Help me get my Deneb above 4Ghz.

[Plimogz]

*stretches* *yawns* *scratches*

It's good to sleep sometimes, ain't it? boy howdy! nn-e-way...

@DrMrLordX+Nemesis 1: Thanks for the imput/advice!

Now:

---While it may be true that my CM Hyper212 is now holding me back (FFS a 4GHz Deneb with 1.575v sucks down about 250W of juice? I doubt that the 212 was designed for this kind of abuse!) BUT I'm not about to spend 60-70$ on a Venomous X right now. A Venomous X would certainly help me at this point, maybe it would even get this 4ghz stable all by itself -- far be it from me to underestimate the value of a good cooler... but... hmmm... maybe I'll buy one in the coming weeks, but for now I'm stuck with the 212, I'm afraid.

---I don't make a habit of running 2T. but sometimes I do when I'm trying to unburden the memory controller for whatever reason (usually for testing, or trying to hit the highest CPU overclock)

---I don't run 1.9v through my DDR3. though, as I have stated, I would in fact be within the manufacturer's specs if I did. (and hey, these sticks do respond to volts a little bit, and I can get within spitting distance of the kits being able to pull off 1333 5-5-5 at 1.85-1.9v...)

---As for the Delta between vIMC and vDIMM, that's an important point because I certainly don't want to go around damaging my IMC too much. but I think that I'm keeping a pretty close eye on these things, and I'll be able to just damage my IMC a little bit... as in: running higher than specced volts, and higher than specced clocks will shorten the hardware's life expectancy, but if proper cooling is applied, and the overvoltage/overclocking is done carefully, the hardware should still be alive and kicking for a good long time after I have upgraded it out my rig...

---Highest stable OC, and just leave it alone: Well. Yeah. This would be the simplest thing. But right now, I'd like to reach, albeit mainly for psychological reasons, that nice 4 GHz mark. I'm ridiculously close: my mobo does 201Mhz base clock (instead of 200, obviously) and when I go for a straight multi overclock (20x201), this amounts to a frequency of 4027MHz. This setting has proven unstable in Linpack at 1.5v and 1.525v. However, settings of 19.5x205 or 19x211, both of which yield 3997MHz, are pretty stable in Linpack @ 1.5v. Will I notice this last 30MHz? Nope. Is spending this much time and effort to try to get a setting over 4000MHz stable cost-effective? Hell no. Why do I do it? because I like it.

@Nemesis1: when you say that running a highest OC 24/7 is expensive, do you mean because of your electricity bill? because if that's the case, don't worry about it. I'm in Montreal, and it's like -10C outside all winter long -- Any power that I carelessly pump through my Overclocked 4Ghz, 250W TDP Deneb will serve me as heat. Plus, the case fans do a good job of circulating it around the room...

Thanks again for the replies guys.

 

[deimos3428]

Is spending this much time and effort to try to get a setting over 4000MHz stable cost-effective? Hell no. Why do I do it? because I like it.
...
I'm in Montreal, and it's like -10C outside all winter long -- Any power that I carelessly pump through my Overclocked 4Ghz, 250W TDP Deneb will serve me as heat. Plus, the case fans do a good job of circulating it around the room...

And there's the solution! PhII's love cold much more than voltage. Duct that -10C directly into the case, and you'll hit 4Ghz!

(Only partly joking; I'm planning on doing some exterior OC'ing here in balmy Toronto this weekend if the weather holds up. )

 

[Plimogz]

And there's the solution! PhII's love cold much more than voltage. Duct that -10C directly into the case, and you'll hit 4Ghz!

(Only partly joking; I'm planning on doing some exterior OC'ing here in balmy Toronto this weekend if the weather holds up. )

Ok. where can I start? I've thought about this, it's true. but... hmmm. the fans and the hard drives don't like it, for starters. Get SSD's and replace the fans when/if they die, on me? hmmm. I'd want this to be a 24/7 machine and I need actual storage (i.e. more space than SSD will get me at a reasonable price) Get a NAS and use that with an SSD for OS? maybe, but this would be going further than I currently am willing to for an extra 30MHz of speed. lol

Ok. where to next? I'm not heating inside the house, and this place is pretty poorly isolated, so I keep my winter coat on, all the time anyway, but I'm not about to start piping -10C, -15C air into the house for 30MHz extra... the pipes will freeze and that'd cost me hundreds, if not thousands of $... Way way way way more than an extra 30MHz is worth to me... I'm not shooting for the top of the O.R.B. here... I'd just like to get this 965 linpack stable above 4GHz...

Thanks for answering though. And don't get me wrong, I know you were kidding to an extent. And while i do appreciate a bit of levity, I really am dead serious in my attempt to get this thing linpack stable above 4GHz... without freezing any more of my dangly bits than I absolutely have to... Cheers mate, and good luck with you sub-zero toronto air overclocking. Remember though, HDDs don't like prolonged use in freezing temperatures very much!

 

[heyheybooboo]

http://premium1.uploadit.org/bobbyboyu//gigabytegama790fxtud5pbiosf3l.zip

Some folks have noted a .05v overvolt with the vDIMM on the -UD5P.

Have you tried taking the RAMs and NB completely out of the equation and focusing on the CPU multi and clock to find the top end?




--

 

[Plimogz]

Have you tried taking the RAMs and NB completely out of the equation and focusing on the CPU multi and clock to find the top end?


--

I've tried running 20x200=4027MHz (because my UD5P defaults to 201.4MHz when set to auto) with the NB-CPU @ 2GHz (stock) and ram at 1333... I know for a fact that these sticks are rock-solid stable at this 1333 setting.

However, in response to your post I've reduced RAM speed to the 1066 multiplier. I consider that running NB-CPU at stock and the memory @ 1066 effectively takes them out of the equation. Furthermore I've increased NB-CPU voltage from the default 1.1v to 1.25v. I know that this will increase heat needlessly, at least within the context of this "focus on the CPU multi and clock to find the top end" test, but I think that this extra little bit of voltage is probably not going to hold me back at this point.

I'm currently testing with IBT v2.3. I've set it to run 25 loops with 6000MB of memory. I know this to be an extremely stressful test -- in terms of used memory, in terms of heat produced and also in terms of "long term" stability (i.e. 25 loops is rather a lot) -- but I'm a stickler for extreme stability. I'm also watching some .avi files while I run the test, because watching stability testing gets to be really boring after a while.

I expect -- based on very similar tests which I've been running since yesterday, that this set-up will BSOD around loop 3-5 -- once there is sufficient heat build-up in the motherboard and power management circuitry.

UPDATE: ah, I was actually too optimistic: IBT returned an incorrect result on loop 2. the test ran for 785 seconds. what should I do know? I'm getting to think that this chip won't do 4GHz with this cooling if I insist on wanting to use Linpack to test it's stability.

Oh and I've tried ACC -4% all cores, +2% all cores, +4% all cores. I guess I'll reboot and try -2% all cores now.

Does anyone think that changing PLL (be it CPU PLL, NB, HT PLL, or SB PLL) voltage might help?

 

[DrMrLordX]

Just to give you fair warning, proper cooling for a huge NB voltage to compensate for high vdimm would probably entail something exotic like dice, ln2, etc. I did "stress" (read: partially damage) the IMC on my K8 by running DDR2 at 2.4v (which was within manufacturer's spec) while the cores/IMC (which, on a Brisbane like mine, are on the same voltage plane) were at 1.3v for a few days. Chances are your memory will survive high voltages longer than your IMC will tolerate a significant vNB/vdimm delta.

If you're going to go over around 1.6v on your DDR3, you should push your NB voltage to the maximum comfortable limits (1.3v may be the most you want). If you don't, really, no amount of cooling (and certainly not that hyper 212) will save the IMC from damage sooner or later, and there will be no way for you to know that the damage is being done until it's too late. Temperature sensors won't give you the information you need to notice that something is going wrong.

What happened to my IMC is that it will no longer sustain high HTT speeds at DDR2-533 or DDR2-667 multipliers, even though the memory speeds obtained through such multipliers are well within the capabilities of the RAM. I'm limited to DDR2-400 and DDR2-800 multis at HTT speeds above around 290 as result. This happened to me while all recorded temps were very low, and I didn't even notice it until trying to tweak settings.

Others have simply had the IMC burn out altogether, essentially killing the entire CPU.

Also, if you want to take advantage of cold temperatures outside, you may find it difficult to do so when using an HSF that utilizes wicking heatpipes. A solid block HSF would work better in sub-zero ambient temperature situations.

 

[Plimogz]

I run the memory at 1.75v. I don't know if that's too high or not -- time will tell, i suppose (and there may well be a 0.05v overvolt on the UD5P, i suppose that to be really sure I'd have to dig up my multimeter and find good mesurement points on the board) but I think that I should be pretty safe.

Here's my logic: High performance, Low voltage DDR3 is specced to run at 1.65v, right? well, 1.75v is close enough to that... for my tastes anyway. Honestly I'd be more worried about this 1.5v I'm putting through the core... and the approximately 200-220W that the CPU will be drawing from the socket and power delivery circuitry. I mean, the max TDP for this generation of AMD chips is 140W. I'm probably asking for trouble right now. Long story short, I'm not so sure that the delta between vIMC and vDIMM is really where my head should be at, not right now anyway.

But yeah... i'm up to my fourth loop of IBT right now. this is with the same stress-test settings as in my above post. The only BIOS setting I changed on my last reboot was setting ACC to -2% for all cores. Which is the default setting on the UD5P anyway. lol.

I'm pretty much loosing hope that I'll be able to get this thing Linpack stable above 4GHz though... Oh well, I enjoy messing around with it, at least.

UPDATE: well! It's gotten through 10 loops w/6000MB MB assigned. Maximum temp is 52C. this is at 4027MHz. with ram @1066 and NB-CPU @2GHz. Of course, these IMC/memory settings yield poor real-world performance... and it may well be that once I turn up the IMC frequency, the additional heat output from the UnCore will destabilize the Core... sigh. But hey, 10 loops of IBT @4.027GHz! maybe i started loosing hope too soon after all.

 

[Plimogz]

TTT

so as per my last update, I'm now able to Linpack right above 4GHz if I turn down my RAM to 1066. I've also experimented and found that I can turn down vCORE one notch to 1.475v if I increase my _NB/PCIe/PLL Voltage Control_ one notch (+0.1v on this UD5P) for a resulting 1.9v.

doing this makes my Core Voltage, as read in CPU-Z, pass from 1.472 to 1.488 when going from idle to load; and thus I can get it linpack stable @ 4GHz with 1.475v set in BIOS... but it really is on the very edge of instability and I can't just add a NB overclock to this because it blows my 4GHz house of cards right over.

Does anyone have any good advice on how I should go about adding a NB overclock to this?

 

[Plimogz]

Ok. Well I give up. Don't get me wrong, if my sole objective was getting to 4GHz stable, I'd have succeeded (these C3's really have a little something extra compared to the C2's); but the idea was to get to 4GHz Core, 2.6-2.8GHz UnCore, with all 4 DIMM slots populated...

So I've failed.

My load temps in my last attempts (4.027GHz, 2.6GHz, 8GBx1333 6-5-5-18-21 1T) after a few loops of IBT were reaching 55C so it would come as no surprise to me that better cooling would probably be enough to get me stable at my current settings.

I suppose I could take out 2 sticks of RAM, or run my IMC @2.4GHz/1.25v thus saving some heat, or just lower the core to 3.9 and vCORE to 1.475 (maybe even 1.45v) and all of those settings would work. But hey. that wasn't the idea, was it.

It may even be possible to get my current 4GHz settings stable if I gave my Hyper212 and IHS a good lapping... I'll have to think about that one. It would be really cool to solve the problem just by lapping...

TTFN

 

[lopri]

@DrMrLordX: I am not familiar with the 'delta' theory on K8/K10's IMC. I mean, as you exampled K8 didn't have separate power planes for core and memory controller, yet they all handled 2.5~2.8V DDR fine.

Also, a major difference between AM2 and AM2+ is a separate power plane for IMC, and what AMD did was lowering voltages for IMC. If I'm not mistaken, current 45nm K10's default vNB is 1.1V. And I know that K10's IMC gets very hot as it gets more juice.

 

[soccerballtux]

@DrMrLordX: I am not familiar with the 'delta' theory on K8/K10's IMC. I mean, as you exampled K8 didn't have separate power planes for core and memory controller, yet they all handled 2.5~2.8V DDR fine.

Also, a major difference between AM2 and AM2+ is a separate power plane for IMC, and what AMD did was lowering voltages for IMC. If I'm not mistaken, current 45nm K10's default vNB is 1.1V. And I know that K10's IMC gets very hot as it gets more juice.

I'm pretty certain it's 1.2v. Also, the transistors need to be designed/spec'd for whatever given voltage you're going to run it at. So if AMD designed it for 2.8, that's fine; but with DDR2 if they designed it for 1.2v or 1.1v, then running it at 2.8v is obviously not OK.

 

[soccerballtux]

Ok. Well I give up. Don't get me wrong, if my sole objective was getting to 4GHz stable, I'd have succeeded (these C3's really have a little something extra compared to the C2's); but the idea was to get to 4GHz Core, 2.6-2.8GHz UnCore, with all 4 DIMM slots populated...

So I've failed.

My load temps in my last attempts (4.027GHz, 2.6GHz, 8GBx1333 6-5-5-18-21 1T) after a few loops of IBT were reaching 55C so it would come as no surprise to me that better cooling would probably be enough to get me stable at my current settings.

I suppose I could take out 2 sticks of RAM, or run my IMC @2.4GHz/1.25v thus saving some heat, or just lower the core to 3.9 and vCORE to 1.475 (maybe even 1.45v) and all of those settings would work. But hey. that wasn't the idea, was it.

It may even be possible to get my current 4GHz settings stable if I gave my Hyper212 and IHS a good lapping... I'll have to think about that one. It would be really cool to solve the problem just by lapping...

TTFN

Your CPU temp is fine, you do not need to worry in the least. Under _full_ load, mine sometimes reaches 63C. Max AMD limit is 71C for these quads. They revised the numbers-- so you don't need to worry about people saying "you need to keep it below 60C".

You could probably go ahead and go to 1.5V or even 1.55V (if you're not going to be running it at that 24/7-- if I were I would play it safe, but that's because I'm risk adverse that's all) and see what happens with your temps. You might need 1.5 if you push the CPU-NB (uncore as you called it) to 2.8-- keeping the cores fed more makes them more power hungry. What voltage do you need on the CPU-NB to reach 2.8Ghz?

I am jealous of your overclock, you have done very well. Grats. 4G on a Ph2 is nothing to sneeze at, you'll be set for quite some time.
I don't quite see how you think you've failed? You've got 8GB RAM + 4.0Ghz + 2.6Ghz CPU-NB? That's great by my standards. 2.8Ghz would be nice, but meh, who cares?

 

[Plimogz]

I am jealous of your overclock, you have done very well. Grats. 4G on a Ph2 is nothing to sneeze at, you'll be set for quite some time.
I don't quite see how you think you've failed? You've got 8GB RAM + 4.0Ghz + 2.6Ghz CPU-NB? That's great by my standards. 2.8Ghz would be nice, but meh, who cares?

Oh, don't be jealous, the _4GHz CPU+2.6GHz CPU-NB+8GB RAM_ setting isn't stable enough to LINPACK-64bit or PRIME95-64bit with 6000MB ram assigned.

Furthermore, the reason I try to keep it under 55C is twofold: It is partly because there have been accounts of Speedfan, Coretemp, etc. underreporting temperatures by about 5-10 C.
So, if you do the basic math, you see that the actual temps would be around 60-65. and max temp for these chips is 63C.... Anyway, I said my reason was twofold; the other reason, beyond possible missreporting by "temperature programs" is that, as guys over at XS, have pretty much established by now, once core temps start hitting 55C keeping an overclock together becomes pretty much a crap shoot.

Now let me try to expand on that last statement: First lets assume that the temps reported in whatever program we use are accurate. OK? So that would mean that my temps ACTUALLY are 55C and not some higher, unknown value. Now, If I was running stock speeds, I'm sure that my temps would not be this high; (afterall, I'm hitting 55C @4GHz with 1.5v vCORE instead of 1.4v) stock under full-load is more like 40C for me with my Hyper212. So, if I was running stock speeds, but changed my HSF for something lesser, like the stock AMD HSF which came with my 720BE (a little aluminum job -- not terrible, but not great) my temps would go up. Lets say they went up to the 63C max. would the chip crash? I don't think so. I don't even think that the chip would fail at stock speeds if it reached 70C. But this is me talking out my ass here: I don't even have anecdotal evidence to support this; I've personally never let any of my K10 silicon go over the 63C max.

But I'm not running stock. not even close. at these speeds, PhenomII really like being kept under 55C. There is no doubt in my mind that I could easily get this chip stable at 4.1-4.2GHz if I had some nice water cooling which would keep it under 35C when at full-load. but I don't.

 

[DrMrLordX]

@DrMrLordX: I am not familiar with the 'delta' theory on K8/K10's IMC. I mean, as you exampled K8 didn't have separate power planes for core and memory controller, yet they all handled 2.5~2.8V DDR fine.

Also, a major difference between AM2 and AM2+ is a separate power plane for IMC, and what AMD did was lowering voltages for IMC. If I'm not mistaken, current 45nm K10's default vNB is 1.1V. And I know that K10's IMC gets very hot as it gets more juice.

To be honest, I can't explain it. You would think that the voltage delta between the imc and core on early k8s would indicate that there would be no problem pushing DDR2 to 2.8v+ on later k8s/k10s or DDR3 to the same voltages on k10s provided the DIMMs could handle it.

It may have something to do with process shrinks but that doesn't really serve to clarify anything.

I have my own theories as to why this might be, but without any kind of evidence beyond my own anecdotes and a few postings by others who have experienced the same problem (or know someone that has, or heard about someone that has), I expect that my ramblings would do more to confuse the situation than anything else. Take it for what it's worth; IMCs may be subject to damage if you overvolt RAM without overvolting anything else. This effect has already been observed on Nehalem and on various K8/K10 chips, though it's less likely to occur on K8/K10 since people typically overvolt more than just their RAM.

All I know is that it came as no surprise to me when it was announced that Nahelem had problems with IMC burnout if you pushed vdimm too far. At least with AMD chips, you can (sort of) prevent that, provided that you're willing to overvolt more than just your RAM.

As far as AMD's decision to lower IMC voltage between K8 and K10, I suspect that the move was made for more reasons than just one.

 

[soccerballtux]

Oh, don't be jealous, the _4GHz CPU+2.6GHz CPU-NB+8GB RAM_ setting isn't stable enough to LINPACK-64bit or PRIME95-64bit with 6000MB ram assigned.

Furthermore, the reason I try to keep it under 55C is twofold: It is partly because there have been accounts of Speedfan, Coretemp, etc. underreporting temperatures by about 5-10 C.
So, if you do the basic math, you see that the actual temps would be around 60-65. and max temp for these chips is 63C.... Anyway, I said my reason was twofold; the other reason, beyond possible missreporting by "temperature programs" is that, as guys over at XS, have pretty much established by now, once core temps start hitting 55C keeping an overclock together becomes pretty much a crap shoot.

Now let me try to expand on that last statement: First lets assume that the temps reported in whatever program we use are accurate. OK? So that would mean that my temps ACTUALLY are 55C and not some higher, unknown value. Now, If I was running stock speeds, I'm sure that my temps would not be this high; (afterall, I'm hitting 55C @4GHz with 1.5v vCORE instead of 1.4v) stock under full-load is more like 40C for me with my Hyper212. So, if I was running stock speeds, but changed my HSF for something lesser, like the stock AMD HSF which came with my 720BE (a little aluminum job -- not terrible, but not great) my temps would go up. Lets say they went up to the 63C max. would the chip crash? I don't think so. I don't even think that the chip would fail at stock speeds if it reached 70C. But this is me talking out my ass here: I don't even have anecdotal evidence to support this; I've personally never let any of my K10 silicon go over the 63C max.

But I'm not running stock. not even close. at these speeds, PhenomII really like being kept under 55C. There is no doubt in my mind that I could easily get this chip stable at 4.1-4.2GHz if I had some nice water cooling which would keep it under 35C when at full-load. but I don't.

Go look at AMD's specifications sheet for these processors-- they've upped the thermal envelope to 71C.

Speedfan is reporting the temp fine-- compare RealTemp and CoreTemp for verification if you like. I decided Speedfan was fine (it always has been in my book, in spite of all the fuss about misreporting temps...) because we know chips tend to idle in the high 20's to low 30's when the voltage is low enough. My 720BE->Ph2 X4 idles @ 35C with fan speed at 25%. Touching the heatsink I have verified the chip is NOT warmer than this-- the heatpipes are very cool to the touch. This would not be the case if it were idling at 40-45C. Besides, what chip idles at 40-45C anyways? None that I've ever heard of, not ones with Ultra-120 heatsinks strapped on...

While Intel has a different litho-process (including Halfium-Dioxide instead of SiO2 for a dielectric), their thermal specifications for their chips are also 71C for quads and 73C for duals. We see this similarity due to the similar physics of the silicon itself.

After I blew through all the misinformation and actually tracked down the AMD spec sheet for my chip (and found it was 71C for quad), I ran my chip up to 69C and it has not error'd out in Linpack or 2 instances of Orthos Small.

I honestly think some of the elite guys at XS and Extreme Overclockers make silly claims like "55C" so they can maintain their overclocking crown-- don't want everyday Joe's reaching the same speeds as they did with their "teh awesome" cooling rigs. That said, I did experience difficulty getting my chip stable at 3.7Ghz no matter the voltage, so there may be some truth to what they say about stability-- all I know is that, on the quantum level, the effect of heat, and higher voltage, is the same-- they can both excite electrons into the conduction band and (with high enough voltage and heat combined) increase quantum teleportation of the electrons to the point where a logical 1 in the silicon turns to a logical 0. This is why the "mobile" variants of chips are spec'd to run at much higher temperatures-- Intel's Core 2 Duo series' thermal spec limits them to 100C for the mobile variant, because they run at a much lower voltage, ~1.1v IIRC. The desktop variants? 73C for duals, because they run at a higher voltage. The math is all there to back up but it's a little beyond this thread; but we covered this and these quantum effects in my Advanced Semiconductor Fundamentals class that I took last year.

In other words, whether you choose heat or voltage is up to you. Empirically we know 1.5v to 1.55v is fine on these chips. If you wanted to run 1.5v up to ~68C, I say that would be fine. This is what I have my fan set to, to kick in @100% speed at-- 68C.
If I had a Ph2 C3 965, I would most certainly do this myself; but I'm happy with my 720BE that I unlocked to a quad.

Choice is yours, I'm just trying to remove some artificial superstitious mental blocks you may have set up for yourself.

 

[Rhoxed]

965 - 62C MAX 140,125W
955 - 62C MAX 125W
945 - 71C MAX 95W
945 - 62C MAX 125W
940 - 62C MAX 125W
925 - 71C MAX 95W
720 - 71C MAX 95W

its all about the TDP, 125W+ = 62C and 95W- = 71C

 

[deimos3428]

its all about the TDP, 125W+ = 62C and 95W- = 71C

Yup. Which isn't very useful because if you're overclocking you probably don't care about TDP. It'd be nice if they posted some values for "above this value the chip will stop working -- permanently".

 

[soccerballtux]

965 - 62C MAX 140,125W
955 - 62C MAX 125W
945 - 71C MAX 95W
945 - 62C MAX 125W
940 - 62C MAX 125W
925 - 71C MAX 95W
720 - 71C MAX 95W

its all about the TDP, 125W+ = 62C and 95W- = 71C

AM2+ quads are 62C, the newer AM3s are 71C. Die process improvement. We determined this a long time ago.

 

[Plimogz]

AM2+ quads are 62C, the newer AM3s are 71C. Die process improvement. We determined this a long time ago.

@soccerballtux: You wouldn't mind posting a link to the source of this information, would you? Because I've been hitting up AMD.com and looking at their specs and from where I'm sitting, Rhoxed's numbers seem to be in accord with the manufacturer's. (i.e 125w, 140w TDP parts are 62C max, 95W parts are 73C.)

Beyond that, did anyone read the thread in its entirety? in the beginning Accord99 was nice enough to offer up a link to a xbitlabs article which indicates that the TDP of a 4GHz/1.575v Deneb would be around 250W. So if we follow the apparent inverse relation between TDP and Max.Temp (as somewhat evidenced by the numbers Rhoxed rounded up) it would appear that the Max.Temp of a 250W TDP Overclocked Phenom would fall well below the 62C max specced for stock chips.

Or do I need to have a bit more coffee before posting further?

 

[soccerballtux]

@soccerballtux: You wouldn't mind posting a link to the source of this information, would you? Because I've been hitting up AMD.com and looking at their specs and from where I'm sitting, Rhoxed's numbers seem to be in accord with the manufacturer's. (i.e 125w, 140w TDP parts are 62C max, 95W parts are 73C.)

Beyond that, did anyone read the thread in its entirety? in the beginning Accord99 was nice enough to offer up this link (http://www.xbitlabs.com/articles/cpu..._10.html#sect0) which indicates that the TDP of a 4GHz/1.575v Deneb would be around 250W. So if we follow the apparent inverse relation between TDP and Max.Temp (as somewhat evidenced by the numbers Rhoxed rounded up) it would appear that the Max.Temp of a 250W TDP Overclocked Phenom would fall well below the 62C max specced for stock chips.

Or do I need to have a bit more coffee before posting further?

Yeah so that's the problem, I was looking at them too trying to find it again so I could source it and I can't find it. The spec sheets I used back when I determined this...aren't there? Maybe I'm mistaken somehow...
I recall other people determining as well that the AM3 quads were able to go up to 71C, but it was the AM2+'s that were 61/62C

At the end of the day I'm not too worried about it-- assuming the heatsink can dissipate the heat, there's no reason a quad can't handle 71C if the dual core can. It's the same silicon as the quad, just with 2 cores turned off.