NB Voltage and overclocking to 3.8ghz

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Z

zealengine

Distinguished
Dec 7, 2009
77
0
18,630
I have the following setup...

Asus M4A79XTD MB
AMD PH II X4 955 3.2ghz BE
Sunbeam freezer with AS5
Saphire Radeon 4890
4gb Gskill PC 1600 DDR3
2x 1 TB Seagate HD (Raid 1)
Antec 300 case with extra fans (all air exiting the case is cool)
500w Thermaltake psu

After much research, I have been able to get a stable overclock of 200 x17 (2200 HT, 2200 NB) at 1.4v cpu, 1.175 cpu/nb (stable prime 95 for hours, max load 52c). However, if I push much beyond that, I prime95 usually wont last longer than a few minutes.

My question is - I have read that I should be running between 1.175-1.25v for my cpu/nb, but the problem is my bios has THREE "NB voltage settings", they are...

CPU/NB Volt
NB Volt
NB 1.8 Volt

Assuming I am aiming for a 3.8ghz overclock, can anyone tell me what THOSE THREE voltages should be? I also imagine I might need to bump up my cpu voltage higher than 1.4v, but I have heard people say they get to 3.8 stable at 1.4 by only bumping their NB voltage...

*BTW do my ram settings matter when overclocking? I am currently 1600hz at 1.5v 9-9-9-24 timings.

Thanks in advance!
 
Sort by date Sort by votes
Dude your RAM cannot be set to AUTO, you need to manually set the TIMINGS/DRAM FREQUENCY and DRAM voltage as specified by the manufacturer. I thought you had done this already.....

Change the frequency from AUTO (666) to 800Mhz

Then change the timings to whatever your RAM states.

And last but not least change the DRAM voltage to whatever your RAM states as well...

Can you specify what model RAM you purchased?
 
Upvote 0 Downvote
This is my ram (all my stuff is listed on the first post)

http://www.newegg.com/Product/Product.aspx?Item=N82E16820231193&nm_mc=AFC-C8Junction&cm_mmc=AFC-C8Junction-_-Memory+(Desktop+Memory)-_-G.SKILL-_-20231193

it appears to have the correct default timings (9-9-9-24), its just underclocked to 666 instead of 800 (according to amd overdrive it is also OVERVOLTED to 1.6v, but I cant imagine that would hurt anything). I can try bumping it up to 800... you dont think thats whats giving me trouble when I go past x18 do you?

*btw prime95 still running strong at the above settings.
 
Upvote 0 Downvote
If you do not set your RAM settings manually YES you will eventually hit a wall because all the values have to be correct or otherwise you will have no stability.

Set you timings to 9-9-9-24

The ram can stay at 666Mhz

Voltage at 1.5v, no need to overvolt the ram for now...
 
Upvote 0 Downvote
But 666hz 9-9-9-24 IS what its at now (at least according to amd overdrive). I suppose I can try manually inputting those values... So what should I try setting everything else too (cpu volt, multiplier, ect...)?
 
Upvote 0 Downvote
As I said 7 posts up, 1.45 cpu volts isnt even enough to run at x18.5 (remember I can do .5 increments, so until we get 18.5 working, dont worry about 19).

I just tested those settings again...

cpu bus freq - 200
pcie freq 100

cpu ratio - 18.5
cpu/nb freq - 2400hz
cpu voltage - 1.45v
cpu/nb voltage - 1.3v
cpu vdda volt - AUTO

HT linkspeed - 2200hz
HT linkwidth - AUTO
HT Volt - AUTO

DRAM freq - 1333hz
DRAM Volt - 1.5v
(under timings there were dozens of settings, however I set the traditional CL, tRCD, tRP and tRAS to 9-9-9-24)

NB Volt - AUTO (amd overdrive shows 1.1v)
NB 1.8v - AUTO
SB Volt - AUTO (amd overdrive shows 1.35v, which is max... odd)

Booted into windows (windows 7 pro 64 btw), ran prime95 for about 2min, got to around 52c (has run stable at 55c on x18), then this is what happens...

