|
Intel Core 2 Quad Q6600 |
|
|
|
Written by Nedeljko Kovacevic
|
|
Thursday, 07 December 2006 |
|
Page 3 of 4 The writer of this article must admit being fulfilled with thrill familiar to any hardware enthusiast, when obtaining one new, very powerful toy. Under such adrenaline rush, we very hastily assembled our test platform, irrationally forgetting the fact that the quantity of thermal paste on our test cooler was everything but significant and even more important „fresh“. Positive side of these circumstances was that we learnt that Kentsfield can work trouble-free even on temperature of 90 degrees. Clearly, when we realized this high temperature, we very hastily disassembled the cooler and slowly according to accustomed manner regulated the issue of thermal paste. With fresh paste on the processor and clean cooler bottom, the temperature during the next starting of Windows was 50 degrees – everything was prepared for beginning of the test.
Synthetic tests in Sandra clearly indicated that, comparing to E6600, Q6600 (mark of one out of two Core 2 Quad processor models) has double raw processor power. This is evidence of irrelevancy attributed to disputes regarding nativeness of Intel’s Quad Core processor. But, Sandra proved something else – drop of anyway poor memory subsystem performances. Although the difference comparing to Dual Core is obviously small, it has been detected and reflects drawbacks in Intel’s approach to inter-core communication relying on QPB for processor’s mutual communication, simultaneously used for communication with memory controller placed in chipsets NorthBridge. Nuclearus, as well portraits all benefits of double-sized processor. Findings obtained through previous two synthetic test programs are illustrated in real application using program for 3D rendering. Replacing Conroe with Kentfield, Cinebench was carried away with enthusiasm. The same happened with packaging used by professionals in domain of video editing. Adobe’s Aftereffect can be very advantageous on additional cores. Unfortunately, the same cannot be stated for Photoshop CS2, which recognizes only one core, out of four.
| 1024x768 | Core 2 Duo E6600 | Core 2 Quad Q6600 | Core 2 Extreme QX6700 | | Sandra 2007 | | | | | Processor (ALU/SSE3) | 22.086 / 15.354 | 44.240 / 30.704 | 49.090 / 32.855 | | CPU-Multimedia (Int/Float) | 131.762 / 71.536 | 263.105 / 142.880 | 292.436 / 158.281 | | Memory (MB/s) | 5.194 / 5.178 | 5.140 / 5.109 | 5.205 / 5.175 | | NucleaRUS (more is better) | 7.261 | 12.138 | 12.811 | | Cinebench 9.5 (more is better) | 725 | 1.280 | 1.337 | | WinRAR 3.60 (more is better) | 969 KB/s | 809 KB/s | 812 KB/s | | Photoshop CS2 (less is better) | 44,1 sec | 42,9 sec | 39,6 sec | | Afterefects 7.0 (less is better) | 229 sec | 145 sec | 134 sec | | 3DMark06 Final / CPU Score | 10.500 / 2.085 | 12407 / 3694 | 12.768 / 3.914 | | AquaMark 3(more is better) | 132.083 | 143.332 | 143.381 | | FarCry (MediumQ) | 168,7 fps | 185,2 fps | 204,1 fps | | Quake4 (MediumQ) | 121,6 fps | 129,2 fps | 131,2 fps | | DivX 6.2.5 (more is better) | 123,6 fps | 148,6 fps | 149,2 fps | | WinRAR+DivX+ | | | | | +Photoshop CS2 (less is better) | 75,3 sec | 70,5 sec | 67,1 sec | | +Quake 4 (more is better) | 51,5 fps | 40,1 fps | 44,2 fps | | Cena (€) | 391 | uzorak | uzorak | | | | | | Testbed | ASUS P5N32-SLI Premium, Transcend DDR2-667 (CL5-5-5-15), ASUS GeForce 8800GTX, WinXP SP2 |
|
But, additional quantity of cache memory provides slightly better results for Q6600 comparing to E6600. Namely, Core architecture uses shared secondary cache; therefore in Kentsfield´s example one core has 8 MB of cache memory at disposal. This advantage was also used by FarCry, which cannot recognize more than one core as well. In certain scope Quake 4 can use two cores, but this is as far as it goes; therefore, it is fair to say that it was faster due to larger quantity of cache memory at disposal. Interesting information that solely test application, beside programs for 3D rendering and video editing, which can one-hundred percent deploy all four cores is WinRAR. We reach special paradox and situations clearly portraying drawbacks of „non-native“ QuadCore architecture. Kentsfield provides poorer results in WinRAR comparing to Conroe due to jam in „Quad Pumped BUS“, created during ultimate (one-hundred percent of WinRAR´s occupation of all four cores, with detached multithreading support) load of all four cores. WinRaR „syndrome“ is more distinguished, because this test fairly relies on memory subsystem, which is an “Achilles´ heel” of Intel’s processors.
Consequently, Kentsfield „chokes a bit“ in one out of two our multitasking scenarios. Test is rather simple, but illustrative. WinRAR runs, then DivX compression, Photoshop in first and Quake 4 in second scenario. Goal of first scenario is to evaluate conformity during serious operating, and second during playing. Measured parameters were time necessary for performing series of filters in Photoshop and number of frames in test demo of Quake 4. Photoshop provides better Kentsfield performances comparing to Conroe. Those are not large, but again, Photoshop anyway recognizes and uses only one core. Quake 4 provides inverse situation and Conroe comes out as victor comparing to its double-size brother. We call attention that Quake 4 can use other processor, however following up of processor’s load in task manager we observed that it never occurred in any testing phase. WinRAR preserved dominance over two cores, and DivX compression (designed in XMPEG) and Quake 4 struggled for capacity of the remaining two. It was Quake 4 which always obtained less (approximately around 80% of one core). In this event, combination of not quite ultimate Windows XP operative system thread management and non-optimized multithread applications additionally exposed aforesaid Achilles’ heel of Kentsfield, which leads us to the final impression.
|
|
Reviews 