - crash 8 min
3.6ghz - x18 - 1.3625 v - 48-49 C - crash 7 min
3.6ghz - x18 - 1.3750 v - 49-50 C - crash 3 hours
3.6ghz - x18 - 1.3875 v - n/a - crash time n/a
3.6ghz - x18 - 1.40 v - 51-54 C - stable 9.5 hours prime(i stopped)


so hmmm 3.6 easy with decent temp - one fan, imo. waiting on fan holder for push pull method on cooler. new goal! 3.8 ghz. cont.

3.7 ghz - x18.5 - 1.40 v - 51-53 C - crash <2 min
3.7 ghz - x18.5 - 1.4125 v - 53-54 C - crash 5 min
3.7 ghz - x18.5 - 1.4250 v - 54-55 C - crash 8 min
3.7 ghz - x18.5 - 1.4375 v - 56-59 C - crash 1/1.5 hour < ambient here was 78 F, had to change rooms
3.7 ghz - x18.5 - 1.45 v - 52-61 C (54-55 avg) - stable 7.5 hours prime <ambient was 78 here for awhile a/c turned on

3.7 ghz - x18.5 - 201 fsb - 1.45 v - 50-54 C - 3 hours stable prime < normal ambient temps again


3.7 ghz easy, cont.

3.8 ghz - x19 - 1.45 v - n/a - crash 1 min
3.8 ghz - x19 - 1.4625 v - n/a - crash 1 min
3.8 ghz - x19 - 1.4750 v - 53-54 C - crash 10 min
3.8 ghz - x19 - 1.4875 v - 55-56 C - crash 9 min
3.8 ghz - x19 - 1.50 v - n/a - crash <20 min (forgot)
3.8 ghz - x19 - 1.5125 v - 59-61 C - crash <20 min
3.8 ghz - x19 - 1.525 v - n/a - crash <20 min
added push pull here
3.8 ghz - x19 - 1.5125 v - 60-61 C - crash time n/a
3.8 ghz - x19 - 1.525 v - 60-61 C - crash 1.75 hour (1hour 45 min)
no go i guess on multi ill lower and add some fsb
3.8 ghz - x18.5 - fsb 206 - 1.45 v - crash <1 min
3.8 ghz - x18.5 - fsb 206 - 1.4625 v - crash <5 min
3.8 ghz - x18.5 - fsb 206 - 1.4750 v - crash <3 min
3.8 ghz - x18.5 - fsb 206 - 1.4875 v - crash <9 min
3.8 ghz - x18.5 - fsb 206 - 1.50 v - crash 30-45 min
3.8 ghz - x18.5 - fsb 206 - 1.5125 v - 2 min test error
3.8 ghz - x18.5 - fsb 206 - 1.5250 v - 1 min test error
these i saw temps round 54-55 C

summary

stable settings:
3.2 ghz - x16 - 1.30 v - stable - 45-46 C - just stable ok
3.4 ghz - x17 - 1.30 v - stable - 45-46 C - prime 1 hour (i stopped)
3.6 ghz - x18 - 1.40 v - stable - 51-54 C - prime 9.5 hours (i stopped)
3.7 ghz - x18.5 - 1.45v - stable - 51-54 C -prime 7.5 hours (i stopped)
i have taken the voltage up to 1.55 but forgot to write down the results but they were all quick crashes as well.
can you guys help me get to 3.8 ? challenge!

[BeepBeep2]

Just a quick note, I had my CPU stable for 4 hours in prime one night, started up OCCT and it failed in 10 minutes. Just a note.


I think your CPU is at its limit, these Phenom II's are apparently touchy and sometimes more voltage isn't a good thing.

What thermal paste are you using?

The CPU sensor isn't completely accureate and your CPU's cores could actually be up around 65-70c. (Unsafe, and unstable)
They seem a little high to me.

[BoNg]

artic silver 5. and dont forgot about the difference between the the voltages i posted set in the bios arent exactly accurate look at the ss's for v after v-drop and v- droop.

[BeepBeep2]

3.8 ghz - x18.5 - fsb 206 - 1.4875 v - crash <9 min
3.8 ghz - x18.5 - fsb 206 - 1.50 v - crash 30-45 min
3.8 ghz - x18.5 - fsb 206 - 1.5125 v - 2 min test error
3.8 ghz - x18.5 - fsb 206 - 1.5250 v - 1 min test error

Thats what I was looking at.

