MP 6,1 Am I Frying My CPU? (MP 2013, E5-2667v2)

[Langdon St. Ives]

[Whoa, just reviewing what I wrote, this got long, sorry about that!]

Hey all, new member here... First of all thanks for all the great resources here in the forum, which helped me a great deal in my recent upgrade to an 8-core E5-2667v2 from the stock 4 core on my MacPro 6,1.

So in general it all went really smooth, working great, and if I check my temps with all but
one tool, it all looks fine and dandy. It's just the cpu temp shown by one particular tool,
Fanny (to which I found a reference here in the forum as well) which looks way above what I would consider healthy. Unfortunately it's seldomly well-documented what sensor exactly corresponds to readouts shown by various tools. Also unfortunately, I didn't perform any pre-upgrade comparisons with all the same tools... The only direct comparison I have is what smcFanControl shows, and on that one, temps have gone down compared to before...

The four tools I used to read temps are Fanny, smcFanControl, iStat Menus, and Intel Power Gadget. I have confirmed that each one shows the same reading with the others disabled to exclude weird interactions that I also read here that may happen if several of them run at the same time.

For loading the CPU, I have used Folding@Home using all cores as well as encoding HD material with Handbrake, which pretty much lead to the same results in my experiments. (F@H on "low" setting interestingly yields higher temps, but I can explain that because the CPU will run at a higher frequency up to 3.6 GHz with only 6 cores busy, while it'll drop to ~ 3.4 with all cores saturated. Side observation.)

So anyway, here are my measurements, all at ambient outside the machine ~ 22°C). Fanny lets me select various sensors, the one that concerns me is only what they label "Die", all others seem fine and in line with other tools. (EDIT: after some sleuthing, see #12, it appears this is what Apple calls the TC0F sensor in the SMC library.) In general, smcFanControl seems to show what iStat Menus labels as the proximity (not VR proximity) sensor, so I'll use that name below. What I call iStat is what iStat simply calls "CPU", which sounds like it could be a die temperature, but it seems more likely to me it's some inferred T_case (AFAIK there is no direct sensor for that).

• CPU idle: Fanny already reports 83° - 85° while proximity is 40° - 42° and iStat "CPU" a comfortable 47°-50°
• FAH 15 cores: Fanny = 120° (!), proximity 65°, iStat 84° - 86° -- at this point the fan steps up slightly to 1000 rpm - 1200 rpm and temp readings stabilize there
• If I now crank up the fan to full blast, Fanny still says a whopping 110°, proximity 40° - 46°, while iStat "CPU" never exceeds 75°C
• Even on a moderate fan setting of 1750 (which I've always set manually for high CPU usage like encoding jobs), iStat won't go above 77°, while Fanny still reports close to 110°.

I have also checked the CPU frequency during all of this, and I can say that during all of the high temp phases, there is obviously a small amount of throttling happening -- but not an extraordinary amount. The freq graph in iStat is not completely flat, but just hovers around 3.4 GHz, dropping or rising only on the order of maybe 50 MHz.

So there are a few scenarios I could see what's going on here:

1. Fanny is the only tool reporting actual (and correct) die temps, and these things really run this hot. (Possibly compounded by the well known IHS TIM issue that started with Ivy Bridge...?)
2. Fanny somehow doesn't scale the raw reading correctly – possibly it doesn't know about the E5-2667v2, since it's not an official CPU for the MP 6,1. Anyone else here with a 2667v2 who could confirm what the tool tells them?
3. I've botched the job of applying my thermal paste (Kryonaut) on switching out the CPU, it's now not cooling properly, Fanny is reporting true core temps, the other tools true external temps (which look fine due to imperfect heat transfer from the cores) and I'm slowly cooking my CPU, even when running idle. This I somehow have a hard time believing: If I go by what Puget Systems have found in this really nice writeup of their thermal throttling experiments on a Core i7 4790 (Haswell, one generation ahead of my Ivy Bridge, same 22nm lithography), I would say if my CPU is really running at a staggering 120°, it should all but shut itself down... To be sure, the E5 v2's throttling temp is 105° (compared to the 100° of the i7), but 120° is so far beyond that, I can't see how it would not throttle itself dramatically at that point! Also, would the SMC not start spinning up the fan at this point even at factory settings? As I mentioned, it barely budges, to around 1000 rpm...

If #1 is the case, I can't really do much, apart from spinning up my fan more aggressively than the SMC factory setting (which I'm already doing now in any case), and hope intel know what they're doing...

In case #2 I could rest easy and simply trust what the other tools tell me and forget about the whole Fanny thing.

But if #3 is the real explanation, I should obviously pop it open again ASAP and reapply my thermal paste, and hope I'll do a better job this time... I don't think so though, this is not the first time I've replaced CPUs/cooler/thermal paste, although it's been a few decades since I've done it regularly. Besides, I've replaced the thermal paste on the GPUs of the MP while I was in there, and those are running quite cool around 60° (the second = main one, the other one even cooler since I'm not using it at all).

Sorry this got so long -- if anyone made it this far: Any thoughts? Am I slowly but surely frying my processor? Or am I worrying about the wrong things, seeing as all other tools (including Intel's own Power Gadget) report temps well within normal ranges? Or maybe I should contact Fanny's developer(s) to see if it could be misreporting die temps...

Any input welcome!

[Edit: corrected some awkward line breaks]
[Edit 2: added some details regarding sensor reported by Fanny]

 

[Tiem]

When was the last time you cleaned the thing and in what environment does it run (do you smoke around it)? Most heat issues are because fans get clogged with dust and inhibit their cooling.

If the machine is clean, it could be the heatsink has warped. I had that happen to my first gen 2006 MBP. I had to source out a new heatsink assembly which fixed it.

I would take it apart (hit up iFixIt) and give it a cleaning. You could try removing the ******, ancient thermal paste from the factory and apply some decent stuff, like arctic silver or something. I did that on mine and my temps went down like 15%. Quite huge.

There is only a few things that affect heat so finding the cause is actually pretty easy.

 

[GoodGuy12345]

Sadly the trash can Mac Pro is "normal" for doing that. Since Apple had thermal issues trying to upgrade it, thus causing the 2019 Mac Pro to revert back to the cheese grater design.

The 2013 Mac Pro will definitely surpass 100 degrees on heavy loads. You might want to take it easy on that computer if you dont want to bake it to death. 50-60 degrees idle is not normal for any computer, and past 100 on load is not either.

 

[Tiem]

I know lots of people had issues. But to say that is normal is reaching. I had one for 4 years. Never had an heat issues.

I don’t doubt many do/did mostly because it was a very eclectic design that, like you say, had a very very tight TDP.

But my cores never shot to 100C. In fact temps were actually just as good as PC (mid 30s)

 

[Langdon St. Ives]

So thanks for the responses so far. I am aware of the thermal history of the trashcan, but the thing that I'm trying to figure out here is whether the cores are really running that hot, or it's maybe a measuring artifact. To repeat: there is only one single tool that reports these (to me) outrageous temps (Fanny), all others, including Intel's own Power Gadget are showing temperatures that are entirely within normal range (<~ 75°C under all-core load, for what I can only assume is T_Case). I'll respond to some specific points raised individually.

 

[Langdon St. Ives]

When was the last time you cleaned the thing and in what environment does it run (do you smoke around it)? Most heat issues are because fans get clogged with dust and inhibit their cooling.

If the machine is clean, it could be the heatsink has warped. I had that happen to my first gen 2006 MBP. I had to source out a new heatsink assembly which fixed it.

I would take it apart (hit up iFixIt) and give it a cleaning. You could try removing the ******, ancient thermal paste from the factory and apply some decent stuff, like arctic silver or something. I did that on mine and my temps went down like 15%. Quite huge.

There is only a few things that affect heat so finding the cause is actually pretty easy.

So as I wrote (but I understand it may have gotten lost in my long post, so I'm not complaining), I only recently upgraded the CPU to this 8-core E5-2667v2. At that point I removed and replaced the thermal compound (I used Kryonaut), and of course I also completely cleaned all the dust out. Right now it's clean as a whistle. And the temps reported by all other monitoring tools I mentioned have gone down for me too after the upgrade, even though I went from 4 cores to 8 (same TDP though).

Only "Fanny", when set to report what it calls the die temperature, shows these outlandish numbers...

Regarding heatsink warping: Hmmm... now I didn't particularly look for that, but I did inspect it carefully for any imperfections after cleaning off the baked-on old gunk of Apple's low grade thermal compound, and thinking back, it did look completely plane to me. As I said, I didn't look for this specifically, but I would think I would have noticed any significant warping.

So again, to sum up my doubts about these >> 100° numbers (only shown by "Fanny"):

1. Shouldn't the CPU's internal thermal management kick in at 120° (or already well below, at that), and significantly throttle the frequency, if not outright shut down some or all cores?

2. This particular CPU model was never offered by Apple for the trashcan (they only had the 1680v2 as 8-core), so possibly its sensors report numbers that Fanny does not understand correctly...? This is right now pure speculation, but seems not far-fetched to me.
[automerge]1591212895[/automerge]

I know lots of people had issues. But to say that is normal is reaching. I had one for 4 years. Never had an heat issues.

I don’t doubt many do/did mostly because it was a very eclectic design that, like you say, had a very very tight TDP.

But my cores never shot to 100C. In fact temps were actually just as good as PC (mid 30s)

Couldn't agree more with the first part. My machine has also been 100% stable – until recently with the stock 4-core, and now with the 8-core that I dropped in two weeks ago or so. I don't see any real world problems even with all 16 virtual cores fully loaded. CPU freq stays at 3.4 GHz, no crashes or anything. Base freq is 3.3, so I really can't imagine cores being too hot and its thermal management still giving them 100 MHz on top of that! The cut-off (T_junction_max) for these Ivy Bridge E5s is supposedly 105°, after which they start to throttle on their own. How could they stay above base frequency at 15°C above that? Just doesn't compute, to me...

But there is this nagging doubt in the back of my head, what if I really am cooking the poor thing to death?

 

[Tiem]

If this showed up the moment you installed your new cores then it’s more than likely it’s them. Recheck you installed them properly. That’s I would do first.

As for Fanny, it’s readings are right. But you grab a trial of iStat Menus if you want to double check.

I would also check the cores themselves and make sure they are 100%.

 

[Langdon St. Ives]

But my cores never shot to 100C. In fact temps were actually just as good as PC (mid 30s)

Oh forgot to address this one: In this particular case, it's extremely important to be more specific about exactly which sensor readings we're comparing. There are so many that could be reported by various tools (under various names), and in some cases computed values using some scaling and/or offset...

For example, as I'm writing this, my Folding@Home client is chugging along on all 16 virtual cores, and iStat Menus is reporting the following temps (the labels are exactly as shown, I have no info where exactly all of these sensors are located):

Ambient 1: 29°
Ambient 2: 32°
Ambient 3: 30°
CPU: 76°
CPU Proximity: 54°
CPU VCC VR Proximity: 61°
[Leaving out 15 more, 2 for each RAM module, 3 for each GPU, and a bunch of board sensors]

When the machine is idle, the proximity sensors report <~ 40° at SMC default fan setting (i.e., fan idle), and in the thirties if I spin up the fan a bit.

Even for these Fanny numbers it's critical to be precise: the default setting that this tool reports is an "Average" of all the CPU sensors it reads, and those numbers also look completely unremarkable in my case. Only if I go into settings and explicitly select what it calls the "die temperature" do I see those numbers in the 100's...

 

[Tiem]

It depends on your system. Different Macs have different thermal profiles.

For the MP, using Fanny, take the average. You can check them all but it really doesn’t matter here all that much because so long as the value is correct you’re comparing oranges with oranges. So long as you pick the same value every time.

 

[Langdon St. Ives]

If this showed up the moment you installed your new cores then it’s more than likely it’s them. Recheck you installed them properly. That’s I would do first.

As for Fanny, it’s readings are right. But you grab a trial of iStat Menus if you want to double check.

I would also check the cores themselves and make sure they are 100%.

Thanks – I do have iStat Menus, actually paid for it just a few days ago after realizing it would replace both the smcFanControl and the MenuMeters I was using until now, and do a better job in both cases. ;-)

I'm not quite ready to tear the machine down again just to redo the thermal paste... it wasn't difficult, but quite laborious... a royal pita...
[automerge]1591214036[/automerge]

If this showed up the moment you installed your new cores then it’s more than likely it’s them.

Oh here's the other thing: unfortunately I do not have the direct comparison for the Fanny "Die" temps from before/after, because I only found that tool after the CPU switch. The temperatures that I can compare (what smcFanControl reports, and which by comparison with iStat Menu's different readings now seems to be the proximity sensor) have gone down after the upgrade.
[automerge]1591214182[/automerge]

It depends on your system. Different Macs have different thermal profiles.

For the MP, using Fanny, take the average. You can check them all but it really doesn’t matter here all that much because so long as the value is correct you’re comparing oranges with oranges. So long as you pick the same value every time.

Yup. So that's one of the reasons I'm posting here: because I hadn't known about Fanny before, so I don't have comparison numbers, so I was hoping someone could check what Fanny reports for them when set to the die temperature – ideally someone with the same E5-2667v2 I have... ;-)

 

[Tiem]

Ya sorry man. I sold mine last year. But I only had the 6-core.

But my temps were never insane. Encoding. Gaming. I stressed that little thing and never saw 100C on any sensor. The 8-core might produce substantially more heat you never know.

 

[Langdon St. Ives]

No sweat, maybe someone will come along at some point. Thanks for the feedback though!

Did some more sleuthing, digging through Fanny's source code on github, and it looks like what it reports as Die temp is the TC0F sensor, which comes directly from Apple's SMC library, and which is documented as, well, the "CPU 0 die temp"... ;-) Of course that does leave the door open to one of my suspicions – wrong scaling of the raw value because the CPU is not "meant" (by Apple) to be running in this mac. (Speaking of which, maybe that's even in some way part of the reason OWC stopped selling upgrades to this E5-2667v2 at some point, but continued to offer other upgrades that weren't part of Apple's original lineup... Hmmm... Or they noticed it was running hot and stopped due to that. Getting very hypothetical here.)

 

[SamPotts]

I also fitted the 2667 and while it does run warm, I've always assumed that was down to the Mac Pro's design. Currently it's ~10c ambient here in Sydney, but I still see ~65-70c at idle which always strikes me as being on the high side. I've never seen it go above 95c under load though. I do have 3 SSDs inside the Mac Pro though which each kick out 45-50c which doesn't help I bet. I've been meaning to take it all apart again and try some better CPU compound although I used Gelid stuff at the time that was meant to be one of the best. Perhaps I didn't use enough? My PC with a 5960x (@4.5GHz) only hits ~30c at idle but that does run an AIO water cooler.

 

[MisterAndrew]

I'm looking at the sensor readings in the Macs Fan Control app and it shows idle temperatures in the upper 30s and low 40s C for my E5-1680 V2. It's about 22 C in the room. When I ran Geekbench it went into the upper 50s. If the E5-2667 V2 is getting that hot then it's consistent with my theory that it's too hot for the 6,1 based on Intel's data. You might want to then switch out to an E5-1680 V2. The performance between those is so close it's basically negligible.

 

[SamPotts]

I think I'll try re-applying the thermal paste before I replace the CPU with another model.

 

[Langdon St. Ives]

I also fitted the 2667 and while it does run warm, I've always assumed that was down to the Mac Pro's design. Currently it's ~10c ambient here in Sydney, but I still see ~65-70c at idle which always strikes me as being on the high side. I've never seen it go above 95c under load though. I do have 3 SSDs inside the Mac Pro though which each kick out 45-50c which doesn't help I bet. I've been meaning to take it all apart again and try some better CPU compound although I used Gelid stuff at the time that was meant to be one of the best. Perhaps I didn't use enough? My PC with a 5960x (@4.5GHz) only hits ~30c at idle but that does run an AIO water cooler.

Thanks for your feedback – could you specify which app exactly you're using to measure,
and in case it can show more than one sensor reading, which one? Just so we're
comparing apples and apples... Also, maybe if you have a few minutes you could install the
Fanny app (it's free and not very invasive), and make sure to set it to show the "Die"
temperature, and watch that in idle and under load? That's the only reading (technically,
it's the TC0F sensor) that goes that high for me. That would be very interesting to compare.
[automerge]1591572993[/automerge]

I'm looking at the sensor readings in the Macs Fan Control app and it shows idle temperatures in the upper 30s and low 40s C for my E5-1680 V2. It's about 22 C in the room. When I ran Geekbench it went into the upper 50s. If the E5-2667 V2 is getting that hot then it's consistent with my theory that it's too hot for the 6,1 based on Intel's data. You might want to then switch out to an E5-1680 V2. The performance between those is so close it's basically negligible.

Ya I was actually trying to get a 1680 instead of the 2667 when I saw that OWC stopped
(at some point) doing upgrades with the latter, but I couldn't get my hands on one (well,
Aliexpress of course but I wasn't ready to go that route, I got my 2667 from Tech Buyer
which seemed a bit more dependable to me.)

But here's the thing that I don't get: if the CPU is really running that hot, shouldn't I see
significant thermal throttling applied? I thought the CPU did that on its own, no intervention
from firmware or the like... So as long as it's not doing that, shouldn't I be fine?

I may just run it till it dies and replace by a 1680v2 once that happens – if it ever happens...

And maybe I will redo the thermal compound after all.

[Edit: closed an unclosed parenthesis]

 

[Langdon St. Ives]

BTW, I also found one other tool that will report this sensor, and that's smckit
(compiled from source), and interestingly, while the absolute value it reports is
exactly in sync with Fanny's die temperature, SMCkit shows this as negative
for whatever reason... e.g. as I'm writing this (cpu 97-98% idle):

#smckit -cdwt
-- Temperature --
AMBIENT_AIR_0 (TA0P) 28.0°C (Nominal)
AMBIENT_AIR_1 (TA1P) 32.0°C (Nominal)
CPU_0_DIE (TC0F) -89.0°C (Cool)
CPU_0_PROXIMITY (TC0P) 46.0°C (Nominal)
GPU_0_DIODE (TG0D) 48.0°C (Nominal)
GPU_0_PROXIMITY (TG0P) 45.0°C (Nominal)
MEM_SLOTS_PROXIMITY (TM0P) 42.0°C (Nominal)
MISC_PROXIMITY (Tm0P) 28.0°C (Nominal)
THUNDERBOLT_0 (TI0P) 50.0°C (Nominal)
THUNDERBOLT_1 (TI1P) 45.0°C (Nominal)

So maybe that's another indication that there is some missing/wrong scaling happening
here because the SMC doesn't "know" about this CPU?

 

[Langdon St. Ives]

I'm looking at the sensor readings in the Macs Fan Control app and it shows idle temperatures in the upper 30s and low 40s C for my E5-1680 V2. It's about 22 C in the room. When I ran

Forgot to mention: so the CPU proximity sensor is also only showing some 45°C for me
idle (like right now, with ambient also almost exactly 22°). But I think I've never seen that
one in the 30s with this chip.

If I load all cores (Folding@Home or x264 with HD material), at factory fan settings
CPU proximity will go into the 60s, but if I set the fan to ~1750, it won't exceed the
low/mid 50s, and fan at full blast will push it down to like 48°C.

So looking only at the proximity temps I probably wouldn't worry, it's only that pesky
>100 die temp that's unnerving.

 

[SamPotts]

I just literally got done replacing the thermal paste. When I took the machine apart, the top fan was caked in dust and would have barely have flown much air. That probably didn't help at all. Gave it a good clean out and replaced the thermal paste with Gelid GC Extreme. I've only just booted the machine but the CPU is idle at ~45c. Proximity ~35c. A big improvement. I did notice I'd not used enough paste last time so I used a little more this time around. As testing has shown, too much isn't anywhere near as bad as not enough.

I'm measuring using iStat Menus which shows all the available sensors. The other one I keep an eye on is the PCIe Switch Diode as that gets pretty toasty and is known to many. I can give the Fanny (nice name haha) one a go later.

Ignore the 960 Pro bar. It's only just been fitted so iStat Menus doesn't know it's "max" yet.

I do also use Macs Fan Control to ramp up fan speed when the PCIe Switch Diode gets hot. That shows similar temps.

I'll give it some load later to see how it copes but so far so good.

 

[MisterAndrew]

That looks good. Cleaning it probably helps a lot. I use an air compressor with an in-line dryer filter with the regulator set below 60 PSI. I carefully hold the fan steady with a pointed tool so it doesn’t move.

And to help dust from accumulating in the first place I keep an air purifier running near the computer.

 

[Langdon St. Ives]

I'm measuring using iStat Menus which shows all the available sensors. The other one I keep an eye on is the PCIe Switch Diode as that gets pretty toasty and is known to many. I can give the Fanny (nice name haha) one a go later.

Thanks, that's helpful – I'm also using iStat Menus as my main monitoring tool. Unfortunately it doesn't
list that particular die sensor (TC0F), and I'm also thinking maybe that's for a reason (because it doesn't
give user-relevant feedback?)... All these points are what's giving me these doubts whether I'm worrying
about a real issue or some artifact here.

So yeah, I'd be really interested in your readings with Fanny (remember, needs to be explicitly set to
die temperature, by default it shows an average of all sensors it knows about).

For fan speed control I use iStat Menus as well, and I've also read elsewhere about the PCIe diode and
was thinking about using that reading to ramp up fan speed, but since it's basically always "warm" I don't
see that much use in it.

That looks good. Cleaning it probably helps a lot. I use an air compressor with an in-line dryer filter with the regulator set below 60 PSI. I carefully hold the fan steady with a pointed tool so it doesn’t move.

And to help dust from accumulating in the first place I keep an air purifier running near the computer.

Dust can't be the issue in my case, since I just replaced the CPU a few weeks ago and cleaned out all the
mess at that time (and redid the TIM). Good idea about the air purifier though, that hadn't occurred to me.
I had been considering adding some filtering to the inlets on the bottom but stayed away from it because I
was concerned they'd also just create obstructions.

 

[MisterAndrew]

I have this air purifier. It was the one they had at my local Costco warehouse. It seems to work well. It's not very loud. It has 4 speed settings.

Winix C545 4 Stage Air Purifier with WiFi With PlasmaWave Technology

Winix C545 4 Stage Air Purifier with WiFi With PlasmaWave Technology

www.costco.com

 

[Langdon St. Ives]

I have this air purifier. It was the one they had at my local Costco warehouse. It seems to work well. It's not very loud. It has 4 speed settings.

Winix C545 4 Stage Air Purifier with WiFi With PlasmaWave Technology

Winix C545 4 Stage Air Purifier with WiFi With PlasmaWave Technology

www.costco.com

Thanks that looks like a good unit. No Costco here in Germany and I won't be coming to the States any time soon under the given conditions, but it seems Winix is readily available locally as well (not at Costco prices but what can you do), so thanks a bunch!

 

[Langdon St. Ives]

I just literally got done replacing the thermal paste. When I took the machine apart, the top fan was caked in dust and would have barely have flown much air. That probably didn't help at all. Gave it a good clean out and replaced the thermal paste with Gelid GC Extreme. I've only just booted the machine but the CPU is idle at ~45c. Proximity ~35c. A big improvement. I did notice I'd not used enough paste last time so I used a little more this time around. As testing has shown, too much isn't anywhere near as bad as not enough.

I'm measuring using iStat Menus which shows all the available sensors. The other one I keep an eye on is the PCIe Switch Diode as that gets pretty toasty and is known to many. I can give the Fanny (nice name haha) one a go later.

View attachment 922411

Ok so that's interesting. Here's how this looks for me, the machine has been woken up
for a few hours but running unattended and mostly idle until I sat down at it now.

What strikes me is that _all_ temps are around 10°C (call it 10 +/- 5) higher in my case, from IHS through proximity to ambient, and also the ones around other components. Also the notorious PCIe diode, which is a full 20° hotter than yours – in fact comparing this to your numbers right now makes me think if that might not be one of the sources of the higher temps overall.

Also weird: my CPU and total wattage is also higher, ~ 10W CPU and ~ 10W GPU, also weird... machine is provably not doing anything much, neither CPU (97% idle) nor GPU (99% idle).

So maybe I really should have another look at the thermal compound and reapply it. But I have to honestly say all of these numbers together don't look to me like I have a TIM issue but something else is going on... or am I delusional?

Edit – PS: The HDs are external in case it's not obvious, so ignore those.
[automerge]1591655065[/automerge]

I do also use Macs Fan Control to ramp up fan speed when the PCIe Switch Diode gets hot. That shows similar temps.

View attachment 922413

I'll give it some load later to see how it copes but so far so good.

I also have that although I don't have it running all the time (iStat Menus does all I need for now), but I can confirm the same picture as for the iStat numbers: ~ 10° hotter for the CPU Sensor 1 through 4 and the GPU diodes as well.

Thanks a lot for your numbers!

 

[Langdon St. Ives]

So I did decide to tear the thing down and reapply the thermal paste once more last night. There is a slight difference in temps that I'm seeing now, maybe 2-5°C or so lower under load on all of TC0F (aka die temp in Fanny, aka I presume T_jct, and the one that was worrying me), iStat's CPU, proximity, and voltage regulator proximity sensors.

So maybe the first time around I really didn't do a very good job with the thermal paste... but the big picture didn't really change, these TC0F numbers still go above 105°C (but now only just), so I still don't quite understand what's going on. Maybe I still didn't do a good enough job, but I'm not planning on redoing it again.

Using more aggressive fan speed settings via iStat Menus, I can now at least keep this TC0F around 107° even under continuous 100% CPU load (F@H), and T_case well below 75°C, so I figure there isn't much more I can do and I'll just live with it and hope Intel knows what they were doing.

If the chip dies, I'll probably just get the 10-core E5-2690v2 instead, which seems to be thermally less challenging (or challenged, if you prefer). Should have done that from the start, I was going back and forth forever between those two. But I have a feeling the chip will outlast the lifetime (for me) of this machine so I'll try to let it go and stop with the paranoia...