Bulldozer and prime95 torture test

[gordy]

Is there any better way to test my overclock for stability? I'll get these errors after 15 minutes of successful testing. Temps are stable. It seems like the FPU is just malfunctioning in some specific way that's only verified at the end of a very long test. Increasing voltage and retest clears it up but things are really slow going at this rate. Any other BD owners having similar errors with p95?

[BeepBeep2]

Yep.
It's either almost always the exact 9E28D8E1 error, or a rounding error, or ILLEGAL SUMOUT.
Other than the rounding error, more vcore and/or less temperature solves it.

Prime95 crashes at about 3 hours on my 8150 @4.8 with 1.49v, most likely due to heat. CPU temps get up to a toasty 78c in Prime95 at that point.

I run 4.8 @ 1.46v 24/7, it never crashes in anything but Prime95, it will run LinX all day and any other program I throw at it.

[gordy]

I've got my temps well under control so I figured p95 was just getting to the malfunctioning fpu part 14 minutes in
I'm testing 4.6 @ 1.48V now and temps are stable at 64c.. I've just bumped the vcore from 1.47 and gave VDDA a nudge to 2.55. This is the first I've increased VDDA, did you have to increase yours to stabilize @ 4.8?

[BeepBeep2]

I'm not sure if it actually changes anything, though I'm running 2.6v VDDA as a "just in case".

I haven't tested very extensively.

[Gener_AL (UK)]

Thread makes me wish I had BD to tweak love to find what makes CPU's tick faster. Interesting if CPU VDDA does affect stability with odd multipliers as is for PLL voltages for CM's in system. CPU PLL Voltage is pretty much same thing.

[demonkevy666]

Thread makes me wish I had BD to tweak love to find what makes CPU's tick faster. Interesting if CPU VDDA does affect stability with odd multipliers as is for PLL voltages for CM's in system. CPU PLL Voltage is pretty much same thing.

I concur.

I mange to get stable with up My cpu VDDA up and getting a lower load voltage about 0.250 drop in voltage.
1.3875 beats 1.4125 for 4.5ghz.

[BeepBeep2]

I concur.

I mange to get stable with up My cpu VDDA up and getting a lower load voltage about 0.250 drop in voltage.
1.3875 beats 1.4125 for 4.5ghz.

What VDDA are you running?

[demonkevy666]

What VDDA are you running?

I've ran all up to 2.9 volts

[PatRaceTin]

wow i never touch VDDA !!!!!

interesting

[PatRaceTin]

wow i never touch VDDA !!!!!

interesting

[gordy]

increasing VDDA to 2.6V got me p95-stable at 5ghz @ 1.56V - passed all 3 torture tests, no warnings no errors after 45min or so, temps were cooling down from 74c. I was very happy and then my system froze under no load. I'm wondering now if my PSU just can't handle it.

[BeepBeep2]

increasing VDDA to 2.6V got me p95-stable at 5ghz @ 1.56V - passed all 3 torture tests, no warnings no errors after 45min or so, temps were cooling down from 74c. I was very happy and then my system froze under no load. I'm wondering now if my PSU just can't handle it.

That voltage is (more than) a little dangerous, don't you think...most aren't even running those voltages benchmarking, despite the fact that the chips scale to oblivion.

[Oese]

u mean the 1,56v?

haha depends. i would like to know vrm temps/cooling

[gordy]

I've got pretty good cooling
nb, mosfets, ram are all covered and water temps are staying low even when the cpu reached its hottest 74c @ 1.56V. 1.56 was still in the yellow according to the indicator in the Crosshair Formula V BIOS. I don't really have an idea of "safe" other than what others are doing and I've seen a few 1.6+ around here.

[BeepBeep2]

The guys doing 1.6v+ (gotvoltage pushing up to 1.7v) are trying to kill their chips...
1.9v is killing chips on LN2 at -192c just booting up the machine.

Most (here anyways) are pushing about 1.525v benchmarking on normal types of cooling, ~1.45v or less 24/7.

[Daveburt714]

Is there any better way to test my overclock for stability? I'll get these errors after 15 minutes of successful testing. Temps are stable. It seems like the FPU is just malfunctioning in some specific way that's only verified at the end of a very long test. Increasing voltage and retest clears it up but things are really slow going at this rate. Any other BD owners having similar errors with p95?

Hi gordy, when I first started tweaking the FX I saw the same thing your seeing, and like you, I assumed something must be wrong with P95!

LinX has always been a good stress test too and I could run it all day ~4.9 without errors.
Sooo, I said screw P95, I'll get it LinX stable and call it good... This worked really well for everyday use, including light gaming (I'm not a hardcore gamer).

Then I tried to encode a movie and my machine crashed!

Long story short, I don't think theres anything wrong with P95. It just pushes the right buttons to find true stability on FX chips.

I'm not saying P95 is the perfect test, just that if your looking for encoding stability get it blend stable first...

** EDIT **
BTW, Thanks for the info on the VDDA voltage folks! I've never messed with it either.
I will now...

[tbone8ty]

I remember when the phenom 9850 came out vdda was key in getting the chip to overclock and be stable. When phenom ll came out vdda sorta went away and wasn't needed.

could the bulldozer be fallowing the same path?

[mongoled]

gotVoltage doesnt kill chips he just tortures them, lol

[Zeus]

Guys, for your information, VDDA reduces Vdroop, that's pretty much it.

[BeepBeep2]

Guys, for your information, VDDA reduces Vdroop, that's pretty much it.

You tested this?

Doesn't seem that way to me.

[Zeus]

Nope, don't need to as my board has LLC option in bios. On older boards that did not have LLC it worked that way.

[I.nfraR.ed]

Yes, I was using VDDA on 790FX based boards for the exact same puprose.
You can still play with it to fine-tune the LLC, e.g. Crosshair IV has only 3 levels (0%, 50%, 100%). I was running VDDA = 2.74V with my Thuban when tested the 8 hours Prime back in the beginning.
Bios had only two possible options - on and off. Not sure what is the effect with newer boards/bioses. Haven't run stability tests from a very long time.

[BeepBeep2]

My multimeter is working now, so I think I'll test this myself.
Interesting that you guys say this, as 2.5v VID - 2.7v don't register in software as higher voltage and therefore does not present itself as reducing vdroop.

[gOtVoltage]

Guys, for your information, VDDA reduces Vdroop, that's pretty much it.

+1 ,,

on Almost every TOP ASUS mobo i have used VDDA .By far one of the best Bios options for curing massive vdroop.
VDDA which in essence is the base line volts of the mobo ,This handy feature always kept Vdroop in check on Asus platforms. It also allows a lower Vcore to be used, seeing the Vdroop does not drop as much @ boot or Under heavy loads(Priming).

...i wish i Still had VDDA option ( UD5 does not have it), so its even harder to control droop.
All i can say ,I have real good cooling and know its limits and a thing or two.

. lucky No, i just take my time seeing what i can get away with.

VDDA is very usefull ( 2.6v -2.8v) Most boards have the same range , small increments will net the sweet spot. Higher volts for Higher Clocks as you would expect. Some Boards of the Same maker ,ya might NOT Even need ta touch it.