Again, it seems your temps are abnormally high. It's probably crashing from heat, and I guarantee that you will need 1.55v+ for 3.8 unless you start getting temps about 5-8c lower.

[BoNg]

Thats what I was looking at.

Again, it seems your temps are abnormally high. It's probably crashing from heat, and I guarantee that you will need 1.55v+ for 3.8 unless you start getting temps about 5-8c lower.

no temps are listed there but these tests were all round 54-55 C as well should posted that in first post will fix, since my thermal paste has set a couple days those 60-61 temps dropped to the 54-55 mark so far.. also if you click the link i put in to my first post with this system youll notice that my temps with a stock cooler @stock speed and volts temps were higher than that

[gOtVoltage]

hello again after about a grueling week of patiently waiting i have decided it is time for my post, and so it starts. 3.6 ghz was original goal....my system is as follows:

phenom II x4 955 BE ?? GHz

thermalright true extreme push/pull

msi 790fx-gd70 bios ver 1.4

8 gigs - OCZ platinum amd edition 1600 MHz @1.65 v and 7-7-7-24-28-1T - OCZ3P1600LVAM4GK

msi 4890 oc edition 955/1150

2x western digital caviar black 640GB

kingwin mach 1 1220 w psu

coolermaster cm690 nvidia edition (LOL)first post with new system

log bios -cpu-vdd voltage ambient temps were normally 70-71 F
v
first post with new system
nb and ht link oc

3.2ghz - x16 - 1.30 v - 45-46 C - stable
3.4ghz - x17 - 1.30 v - 45-46 C - 1 hour prime(i stopped) i know 3 hours...cont.

3.6ghz - x18 - 1.325 v - n/a - crash prime 95 start
3.6ghz - x18 - 1.3375 v - n/a - crash 2 min
3.6ghz - x18 - 1.35 v - 47-48 C - crash 8 min
3.6ghz - x18 - 1.3625 v - 48-49 C - crash 7 min
3.6ghz - x18 - 1.3750 v - 49-50 C - crash 3 hours
3.6ghz - x18 - 1.3875 v - n/a - crash time n/a
3.6ghz - x18 - 1.40 v - 51-54 C - stable 9.5 hours prime(i stopped)


so hmmm 3.6 easy with decent temp - one fan, imo. waiting on fan holder for push pull method on cooler. new goal! 3.8 ghz. cont.

3.7 ghz - x18.5 - 1.40 v - 51-53 C - crash <2 min
3.7 ghz - x18.5 - 1.4125 v - 53-54 C - crash 5 min
3.7 ghz - x18.5 - 1.4250 v - 54-55 C - crash 8 min
3.7 ghz - x18.5 - 1.4375 v - 56-59 C - crash 1/1.5 hour < ambient here was 78 F, had to change rooms
3.7 ghz - x18.5 - 1.45 v - 52-61 C (54-55 avg) - stable 7.5 hours prime <ambient was 78 here for awhile a/c turned on

3.7 ghz - x18.5 - 201 fsb - 1.45 v - 50-54 C - 3 hours stable prime < normal ambient temps again


3.7 ghz easy, cont.

3.8 ghz - x19 - 1.45 v - n/a - crash 1 min
3.8 ghz - x19 - 1.4625 v - n/a - crash 1 min
3.8 ghz - x19 - 1.4750 v - 53-54 C - crash 10 min
3.8 ghz - x19 - 1.4875 v - 55-56 C - crash 9 min
3.8 ghz - x19 - 1.50 v - n/a - crash <20 min (forgot)
3.8 ghz - x19 - 1.5125 v - 59-61 C - crash <20 min
3.8 ghz - x19 - 1.525 v - n/a - crash <20 min
added push pull here
3.8 ghz - x19 - 1.5125 v - 60-61 C - crash time n/a
3.8 ghz - x19 - 1.525 v - 60-61 C - crash 1.75 hour (1hour 45 min)
no go i guess on multi ill lower and add some fsb
3.8 ghz - x18.5 - fsb 206 - 1.45 v - crash <1 min
3.8 ghz - x18.5 - fsb 206 - 1.4625 v - crash <5 min
3.8 ghz - x18.5 - fsb 206 - 1.4750 v - crash <3 min
3.8 ghz - x18.5 - fsb 206 - 1.4875 v - crash <9 min
3.8 ghz - x18.5 - fsb 206 - 1.50 v - crash 30-45 min
3.8 ghz - x18.5 - fsb 206 - 1.5125 v - 2 min test error
3.8 ghz - x18.5 - fsb 206 - 1.5250 v - 1 min test error

i have taken the voltage up to 1.55 but forgot to write down the results but they were all quick crashes as well.
can you guys help me get to 3.8 ? challenge!

