星期三, 1月 04, 2012

Pre-test of Latest Intel Top-Level 4-Cores CPU i7-2700K



Intel
released the new generation framework of Sandy Bridge at early January, 2011;
meanwhile, LGA 1155 is provided for support.
Compared with the
specifications of several platforms in the past, the LGA 1155 is provided with a
lot of modifications in CPU overclocking (OC) at this time.
Firstly, when a
CPU with the external frequency of 100 MHz is overclocking, the other
surrounding clock rates will be increased accordingly, thus the upper limit will
fall around 110 MHz.
Although the new 32nm process is introduced into the
new generation 4 cores CPUs, overclocking can be only achieved by using
frequency multiplication.
The model names of i5-2500K and i7-2600K that are
two kinds of CPUs without lock frequency multiplication have been released for a
long time, which are considered as the representatives of the overclocking
line.
However, the overclocking can be only done by working with the chipset
of overclocking frequency multiplication, i.e., P67 or Z68.

The
overclocking environment for LGA 1155 is simply declared above, hoping the users
can aware of the hardware for the equipment clearly.
The middle-level CPU
featured by overclocking is Core i5-2500K, while the advanced-level is Core
i7-2600K.
The main difference between them is that the 2600K is provided with
Hyper-Threading (HT) technology and an 8MB L3 cache, while the 2500K is only
provided 6 MB L3 cache without HT.
This time I pursue Core i7-2700K which
will appear on the market very soon and replace the advanced 2600 K.

The
total clock rate of 2700K is 3.5 GHz, which supports the new generation Turbo
Boost 2.0 and can reach the maximum performance of 3.9 GHz.
The real 4 Cores
are incorporated into the Hyper-Threading technology, so that eight threading
can be executed, 4C/8T for short.
With the 32 nm process, the power
consumption of 95W and the 8-MB L3 Cache, the 2700K, surpassing the 2600K
100MHz, becomes the most advanced CPU of LGA 1155.


The
rear side of the 2700K is still D2 Revision, so that there is little difference
in appearance between the 2700K and the 2600K at this part.


The MB
is the latest Z68 high-level chipset of LGA 1155, i.e., the latest version
TZ68K+ released by BIOSTAR.
The appearance and the specification are the same
as that of the TZ68A+ shared before and the primary difference is supply voltage
of CPU.


Although
the TZ68K+ is the new edition of BIOSTAR Z68, the price is still about US$99,
thus it can challenge the low-price market of ATX Z68.
By using the
well-marked red-black matching, it has the most specifications as that of the
Z68 series with the built-in Power/Reset button and a simple debugging
LED.


With
respect to IO, there are three types among which the two blue USBs 3.0 can
provide high-speed transmission.
If provided two more USBs 2.0, it would be
more perfect with respect to the scalability.


The
previous TZ68A+ is designed with four-phase power supply only, while the new
edition TZ68K+ is provided with eight-phase power supply, which is the primary
difference.
However, as mentioned before, one-phase power supply can provide
about 30-40 W, so , the four-phase power supply is already quite enough for
overclocking of power consumption of Intel CPU.


The
heat sinks above the Z68 chipset look good because of being cut by a special
manner.
Two native SATA3 units of Z68 are positioned at the right side, which
are used at a general environment or in a Raid0 state.


Test
platform
CPU: Intel Core i7-2700K
MB: BIOSTAR TSERIES TZ68K+
DRAM:
CORSAIR CMZ8GX3M2A1866C9R
VGA: Intel HD Graphics 3000
HD: Intel 510 Series
120GB
POWER: Thermaltake TR2 450W
Cooler: CORSAIR Hydro Series H80
OS:
Windows7 Ultimate 64bit


Firstly,
a performance test is performed by using the preset value of CPU.
Preset the
performance
CPU 100 X 35 => 3500 MHz (booting Turbo Boost, and then
C1E)
DDR3 1599.8 CL7 8-7-22 1T

Hyper PI 32M X8 => 15m
21.977s
CPUMARK 99 => 597


Nuclearus
Multi Core => 23908
Fritz Chess Benchmark => 27.86/13374


CrystalMark
2004R3 => 264699


CINEBENCH
R11.5
CPU => 6.93 pts
CPU(Single Core) => 1.57 pts


Experience
index of Windows- CPU 7.6



PCMark
Vantage => 19386


The
frequency multiplication of the 2700K in the mode of Turbo Boost 2.0 is 39, 38,
37 and 36 respectively; and the basic clock rate is 3.5 GHz.
The above
specification is 100 MHz higher than that of the 2600 K, the performance of this
part is presented in the test, but the improvement is not much.
Practically,
both the 2600K and the 2700K are the highest-level CPUs in the framework of the
Sandy Bridge, and the data obtained without overclocking is fairly high
already.
For a general use environment, most of the software requirements can
be met by using the high performance preset in 2700K.

