FX Vishera series and CPU-PLL

[RGone]

A little after 6 PM CST and NOW I finally get to fire up the CHV and test CPU_PLL. Whooopee. RGone...ster.

 

[tungureanu]

hmmmm. i'm gonna see if this works on BD chips aswell. i allready had to give it 2 bumps when i used only multi to OC.

PS: i found this yes it's intel, but i think that the explanation stands

 

[juane414]

PS: i found this yes it's intel, but i think that the explanation stands

That's way over my head. All I really care to know about CPU PLL voltage is whether or not it improves my overclock (temps, stability, etc.) and whether or not it has a negative affect on my CPUs lifespan. I haven't been able to find any forum post, article, or website that clearly answers these questions in plain English. I want a good overclock but I don't want to be a guinea pig and blow my CPU trying something new and unproven. That being said, I might monkey with the CPU PLL just to see what happens but I don't think I'll run it 24/7 with a huge CPU PLL voltage increase.

 

[RGone]

I haven't been able to find any forum post, article, or website that clearly answers these questions in plain English. = There is none and likely will not be. And for sure the use of up to 2.695 VDDA is only just now been discussed for AMD and there is not nearly enough time passed nor enough tried such to determine if there will be negative effects. No way. No how for such information today.

 

[RGone]

Well crapOla. I do not think that though I am tweaking harder with CPU_PLL that it is going to be of help to me at all. Well not much compared to some others. When I first pushed the 8350 to 5.0Ghz, I was using VDDA of 2.55 and barely got 5.0Ghz stable at just over 1.5Vcore. So I had settings for 5.0Ghz and went back in bios a little while ago and upped VDDA to 2.64 and dropped VCore to less than 1.5V. and started stress testing. Ran 11 of 25 loops of Lin-X and failed with the lesser Vcore so I upped VDDA to 2.65 and it failed faster. Upped VDDA to 2.675 and failed faster still when stressing. So am now with VDDA back down to 2.6V and stressing is running longer again.

So CPU_PLL aka VDDA is by nO means going to be add "A" voltage amount and plug and play and results seem likely to be very varied.

 

[ssjwizard]

I wager it has to do with the type of VRM. You are pushing an 8 phase full digital VRM. Most gigabyte boards(the ones who have mostly tried this) are using analong VRM still. That is a complete guess though, I really have no idea why this is doing what it does.

I fully agree that it is not a magic bullet, as I noted in both my results post and the initial post, this has been reported to work with some motherboards. Its to bad you were not able to get the same kind of results we have been.

In my case the voltage that I needed to be completely stable VS what I need to pass P95, or any standard stress test are pretty different. I can get past P95 at 0.05v less than I can get my machine 24/7 stable on. So, for me I was able to take a needlessly high voltage and drop it back down into a more typical range for this chip to do these clockspeeds at. I have considered upping my OC since I have a bit of wiggle room on vCore and temps again, but have elected to instead put my fans back on auto speed(PWM driven via CPU fan header to smart controller) and my rig is now pretty close to silent while folding.

 

[MediaJunky]

For me its two clicks of vcore down with VDDA 2.6+, two clicks on the Sabertooth = 0.0125 which in temperature wise means at max 2 degrees better off.

I had previously thought there was a 4 degree difference but then my girlfriend very wisely pointed out that the heating wasn't on where as in my previous tests it was, hence the additional 2 degrees.

Has anyone tried dropping the VDDA / CPU-PLL lower than 2.5v?

I won't leave it at 2.6+, with there being no information on the long term effects its not worth the 2 degrees and I have no intention of running at 4.8GHz full time anyway.

 

[RGone]

You did what?

You might be correct in the difference being digital VRM vs analog VRM, "ssjwizard".

I also hear clearly "MediaJunky" when he says 0.0125 is not enough lowering of voltage to justify using a raised voltage that the full results of so so doing have not been determined. Crap we are not even sure what the raised VDDA is is actually changing on AMD that seems to bring good results in some cases.

I mean my gawd the how to overclock a processor is nearly anywhere on the internet that you want to look and this AMD forum section is FULL of how to overclock and yet we get post after post wanting to know how to overclock somebodies processor. Can you imagine trying to help some newb that already has more information available than one can begin to imagine and try to coach that type of user thru the use of VDDA?

I linked "Johan45's" use of VDDA to help get a 'presentable' overclock of an FX-8350 on an ASUS M5A99FX PRO R2.0 motherboard. His was a very special situation it seems and he was far and away from the type that posts "how do I overclock my X". He spent hours tweaking VDDA on his own without posting anything and came back after he had results. What he got was a low enough Vcore to be pleasing to him but ONLY about 50 more Cpu Mhz increase but the increase had him dead in the middle of 4500Mhz cpu speed and he was happy since he "seems" to have one of the newer Vishera's that just don't want to overclock like one expects after the BullDozer cpus.

