Is 1.6v volts ok if cooling fine for AMD-8120

[indigoblades]

[RESOLVED]

I am new here .... not sure how to set my signature up but below is my system. I used the great Bulldozer guide by Stevenb to OC.

http://www.overclockers.com/forums/showthread.php?t=688663

My AMD-8120 bulldozer needs lots more volts than most (1.6v), i needed 1.45v to boot into windows maybe one step more @ 4.5ghz .... i dont think mine could 4.5ghz Prime95 on air or a small water system. For my mobo and bulldozer i get a consistand 0.075 droop DURING PRIME so 1.6v set sees 1.5 to 1.525v. i can post a pic of my run. Any tips on how to lower core volts appreciated. Or if high volts dont matter if temps fine and not big voltage overshoots tell me. i am more interested in how to keep the volts down day to day than what a see on validation run. i haven't spent time uppin my fsb coz my volts are about at the limit i think. if others are runnin higher volts please tell me. i have seen others running much higher temps than i am.

I also have DFI-Intel Q6600 system and a Z77-DS3R/i5 system (hackintosh) but neither is overclocked. i used to overclock the old AMD X2 and opterions with a DFI mobo. ... probably about 20 to 30% depending on how good the cpu. on them i would lower the memory multiplier and the hyper thread bus multiplier and find what the cpu could do first using the fsb. then bring the hyper back up and try to find a good memory match. i am new to bull dozer and never over clocked an INtel. In the old days i never had enough cooling head room to add many volts ... for my bulldozer i used ur guide just the first steps only changing multiplier and Vcore... once i decide what clock i want and what volts are good ill do the next step.

I have got a 6hr PRIME95 stable 4.5Ghz OC on my bulldozer from using Stevenb's guide ... i just did the multipler not the fsb coz i feel i am at voltage limit (1.600 vCore set). i basically turned off turbo, saved .. and found out what multipler is stable at what volts thur his methodolgy. just upped volts each multiplier step. then did final 6hr run when i passed a 30 min PRIME run before i played with fsb, northbridge and other multipliers ... on my system PRIME would have a core, usually #7 fail the test, or i would see the thread die in resource monitor & Prime act funny. my pc never crashed. .... my volts did go up a lot to pass PRIME95 going from 4.2 to 4.5. it will run most stuff at 1.475v but prime needs lots more. my feelings are MY 8120 needs lots volts and it droops alot during prime ... i did some runs with power savings off but didnt see a huge diff, i left most power saving on in the end because day to day i would just soon it throttles down .. i didnt see huge difference with em on, maybe a voltage step or 2.... my volts required are high (set 1.6v) but i have excess cooling, one big heater core and a thermalchill triple all getting cool outside air, two of my five fans are 5.5 inches ... its also not air conditioned so my house stays in the 80's except at night .. my highest core temp is 45C spikes very occasionally (if volts spike to 1.625) & typically 38C if volts are bouncing 1.58 to 1.6, with prime it stays about 39C since prime95 has 0.075v droop (1.500-1.525)... but if i play games, theres less load on cpu than prime, but then there's no vdroop and it will stay 1.58 to 1.600, typically bouncing 1.58 to 1.600 ... . ... stays about 37c tho when vcore is 1.6 .... HWmoniter says my highest volts are 1.632v with 50C my high temp. those are spikes. I am at the point i should ask some questions now. My volts are so high i am not sure if should drop to 4.3 GHz & optimize fsb (i could drop vcore 2 steps to 1.55 at 4.3Ghz) . i got coolin headroom at 1.6 set, but i know i will use it up if i go higher. i feel im at max volts i wanna put in. .... normally when i over clocking after a long final stability run i would up my volts 1 or 2 steps for stability but my volts are so high now .... i could use LLC set to gigabytes 'extreme' (its on regular now) but my feelings are ill turn the processor down 2 volt steps and LLC will put in 2 more steps and the cpu see the same volts. i would only use LLC if got rid of droop for PRIME95 runs only but day to day it was lower ... in the old days overclocking 2 core cpu's, i feel getting 2 cores maxed is very believable in real world ... maxing 8 threads 100% i think is very artificial so i feel if PRIME passes for hours that's stable & no need for safety margins ... most software just cant multitude that many threads 100% ... as for stability my experience with windows is it crashes randomly with a stock system so a crash every 2 weeks isnt a good measure FOR ME of if my overclock is stable.... i am forced to stress test for stability. Anyway unless i persistently crash its hard to tell if that's cpu stability or just windows.

IF One has lots of cooling which i do, how much would u worry about volts that high. Again i never had lots of cooling before so i never over volt'd much. also am i correct to worry about the volts it sees in HWmonitor not the volts set. i have noticed my temps now are more proportional to the voltage changes not the threads used or cpu load. i have been watching it closely in will playing skyrim with other stuff running in the back ground. My skyrim is GPU bound mostly but i can go to places that are CPU bound too. its got lots of mods and ENB so it loads up 4 cores pretty good. Also i assume TMP2 on Hardware Monitor is cpu socket temps (about core +12C on my motherboard ) i trust the core temp more, so my standard is 65C is the limit on core temp and the other isn't critical.

again, it would boot fine into windows at 4.4, 4.5ghz at 1.45 volts. one more volt step & it ran skyrim, & simpler stress test and like super pi. I didn't have OCCT at the time i over clocked but i followed stevenb's guide up to the point where u traded multiply for fsb speed ... my prime95 seems to stay consistent 1.500 to 1.525 with 1.6 set ... if i lower it prime will fail a core.


Any tips how a typical my volts are is appreciated. also i read AMD bins differently now... my AMD-8120 was bought in dec 12, Rev OR-B2, so maybe they don't overclock as well now. i am not unhappy, i got the cpu for about $125 and the mobo was free in a microcenter special. i can get 4.0 ghz with moderate volts which is fine, and that's what my guess what others could get with moderate $50 to $100 after market air or water cooling set up.

Also, since i am new, I am pretty good with Water Cooling, if anyone has questions ... i have a masters in Mech Engr with specialization in Thermal-Energy which helps a lot. Most to the data in reviews is insufficient to CALC what a Water cooling parts will do in a system(before u buy it). But my view is 80% is just how many fans u have if other components decent.

Thank You for the help.

GA970A-UD3, AMD-8120 4.5ghz (1.6v set on Vcore), HD-7850 1050Ghz, TX850 watts, 16Gb mushkin ram set at 1333, water on CPU & GPU, 2 rads, 5 fans, MPC355 pump

 

[RGone]

You were thorough in your description and that is good although many will not read more than a few sentences. Hehehe.

Going to just be as straight as I can be so we get the high spots out of the way.

1. A Sabertooth FX990 Rev2 (nonGEN3) or the CHV-z or non-z would handle that cpu better without the big Vdroop. That would have lowered temps.

2. A couple of us in here have done some pretty complete testing and the earlier BullDozer processor actually does not need much more than 4.2/4.3Ghz to do about as much real work as it "will" do. So you might consider dropping the speed a little and allow for less Vcore with all that board Vdroop.

3. There is a good article > let me find the link > http://www.overclockers.com/forums/showthread.php?t=384756 > that talks pretty straight about over-volting. I did not argue with what was written since when I followed a couple of the links, I came across information that I had seen previously in something else I was investigating. Suffice it to say, I would not run my stuff over 1.525Vcore day in and day out which is 24/7 use. YMMV. I know of a very few that have gotten by with more than 1.525Vcore for a good while. But they are not asking if it is safe. They own the stuff and made up their own minds.

I will put it this way. I am not personally interested in 4.5Ghz if it takes so much voltage. If I could get much closer to 1.525Vcore with the generally powerful 4.3Ghz, then my cpu would just be at 4.3Ghz for the long term.

