Gigabyte P35-DS3L fsb overvolt question

[ferencster]

Hi, with my Gigabyte P35-DS3L and q9300 i can only reach 400 fsb if i add +0.3 v to the fsb. My question is that that kind of voltage is safe for the chipset? Will it degrade acceleratedly? I am planing to keep this config for 3-4 years.

 

[Lupiron]

Uhh.. that really depends on your CPU, and boy, that poor 45 MN processor of yours! With that FSB term Voltage up that high, it may not last long!

You need North Bridge voltage for maintaining a higher FSB, and a lil PLL if it gets higher than 400.

You may wanna list the rest of your gear here.

With VTT (FSB termination voltage) up that high, your poor processor will receive 1.4 volts of power to the chip anyways, each and every time it switches from a low logic state to a high logical state.

A special device within the chip "pulls" up the voltage to VTT each time it wants to enter a high logic state.

Try NB to +.2 (gmch?) And PLL to 1.55.

Like I said, you may wanna mention your gear, and what else you have tried!

--Lupi

Basically what you did is widen the margin of error the chip can accept before making a wrong calculation. You didn't address the actual problem.

 

[th_username]

100% safe

 

[Lupiron]

Smart fellow there! ^

Thats right, shove voltage through the processor to support your FSB speed!

To hell with that North Bridge!

All he did is widen his threshold to accept errors, with Voltage!

Not help stabilize the FSB speeds with the correct things!

NEED the rest of his gear!

--Lupi

 

[ferencster]

So My board is Gigabyte P35-DS3L and cpu q9300, ram 2x2 gb kingmax 800 mhz, ati hd4850, Seagate 500 gb hdd, Samsung dvd rw, Thermaltake 470w power supply. I have tried 400 fsb with fsb +0.2 v, gmch +0.2 v, ddr +0.2 v comp freezes after 2-3 hours of stressing. I tried adding +0.1 v to pci-e but that didn`t help. I tried raising my cpu voltage up to 1.25, no stability with that voltage either. I am rock stable with 380 fsb 1:1 ratio ddr2 with +0.1 v, fsb +0.1 v, and every other voltage at normal/stock(not auto) including cpu. But if i try 385 fsb even with raising fsb, gmch, cpu voltages i can`t get stable. I have a cheap motherboard, so it doesent have option for raising vtt, pll. (i have the folowing options: ddr overvolt max +0.6, fsb overvolt max +0.3, gmch overvolt max +0.3, pcie overvolt max +0.3, cpu overvolt max 2.0 v) I have an Arctic freezer pro (pretty good cooler), and for the northbridge Cooler Chipset Thermalright HR-05 SLI IFX 2 Heatpipes, Cooperbased, 0db, 87X62.5X105 mm.

 

[Lupiron]

Can you run Core Temp or real temp and list the VID of your processor?

--Lupi

 

[ferencster]

My vid is VID = 1.2250

 

[Lupiron]

You will more than likely have to raise your CPU voltage up more, since your starting VID is 1.2250

You need test VDrop and droop... First, enable load line calibration if your board has it!

Then boot into windows at stock speeds, using 1.2250 as your VCore, manually entered, and make sure speed step is off, as well as eist!

Then using CPUz, after booting into windows, look to see what it lists your Core voltage as, list that here.

then run Prime 95 small FFTs torture test on all the cores, and list the core voltage CPUz lists after running it for a couple minutes, the lowest you see it flicker too, even if it ends up higher.

After that, we can go for a new CPU voltage, and lower that FSB voltage. (FSB voltage is VTT, The front side bus termination voltage.)

--Lupi

 

[Lupiron]

Let me put it this way. If your FSB voltage is up .3 volts from its 1.1000 then that means your processor is receiving 1.4000 volts to it no matter your VCore.

So we wanna at the least balance them out.

--Lupi

 

[houstonserenity]

You said that your Computer freezes after 2-3 hrs.
What temps are you getting?
Your memory how are you strapinging it? you mite need to just OC the cpu first and get that stable be for you try your GPU and Memory.

 

[ferencster]

I have done everything you asked: my idle voltage is 1.216 an load voltage is 1.184v

 

[ferencster]

My temps in real temp are 60 c under load for all 4 cores, and my northbridge is max 43 c

 

[Lupiron]

Yup, can you check and see if you have Load Line Calibration in your Bios? if you do, enable it, and it should take care of some of that voltage drop.

if not, you prolly just needed more voltage. You said you placed it at 1.2500 in the Bios? It's VID is two notches under that, so it prolly needs a bit more.