I mean really the only reason I re-opened the efforts for testing VDDA myself some more is because the 'idea' is being thrown around with little real results being spoken about here in the AMD Cpu section except for mostly 'ssjwizard', 'Anonaru' and myself I guess to begin with. All three of us and "Johan45" have been testing raised VDDA on both generations of FX cpus and now also "MediaJunky".

"Johan45" has the ASUS M5A99FX PRO R2.0 which is a 6+2 DIGI+ VRM motherboard so in reflection "ssjwizard" I may not agree with VDDA not working so well on my CHV since it is using Digi+ VRM. "ssjwizard" I only realized what I just wrote after attemping to bring together in one place what we 'think' we know. Funny way to speak I know. Not sure I know anything.

I only want to put as much real information into one location as we can so that those that actually TEST can do so. I begin to surmise we might be able to generally say:
1. NO clear answer as to why raised VDDA has helped anyone lower the Vcore on some FX type processors has been determined.
2. Long-term effects of raising VDDA has most certainly NOT been determined.
3. Increased CPU_PLL aka VDDA seems currently more likely to work on some Gigabyte analog VRM circuit motherboards.
4. The Maximum VDDA has not been determined anymore than #3 above.
5. Increasing VDDA to try and lower temps and Vcore IS NOW and will always require user performed trial and error testing to determine if there are any results to be obtained from raising VDDA.

Do those five points seem fairly accurate to date?

I will say now that before I got up this morning and saw MediaJunky ask if anyone had tried lowering VDDA, that it had crossed my mind to try such. I previously tested my FX-8350 with NO more than 2.52-ish VDDA and when I started to test last night I jumped right up to 2.64 VDDA and got poor results. Upped VDDA again and even poorer results. I am now in the 2.55V VDDA voltage range and my results seem more promising. I think if I find any good result it will be with less than 2.6V VDDA and my temps and Vcore will only be lessened by an amount similar to what "MediaJunky" is reporting in his post just above. In other words not nearly the seemed good results that some Analog VRM circuit Gigabyte motherboards appear to have as a result of raised VDDA.

I am of the mind right now that if one wants to try to raise VDDA to see what happens, there is no need to ask about it but rather to do it and test it out for themselves and only then after complete testing...report what the result is. We have no idea overall who maybe helped and certainly not how much the help will be nor how much VDDA to add to get a positive result. Newbs likely need not attempt this at home.

Maybe we are current to date now. Off to test some more. RGone...ster.

 

[Johan45]

Hi guys,
Just like to put in my 2 cents,
Thanks for the kind words Rgone.
My board or processor has a ceiling of 2.6 on the VDDA anything above with all other settings the same my workers would drop off in Prime 95.
You were right Rgone I did spend hours messing with different combinations to attain my goal which I owe in part to your suggestion of the CPU_VDDA adjustments. I never would have achieved 4515 MHz @ 1.392 V_Core. Without tweaking the VDDA.
If anyone was curious the FX 8350 I have is "0" stepping Rev "OR-C0" manufactured in week 44.
@ MediaJunky I will PM you my settings, I see you have the Sabretooth board and can't see any issues with you using them. But you do understand I take no responsibility for what may happen.

 

[RGone]

Thanks man for chiming in. I am finding as well my CHV with Digi-Vrm hates more than 2.6VDDA. And looking for a sweet spot is getting to be a "process" for sure. Dropping off of workers is a good wording. That is precisely how my setup is working. Right now at about 2.57, it is just running and then sneaks up on me and is frozen when I go over and look at it. So I am likely close to the good VDDA but it is taking sometime for each run to complete or freeze and me catch it and go adjust some more. Sneaky shett. Hehehe.
RGone.

Hi guys,
Just like to put in my 2 cents,
Thanks for the kind words Rgone.
My board or processor has a ceiling of 2.6 on the VDDA anything above with all other settings the same my workers would drop off in Prime 95.
You were right Rgone I did spend hours messing with different combinations to attain my goal which I owe in part to your suggestion of the CPU_VDDA adjustments. I never would have achieved 4515 MHz @ 1.392 V_Core. Without tweaking the VDDA.
If anyone was curious the FX 8350 I have is "0" stepping Rev "OR-C0" manufactured in week 44.
@ MediaJunky I will PM you my settings, I see you have the Sabretooth board and can't see any issues with you using them. But you do understand I take no responsibility for what may happen.

 

[ssjwizard]

Im glad were finding some time to actually put the microscope over this question. I still cant explain it, but I'm quite happy with my result sofar. At least now we have some idea when this is a good recommendation for people, and it looks like it falls into the "Oh you have one of XX design mobos" category. I wish we could get someone with a UD7 to test this out since they have essentially the same VRM design as the newer Asus boards. If it fails on that we know it works with older VRM designs, but for stuff thats newer and has exceptional voltage regulation already its a moot point.