4. What have you set the CPU_LLC to? That wonky CPU_LLC is what caused many of the big boys that are in here, to swap to better boards. The CPU_LLC was so all over the place that they too had to set more Vcore than was really necessary. That is and has been an issue with Rev 1.0, 1.1 boards and caused some to leave the board. Plus the VRM circuit was just too underpowered to keep up with the 8 core processors and the result was the large Vdroop.

5. That is about it. Run it up there with the Vcore you must or slow it some. Not the first one of us knows if your cpu is prone to begin electron migration or deterioration. You did not pay much for it, so you might run it and if fails get another better cpu.
RGone...

EDIT:
Crap I always forget a point that might be of interest. Giga came with the painful Rev 3.0 boards because it was said the Mosfets were dying on the earlier Rev mobos. The Rev 3.0 was a painful experience for many since the only bios that had APM = Disable in the bios menu to stop the cpu from throttling when the VRMs were loaded was not available except in a beta bios "finally". Many had to dump the Rev 3.0 board because it took forever to unearth the hidden beta bios for those that wanted to push that cheap UD3 board. So it seems Giga came now with the Rev 4.0 970 UD3 board, but I have heard nothing about how it acts. Most of the really cheap boards have issues with FX 8 cores. Period. Not enough muscle in the VRM circuit.

If yours is drooping that much, you may also blow a mosfet. Then it could take other stuff out with it. That is a thought.
END EDIT.

 

[indigoblades]

Thanks Rgone ... think i used to read ur stuff, because i had DFI's in the old days .... or it could be different Rgone.... theres not not much performance diff between 4.3 and 4.5 so maybe think ill drop it back ... also i may go in OCCT and watch the voltage with different stess test running and play with some settings. i got a lot reading ahead and may drop back in the mean time. I have bio F7, i think i needed that to to run bulldozer. My Mobo is rev 1.2.

i have LLC on regular, my understanding is it tries to smooth spikes underload ... my mobo has an extreme and auto option. in theory extreme might work better, just from my reading some like LLC some dont. ill just turn it on an watch it. Think i will lower multipler then watch for 30min with extreme LLC and see if Prime and OCCT keep voltage closer to what i set. if theres more voltage spread ill leave it off. if it helps ill use it. i know i shouldnt in theary need to set it so high but ill just watch it. ... ill watch it if its all over the place i wont use it. there is no off for LLC but regular may be off. If its a cheap Mobo its the mobo i am stuck with lol. i have got some bad $200 plus mobos with lots of options i dont need so trying to go with $100 mobos ... the CPU's i am getting are $150.

Just One question .. when you or others say 1.525 do u mean thats about the average volts the processors sees or set at 1.525.

2nd question is Vdroop mainly of function of the Power supply and Mobo not the cpu.

thanks again.

 

[RJARRRPCGP]

1.6v set sees 1.5 to 1.525v.

A little OT, the Vdroop reminds me of my MSI 845E Max. Setting to "1.6 V" or close to that in the BIOS was actually 1.50 V to 1.52 V, which was the stock Vcore for my Northy 2.4 B0.
(That was a freebie I was testing out in 2007, when a Barton was my newest rig, LOL. )
(I was already behind even back in 2007, LOL.)

Definitely was more droopy than my Asus A7N8X-X with my Barton 3000+ AQZFA!

I would enable CPU load line calibration, to avoid major Vcore swings like that.

 

[indigoblades]

Thanks guys .... i just dropped back from 4.5ghz to 4.4ghz and so got to drop set voltage 3 steps lower ... now its set at 1.525 and prime stays about 1.424 to 1.488 ... i guess some droop is safety feature.. now i will try More LLC see what happens, hopefully less droop and somewhat stable voltage. if thats good and less ill try a step less too next ( i assume LLC will reduce the droop).

 

[RGone]

Same RGone you used to read. DFI quit selling DIY motherboards. No job.

Just One question .. when you or others say 1.525 do u mean thats about the average volts the processors sees or set at 1.525.

