Overclock 6700K Max Temps Prime95 Small FFT

On 2017-01-26 00:59, Paul wrote:

B00ze wrote:
On 2017-01-25 05:15, Paul wrote:

Your curve seems to have a longer rise time than mine,
almost as if the cooler is being "overwhelmed".

It's a "bug" in SpeedFan - I tried again and this time I added the CPU
Cores to the graph: They shoot-up to 88C immediately, while the line
for "CPU" slowly slowly slowly rises.

[snip]
They barely get hot to the touch tho, which I find strange - on my
laptop the exhaust vent becomes really hot immediately when I run
Prime95, not so on the Noctua cooler in my rig, it just becomes
warm'ish - I mean it's higher than room temp, but I can leave my hand
there forever and never feel the need to remove it because it's too
hot...

You could try a more powerful fan on the cooler.

Yeah, I thought of that too - the Hyper EVO has a much faster fan, but
I don't know, Noctua fans are supposed to be good. Oh well, I guess I
will live with it. The only thing that worries me is whether I have
bent the CPU's PCB or not - if not bent then fine, it's how it is. And
if I did bend it, well I'm not buying another one, lol, so I will just
accept that it gets hot under Prime95.

The other issue with these designs, is "no cowling".
Mine doesn't have one either. A cowling on the heatsink,
would ensure that any static pressure the fan can develop,
goes through the fins.

Since I own a 37.5mm thick fan, if I had to attempt a retrofit,
I'd do a sheet metal cowling, plus fit the 37.5mm thick fan.
That's to get as much air velocity as possible through the
fins. I've only done one cowling/shroud for a CPU cooler here,
in the time I've worked on computers.

Air cooling goes asymptotic over about 800LFM (linear feet per minute).
Run of the mill electronics, might use around 200LFM. So there
is a point, where "mo fan" does "no good".

Regards,

There's probably a couple ways the pipe wouldn't get hot.
Likely leaving no choice but to take it apart and verify things.

Yeah but for now I'm happy with it, I'm not likely to do anything close
to Prime95 in normal use.

I agree that the silicon die on your processor, is probably
pretty tiny. (All that FINFET goodness.) Which puts an extra
responsibility on the sole plate of your Noctua. If you had
one of those coolers where the bare pipes touched the processor,
then a couple of the pipes might be carrying all the heat. Yours
has a sole plate, so that gives a bit of spreading. The newer
processors have a thicker lid, which was added for spreading
angle too.

6700K - de-lid
https://hardforum.com/threads/6700k-...sults.1878870/

That guy hits 87C with a water cooler, so my hitting 92C (max) on air is
ok, even tho he reaches 1.4V to my 1.360V maximum. On Large FFTs @
1.344v I hover around 67C which is fine. He dropped 20C with the de-lid,
amazing. Btw, I fixed the Watts reading not working by reEnabling SVID;
HWMonitor says I pull 135W @ 1.360V when I hit 92C (nowhere near the
177W that guy is burning).

Technique demonstrated on a 3770K
https://forums.anandtech.com/threads...-tims.2285595/

The first one, shows the motherboard measurement tool showing
177W when a bit more than your voltage is applied.

177W that is crazy; I can never reach that, my cooler setup would never
work.

The second article, shows just how thick the IHS is getting
on these processors. It's there to get spreading angle
and make the thermal footprint larger on the lid surface.

DeLid is just too much trouble, even if it is extremely efficient.

For best results, the heatsink should be oriented such
that it had the feet of the pipes going across the
thin dimension. If just one pipe runs along the die, that
would give poor cooler response.

I think that's how it is for me...

In this picture, you could compare the three screws of the
normal socket mounting hardware, to the orientation of
your setup. I would guess they don't arbitrarily rotate
the die, between 3770K and 6700K :-) But who knows.

http://i272.photobucket.com/albums/j...0kinsocket.jpg

Yup, that's how it is on my MB as well.

My CPU would be a bit different, as the die is square
on mine. And the lid is soldered down. I believe at the
time, that was part of my buying decision (4930K or 5930K).
Too bad I was such an idiot on motherboard selection, and
am now limited by a gutless VCore.

http://i.imgur.com/ZnOF6xh.jpg

Paul

--
! _\|/_ Sylvain /
! (o o) Memberavid-Suzuki-Fdn/EFF/Red+Cross/SPCA/Planetary-Society
oO-( )-Oo If there is no God, who pops up the next kleenex?

On 2017-01-26 11:26, hanky liu wrote:

Have you considered liquid cooling? What mobo are you using, and the build?

MB is a Maximus VIII Hero. I'm fine with Air - I just wanted to increase
the MHz as much as I could on Air and still get a 100% stable system, I
was just surprised at the high temps on some tests. I tried 4.6GHz
without touching my current setup but Prime95 fails; I would have to
increase base voltage from 1.30V to 1.35V and tweak LLC again, but
4.5GHz is fine.