Note: Your CPUvcore is scaling Fine for using Air and 8gigs.



What you didnt tell us,,
Was if the NBvid or NBvolts are still stock? Thats abiggy for you to go further in your New adventure.

Have you set the NBvid (CPUnb or AKA IMC)to 1.2v or even 1.3v to make the IMC and Ram stable @ 3.8ghz.


Nb@1.1volts(default) will only get you so far with 8gig of ram and four slots filled once you pass 3700mhz. This Also lets you run higher NB too.

The NBvid is what makes things happen past a given point. Vcore alone will not keep the IMC stable past this given point. You must then add NBvolts to compenstae even if you are only running 2000nb @3800mhz.

Sorry if i sound harsh,

Your on the right track with CPUvcore and the MHZ scaling , just tweak the NBvolts to run on 8gigs an youll most likely become stable @3800.

@T1 for all that ram and IMC,, Even though thier LVA's It might be a lot more stable on 2T with Cas 6 or 7 and raise the ram freqency a little more with 8gigs to gain the bandwidth back. I like 1T but ive found 2T to be the best at stability when going for that nice 24/7 OC.

gOtVolTage

Maybe some one else will chime in about the Ram,, im no expert in that department LOL

[BeepBeep2]

Then look closely here:
3.8 ghz - x18.5 - fsb 206 - 1.50 v - crash 30-45 min
3.8 ghz - x18.5 - fsb 206 - 1.5125 v - 2 min test error
3.8 ghz - x18.5 - fsb 206 - 1.5250 v - 1 min test error

1.50v is your voltage cut-off, at least for scaling on that multi.

[BoNg]

@ gotvoltage zomg i forgot ... my cpu-nb vdd = 1.125 all else unsaid is on auto and dram v is @ 1.72 which i fixed in original post

[gOtVoltage]

@ gotvoltage zomg i forgot ... my cpu-nb vdd = 1.125 all else unsaid is on auto and dram v is @ 1.72 which i fixed in original post

Try using 1.2v or 1.3v on the NBvid....I believe thats all your 3.8ghz problem is.

1.1 -1.25v i think is too low of a default when pushing over 3.6mhz.

720/550/955 all use the same NBvid voltage @1.1v default and actually clock somewhat identical except for temperatures up to 3900mhz.

Ive seen lots of CPU's stuck @ 3600-3700mhz because this Nbvid value was not raised to 1.2v or even higher in the beginning.

Ive tried this with Air and Water and the the only difference is that you will use slightly less Volts for water.

My 720 and 550 used the same volts for the same NBclocks on Air and water

volts used below = NBvid value

1.15v is good for up to 2400nb
1.2v is good for up to 2500nb
1.35v is good for up to 2700nb
1.4v is good for up to 2800nb
1.45v is good for up to 2850nb

Each Mobo will varry how much volts is reqiured for the CPU NBvid, but you get the Idea.

[BoNg]

that brings another question then should i run the nb at 2400 or 2600 ... ie 2400 at 1.2 + or 2600 at 1.2/3 and should i be worrying about temp too much for those v's on air. btw all those ss had a 2400 nb and 2200 ht link not stock 2000 if it was confused

[gOtVoltage]

that brings another question then should i run the nb at 2400 or 2600 ... ie 2400 at 1.2 + or 2600 at 1.2/3 and should i be worrying about temp too much for those v's on air. btw all those ss had a 2400 nb and 2200 ht link not stock 2000 if it was confused

Try it with 2400mhz 1.2v first..then 2600 @ 1.2v
If it fails @ 2600 but not 2400 then try 1.3v for 2600.

You never know ,you might get 2600 with just 1.2v.

Run your stability test and see the temps.

Your on Air and you can go up to 1.35v easy on the NBvid. Use the least amount to get your self stable and then call it the day.

