tistou: Glad to see you again. And thanks for that link, that provides me a contact at Intel I was hoping for willing to address this issue, since no one's responded to mine yet. I would still like to talk to them.
Anyways. Did a whole bunch of testing in XTU, and I don't think it's capable of deciding the source of the issue one way or the other definitively. As scone himself noted, the mode you set in the BIOS (auto or adaptive) affects how XTU applies the voltages. That means, regardless of how XTU addresses the MSR, the initial settings it's reading are downstream from the BIOS, and any issues there are capable of contaminating what XTU does.
And jmpboy: To your earlier question, I think I can get my 4.2Ghz overclock working with an application of 0.12v offset. This would still net me a good 0.15v lower idle voltage than offset does. Unfortunately, because we're being stopped from applying our desired additional turbo voltage, we have to make up the slack some other way, and 0.12v offset is the lowest amount I can use (with 1.92v set in the add'l turbo field) to get my system to work at the 1.212v I need for 4.2Ghz under load. I'll give it a shot later.
Did I say "we're being stopped" from it? Yes, I did. Because I think I've figured out what's going on, and I don't think I've been treated with enough courtesy to refrain from stating my hypothesis in public, even though it's very clear to me that those exhibiting the lack of courtesy really wish I'd do so, to the point of thread locking where they have the power to do so. At this point, too bad, and I'll explain why.
Occam's razor is telling me the following: I believe the nerfing of the additional turbo voltage is deliberate. It's possible Raja and Praz don't know that, but the tone of their responses indicates to me that they do. I think when the executive decision was made by Asus to not support adaptive cache voltage, someone went in and deliberately forced the additional turbo value to be unable to be set high enough to cause any possible damage.
And that's actually perfectly legitimate! I see nothing at all immoral or unethical or unwise about that initial decision, as long as Intel's side of things was acknowledged to be broken. It's not only fair, I'd *recommend* putting that nerf in under those circumstances. And it was perfectly fair for them to continue to say, all these years, "It's Intel's problem, they need to fix it." For at least as long as Intel acknowledged the problem. (I never *did* actually see anyone from Intel acknowledge it was broken on their end, btw, but despite everything I'm willing to take it on trust that in previous versions of the microcode they did acknowledge it, as I have been told here and elsewhere).
It's *perfectly natural* that a nerf to the entered value, which I would have implemented exactly in the way we're observing this is functioning now, would have been introduced into the BIOS once the decision not to support it was made in light of broken functionality. Not putting it in would've arguably been negligent!
But now, Intel seems to be stating they *have* fixed it, and that changes the game somewhat.
I said I'd explain why I'm not inclined to respect the requests to shut up about this. It's because of four mistakes, one tactical, two ethical, and one argument just based on a really bad premise.
1) Tactical mistake: When tistou managed to get adaptive working at all, and did get a drop in idle voltage with it, Praz dismissed this, fairly condescendingly, on the basis that he was using a significant offset value. That's one hell of a challenge to lay down in front of your audience. And I did it. If Praz's interest in the subject was technical, if this was a problem he *wanted* solved, then the natural response to someone actually accomplishing what he himself set up as the bar for success in getting it working would naturally be, "Hey, nice work". Instead, it got more condescension. That means technical matters or solving problems is not the driving force behind the reaction. Justifying a political/business agenda is. And it's very revealing.
2) Ethical: "Intel needs to fix it" is still being played, despite it being pretty blazingly obvious that it hasn't been tested on the newer microcode. I mean, can anyone find anything in any ASUS post in response to me or tistou's findings that suggests that there was any interest by anyone at Asus, at any point in time since Intel claims to have fixed it, to test or check the new microcode, for even one instant? (scone claims he tried on 37... given what he thinks "proved it isn't ASUS's fault" that does no such thing, and his general demeanor, I have zero faith in his debugging abilities.) No, they're actively discouraging any testing of it at all, stating flat out they have no interest in doing so, and shutting down a thread where the request is made where they have the power to do so.
3) Ethical: In order to cover up the fact that the *real* agenda is that you don't feel like dealing with it *even if Intel fixes it*, you've blown smoke up my ass about vid tables and so forth and basically relied on accusations of ignorance to deflect from that fact. Now, to their credit, both Raja and Praz have acknowledged flat out that they don't want to deal with it. It would have been honest, and not unethical, but possibly embarrassing (which I think is what's driving this whole thing) if they had openly appended "even if Intel fixes it" to that. But they didn't want to admit that last part. That response would have annoyed some folks, but it's a business decision I could respect and I would have dropped it. But to avoid saying it, they took the tactic of continuing to blame Intel and of insulting my intelligence (and I think tistou was insulted as well - "Another myth LOL!") to cover up their nerf in the BIOS that disables us from being able to take advantage of the feature.
4) The bad premise.
Originally Posted by Praz
No they shouldn't. There's nobody outside of Intel that knows the repercussions of setting this type of voltage value.
This is true of every single voltage setting sent by the BIOS to the CPU. We know how they should work, but Intel's the ones with the magic sauce. The processor works because Intel tells us it does, and when we try it, it seems to work. We take a leap of faith in Intel in doing that, and that faith is buoyed by successful testing of what they do. Now, it was broken on their side once upon a time. That breaks trust. And who knows, maybe someone at Asus who tried to get it to work when it was broken on Intel's side blew up an expensive piece of hardware. Maybe some heads rolled. Maybe it was such a colossal CF in the Asus hallways that really, NO ONE WANTS TO DEAL WITH IT EVER AGAIN.
And that would perfectly justify the initial decision to abandon support for it and to nerf that setting. The justification is weakened, but still arguably okay, when Intel comes out and says it's working now, to say "Even if Intel did fix it, we still don't want to support the feature, so we're not going to do anything." Asus doesn't have to support anything they don't want to, period. Free country and all that.
And if you're embarrassed to flat out deny support for an Intel feature we paid them for because you don't think it's worth your OR OUR time, and we're simply too ignorant to know better (a theme that keeps coming up over and over in the responses), okay, fine. Public relations and all that.
But it is NOT okay, when someone who is being deliberately fed false information manages to be intelligent enough to figure out how to get around the block you erected to shut down the functionality, to respond by harping on their continued ignorance and deriding their intelligence. It is NOT okay to actively establish the bar for success (do it without offset) in getting that feature working, and respond to *that* by not acknowledging the accomplishment at all (especially when taken at great risk) and instead insulting the intelligence of the person who figured it out. It is NOT okay to continue to say "Intel needs to fix it" when you can't deny you have no bloody idea if they already did what they claim to have, and you don't want to know.
Originally Posted by Praz
Really? So you edited a post that was thought to be non-existent also? This statement and all the ongoing edits is starting to let all the pieces fall into place.
The final tactical mistake that leads me to completely ignore your requests that I drop the matter. Should have treated me with at least some small fraction of the same politeness and respect - hell, I was practically obsequious when I first raised this issue - that I treated you with.
Now, cue the responses of "what a ridiculous conspiracy theory!", regardless of how entirely plausible it is, and how consistent it is with all the stated responses so far, and the behavior of the settings as demonstrated, and the fact that *I got it bloody working under highly variable load with offset 0 which you set as the single bar for success*. I'd like to posit, in advance, that since I am pointing out an embarrassing decision they'd like to avoid being made public (too bad, shouldn't have insulted my intelligence, coulda emailed and been honest with me too instead of thread locking and insulting me as ignorant), that response is inevitable regardless of whether I'm wrong or if I've hit the target and split four consecutive arrows with my next four shots with this theory. It's how this crazy internet game is played.
I also wouldn't be surprised if we see Adaptive cache removed as an option in the next BIOS release, instead of attempting to validate if it's functional and just removing the nerf if it is. I do wonder why they didn't just do that once the decision to not support it was made. Guess they wanted to continue to be able to list it as a feature they *want* to support but can't because of big bad Intel. Tactical mistake #3, I guess.
Cause the last relevant thing I learned from being a programmer for 25 years is - 98% of IT geeks are completely incapable of admitting when they're wrong about *anything* in relation to IT matters. In most cases, IT geeks got/get picked on in their daily life on all matters outside of IT, and their egos become entirely dependent upon being Godlike in their technical field. Condescendingly arguing from authority and lording ultra specialist knowledge over others every chance they get is as necessary as breathing. Industry hazard. I'm way beyond the bad old days of being picked on in all other matters of life, so I can and do admit when I may be wrong pretty much whenever it's called for, once a plausible counterargument that fits all the available evidence (both technical and in terms of human motives).is presented. That certainly hasn't been done here to my satisfaction yet (and in this case it'd be pretty hard to get there now, since I'm not the one insisting a negative has been proved and no further evidence of absence is required.). But I've almost never had the luxury of being afforded the same sense of fair play from the rest of the egos in the industry.
But the great thing is, despite it all, I'm personally having a blast. Between being the only person in the world who has adaptive cache fully working at 38x with 0.735v idle voltage despite the "shut up!" of the experts, and further sleuthing and discovering the exceedingly likely truths and hidden motives in the face of insults and BS... well, it really doesn't get much better than that on the internet.
/end Epic Rant.
*dances merrily on all the eggshells on his way out the door*