Btw, I installed HandBrake and tried to transcode some 4K video -
Instant HandBrake crash. So I dialed down to plain vanilla 4.0GHz and
still crash. A few more crashes later, I installed it on the laptop and
tried there, and it still crashes. Works fine for 1080p video but no at
all with a 4K source. Suspect HandBrake problem rather than both my
machines having issues.

Best Regards,

--
! _\|/_ Sylvain /
! (o o) Memberavid-Suzuki-Fdn/EFF/Red+Cross/SPCA/Planetary-Society
oO-( )-Oo Kirk to Enterprise, Kirk to Enterprise! DAMN, It's BUSY!

On 2017-01-23 20:55, B00ze wrote:

Good day.

So I've overclocked my 6700K to 4.5GHz; this required 1.3V base and with
LLC=6 it goes up to 1.360V in extreme situations, and this is where I am
wondering if I screwed-up the heat-sink/termal-paste somehow. What temps
are you guys seeing when you run Prime95's Small FFT (Maximum Heat) on
your overclocked CPU? Because on mine that particular test can make the
temps climb all the way to 93C. The heat-sink is a really good one
(NOCTUA NH-U12S) but it doesn't even get hot to the touch, it is barely
warmer than room temp, and I used good Arctic Silver TIM. While I may
have placed a tiny tiny bit too much I was real careful with the stuff
(maybe I didn't put enough?) Under normal testing I have good temps, but
under that extreme Small FFT Prime95 test, my CPU really cooks - do you
all get the same temps on air?

Good day Paul, Hanky Liu, all.

So I took the plunge - I had to open the desktop and move some hard
drives around, so I took the opportunity to remove the CPU heatsink and
have a look: Good news; I had, as I expected, just a tiny tiny tiny bit
too much TIM and the CPU PCB was NOT bent. Cleaned-up with 100% alcohol
(NOT rubbing alcohol) and re-did the TIM and put the heatsink back
(pretty much with the same amount of TIM). Since this is Arctic Silver
5, I need to wait a couple days for it to settle before I test again,
but I ran OCCT real quick and I get the same temps as before (70-80
degrees under the CPU stress test (NOT the Linpack)). We'll see in a few
days, I will report back.

PS: I had to move a drive because it was heating-up, it's a 6TB with
many platters so it gets hot. Turns out what I thought was a really good
feature of my Thermaltake Urban T81, those two 200mm front fans, is also
a problem: The cylinder in the center of the fans that holds the blades
is really big, and the hard drives that sit at that position do not get
any air because of that. Seems smaller fans are better...

Best Regards,

--
! _\|/_ Sylvain /
! (o o) Memberavid-Suzuki-Fdn/EFF/Red+Cross/SPCA/Planetary-Society
oO-( )-Oo Is there another word for synonym?

B00ze wrote:
On 2017-01-23 20:55, B00ze wrote:

Good day.

So I've overclocked my 6700K to 4.5GHz; this required 1.3V base and with
LLC=6 it goes up to 1.360V in extreme situations, and this is where I am
wondering if I screwed-up the heat-sink/termal-paste somehow. What temps
are you guys seeing when you run Prime95's Small FFT (Maximum Heat) on
your overclocked CPU? Because on mine that particular test can make the
temps climb all the way to 93C. The heat-sink is a really good one
(NOCTUA NH-U12S) but it doesn't even get hot to the touch, it is barely
warmer than room temp, and I used good Arctic Silver TIM. While I may
have placed a tiny tiny bit too much I was real careful with the stuff
(maybe I didn't put enough?) Under normal testing I have good temps, but
under that extreme Small FFT Prime95 test, my CPU really cooks - do you
all get the same temps on air?

Good day Paul, Hanky Liu, all.

So I took the plunge - I had to open the desktop and move some hard
drives around, so I took the opportunity to remove the CPU heatsink and
have a look: Good news; I had, as I expected, just a tiny tiny tiny bit
too much TIM and the CPU PCB was NOT bent. Cleaned-up with 100% alcohol
(NOT rubbing alcohol) and re-did the TIM and put the heatsink back
(pretty much with the same amount of TIM). Since this is Arctic Silver
5, I need to wait a couple days for it to settle before I test again,
but I ran OCCT real quick and I get the same temps as before (70-80
degrees under the CPU stress test (NOT the Linpack)). We'll see in a few
days, I will report back.

PS: I had to move a drive because it was heating-up, it's a 6TB with
many platters so it gets hot. Turns out what I thought was a really good
feature of my Thermaltake Urban T81, those two 200mm front fans, is also
a problem: The cylinder in the center of the fans that holds the blades
is really big, and the hard drives that sit at that position do not get
any air because of that. Seems smaller fans are better...

Best Regards,


If I look at image 28 here, how is the air supposed to get
through that drive tray and rack design ?

http://www.thermaltake.com/products-...?id=C_00002239

The air from the top 200, it going to squirt over the top rack,
instead of going through it.

http://www.overclockersclub.com/site...t81/t81-21.jpg

And I'm surprised there isn't a noise problem, from the
200mm fan being so close to the metal of the drive racks.

On a high velocity air cooled design at work, when a slot
is not filled in the chassis, we provide a "filler" card.
That's a box that fills the slot, and prevents cooling air from
leaking through the hole. (The air is then forced to go
up through the powered and operational cards.) The I-81
design looks like it's in need of both "dremeling and filling" :-)
Of course, I say that, because I love to mod stuff like
that.

Paul

On 2017-02-11 01:22, Paul wrote:

If I look at image 28 here, how is the air supposed to get
through that drive tray and rack design ?

http://www.thermaltake.com/products-...?id=C_00002239

Well there /are/ holes in the racks, and slits between each drive. But
don't get me started on that case. In the last 2 pictures on that
webpage, you see cool blue arrows showing incoming air on the side of
the front panel (where the 3 rows of rectangular slots are in the
plastic, on the side of the front panel). Well, air WOULD come into
those slots if the fans were further inside the case, but they are not,
they are further to the front of the case. Those air slots are mostly
useless. Then there is the front door - there is no real way for air to
come into the case when the door is closed (except from the air slots at
the very botton of the front panel) - I am using a bent paper clip to
keep the door ajar. It's a great case because it is huge, but that's
about it; too many stupid design choices all over.

The air from the top 200, it going to squirt over the top rack,
instead of going through it.

http://www.overclockersclub.com/site...t81/t81-21.jpg

If you look at that picture, the problem HD slot is the 3rd one from the
top (bottom of first HD cage). It is now empty, the 3 drives for the
RAID are all in the 2nd cage and staying cool. That 3rd slot from the
top just so happens to be right smack in the center of the first fan...

And I'm surprised there isn't a noise problem, from the
200mm fan being so close to the metal of the drive racks.

Nope, super quiet.

On a high velocity air cooled design at work, when a slot
is not filled in the chassis, we provide a "filler" card.
That's a box that fills the slot, and prevents cooling air from
leaking through the hole. (The air is then forced to go
up through the powered and operational cards.) The I-81
design looks like it's in need of both "dremeling and filling" :-)
Of course, I say that, because I love to mod stuff like
that.

Lol, the HD trays do fill the slots somewhat. It's hard to choose a good
case! Looks easy, but there are many variables! I would not buy this
again, even tho I'm somewhat OK with the purchase. I'd look for some
kind of server case instead. Or just give up and get the Cosmos II lol
(but I didn't like the fact that the top of that case is not flat; can't
put USB drives on top).

Best Regards,

--
! _\|/_ Sylvain /
! (o o) Memberavid-Suzuki-Fdn/EFF/Red+Cross/SPCA/Planetary-Society
oO-( )-Oo Windows-NT is the O/S of the future (and always will be.)

On 2017-02-10 22:42, B00ze wrote:

So I took the plunge - I had to open the desktop and move some hard
drives around, so I took the opportunity to remove the CPU heatsink and
have a look: Good news; I had, as I expected, just a tiny tiny tiny bit
too much TIM and the CPU PCB was NOT bent. Cleaned-up with 100% alcohol
(NOT rubbing alcohol) and re-did the TIM and put the heatsink back
(pretty much with the same amount of TIM). Since this is Arctic Silver
5, I need to wait a couple days for it to settle before I test again,
but I ran OCCT real quick and I get the same temps as before (70-80
degrees under the CPU stress test (NOT the Linpack)). We'll see in a few
days, I will report back.

Alright, I tested the PC with Prime95 again, after having redone the TIM
and reseated the CPU - Same results; Prime95 (the latest version) can
make some cores climb to 94C for some tests; we're still hovering at
around 80C in OCCT's CPU test. One thing that HAS changed is that OCCT
crashes the first time I run it and start the CPU test. Run it again, no
matter how many times you like, and it works fine, but start it for the
first time that day and it crashes. So the situation is actually worse
than before. I'm not going to take the cooler off and reseat the CPU
again tho, unless I find some other symptoms/programs crashing
(Thunderbird did start crashing randomly last week but I re-installed it
and it's been doing fine since). I will re-install OCCT and see what
that does; maybe I will have to increase base voltage from 1.30 to 1.35
to fix this; I shoulda left well enough alone, it was rock solid before!

Anyway, except for stability issues, I'm giving up on this strange
cpu/heatsink combo, where the HS base doesn't really get hot even when
the CPU is running 90C (it becomes "warm" only). Looks like it is how
things are - unless I run a water cooler, I must expect very high temps
in Prime95.

Best Regards,

--
! _\|/_ Sylvain /
! (o o) Memberavid-Suzuki-Fdn/EFF/Red+Cross/SPCA/Planetary-Society
oO-( )-Oo Double your drive space! Delete Windoze!

Once upon a time on usenet B00ze wrote:
On 2017-02-10 22:42, B00ze wrote:

So I took the plunge - I had to open the desktop and move some hard
drives around, so I took the opportunity to remove the CPU heatsink
and have a look: Good news; I had, as I expected, just a tiny tiny
tiny bit too much TIM and the CPU PCB was NOT bent. Cleaned-up with
100% alcohol (NOT rubbing alcohol) and re-did the TIM and put the
heatsink back (pretty much with the same amount of TIM). Since this
is Arctic Silver 5, I need to wait a couple days for it to settle
before I test again, but I ran OCCT real quick and I get the same
temps as before (70-80 degrees under the CPU stress test (NOT the
Linpack)). We'll see in a few days, I will report back.

Alright, I tested the PC with Prime95 again, after having redone the
TIM and reseated the CPU - Same results; Prime95 (the latest
version) can make some cores climb to 94C for some tests; we're still
hovering at around 80C in OCCT's CPU test. One thing that HAS changed
is that OCCT crashes the first time I run it and start the CPU test.
Run it again, no matter how many times you like, and it works fine,
but start it for the first time that day and it crashes. So the
situation is actually worse than before. I'm not going to take the
cooler off and reseat the CPU again tho, unless I find some other
symptoms/programs crashing (Thunderbird did start crashing randomly
last week but I re-installed it and it's been doing fine since). I
will re-install OCCT and see what that does; maybe I will have to
increase base voltage from 1.30 to 1.35 to fix this; I shoulda left
well enough alone, it was rock solid before!
Anyway, except for stability issues, I'm giving up on this strange
cpu/heatsink combo, where the HS base doesn't really get hot even when
the CPU is running 90C (it becomes "warm" only). Looks like it is how
things are - unless I run a water cooler, I must expect very high
temps in Prime95.

I haven't been following your progress but from that last bit you wrote it
sounds to me like the factory TIM between the CPU proper and its IHS
(Integrated Heat Spreader) could be faulty. Apparently it's not that
uncommon a thing...
--
Shaun.

"Humans will have advanced a long, long way when religious belief has a cozy
little classification in the DSM*."
David Melville (in r.a.s.f1)
(*Diagnostic and Statistical Manual of Mental Disorders)

On 2017-04-06 23:03, ~misfit~ wrote:

Alright, I tested the PC with Prime95 again, after having redone the
TIM and reseated the CPU - Same results; Prime95 (the latest
version) can make some cores climb to 94C for some tests; we're still
hovering at around 80C in OCCT's CPU test. One thing that HAS changed
is that OCCT crashes the first time I run it and start the CPU test.
Run it again, no matter how many times you like, and it works fine,
but start it for the first time that day and it crashes. So the
situation is actually worse than before. [...]

I haven't been following your progress but from that last bit you wrote it
sounds to me like the factory TIM between the CPU proper and its IHS
(Integrated Heat Spreader) could be faulty. Apparently it's not that
uncommon a thing...

Quite possible. For CERTAIN de-lidding would make this CPU run much
cooler, but I'm not going there lol. Thunderbird crashed again upon
starting by the way, so I now had a few crashes of applications "upon
starting" which told me LLC was not boosting voltage fast enough. So I
changed by base voltage from 1.300 - 1.310 (and kept LLC @ Level 6).
This has the effect that unless the CPU is doing absolutely nothing (and
running Windows is /something/) then it runs at 1.328v (instead of
1.312v as before). So far so good. The downside of course is that if I
run Prime95 v28 small FFT I now see a Max Temp of 97C because the volts
boost-up to 1.360v. I installed the recommended Prime 95 (v26) and I
stay around 80C on small FFT's - v28 uses FMA3/Avx and together with a
high LLC it means a lot of energy is used. I tried LLC @ 5 but then LLC
*lowers* the voltage when the CPU gets really busy. That's what LLC is
supposed to be for actually, but it doesn't work when you overclock.

I agree with you my CPU sample isn't great. The best overclock I get
without boosting volts too much, like I am doing now, is 4400MHz. At
that speed all I need is 1.260v base volts (an increase of 0.010v) and
LLC @ 7. Getting 4500MHz stable requires a lot more power.

If I become unstable as the years go by, or if I discover that I can't
encode a movie or play a game without the CPU climbing to 95C then I
will just fall back to 4400MHz.

Best Regards,

--
! _\|/_ Sylvain /
! (o o) Memberavid-Suzuki-Fdn/EFF/Red+Cross/SPCA/Planetary-Society
oO-( )-Oo Soon to be a major motion picture!