FX Vishera series and CPU-PLL

Ok so I have seen mentioned in a few threads now on OCF that increasing the CPU-PLL voltage on some motherboards can help with overclocking the new AMD FX chips. I have been happy with my systems stability running F@H CPU + GPU for the entire moth without issues so I have not bothered to test it. I decided to poke around the net a bit and found THIS thread @ OCN where it was discussed in alot more depth than has been here. Based on the recent improvements and discussion in THIS thread here on OCF I decided to go ahead and give it a go.

Right now I am testing @ 2.6v as I didnt want to jump directly to nearly 2.7v in one shot without testing stability first. My temps all look fine; In fact my max CPU temp has gone down ~5C and the average loaded temp ~8C as a result and I am seeing the same stability on my system with .02v less after LLC currently.

I plan to look into this a bit further with a jump up to the generally suggested 2.675v and attempt to lower m vCore a bit more.

Edit. I have just changed to 2.675v CPU-PLL and dropped LLC 1 level. I am currently burning it in using P95 small FFTs to test for CPU stability before I get back to folding. Id rather loose the processing time upfront rather than let a WU crash part way though. Ill update in the morning.

Edit 2. After passing 15 mins of P95 I decided to reboot and drop the CPU vCore one notch and retest. Sofar its looking good.

This leads into the second page of thread by Johan45 and his quest to clock up an FX-8350 on a mid-range Asus 990 board. The second post on that page #2 has my suggestion about CPU_PLL and what he did with it and the results.
You can go back to page #1 to see just how BAD he was hung out to dry with a less than stellar FX-8350 or the lesser board.

I say little about CPU_PLL because it is just another setting that these newbs think that there are exact voltages to plug in and GO. It is not in anyway so. It is strictly trial and error and it works for some and then helps others none at all it seems. Don't know if that is a board problem or cpu problem when it does nothing.

The CPU_PLL voltage is said to be the voltage for the frequency generating circuit. Pretty hard to sit here at the keyboard and write how upping that voltage could lower temps any or allow VCore to the cpu to be lowered. Not that I am willing to type yet anyway.

Because I was going to rate my new FX-8350 as middling, good or excellent based on how much Vcore it took to each 5.0Ghz stable, I had the first boot, yes the very first boot of 8350 at 4.5Ghz and then JUMPED straight to 5.0Ghz which is something that should not be tried at home. Get into windows and fail P95 Blend quicker than my cat howls when I step on it's tail. This because I was nursing the Vcore up and juggling the CPU_PLL to try and reach 5.0Ghz at only 1.5Vcore. FAILED in that quest of 5.0Ghz what I am sayiing is stable for today after Lin-x which a weakling compared to P95 Blend and had to use 1.512 VCore on High LLC not the last two notches up on a CHV.

Was ticked off at the 1.512 Vcore but now I see it is pretty good. Whoopee. Anyway I have very good cooling and I only run @4.5Ghz with 1.38 Vcore and nearly no temps to speak of and I put the thing to work and have not played with it and CPU_PLL higher than the 2.56V I think I tried as Max. I may go back and give it a testing as you have done "ssjwizard" and see if it will lower the Vcore at 5.0Ghz and greater. I was not really worried one way or the other since I know enough to know a max overclock is very very seldom any good for 24/7 use.

RGone...ster.__________

I'm going to play around with the PLL and see what happens. So far I have been limited by temps. When I hit 4.5GHz I need a vcore bump to be stable and that causes my temps to skyrocket. If I can get it stable at 4.5GHz with a PLL voltage bump and no temp increase that would be just swell.

Ok well I have dropped the vCore one notch so now my chip is full load 0.045v lower after LLC fully engages. Im considering dropping it one more notch and kicking LLC up one to pull the vcore down another 0.15, but not until at least 24 hours F@H stability. Right now in total my CPU core temps are maxing out almost 15C lower as a result of the reduced volts. If it holds out stability Im just going to be giddy about the entire thing. Freeing up that vCore actually could allow me to push this chip to a higher clock for my 24/7 speed given the amount of processing it does everyday It will actualy serve me well.

ssjwizard said:

Ok well I have dropped the vCore one notch so now my chip is full load 0.045v lower after LLC fully engages. Im considering dropping it one more notch and kicking LLC up one to pull the vcore down another 0.15, but not until at least 24 hours F@H stability. Right now in total my CPU core temps are maxing out almost 15C lower as a result of the reduced volts. If it holds out stability Im just going to be giddy about the entire thing. Freeing up that vCore actually could allow me to push this chip to a higher clock for my 24/7 speed given the amount of processing it does everyday It will actualy serve me well.