Raising the NB freqency 200-400mhz will add more heat Vs bumping the NBvid .200v's to make 1.3volts

Start with 1.2v youll probally be stable or may need 1.25v . I just bump it up a little at a time ,,till stable .

Make shure the temps are with in the limits or under the Max Loaded CPU threshhold and youll be fine. Usually @ or under 59-60c for Air full load..

Im using only 1.35v on my NBvid ATM24/7 for 2500nb @ 4ghz @1220 DDR2 im running looser timing than you and 2T.

Ive tested my NB to 2900+ @ 1.45v but my NB i limited by my Ram @ 4ghz

[BoNg]

well since i already had 2400 stable at 1.125v im just gonna jump to NB 2600 @ 1.2v and ht link 2400. stressin now gonna be a long night again woo hoo

[Rezag1000]

well since i already had 2400 stable at 1.125v im just gonna jump to NB 2600 @ 1.2v and ht link 2400. stressin now gonna be a long night again woo hoo

don't raise the ht link above the stock 2000; this actually decreases performance

check this out: http://www.ocztechnologyforum.com/fo...ad.php?t=56715

[gOtVoltage]

well since i already had 2400 stable at 1.125v im just gonna jump to NB 2600 @ 1.2v and ht link 2400. stressin now gonna be a long night again woo hoo

You shouldnt need more than 1.3v for what you want to do .

As Beep Beep posted ,, it seemed like 1.50v was a voltage cut off for that multi(Cpuvcore) ..

I dont think it is. @1.50v it was just running cooler and not erroring as fast. As soon as the CPUwent above 1.51v it would error faster. Meaning Heat or unstable IMC.

3800mhz @ 1.35vcore with 1.3NBvid (On Water)
Air will use more Vcore and maybe 1.35vnbvid max depending on your NB speed.



1.25nbvid on Air mostlikely and you may be able to (lower the Vcore) some..You wont know till ya find what NBspeed and NBvid volts work for the given NBspeed.


A lot of times when the NBvid is Raised just a little over stock it will allow you to use less Vcore because the IMC is more stable for the Same OC.

[BoNg]

interesting is there any explanation to this im gonna stop my test i got going now with ss and move ht down to 2000 again and start again

[BoNg]

run @ stable settings 3.7 ghz @ 1.45v nb 2600 mhz @ 1.20 v, ht link 2400 mhz prime 1 hour 51-53 C

backing ht link down to 2000 mhz

[gOtVoltage]

If you look at my Screen my NB is Only @ 2200mhz...Its not always about the NB but how everything works together.

Your ram has the High MHZ and can be tweaked real good. thers a chart floating around that shows NB @ 2500mhz is ideal for 3900-4000mhz cpu's and 2200-2400mhz for around 3600-3800mhzcpu's..

Ill see if i can find the post and chart for ya..

Also i usually run my HT @ around 2200mhz or even lower too.


Update ..
I found the Chart,, i just cant remember the dang link i copied it from.. thanks to who made it becuase my 550 scale exactly like this.




Ok PHII's 720/550/955...

What ive found by testing my 720/550 that NBvid @ 1.125v is actually set low for what most Overclcokers are doing with thier PhenomII's. Its even set low for stock NBfreqency. I guess to help keep TDP down.

This why i allways suggest 1.2-1.3volts for NBvid. It allows you to run the NB in the range of a optimal Freqency that matches the CPUclock for clock past 3500mhz.

10xNb is Default(2000mhz) Based on the chart and what ive found this usually uses 1.225v or should be the defualt all the way up to running the NB@2600mhz . But its not and With Air you sometimes need 1.25 or 1.3volts to achieve 2600mhz NB.

I have Always bumped this NBvid Value so it covers the range of the Operating freqency Im shooting for example 2600mhz i use 1.3volts and sometimes the next value if you are running 2600mhz+

You see how this works volt wise. This scales about the Same for Air and water. The only difference is the Temperature on Air will make it a little harder to tweak and may use slightly more volts on the NBvid.

Even when PHenom B2 was New I always set 1.55vcore and 1.45nbvid just to see what they would do. I have not yet Burnt up a Phenom CPU from Volts Only PWM's lol.

Even this little bugger handled volts ,, It was Coldbugged.
NB and CPUvcore Are both 1.55v with water(testing IMC).