REply >> Most good boards for FX processors have a pretty good LLC that you can adjust so that Vcore setting in bios is mighty darn close to supplied Vcore seen and logged in HWMonitor. So if I set 1.525Vcore on my CHV board it sees that 1.525Vcore when loaded and drops a little when not under any load. That is how I expect my CPU_LLC to actually work and does on my CHV.



2nd question is Vdroop mainly of function of the Power supply and Mobo not the cpu.

REply >> That is two questions. Hehehe. 1. It is less a power supply to my mind. 2. It is not the cpu but the voltage supply circuit built onto the motherboard that regulates the voltage from the power supply to the cpu or generally the VRM circuit.

What you are experiencing with that 970 UD3 Rev 1.2 board is not new. It is why many left the board for better ones with heavier VRM circuits.

You are correct to try the various CPU_LLC settings. They might work for you where they did not work for most of the other power users. But it is worth a try for sure.

Some find that using some FSB "and" the multiplier can often get a little more cpu mhz with slightly less Vcore. But it is a "big" test and try sort of thing. Not do A and B and C is the result for sure.

I wrote up a short how to on using Offset Vcore on Asus boards for some that might have benefitted. Then about 4 months later I visited the idea myself. With my CHV, I can run 4.8Ghz at 1.48Vcore and then when unloaded and idling the Vcore falls to < 1.00 Vcore most of the time. You might try to use the Offset Vcore setting if that board has it.

There are many methods you might try to work around a cheaper board or you could get an Asrock 990FX EXT9 board since they will do the job. Of course the Asus Sabertooth 990FX Rev 2.0 and the CHV-z will do the job as well. I have heard of fair success with the Giga 990FX UD5, but they have now moved into Rev 3.0 also and nothing much has been heard of how it operates. For the money the Asrock 990FX EXT9 will do the job.
RGone...

 

[indigoblades]

Yes after some thought 4.3 vs 4.5 is only 4% which i should never notice. 4.5 sounds good but id rather have stable OC six months from now so i will end up keeping it close to 1.5v for final 24/7.

i tested last night and slight gains with one less step volts and LLC extreme .. i saw less volts in general for a given freq. But on my windows machine i ended up with random windows error at least 3 of 4 times for every hour tested. small stuff but i think the voltage jumps too much ( i have power saving functions on) it does have volts all over the place but does have less droop under full load. with one step less volts i saw less min and max volts and ran prime stable 30min .... again i feel its too unstable in windows to tweak on the edge. For me i will keep LLC at regular not extreme. (Giga has no OFF option in bio's just AUTO, REGULAR AND EXTREME) but to me it looks like regular is off.

also made volts vs Mhz plot and realized i was putting in 3 or 4 volt steps for .1 mhz ... so ill prolly end up at 4.3 or 4.4Mhz at 1.5 .... both 4.3 and 4.4 need 1.5 volts to prime. 4.3ghz @ 1.475 not quite prime 30mins. for me everytime i pass the 1st and 2nd group of prime test i have passed an over night prime. whether i stay 4.3 or 4.4 depends on what see when i play with the multiplier and if get any significant gains moving HTT up and multiplier down........ same goes for playin with NB, HT and mem mult.

next guess i will make 1 step multiplier jumps at 4300 and see if any significant gains and how volts do. then with that worked out ill play with memory speed, HT and northbridge ... me i doubt ill notice any diff. but we will see .. my memory is mushkin blackline 1600, i think it may be cheap stuff. i know when i got i had no intent of overclocking mem.

thx and i will ask for advice next MOBO buy.... again this one was free & the processor- mobo a steal i thought, microcenter had Xmas specials last fall ... i got a Z77-ds3h, and an i5 (ivy league), i got that on special too but it was about $75 more, but prolly wont OC anytime soon coz its a hackintoh project & a small power supply. but its probally the better processor/mobo combo tho ..... i finally ended getting this on water last weekend with parts layin around so i decided to over clock.

thx again.

 

[RGone]

The sad thing after I read your signature is that you have pretty darn good cooling. But that wonky motherboard with too much Vdroop. All you can do is get the best you can within reason or get a good board to go with the pretty good cooling.
RGone...

 

[indigoblades]

well i am pretty happy ..... i got margins with all this cooling and its whisper quiet ... i have a good base OC and my core & gpu temps in the 30's durin skyrim ... with stock colors every computer fan was kicked in before, nothin overclocked and my core temps in 50's and gpu close to 60's.

the motherboard is kinda wonky .. i never like motherboard much ... i like my first dfi a lan party for amd X2/opterions, that one i loved and never a had a problem, it was on water but had a cheap double radator that worked good until i put volts in then the fans got to loud for 24/7 ... still quiter than air tho .... my 2nd dfi for an intel Q66 would randomly crash and need a bios clear. i had a few asus too.. i used to get top of the line mobo's but i dont use SLI ... newer stuff i assumed doesnt need the HTT/FSB as high ... guess good power regulation is good tho. But all that comes at a price.

well i was happy but i got no voltage margins to over clock HTT/Memory i found last night i ran these last night:

memtest (mem v=1.5+0.1) (HT and NB effect strongly and set more for stab this run) all 8x multiplier (mem @ 14-14-14-40 will go high with HTT alone) Mem VERY LOOSE at 14-14-14-40 its rated at 1600 and i could not get it boot at spd unless i upped volts to 1.55 so its not the greatest memory. (this is my first memory overclock other than raising HTT with very good memory on old X2/Opterions.
-----------------------------------------------------
1680= 8x210*21=4410 PRIME STABLE @ 1.550V (at least 8 housrs)
1760=8x220*20=4400 almost prime stable (20 mins)not sure v=1.550 or 1.575
1848=8x231x19=4389 close to prime stable (15 mins) @ 1.575
1904=8*234x18.5=4403 NOT prime stable (5 mins) @ 1.575
1960=8*245x18=4410 NOT prime stable (2mins) @ 1.575
2000=8x250x17.5=44375 NOT prime stable (5 secs) @ 1.575

initially my cheap memory/mobo combo wouldnt boot at 9.33 memory multiplier so i wanted see what the memory would do with HTT alone and the 8x multiplier .... if i could get the NB multiplier up it scaled well but again i am out voltage margins so think ill drop to around 4200 Mhz and try memory 1850 IF i can get the 9.33 multiplier to work. Then ill see if memory gains offset lost pure cpu speed.

 

[indigoblades]

I settled on 4.4 @ 1.525v. i couldnt overclock the HTT any and pass prime95 till i upped the NB vid to just 1.3375 and NB one step. Now i am at 220 HTT, with a 20x multiplier for 4.4 ghz. HTT is at 12x, NB is at 11x. i wanted to match those but thats as high as my NB will go. Memory is at 1.6v on the 8x multiplier, it wont boot on the 9x. i got memory to 9-9-9-30 but to be prime stable i had to loosen it to 10-10-10-32. i seem to be prime stable. I am happy, volts are high but i think ok since my temps so low .... light load at 27C fully loaded 37C so i am not worried.

 

[RGone]

It is what it is. Have some fun now with computering. Thanks for bringing us up to date. You might go back into your very first post of this thread and add (RESOLVED) to the front of the title. Thanks
RGone...

 

[indigoblades]

thanks for the help everyone.

 

[caddi daddi]

I don't see an issue with high voltage if you have the temps in line and are prepared for what ever happens.
my 8120 takes 1.476 @4.6 before llc kicks it up and heaven knows I have been well over 1.6.
the things to do are to get the cpu/nb voltage up to 1.4 and the cpu/nb and ht speed up to 2500-2600 even if you have to give a little cpu speed up to get it.

make sure you have the ram voltage up also.

having to support the cpu with very high vcore might be helped some with higher cpu/nb voltage and dram voltage.

 

[indigoblades]

i have ram at 1.6v, i got better results when i boosted the cpu/nb voltage, especially over clocking the HTT. i might get the HTT higher if i boosted NB CPU volts but my bios steps the volts so i dont want to make the next step. Not sure if my board bios is working right because the steps seem to go up in increments too on my cpu/nb voltage.

starting NB cpu volts step resulting volts
------------ ------ --------------
1.175 none 1.175
1.200 0.25 1.200
1.200 0.50 1.250
1.250 0.75 1.325 <- i am here
1.325 0.100 1.425

any way the steppins in voltage above confused me and why on earth does it increment the step ... at higher volts if anything i would want a lower step to fine tune it. i just wish it stepped 0.025 like the cpu.

my NB is 2420 (11x220), and thats the fastest i can get it. for me if NB multiplier too high i cant boot. my HT is 2640 (12x220). i wanted those to match but i cant get my NB any higher. i tried bench marking stuff but superPI & Geekbench had so much varience between test with slight changes cpu ghz i could see no difference. I have benchmarked with MaxxMem for the HTT and memory speed and latency effects. A Skyrim test in a cpu limited section (if u play, if u go the dragon keep but stay outside, turn around and walk till u look out over the steps at white run. Your frame rate should drop by more that 50%, probally more if ur not runnin an ENB. i play with lots of mods and a full ENB so i am usually GPU limited) Anyway fps will show good scaling there with Cpu speed and memory too. even changes small changes showed up.

if its really hot indoors say 85, my core cpu temps with 1.6v cpu volts are max 43C and typically 38C. there so low coz i have 5 fans cooling one cpu and one 100 watt graphics card. HWmonitor TMP2 reports something 12C higher that follows core cpu temp perfectly so i assume its the cpu socket temp of the motherboard. For 24/7 i decided to be safe and bring it down to 1.525 cpu volts. I suspect Lots of people running around 1.450 to 1.475 on air or a small 1-2 fan water cooler 24/7 with way higher temps than i have, my guess mid to high 50's on Coretemp after reading some post, and if there fine, my feelings are i would be fine at 1.6v. But i decided to be safe, and the performance diff was small. also i might need to later boost to 1.625, i just passed a few Prime95's for 8 hours & seems generally stable, but my experience is when i think i am stable, a week later it goes crazy and i gotta drop back or raise volts a hair.

my be someday if new processors cheap ill take it up higher, and put the cpu at more risk. i assume the ram could go to 1.65v, but again i am being safe.

my mobo is has 8-2 analog VRM with what look to me to be decent heat sinks which dont feel super hot... i dont have any temp on HW monitor other than TMP2 above 40C. amd processors are rated to 65C on core so i assume mobo cpu sockets are rated around 75C (they may track closer at high temps but still the socket temp cant be 65C, its further from the cpu and its cooling... id never run 24/7 above 50C core anyway & prefer to keep lower, just my feelings but i think temps are more critical than volts.

Ty for the help.

 

[RGone]

but my experience is when i think i am stable, a week later it goes crazy and i gotta drop back or raise volts a hair. = If your testing has been done with Prime 95 in Blend mode and then it does okay for a while and then needs more voltage it generally could mean some component on the motherboard is drifting out of spec or the voltage was borderline at previous setting. In your situation where it takes more Vcore than just about any FX-8120 that we have seen posted about in this forum section; it could be the cpu itself is just acting very very strangely.

Does not seem like there is much you can do. Results like you are seeing is one of the reasons I could never find myself comfortable suggesting that board to any of my buddies. It seemed the 970 UD3 was at first stout enough for the 8 core cpus but the 1.1 and 1.2 boards were a real pain to get adjusted for a decent overclock. I heard one user in this forum section say that a Sabertooth 990FX compared to the 970 UD3; that the 970 UD3 would take 50 more times of adjusting than the Sabertooth and this user sold and used the early 1.1 and 1.2 Gigabyte motherboards. So you may just have a case of two parts that together are a pain to get adjusted exactly.

It is what it is and all you can do is use it or replace some parts. At least you know you have tried to dial it all in and that is about all you can do without changing out some parts. Luck man.
RGone...