Look here: My top 10 "quick facts" about AM2/AM2+/AM3/AM3+ platforms.

[RGone]

Did you put information in there regarding killing the IMC at speeds over 1600Mhz (in that it DOESNT do that?). Sorry if I missed it.

Great post!

Yeah, I think #9 under the "10 quick facts about AM2+/AM3 (Sempron 1xx/Athlon II/Phenom II X2/X3/X4)" category.

 

[BeepBeep2]

[/COLOR]

Some valuable info here but I disagree with you at some points, Sam, as you see in the lavender-colored comments.

I have measured temps with a probe

 

[BeepBeep2]

Here's a question that I would like to get some kind of definitive answer on, and maybe one of you guys know as I'm not too sure myself. If you pump a high voltage (2.7v+) into ddr2 on a AM3 chip will it likely damage that chips IMC ? And what is the rated speed Imc for ddr2 on AM3 (ie it's 1333 ddr3) is it the same for ddr2.
Thanks
Mark

I haven't seen this in awhile, but when the AMD IMC first came into play, the high Ram voltage could and did damage some IMC's. Remember the IMC was made up of part of the NB that used to reside on the mobo. Way back then we said the IMC was damaged since the ram worked fine in another motherboard but the ram gave issues in the AMD board with IMC that had had high Ram voltage given to it. I think the term used was back EMF causing the damage.

Scratching my head trying to remember all the surrounding circumstances. Mainly trying to remember when the AMD started to use or call out a dual power plane as in use. What we found to do was to use more Vcore than necessary to run the cpu mhz but the extra voltage saved the IMC. You got to remember this was a few years ago now.

When you look at most AM3 and AM3+ boards, you will see a memory spec of 1.5V in most I have read lately. NO I have not read every AMD motherboard spec for memory manufactured in the world. Just saying 1.5V ram volts seems the standard today in AMD world.

I am looking toward PD and had to make a NewEgg order the other day and ordered me some 1.5V ram because " I " did not want to take a chance on hurting "my" IMC. I still have quite a bit of older DDR3 but it takes 1.65V in general.

So when we type into a forum whee gawd himself only knows who is reading, we need to almost always err on the side of caution. IMO only. YMMV.

RGone...ster.

To address another myth, only of which I have heard on Overclockers.com, that much over 1.5v DRAM Voltage can kill your CPU? Nah. C2 and C3 (before 2011 weeks) non X6 or "T" model CPUs can take up to 2.4v without issue. However, there is a great chance that your DDR3 can not. Most DDR2 can, though! For later X4 and X6 models, however, stick to 1.8v or less.

(Fixed typo to make it clear it is regarding DRAM)
I would keep it under 2.5v on DDR2 to be honest for models made before 2011. (I have tested as late as 1045 with 2.4v on DDR3 and no ill effects)
The "keep near 1.5v" myth makes no sense as the same controller is dishing out voltages for both DDR3 and DDR2.

HOWEVER the Phenom II X6 and later X4s are much more sensitive, when on DDR3 you will actually notice maximum clocks going down with voltage above 1.8-1.9v, which is a hint at future damage to the IMC. On DDR2 I would keep those under 2v.

 

[BeepBeep2]

Sorry for the super-uber bad triple post here too.
Yes, I am not perfect and some may have experienced differently...

However these fit under "most case scenario", and are not solely my opinions. Some say that it must either be "fact" or "opinion", and had I labeled the thread my "opinions" people would be hounding me about every little thing for credibility.

@IMOG
I'm not totally sure about Bulldozer under LN2 however I have not seen gains with Phenom II under LN2 with CPU-NB voltage. If you have, (and you probably have much much more experience with them than I do,) that is an interesting point to make. On air cooling and DICE I have pushed chips over 5 GHz / near 6 GHz with CPU-NB voltage at a mere 0.87v. (Minimum working voltage) though on LN2 when I tried validations it was left at 1.1-1.2v.