Thats why im gOtVolTage

[BoNg]

this one shows 2400 for 3.8 idk check it out http://www.ocforums.com/showthread.php?t=596023 bout half way through the page... doh .... a post behind.. i think thats your link though

[FlanK3r]

dont use NB about 2200MHz, its very bad performance...I tested it with diferent CPU clocks and diferent NB clocks in practise dekompression, video encode and benchmarks. HT is the best between 1800-2200 allover.

[gOtVoltage]

this one shows 2400 for 3.8 idk check it out http://www.ocforums.com/showthread.php?t=596023 bout half way through the page

Haha thats the thread i was looking for too,, good old DOLK.

He made a nice one..

In all aspects those values hold true with how my Chips clock on volts. Even though I am using water i see Air taking a little more juice but not much,, if the right Aircooler is used.

The main thing is how your NBvid is set relative to the NB and CPU clock freqencies you plan to run. (That is the Key.) Also Higher freqency ram such as DDR3 and lots of Gigs will push the IMC harder causing it to use MORE voltage in some cases.

Thers a lot of factors and you are already well on your way for clocking AMD .

I hope this was a good help for ya .

Yup im still learning after 15years of AMDtweaking.

[gOtVoltage]

dont use NB about 2200MHz, its very bad performance...I tested it with diferent CPU clocks and diferent NB clocks in practise dekompression, video encode and benchmarks. HT is the best between 1800-2200 allover.

Hey Flanker ,, You mean dont use BELOW or @2200mhz NB right?

In other words it should be higher than 2200mhz.

I wish AMD would just give us chips set @2400nb already or even 2600

[FlanK3r]

i tested it and for "lower" clock between 2800-3200MHz is the best northbridge 2000MHz or 2400MHz. For higher clocks 3600MHz is great NB clock 2600MHz and for highest clocks CPU about 3800-4000MHz is the best 2800-3000MHz for northbridge. 2200 is very strange bug, cpu at 3400MHz and 2200MHz for NB is whorse than cpu with 3200/2000MHz in performance.

[BoNg]

here is my progress so far i have nb @ 2400 and ht link @ 2000

cpu-nb vdd 1.20v here
3.8 ghz - x19 - 1.45v - crash start
3.8 ghz - x19 - 1.4875v - crash <10 min
3.8 ghz - x19 - 1.50v - crash <10 min
3.8 ghz - x19 - 1.5125v - 56-59 C - crash 30-45 min

cpu-nb vdd 1.25v here
3.8 ghz - x19 - 1.50v - crash <5 min
3.8 ghz - x19 - 1.5125v - 56-58 C - crash <10 min

cpu-nb vdd 1.30v here
3.8 ghz - x19 - 1.5125v - crash <15 min

this is where my batteries ran out last night, this morning i woke up and did a little re-reading as somtimes i dont catch on to things first shot all the time

You shouldnt need more than 1.3v for what you want to do .

As Beep Beep posted ,, it seemed like 1.50v was a voltage cut off for that multi(Cpuvcore) ..

I dont think it is. @1.50v it was just running cooler and not erroring as fast. As soon as the CPUwent above 1.51v it would error faster. Meaning Heat or unstable IMC.

3800mhz @ 1.35vcore with 1.3NBvid (On Water)
Air will use more Vcore and maybe 1.35vnbvid max depending on your NB speed.



1.25nbvid on Air mostlikely and you may be able to (lower the Vcore) some..You wont know till ya find what NBspeed and NBvid volts work for the given NBspeed.


A lot of times when the NBvid is Raised just a little over stock it will allow you to use less Vcore because the IMC is more stable for the Same OC.

putting this in practice atm to test... stressing 3.7 ghz @ 1.4375 v with cpu-nb v @ 1.30 v.

reminder: my stable settings for 3.7ghz is originally 1.45 v.

[BoNg]

1 hour crash

- one thing i would like to note. since increasing cpu-nb vdd i have noticed that i see less fluctuation in corespeed in cpuz and no more underspeed either i.e
x18.5multi x 201 fsb = 3718.5 /3718.4 <-- what im seeing now. with my original cpu-nb vdd i was seeing 3717.6-3718.8

[FlanK3r]

the problem with 3.8GHz up are:I mean, u have x64 OS (its here harder get 3.8GHz stable than in x86 OS) and second, u have full 4 RAM slots, with 2 slots are OC better.