DRAM bandwidth
test

DDR3 1599.8 CL7 8-7-22 1T
ADIA64 Memory Read - 19618
MB/s
Sandra Memory Bandwidth - 21335 MB/s
MaXXMEM Memory-Copy - 21168
MB/s


The
DDR3bandwidth is much higher than that of our own platform or a competitor’s
platform, which is the advantage of the improved LGA 1155.
The high
transmission performance of nearly 20000 MB/s can be achieved only by using DDR3
1600, and the voltage is lower than that for the previous
platforms.

Temperature performance (about 26 DEG C at room
temperature)
System standby - 23~33


CPU
full speed - 40~46
Intel Burn Test v2.4,Stress Level Maximum


The
temperature performance when system standby is good by enabling the
energy-saving function of C1E.
With the latest Hydro Series H80 of CORSAIR
and the high thermal diffusivity of the hydro-cooling framework, the temperature
is only 46 DEG C at full speed.
Only by working with excellent air-cooling or
hydro-cooling equipment, the state of the preset value on 2700K can still be
lowered to a low temperature.

Power consumption test
No software
running on the OS desktop - 45W


CPU
full speed- 103W


Due
to introduction of 32nm, the power consumption is much lower than that of the
previous generation Core i7-870 with the process of 45nm.
The power
consumption of the clock rate at the environment of C1E is only 58W higher than
that of the preset maximum clock rate.
Based on several articles about AMD
APU shared by me recently, in a test with the same 32 nm and 4Cores, the power
consumption of the APU platform is nearly twice that of AMD.
It is thus clear
that the performance and the power consumption under the Intel’s framework have
good advantages under the same process.

Performance test for Intel HD
Graphics HD3000
The preset value of GPU at BIOS is 30 which is equal to 1500
MHz after conversion.

3DMark Vantage => P2500


StreetFighter
IV Benchmark
1920 X 1080 => 30.58 FPS


In
the Sandy Bridge, two GPUs, i.e., HD2000/HD3000 are built in by Intel, thus the
performance is much advanced than the previous edition.
The detailed test
data are shared in several related articles shared by me recently
(windwithme).
The HD3000 has about 6-7 times 3D performance of the previous
platform built-in GPU of Intel, thus it is believed that the applicable range is
wider.

The following part is about sharing of the overclocking for
2700K.
New-style UEFI graphical interfaces are introduced into the BIOS items
by BIOSTAR.


O.N.E
debugging page
By changing 2700K at Turbo Mode to 50, the clock rate, i.e.,
100 X 50 => 5GHz, will be executed.
The Internal PLL Voltage Override is a
special Intel overclock option, which effectively strengthens the stability of
the CPU after substantially overclocking.
It is suggested that the C1E should
be powered off when overclocking, thereby facilitating control of clock rate and
voltage of the CPU.


When
setting clock rate and parameters for DRAM, the smaller the numerical values
are, the higher the performance is.
The following example is DDR3 1866 CL8
10-9-27 1T, and the set values for the other detailed options are adjusted for
optimization.
User can perform miniature adjustment on the clock rate or the
parameters according to the condition of the DDR3 on hand, so as to increase the
performance of the DRAM.


The
CPU voltage adjustment is one of the key points for overclocking.
It can be
adjusted according to the condition of CPU and the heat-dissipation system of
the platform; and the CPU VCoce LoadLine is Enabled.


The
DRAM voltage is another option to be adjusted for
overclocking.
Substantially, the overclocking at the Sandy Bridge platform is
much simpler than that of the previous platforms,
so the overclocking can be
realized by grasping CPU frequency multiplication, DDR3 parameters, and voltage
adjustment for both.


PC
Health Status
It’s still felt that the temperature test result for CPU is on
the much higher side;
actually, the wind temperature of the heat sinks or
the temperature of the software for measuring temperature are much
lower.


The
above are personally stable set values for 2700K OC 5GHz, which may be
readjusted for overclocking according to different hardware or difference
between conditions individually.
However, the general direction is at least
right, which can be considered as reference for setting BIOS for the
overclocking users.

Overclocking performance
CPU 100 X 50 =>
4998.9MHz, at full speed is 1.416V.
DDR3 1866.4 CL8 10-9-27 1T
1.600V

Hyper PI 32M X 4 => 11m 52.875s
CPUMARK 99 => 771


Nuclearus
Multi Core => 32119
Fritz Chess Benchmark => 38.95/18694


CrystalMark
2004R3 => 341127


CINEBENCH
R11.5
CPU => 9.74 pts
CPU(Single Core) => 2.03 pts


Windows
experience index - CPU 7.8



PCMark
Vantage => 24126


After
overclocking, the performance is increased by about 30% at the single-core
threading, and by over 40% at the multi-core threading, which is quite
satisfactory.
The 5GHz which is the performance of 2700K should be the acme
of the 4 Cores CPU on the market currently.
In the future, it may be
surpassed by Sandy Bridge-E or Ivy Bridge to be released next year which uses
its own LGA 2011 platform.