Click to expand...

@ ssjwizard, that is a huge difference in temp for sure. Not a shabby reduction in Vcore either. Good idea that full testing for results before another drop, because we have n0 idea what increased CPU-PLL is really doing. Nor why it can allow a reduced Vcore. Does it for some...yes. Why? I have heard nothing but speculation.

I used to raise CPU_PLL all the time when pushing for a max clock for benching but only talked of it between buds of that time. Quit benching at HWBot over 5 years ago so I have raised CPU VDDA myself for years never mentioned it in public. That was why when "Anonaru" mentioned it I did not have to ask WTF like many. I just knew to try it. I do not think I raised it enough for the FX type processor I was using it early Athlons, P1's and P2's.

Be interesting to see if you get to drop anymore Vcore and if then having 'room' for more Vcore will allow for a higher cpu Mhz but stable still. Time and your testing will tell the story.

Good luck man. RGone...ster.

Ok so a small update, around 6am this morning the system did hang up once. I have bumped LLC up one setting and I am now ~0.035V below what It took for this CPU to be stable previously. F@H is up and running CPU+GPU again so we will see.

Honestly its a voltage I have adjusted in the past but I have no idea why this works how it does on these chips. At the moment Im just interested to know IF it works, and then if warranted Ill try and find the cause.

Maybe people will finally stop calling me a crock for suggesting that they use the PLL more

These are fascinating results-- I've not been able to emulate results quite this extreme on my own parts, but I'm sure I could with more tweaking and effort. Typically I get a .02~ reduction in overall needed vCore and a 5-9c drop, but man you're rocking it

Anonaru said:

Maybe people will finally stop calling me a crock for suggesting that they use the PLL more

These are fascinating results-- I've not been able to emulate results quite this extreme on my own parts, but I'm sure I could with more tweaking and effort. Typically I get a .02~ reduction and a 5-9c drop, but man you're rockign it

Click to expand...

a crock? Heck no never that. Other names? Maybe. Hehehe. You have too many toys to play with and find out the good shett. Hehehe.

Like I said that voltage was not foreign to me at all, but to really raise it, well that was a little foreign. And I don't think I have tested mine high enough yet. I was just going by my old ideas from years ago and did not kick it to 2.695 by any means. I will tonight likely though.

Not a crock, but a real thank you "Anonaru". RGone...ster.

CPU PLL Voltage Override (Overvoltage): What the Heck does it do?
So I asked that question to an Intel Overclocking Engineer his explanation was roughly: We went through the BIOS settings trying to find setting that if changed could help overclock our CPUs further. We came across this setting. Think of the CPU PLL voltage as a voltage that is provided to the CPU, but then “clipped” down to an approximate voltage. No matter what that input is whether 1.3v or 1.9v it is clipped (hypothetically let’s say 800mv after clipping (he didn’t say how much)) that way other devices can use the PLL voltage and clip to what they need. The CPU PLL Overvoltage allows for less clipping of that voltage. It can also reduce the lifespan of the CPU, but nothing noticeable

Click to expand...

I am also currently reading THIS article on how the CPU-PLL works from the mouth of an actual CPU engineer.

Since I am back at ground zero on my testing(ok 4 hours into it) I figured Id do a bit of reading. I plan to read both of the articles written at that site in effort to understand this a bit better, but It may not be useful here. I would note that in searching for info on this It seems that MOST heavy SB overclockers are very familiar with this setting and have no issues with jacking it up to help stability.

ssjwizard said:

I am also currently reading THIS article on how the CPU-PLL works from the mouth of an actual CPU engineer.

Since I am back at ground zero on my testing(ok 4 hours into it) I figured Id do a bit of reading. I plan to read both of the articles written at that site in effort to understand this a bit better, but It may not be useful here. I would note that in searching for info on this It seems that MOST heavy SB overclockers are very familiar with this setting and have no issues with jacking it up to help stability.

Click to expand...

Aye, then comes the freak chips-- My own 3930k OC'd higher with less vCore when I LOWERED my PLL. . . It needs to be increased if I enable HT though. Weird *** chip

I might jack up the PLL on my shop 6300 here in a few minutes, I'm curious as to how high it can go, how safe it is to increase, and the sort. As far as I can tell, up to 2.695 causes the chip no pain, but I've yet to try anything higher or had the time to really test the effects of PLL.

