|
 |
Geek warning!!!
This weekend, I decided to give an honest attempt to see how much I could
overclock my new cpu. For reference, the core i7-860 stock speed is 2.8GHz.
Long story short, I was able to hit speeds of 4.4GHz and it would still
post, but at that point, I had the voltage set as high as it Intel says to
go (1.4V) and it would have run so hot under load that there was no way it
could be useful, so it was more of a "Let's see how high I can go!" setup.
Ok, before I go further, I'll have to explain a bit. The 4.4GHz was really
a 4.2GHz overclock with "turbo boost" or whatever the heck it's called.
Apparently, the OS can change the clock speed dynamically, alternating
between a multiplier of 9 and 22. The 860 is normally set at 21. At lower
speeds, I actually ran some tests and with a simple POV-Ray scene (low
memory usage), there was no difference between running at, say, 4GHz with
turbo boost turned off (21x mult), and 4GHz (22x mult, 3.8GHz with 21x mult)
with turbo boost on. i.e. 4GHz is 4GHz regardless of how you got there.
I was able to achieve 4.2GHz, but heat became a serious problem. I was
using Prime95, and temps were approaching 80C. I turned away for a couple
minutes, then Prime95 started the next test and YIKES! Temps were in the
mid-80's. STOP! STOP! You're not supposed to take the chip above 78C
according to what I've read. But it otherwise seemed stable. I tried a
more powerful CPU fan, but it didn't seem to make any difference. I think
you can only conduct so much heat. At some point, I think it doesn't matter
how big the fan is. I ran my POV-Ray benchmark (not The Benchmark) scene at
3:55. That was the lowest I ever saw. Any higher speeds would crash
Windows.
Ok, so I tried 4GHz. Still too hot. Core temps would occasionally spike
into the low 80's under full load and that's just not cool (pun intended).
So more tweeking and lowering voltages, and I found that 3.88GHz (incl.
turbo boost) seemed to be the highest stable speed I could get. I ran
Prime95 with maximum heat stress, and it would run (core temp) 75-76C.
That's pretty warm, but still under the reference limit of 78C. I ran
Prime95 for over 4 hours while occasionally playing Quake 3 and solitaire
too, and it worked well. My POV-Ray benchmark ran at 4:20 and temps never
went above 70C.
Ok, 76C core temp may not actually be 76C. One of the cores, Core #0, seems
to run as much as 5C hotter than the rest at all times. Apparently, that's
normal, and some have speculated that it's Intel's way of scaring people out
of overclocking to high, or perhaps to cause a thermal shutdown.
Regardless, if Core #0 hit a max of 76C, the other cores were between 71C
and 73C. Idle temps were always 20's and 30's, which is really cool.
Summary: 4.4GHz would post, but couldn't load Windows. 4.2GHz might have
been stable, but way too hot. 4GHz seemed stable but still too hot.
3.88GHz, stable, and warm. 3.5GHz very cool, very stable.
Post a reply to this message
|
|
|
|
"Jeremy "UncleHoot" Praay" <jer### [at] questsoftware cmo> wrote in message
news:4b8bc8f9@news.povray.org...
> Ok, before I go further, I'll have to explain a bit. The 4.4GHz was
> really a 4.2GHz overclock with "turbo boost" or whatever the heck it's
> called.
Woops. "Turbo Boost" is something else, which I have turned off. I think
it's "Speed Step" that automatically jumps to the higher clock multiplier.
Turbo Boost is apparently for single threaded apps, so that they can get
more power from one core, and keeping the other 3 essentially turned off.
At stock voltages, turbo boost would take a 2.8GHz system and turn up the
clock on one of the cores to reach 3.5GHz or so. I haven't played around
with it, though. Maybe later.
Post a reply to this message
|
|
|
|
On Mon, 01 Mar 2010 16:52:36 +0100, Warp <war### [at] tagpovrayorg> wrote:
> "Jeremy \"UncleHoot\" Praay" <jer### [at] questsoftwarecmo> wrote:
>> I was able to achieve 4.2GHz, but heat became a serious problem. I was
>> using Prime95, and temps were approaching 80C.
>
> I thought modern CPUs automatically throttle to lower clock frequencies
> when they get too hot.
They do, but the trigger temperatures are higher than that. I think
The main reason for keeping the temperature below a much lower threshold
even the parts of the CPU that are not close to a thermal sensor stay
> Or does overclocking mean that you are actually bypassing that protection
> mechanism and forcing the CPU to run at a given frequency regardless of
> temperature?
No, but there are other things that can cause a CPU to fail (or even
outright destroy it) when over-clocking, e.g.:
- Local overheating distant from a thermal sensor.
- Voltage drops.
- Excessive leakage current.
--
FE
Post a reply to this message
|
|
|
|
"Warp" <war### [at] tagpovrayorg> wrote in message
news:4b8be2c4@news.povray.org...
> "Jeremy \"UncleHoot\" Praay" <jer### [at] questsoftwarecmo> wrote:
>> I was able to achieve 4.2GHz, but heat became a serious problem. I was
>> using Prime95, and temps were approaching 80C.
>
> I thought modern CPUs automatically throttle to lower clock frequencies
> when they get too hot.
>
> Or does overclocking mean that you are actually bypassing that protection
> mechanism and forcing the CPU to run at a given frequency regardless of
> temperature?
>
Good question. I used "Core Temp" which allows you the option of seeing the
CPU frequency in real-time. Even though it was getting hot, the
CPU-frequency wasn't changing under full load.
Sitting idle, the computer would drop down to a 9x multiplier and would
operate at less than half speed. With the 3.88GHz OC, it was actually only
running at about 1.6GHz while sitting idle and consuming only about 100W
with the monitor off. For me, that's about a penny an hour sitting idle. I
have an external device that measures voltage, watts, amps, etc. At full
load, it was between 250 and 300+ Watts.
Post a reply to this message
|
|