 

[I.M.O.G.]

You probably have much much more air cooling experience, so I think your input on that front about CPU-NB volts is accurate, and its probably more important for 99% of people than my experience... I hardly ran any PHII on air, but on cold I have done it a lot, learned on it and spent a lot of time going through PHII stuff pretty meticulously. It helped for max CPU clocks... I ran it lower for everything else, just the right voltage to stabilize whatever the CPU-NB frequency was I was trying to hit.

 

[BeepBeep2]

The issue here within the forum is that CPU-NB has apparently helped several guys with stability even with stock CPU-NB frequency.

That is not a normal occurrence over @ XS, having learned a lot from the guys over there. I haven't seen it anywhere else either, like overclock.net or TPU.
I've OC'ed models from 2009 up through the last batch 1212 and I can show three different CPUs doing 4.1-4.3 GHz on air with stock CPU-NB voltage or less...
All the new Phenom IIs made in 2012 have a 1.0v CPU-NB voltage. AMD does not even tweak them for TDP anymore, only vcore.

I'd like to see proof so I can recognize what kind of instability there was.

Certain BSODs correspond to different parts of the CPU/DRAM, and Prime95 will only fail certain ways too for certain parts of the CPU/DRAM.

 

[I.M.O.G.]

That is not a normal occurrence over @ XS, having learned a lot from the guys over there. I haven't seen it anywhere else either, like overclock.net or TPU.

I'd like to see proof so I can recognize what kind of instability there was.

Certain BSODs correspond to different parts of the CPU/DRAM, and Prime95 will only fail certain ways too for certain parts of the CPU/DRAM.

On the first part, I don't think its a normal occurence here either. It seems like there are a few rare examples of where it has helped. Also it is hard to extrapolate information from XS, OCN, and TPU. The signal to noise ratio of each of those forums compared to OCF is different.

Any idea what might qualify as proof? We currently have one guy in this thread that gave testimony in normal usage, and we have my example. I also haven't ever binned AMD - I've owned one of each chip I've benched, and I didn't just get lucky each time I was able to pass 7GHz or when I went over 8GHz... I found something that worked that most did not. I can't exactly repeat any of my runs though to test, I sold or killed all that stuff (I have a nice 555BE left)... The guy in this thread could perform a stability test at his old settings, then return to his new settings, and maybe that would be a fair test? Not perfect, but one well tested example at least - if he's willing.

On the last part, I've seen that repeated a lot around (that list of common BSOD codes and what they correspond to). I know people here who go by those rules of thumb that they've seen repeated quite often. In my experience, they haven't held that true... I've often wondered where the documentation is that validates those or how its determined that hardware BSOD's are linearly connected to specific BIOS settings. It all seems too neat and convenient to me, and there was never talk of such things a few years ago... Always felt like a new fad to me, or the latest snake oil. I still refer to the list when I'm benching, but its wrong as often as it is right.

 

[Frakk]

I agree with RGone. I've always said the 1.65v ram makes some AM3/AM3+ systems unstable when the ram is run in that upper frequency/voltage range. Somehow it seems to spill over into the CPU.

I did once set my RAM to 1.65v in AI which resulted in instant BSOD, after that it simply would not boot for a while even after clearing CMOS, i left it standing for a couple of hours with the PSU turned off and it seemed fine again.

Yet when i fist got the chip i could set the CPU-NB to 3Ghz no problem, even with the HT Link @ 2.2Ghz.
Some time after the 1.65v BSOD my CPU-NB has become troublesome, i don't know and can't say the RAM volts were the cause as i can't remember exactly when it started to happen, only that it was solid before that.

Now if i set the CPU-NB to 2.2Ghz and 2Ghz HT mostly it will boot but not always. HT 2.2Ghz / CPU-NB 2.4Ghz is better and HT 2.2Ghz / CPU-NB 2.6Ghz is very argues but will stick eventually.

What happens is i set it in the BIOS save and reset, the MOBO runs through its test cycle and cuts off as soon as the CPU LED lights up, i press the power button and the same thing happens... this over and over again.

If i turn the PSU off and drain the power out of the MOBO, try again i get the Overclock failed press F1 message, if i save it immediately some times it will then boot and stick, most of the time i have to go throught that over and over again just to get a 2.6Ghz CPU-NB.

I think my Thuban is now damaged, luckily it still works reliably just as long as i don't touch the CPU-NB once i had managed to get it to stick.
I have even considered getting an FX-8 to replace it, as i can't find any new Thuban's any more, and Piledriver seems a long way away...

I will never run 1.65V on the RAM ever again.

 

[I.M.O.G.]

I've ran 1.65V-1.8V on RAM in every AMD PHII rig I've ever ran. It sounds like you have other problems on your rig that aren't limited to just ram voltage.

Running 1.65V on ram is in no manner dangerous on PHII.

 

[keny]

I run 2.1v ram on my thuban and it makes 3ghz plus on northbridge no problem, and in currently running 2.6v ram voltage on a AM3 athlon ii with no I'll affects "as of yet" so I am doubting the "anything over 1.65v will destroy the mem controller" theory

 

[Frakk]

I've ran 1.65V-1.8V on RAM in every AMD PHII rig I've ever ran. It sounds like you have other problems on your rig that aren't limited to just ram voltage.

Running 1.65V on ram is in no manner dangerous on PHII.

Well ok

I have a brand new Motherboard and the problem persists, what do you think it is?

Genuine question as i'm getting increasingly frustrated with this and if i get pushed to far i end up reacting, which means i will probably end up with an FX-8 and then regret it.

 

[I.M.O.G.]

You have CPU-NB issues it sounds like as well - thats not normal to be that hard to get stable changing CPU-NB or HT link (ht link doesn't really matter, I always kept it at 1800-2000).

Those settings have nothing to do with your mobo though, so it'd be a problem with your chip.

 

[trents]

I never said it would destroy the ICM, never. What I said was that some systems don't seem to be stable when running that high voltage ram.

In my own experience I have had more than one system that would not be stable when the ram voltage was set to the amount specified in the XMP profile but with the frequency set to a lower profile. When I lower the ram voltage a bit it becomes stable. And I have seen that with others I have helped. For instance when the ram is rated for XMP-1600 at 1.65v but actually running at the 1333 frequency profile and timings. When I have them lower he ram voltage to under 1.6 it becomes stable. I first started running into this a few years ago on an AM2+ system I had that would not be stable when running DDR2 1066 ram at 800 mhz if the voltage was set to the 2.1 required for 1066. If I lowered the voltage to 1.9 at 800 mhz it was fine.

Yet, there are other systems that don't seem to mind the high voltage ram.

 

[Frakk]

You have CPU-NB issues it sounds like as well - thats not normal to be that hard to get stable changing CPU-NB or HT link (ht link doesn't really matter, I always kept it at 1800-2000).

Those settings have nothing to do with your mobo though, so it'd be a problem with your chip.

*Screams like a girl*

i'm agonising about what to do now, i love this chip but its clearly not as it should be.

 

[BeepBeep2]

*Screams like a girl*

i'm agonising about what to do now, i love this chip but its clearly not as it should be.

If you think it is the chip that is the problem I'd be happy to help validate the issue with you.

@IMOG about BSOD codes
They are not always 100% correct, but some BSODs are common to every platform to be an issue with the memory subsystem IE. IRQL_NOT_LESS_OR_EQUAL, and on Phenom II certain BSODs have shown up time and time again for many users when pushing cores/CPU-NB, also when stressing FPU or Int. It is the same reason that when the GPU goes unstable it is usually the GPU driver listed in BSOD Really it is to point in the right direction, they are like hints, not exact answers

 

[Frakk]

If you think it is the chip that is the problem I'd be happy to help validate the issue with you.

Thank you, please do