Until we can get more people to test and report here we cant really figure out what the heck is going on under the hood with this trick.

 

[MediaJunky]

If anyone was curious the FX 8350 I have is "0" stepping Rev "OR-C0" manufactured in week 44.
@ MediaJunky I will PM you my settings, I see you have the Sabretooth board and can't see any issues with you using them. But you do understand I take no responsibility for what may happen.

Thanks for the settings Johan, I want to try a few things with VDDA including less voltage than the stock 2.5, I saw you have a similar setup and thought your config maybe a good base for me to try when looking for a different approach.

I also have practically the same CPU, although not sure on the production date but i will confirm later.

 

[Johan45]

No probs MediaJunky.
Rgone were you changing anything with your HT or NB freq. when it froze. That was what I noticed with mine, manipulating that a bit would fix it.

For me dropped workers was power and freezing was my frequencies not playing together.

 

[RGone]

No I maintained the CPU_NB and HT Frequency the same. Only changes were to the VDDA.

I run a rather high CPU_NB that is matched with the HT Freq and that is both at 2570Mhz. But it was staying at that speed thru-out.

What seems to be the case is that I did not use enough VDDA at first and then jumped to more VDDA than a full Digi-Vrm board wants to have and am now having to back up and test in small increments of VDDA since VDDA above 2.6 does not work for me and freezes or fails Lin-X. Was failing at first and then began to only freeze as I guess I am getting closer to a working VDDA. But like "MediaJunky" my Vcore is not dramatically dropped like "ssjwizards" Vcore drop on analog Gigabyte board. So mainly I guess VDDA is not going to benefit me much on CHV.

No probs MediaJunky.
Rgone were you changing anything with your HT or NB freq. when it froze. That was what I noticed with mine, manipulating that a bit would fix it.

For me dropped workers was power and freezing was my frequencies not playing together.

 

[Daveburt]

Just for reference, I tweaked the CPU VDDA on my C5F for a few hours today as well, and saw no real improvement in clocking or the ability to lower Vcore...

I agree with RGone, it seems like boards with Digi+ VRM's don't benefit from this trick.

 

[RGone]

Well I am off and on over 2 hours of VDDA tweaking on my CHV, like "Daveburt", and I did capture image to show what my results are and my thinking.

Link >> http://www.overclockers.com/forums/showpost.php?p=7377991&postcount=46

 

[Anonaru]

Well I am off and on over 2 hours of VDDA tweaking on my CHV, like "Daveburt", and I did capture image to show what my results are and my thinking.

Link >> http://www.overclockers.com/forums/showpost.php?p=7377991&postcount=46

Aye, this is what I'm thinking-- It is a digital vs analog VRM thing.

I have another 6300 running on a M5A99X EVO in my shop, a decently awesome board with a very nice digital VRM. Was inspired to begin tests on this board by this thread and the success another user on the forum, "Mandrake4565" has had with OC'ing it.

I've come to find that I haven't had to take CPU-PLL anywhere on this board for stability, but at 5.0GHz I WAS able to lower vCore one tick with PLL set at 2.565. Further changes did nothing for stability, temps, voltage, and it actually wouldn't POST if I set the PLL to anything higher than 2.625.

I left the 6300 on my test UD3 board running @5.0GHz with that 2.745 PLL value running all weekend on P95 blend. The poor ol chip is still running fine. I went ahead and gave it a break for right now, but it looks like there's no immediate danger from running a value even as high as 2.745.. But I don't think it is needed. Cool beans!

 

[Bobnova]

A higher PLL voltage would give you a longer swing from bottom to top of its clock signal output. That'd be useful for powering through noise on the line, including noise from the CPU power bits. Whether it's the digital vs analog or simply the parts used I can't say as the digital power bits generally get better filters as well, but that power should be cleaner for a few different reasons. It should also generally less EMI in nearby things, like the PLL line.

PLL on Intel is a very, very different beast. It varies by chip whether it wants more or less.

Are y'all sure the VDDA only runs the PLL?

 

[giulio]

hi guys, i'm italian and sorry for my bad english. Last night I read this forum because I wasn't able to overclock my fx8320 to 4.5 Ghz. After reading your posts i'm now able to overclock to 4.5ghz. I want to say "I love you" and many thanks. My specs are: mobo: ga990xa-ud3 rev.3. fx 8320, cooled by antec kuhler 650 on water (very noisy but a very good cooler) 8gb ram corsair 1600 and geforce gigabyte g1 970 gaming.

PLL: 2.55
vcore (hwmonitor) max 1.392
llc medium
fsb stock
multiplier 22.5
temp are 58 (max 59) for cpu, after 20 min of occt, and 59 (max 61) for northbridge.

 

[RGone]

I wasn't able to overclock my fx8320 to 4.5 Ghz. After reading your posts i'm now able to overclock to 4.5ghz. .

Happy computering to you sir.

RGone...