Thanks you two guys. RGone

Just as a quick and dirty test I lowered the vcore on my 8350 to 1.45 @ 4.7GHz with the VDDA increased to 2.695 as per the threads you linked to, 20 mins of P95 blend and it was stable with no workers dropped. Previously had to run over 1.47v to achieve that.

Going to do a little testing myself tomorrow to see how low I can get the vcore @ 4.7GHz with various VDDA settings.

Many thanks for the heads up.

So I have been running with these settings all day and I am convinced that it works. I cant explain it yet, but this trick does something on these CPUs atleast with my motherboard. It has been Folding on CPU and GPU all day ontop of my other regular routine(video transcoding, various programming tasks, photoshop and premiere, browsing, playing games) without a crash. Im seeing a vCore of 1.46-1.47 at max loads @ 4.5Ghz right now. Previously to keep this particular type of stability I was using 1.51v. Right now I am transcoding 60 episodes of my sons favorite cartoon to 720P h.264 M4V files, while folding of course. Im headed off to bed but Im expecting that it will be up and running when I come out in the morning.

ssjwizard said:

So I have been running with these settings all day and I am convinced that it works. I cant explain it yet, but this trick does something on these CPUs atleast with my motherboard. It has been Folding on CPU and GPU all day ontop of my other regular routine(video transcoding, various programming tasks, photoshop and premiere, browsing, playing games) without a crash. Im seeing a vCore of 1.46-1.47 at max loads @ 4.5Ghz right now. Previously to keep this particular type of stability I was using 1.51v. Right now I am transcoding 60 episodes of my sons favorite cartoon to 720P h.264 M4V files, while folding of course. Im headed off to bed but Im expecting that it will be up and running when I come out in the morning.

Click to expand...

My 6300 is on the bench table in the other room folding @ 5.0 GHz, 1.490 (Max 1.495) vCore, and 2.745 CPU-PLL, been running for 2 hours now. I think I might take this 6300 home with me if it survives this test-- I've been torturing the crud out of it ever since I got it (Came into the shop damaged, so free for testing purposes ), and it has been nothing but a gem

Woo

Good show guys.

Crap I wish I had time to get back to testing my 8350. Maybe later today. Who knows? But I surely will use more than 2.59 CPU_PLL just to see what happens.

RGone said:

Good show guys.

Crap I wish I had time to get back to testing my 8350. Maybe later today. Who knows? But I surely will use more than 2.59 CPU_PLL just to see what happens.

Click to expand...

Awesome The 6300 is still holding strong, temps are great, no errors, everything's doing well. Switched it to P95 and encoding runs. I'm not sure what a PLL of 2.745 is going to do to it, but so far the chip's chugging away happy as a clam

Anonaru said:

Awesome The 6300 is still holding strong, temps are great, no errors, everything's doing well. Switched it to P95 and encoding runs. I'm not sure what a PLL of 2.745 is going to do to it, but so far the chip's chugging away happy as a clam

Click to expand...

Anything above 2.72 reboots my system with the 8350 but at 2.693 I can def drop vcore two steps, I save around 4 degrees with those two drops although struggling to push any frequencies higher than 4.8, its still early days for this rig though

MediaJunky said:

Anything above 2.72 reboots my system with the 8350 but at 2.693 I can def drop vcore two steps, I save around 4 degrees with those two drops although struggling to push any frequencies higher than 4.8, its still early days for this rig though

Click to expand...

Might only be working for me because its running at 5GHz haha. I am afraid going over 2.7 might be dangerous in the long run, but my poor lil 6300 is still crunching away, like a true soldier

Edit: If you're working with an actual rig I'd definitly not suggest going over 2.7-- Nobody understands what might happen or become of chips when over volted here.. This 6300 is a test chip on a frankenrig at my shop haha.

Yep so the system is crash free this morning. It finished up all 60 eps no issues and is still folding like a champ. I might try and play with it some more but for now Im happy to have my chip @ max 45C while CPU and GPU are loaded.

ssjwizard said:

Yep so the system is crash free this morning. It finished up all 60 eps no issues and is still folding like a champ. I might try and play with it some more but for now Im happy to have my chip @ max 45C while CPU and GPU are loaded.

Click to expand...

Haha, it feels like cheating with PD chips once PLL starts being understood-- We very well may be front-runners in discovering which voltages for what frequencies, the limits, and the benefits for using this setting proper