DRAM bandwidth test
DDR3 1866.4 CL8 10-9-27
1T
ADIA64 Memory Read - 22805 MB/s
Sandra Memory Bandwidth - 24680
MB/s
MaXXMEM Memory-Copy - 25963 MB/s


The
Sandy Bridge framework improves the bandwidth of DRAM a lot, thus a dual-channel
platform can already compete with the performance of a tri-channel platform of
X58.
Furthermore, the bandwidth improvement for DDR3 is in clock rate, there
is 10-20% bandwidth improvement after being increased from 1600 to 1866.
The
high bandwidth performance for LGA 1155 is helpful to improvement for system
performance to some extent.


Temperature performance (about 26 DEG C
at room temperature)
System standby- 36~42


CPU
full speed - 76~80
Intel Burn Test v2.4,Stress Level Maximum


The
temperature increases a little when system standby because of overclocking,
which is substantially within an acceptable range.
The temperature increases
a lot at full speed, which will also happen to the Sandy Bridge framework at the
high clock rate of 5 GHz.
Therefore, it is quite important to lower the
temperature which is also a challenge for the capability of the heat sinks.

The temperature is about 90 DEG C when top-level air cooling is employed
previously.
The hydro-cooling performance of CORSAIR H80 is excellent, in
which the temperature is below 80.
It is an acceptable overclocking
temperature for most users.
For BIOSTAR, the voltage control of CPU can be
strengthened, 1.512V at standby and 1.416V at full speed are helpful to
stability after overclocking, but it fluctuates strongly.

Power
consumption test
No software running on OS desktop - 103W


CPU
full speed- 239W


After
powering off the C1E, the power consumption after overclocking is increased a
lot simultaneously, which it is over 100W at standby.
The power consumption
will be increased by about 136 W at full speed.
From the clock rate and
performance of 5 GHz, the power consumption is a little higher than 95 W which
is the official preset data.
However, compared with the high power
consumption of competitor’s 32 nm 4Cores/8Cores, the power consumption for 32nm
process of Intel is excellent.


It is necessary to select two
chipsets, i.e., P67/Z68 and the like to play the overclocking performance of the
three frequency-unlocked CPU, i.e., 2500K、2600K and latest 2700K.
P67 has
lowered to a reasonable price on the market, so that it is a MB choice with
light economic burden for consumers with limited budgets.
But Z68 has the
primary functions of two chipsets at the same time, i.e., overclocking frequency
multiplication of CPU of P67 and display output of H67,
so that it is the
chipset of LGA 1155 that has the most comprehensive functions
currently.
Unlike LGA 1156 or LGA 775 at a quite high price when just
released in one year, Z68 of the LGA 1155 platform soon after being put on the
market has a reasonable price.
Additionally, Z68 has exclusive SSD
acceleration, so that it is a good function for accelerating the performance of
system for the users with small-capacity SSD on hands.
If selected by myself,
I prefer to directly selecting Z68 MB optimally through adding some
budgets.



The
price of TZ68K+ of BIOSTAR is about the same as that of P67 of the other brands;
from this part, the value of C/P is obviously enhanced.
With the overclocking
capability of a rather high level and the advantages of supporting UEFI user
interface and eight-phase power supply, the Z68 should be taken into
account.
If the number of USB 2.0 can be increased or the voltage of CPU can
be suppressed, it will be a valuable high-performance Z68 MB.

Intel will
release 2700K in Q4, which should be used to replace 2600K that is the
highest-level CPU of LGA 1155 currently.
Although, 2700K is only increased by
100 MHz, the price is only tens of dollars higher according to information on
Internet.
The voltage for overclocking at 5GHz is about the same as that for
2600K which is shared by windwithme, and is about 1.4V.
At Nov., Intel will
release a higher level LGA 2011 platform, in which the configuration of X79 and
CPU will be much more expensive than that of 2700K and Z68.
On the current
market, the middle-high level platform with higher value of C/P consists of
several combinations, i.e., Z68+2500K or 2600K/2700K.
Personally, I think it
is unnecessary for the users who already have 2600K to upgrade, and it would be
better for those who want to buy 2600K recently to wait the market information
of 2700K.
The above is about sharing my experience for overclocking of
i7-2700K, which is provided for users in need as references.

沒有留言: