Home  |  News  |  Reviews  | About Search :  HardWare.fr 



ProcessorsMotherboardsGraphics CardsMiscellaneousHard DisksSSD
Advertise on BeHardware.com
Review index:
Report: The Intel Core i7, in practice
by Marc Prieur
Published on December 3, 2008

Power consumption
We suggest two parameters to measure power consumption. The first is power consumption of the whole platform, measured at the wall socket with the aid of a wattmeter. The second is the value sent by the power supply to the ATX12V, which is essentially confined to processor power supply.

Here to begin with are the platform power consumption results, with the 790GX for AMD with 2x2 GB of DDR2, the P45 (Socket 775) for Intel with 2x2 GB of DDR2 and the X58 for the Core i7 with 3x1 GB of DDR3. The rest of the set up was identical (4870 X2, Raptor 74, Raptor 150 GB, Audigy).


Although at rest power consumption was limited on the Core i7, although slightly higher than that of a QX9770, in load it went to new heights especially if the CPU is loaded with 8 threads. Of course, the Core i7 does then presumably give better performance but we musn’t lose sight of power consumption.


Measuring power consumption with the ATX12V should give a better indication of processor consumption. Given the figures obtained here on the Core i7 comprared to those of the platform however, it seems as if the “uncore” part of the processor is not supplied by the vCore dedicated ATX12V, but by the VTT which comes from the 3.3V of the standard ATX plug. For information, disactivation of the Turbo mode on the Core i7-920 gives a gain of 3.6w with 2 Prime, 9w with 4x Prime and 12.8w with 8 Prime.
Overclocking
Before anything else, we need to put a stop to two received ideas and rumours about the Core i7: firstly, it is possible to overclock standard versions and secondly, the 1.65V limit on memory voltages, beyond which there is a risk of curtailing the life of the CPU, is no longer valid on the definitive versions of the Core i7.

Like on the Core2, there are two ways of overclocking the Core i7: by increasing the base clock (133 MHz by default), or, only only on the Extreme, by increasing the multiplier in Turbo mode.

On the Extreme Edition, Turbo mode can be configured as you like, whether in terms of maximum power consumption of the processor or the multipliers. This is the simplest method for overclocking the CPU, even if you have to raise the maxiumum power consumption to a ridiculous level so as to shortcircuit this protection.

In any case, increasing base clock frequency is always possible. Careful however, in addition to the frequency of CPU cores, the base clock is also used as the basis for QPI frequency, as well as for the “Uncore” part of the CPU and as well as for memory frequency: when you increase it then, you sometimes need to lower the multipliers used to obtain each of these frequencies so as to retain platform stability.


Here we have increased the base clock to overclock our Core i7-965. Without increasing the vCore, we were able to reach 3.6 GHz (24x150) instead of the default 3.2 GHz, then 4 GHz (24x167) by increasing voltage by 200 mV. Our Core i720 did not want to go beyond 20x175 (3.5 GHz) stable however, whatever the voltage.
The test
We now move on to more classic applied tests which aim to compare the Core i7s to their competitors. This article inaugurates our new processor test protocol.

3ds studio max has been moved up from the two year-old version 9 to the brand new 2009 version and Maya, from the same publisher, has now made way for Cinema 4D version R11. We are now including a test on a compilation, that of the MAME emulator source under MinGW. We still do a file compression but move up from version 3.7 of WinRAR to version 3.8.

DivX compression is also still on the agenda, with version 6.8 of the codec and maximum quality parameters. TMPGEnc and MPEG-2 encoding have been replaced by H.264 encoding under AutoMKV and the free codec x264. After Effects is back, still in version CS3, version 4 no longer available. Nuendo 3 has been replaced by version 4, while Crysis patch 1.2 has been installed, allowing us to use the new built-in CPU benchmark which is even heavier. We have kept World In Conflict whose built-in benchmark is also currently one of the heaviest.

The test configurations are as follows:

- ASUSTeK P5QC + 2x2 GB DDR2-1066 5-5-5 (LGA775)
- ASUSTeK M3A78-T + 2x2 GB DDR2-800 4-4-4 (AM2+)
- Intel DX58SO + 3x1 GB DDR3-1067 (LGA1360)
- GeForce GTX 280 + GeForce 180.84
- Raptor 74 GB + Raptor 150 GB
- Creative Audigy
- Windows Vista 32 bits SP1


<< Previous page
Vista x64 & EM64T

Page index
1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12
Next page >>
3D Studio Max 2009, Cinema 4D R11  




Copyright © 1997- Hardware.fr SARL. All rights reserved.
Read our privacy guidelines.