If you have Load Line calibration, you can just test with it enabled, but if you don't, you have to take into account that of that 1.2500 you set in the Bios, all your CPU is getting while under full load, The time it needs the most, is 1.184.

See how that could be too low for your speed you were running at?

If you have the Load Line, you can run the Vdrop and droop tests again to see the new results, and they will be much better than last time.

We use that lowered during testing voltage, Loaded, to mean after YOU have corrected for your mother boards drop and droop, it needs that amount at whatever speed to run small ffts on all the cores for 8 hours.

If you dont have it in the Bios, try 1.3250 in the Bios, for a lil over 1.25000 while loaded and then just up the FSB till it fails after a while of testing.

So you make the speed equal the voltage.

--Lupi

 

[ferencster]

no i got the 1.184 v with 1.225 v set manually in bios not 1.2500 v

 

[ferencster]

i have no loadline calibration in my bios

 

[Lupiron]

Try lowering the VTT to 1.2000 or so, and then raising the VCore +.125 (so if its a 1.2250, it should be 1.3500)

gMCH to +.2

Then the 400 FSB.

--Lupi

 

[Lupiron]

Then try 1.3250 for about 1.28 loaded.

 

[ferencster]

No luck, after windows starts imadiate freezing

 

[ferencster]

i can`t set vtt just to normal or +0.1, +0.2, +0.3 v and that`s it, I don`t know the normal value for my fsb speed

 

[ferencster]

I don`t know the normal value for my fsb voltage

 

[Lupiron]

Try just +.1 I believe the norm for 45 NM is 1.1000, so that should be 1.2000. Weird, huh?

--Lupi

 

[ferencster]

I leaved cpu overvolt on stock, fsb +.01, ddr +.1, gmch +.2, stable after 2 h of occt (still running) 390 fsb, not 400 but almoust. Thank you all for your help specially for Lupiron. (i have learned today new things with this overclocking)

 

[zipz0p]

I've got the GA-EP35-DS3R, and have similar settings. 400FSB seems to be about the max I can get stable.

I have MCH +.3V, FSB + .2V, DDR +.3V (which is the rated voltage for my units: 2x2GB G.Skill 1066 5-5-5-15 @ 2.1V).
I have the Q9450 and have to volt it up to 1.26 or so in bios for stability.
Coretemp listed VID = 1.2375V.

Under full load, CPU-Z lists core voltage at 1.200V, and at idle it gets as high as 1.232V with the multiplier dropped fro 8 to 6 - Speedstep is on.

 

[zipz0p]

Even if you have Loadline Calibration, you wouldn't want to turn it on - read this Anandtech article: http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3184&p=5.

I would think that the VTT voltage would only put extra voltage on the memory interface portion of the chip - I don't know if this portion of the chip is more or less susceptible to electron migration, but I would think less, as the circuitry should be much less complicated.

Edit: I found that I could reduce my FSB to only +.1V and keep stability. Thanks for the info Lupi - it got me thinking about the problem at a more fundamental level.

 

[Lupiron]

Yes, that article describes what could happen if you enable load line calibration, and it disables all of vdrop and droop. Most don't, and also, as over clockers, especially higher, you don't take the "norm" ideal for what is good with that chip of yours. You want more speed for free!

Also, most mother boards design their load line calibration in such a manner where it is resistant to power serges.

I use it all the time, but I also manually monitor my current voltage to the cores, and it records the max highs and min lows.

The only board I have ever seen a supposed power spike on was an XFX 780i Ref board. HWMonitor would record a Max high of 4.008 volts to the CPU. Dunno if that was a sensor error or not, and no damage was done that I could tell.

But it was also the first and only time that I have seen a recorded power spike. And that board doesnt even have Load Line Calibration, hehehe!

And to FerenCster, if you have any BSODs or anything like that, just up the VCore a notch, and re test!

Glad it works out, almost 400 FSB! better than what it was, especially with that VTT so high. 45 NM chips natural VTT is 1.1000, so 1.3000 would be the max you'd wanna use without having any other choice.

IE, if it doesnt get a high speed any other way, you are left with no choice, if you want a higher over clock, huh? But if there is any other way around it, use that first.

--Lupi