Graphics on the screen get all garbled for a second (kinda washed out rainbow colors). Then a blue screen comes up, and says "A problem has been detected, windows shut down to prevent damage.... blah blah blah....", and the last line it says "Dumping system memory to disk ....". Then my machine auto resets.
 
Upvote 0 Downvote
This way we will be able to find the reason why it is acting up by raising the ratio alone
Click to expand...

Can you elaborate on that? What is wrong with using multipliers? How will raising just the "CPU Bus Freq" (remember there is no fsb) how will that tell us what the problem is?
 
Upvote 0 Downvote
And a minor update - I ran the windows memory diagnostic at the above settings (200x18.5), and it reported no errors. I also turned off auto reset on error, so I could take a closer look at that bluescreen. Sure enough, after 3 minutes of prime it did the same thing, and I recorded this form the bluescreen...

"A problem has been detected and windows has been shutdown to prevent damage to your computer" ... blah blah blah...

"Technical information
*** STOP : 0x00000124 (0x000000000, 0xf... 0x.... 0x...)

Collecting data for crash dump..."

At first I thought the 124 meant nothing, but I am googling it now, maybe it means something...
 
Upvote 0 Downvote
In other words that means that you have failed the overclock = BSOD

Sometimes CPU's that are paired with certain boards do not like overclocking by raising the multi alone. I have a board that does not like anything past x17 but it loves when I raise the BUS...That is why I suggested to forget about the multi for now and concentrate on raising the bus.
 
Upvote 0 Downvote
Ok I will try raising the bus.

On a side note, while researching the 124 error I came across this...

http://groups.google.com/group/microsoft.public.windows.vista.hardware_devices/browse_thread/thread/458382424dfea394/f90096518a4eab5a%23f90096518a4eab5a

Which suggests it might be related to "32 bit HD access" feature in bios. I went into my bios, and the problem is, since I am running raid1, non of my hd's show up (sata ports 1 and 2 show "Not Detected", even though they are clearly plugged in and working), however even though they are not detected, I went into the features for those ports, and it shows this "32 bit HD access" IS enabled. Who knows for sure though...

While in bios I also double checked things like you mentioned (cool n quiet, ect...) they all looked good, but I remembered turning on this one feature called "Anti Surgy Support". The description of it was vague, but it sounded like a good thing to have turned on. You dont think thats whats causing this do you?

Anyway I will try playing with the fsb now...

*oh another thing I noticed while in bios - bios was reporting my vcore at only 1.431v, even though I have it set to 1.45v. So maybe amd overdrive was WRONG when it said my cpu voltage was 1.45. Maybe we need to set it to 1.47 or .148, but I will try your fsb idea first.
 
Upvote 0 Downvote
Ok 220x16, cpu 1.45v, cpu/nb 1.3v, nb & ht both at default x10, all other settings are unchanged (see above settings).

Been running prime stable for about an hour. Not that big of a deal though, since were only at 3.5ghz (I could get this high with just the multiplier).

I am going to try upping to 225x16.5, since that will put us close to the equivalent 200x18.5 (thats where the trouble stats)..

Cross your fingers.
 
Upvote 0 Downvote
No luck 🙁

225x16.5 crashed within three seconds of prime 95.

I also just tried disabling that "anti surgy support" option in bios, and attempting to run the old 200x18.5 again, still only lasted about 3min.

Only idea I have left - when running 200x18 amd overdrive seems to report pretty accurate 1.45v cpu *Well I take that back, even now 200x18 it still reports vcore at only 1.4 to 1.41.., This could just be amd overdrive sucking though... however even bios reported my vcore at only 1.431v... So maybe I just need to try something like 1.46-1.47v cpu? As long as my temps dont go above 55c there shouldnt be any danger in upping the cpu voltage right?

Unless you have any other ideas...
 
Upvote 0 Downvote
Hey I got an idea... So i have always wondered what cpu vdda does.. googled it and found this...

http://www.overclock.net/amd-motherboards/441042-what-world-cpu-vdda-voltage.html

Currently I have this vdda option set to auto (0 I assume) in bios. Sure enough, amd overdrive reports 1.45 cpu v when idle, BUT as soon as I turn on prime 95, my cpu volt drops to 1.4v! Hrmmm think I should try turning this 'cpu vdda' thing on?
 
Upvote 0 Downvote
According to this one guy...

basically what happens to the voltage when a load is applied on the CPU is the voltage actually drops. so if you set your voltage to say 1.45v in BIOS and the CPU VDDA is set to it's lowest setting, under idle conditions, the voltage should be around 1.45v. however, under load, the voltage will actually drop below your setting of 1.45v. how much depends on the make and model of your motherboard.

setting the CPU VDDA to it's highest setting will cause the CPU voltage to be (alot) more than what you set it to in BIOS. for instance, if i set my CPU VDDA voltage to 2.80v (Auto, 2.5, 2.6, 2.7, and 2.8v are the settings on my ATI chipset), under idle conditions, my CPU voltage will be about 0.04-0.05v more than what i set it to in the BIOS. while under load, it can go as high as 0.10v over my BIOS settings.

it does this to make sure that your CPU is always getting a good amount of power so that the power doesn't dip down and cause BSOD's. usually setting it to AUTO is fine as it will use the lowest settings while under idle and use the highest settings while under load. this makes it so that your not wasting too much power and/or over-volting your chip at idle while making sure under load, you're receiving enough power.
__________________
Click to expand...

However I have found it doesnt seem to be working as advertised. I just tried setting it to 2.8 (max) and I STILL dip from 1.45 to 1.4 when I launch prime 95. I am pretty confident this is my problem. Perhaps my only option is to raise the voltage to something crazy like 1.48 (so it will balance back 1.45v under load), and let cool'n'quiet reduce voltage during idle times..

what think?
 
Upvote 0 Downvote
No, what I am seeing is called vdroop. Rock solid stable at 1.45v idle, then drops to about 1.4v under load. I am sure this is why 1.4v lasts about 3 seconds before a crash, 1.45v lasts about 3 minutes, and 1.46+ can sometimes go for an hour.

Unless you have a better idea, my problem is vcore. The question is, is there any other way to stabilize my vcore so it doesnt drop so much under load (other than just upping the base voltage to something really high like 1.47-1.48). Upping that cpu vdda to max (2.8) didnt do jack.
 
Upvote 0 Downvote


No, on my Asus I have raised the voltage to 1.58 and there was still some headroom for more...Try 1.48/1.49 and see if you can get more stability....

The voltage thresold depends on how hot it gets. For example the 965 hit 7Ghz on Ln2 and the CPU voltage was around 1.85v

WoWzers.png



 
Upvote 0 Downvote
Well last night I bumped the cpu voltage to 1.4875v (it dipped to 1.45-1.44v under load), and at first I crashed withing 3 minutes. I then realized I still had my nb and ht at 2400/2200, so I set them both back to 2000/2000, and then I was able to run prime95 for over an hour (for the first time ever at 200x18.5x). Got kinda hot though (would idle around 40-42, and load around 55c). However if that is the MAX load under prime95, I am not too worried (since 'normal' apps will NEVER put that much strain on the cpu). Plus turning cool'nquiet back on should help when idling.

So I guess this is proof that what my system needed all along was more vcore, so that leads me to believe the only thing that will help when trying for 200x19 is even MORE vcore... Is it safe to bump to 1.5v JUST to get 3.8ghz?

*one interesting thing I noticed in bios, it only reports my vcore at 1.46v, so maybe it isnt really running as high as we think?
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom