Overclocking Results
Our Overclocking ResultsNow
that everything has been defined and we know all of the tools we have
available to us for overclocking these processors, what kind of results
did we have?
For my testing, the overclocking scenario was actually quite simple: raise the Turbo Ratio Limits higher on the Core i7-965 EE processor above the 25x and 26x defaults. I started out with a 30x multiplier across all four core settings giving me a theoretical frequency limit of 4.0 GHz. The results were actually pretty impressive:
For my testing, the overclocking scenario was actually quite simple: raise the Turbo Ratio Limits higher on the Core i7-965 EE processor above the 25x and 26x defaults. I started out with a 30x multiplier across all four core settings giving me a theoretical frequency limit of 4.0 GHz. The results were actually pretty impressive:
1 core being loaded
When
loading just a single of the Core i7’s quad cores (using CineBench 10
and manually setting processor affinity) the clock rate of the CPU was
hitting the 4.0 GHz mark almost immediately and consistently. Our
system was effectively overclocked to 4.0 GHz for single threaded
applications with raising voltages and simply adjusting the Turbo
Ratios.
All 4 cores being loaded
When
all four cores were being loaded, the results weren’t quite the same.
Rather than seeing the CPU quickly peg up to 4.0 GHz over the default
3.2 GHz clock speed, I instead saw the frequency fluctuating quite a
bit between the 4.0 GHz speed and 3.2 GHz or so. In this screen shot
of CPU-Z above, you can see at this instant I was running at a 26.5x
multiplier resulting in a 3.5 GHz clock rate. Sometimes this clock
rate would in fact hit 4.0 GHz but it wouldn’t stay there reliably.
Obviously the PCU on the processor die was deciding that the heat and
power being used by the CPU was too high and kept the system from
staying at the 4.0 GHz overclock that the Turbo ratios enabled.
All 4 cores being loaded with overclocked system clock
Next
I attempted to push up the host system clock above the 133 MHz default
– hitting 155 MHz was easy but seemed to be an upper limit for our
testing. I kept the same 30x Turbo Ratios this time and saw similar
results to what I described above. The clock rate varied quite a bit
when all four cores were loaded but were stable when loading with a
single thread only.
You
can see in this screenshot that the memory speeds are indeed affected
when the host clock is increased. With an 8x multiplier shown above
and the 155 MHz system bus we were using we get the 1240 MHz DDR3
memory speed shown here.
The result of all of this is kind of what I expected after initially hearing about Core i7’s Turbo Mode at IDF earlier this year. Basically, single threaded performance with this processor will be faster than many-threaded performance relative to the existing Core 2 products. That is possible because Turbo Mode allows the CPU to automatically overclock more when only 1 or 2 cores of the CPU are being stressed.
This sample CineBench 10 result above shows this to be the case: compared to the standard scores from our earlier testing of the Core i7-965 processor the multi-threaded score is only 0.4% faster while the single threaded score is 16% faster!
Obviously overclocking with Nehalem is a different sort of beast but in my opinion it is a better kind of beast. Intel has done a good job of offering flexible options for hardcore overclockers while offering better and easier overclocking options for those beginners looking to just push a little more out of their investment.
this very interesting
this very interesting information I’ve learned so much from intel over the years i been studying computers alot its ny passion and i will continue to full fill my dreams
Grammar and punctuation are
Grammar and punctuation are important things to know too.