Intel^® Core™2 Extreme Quad-Core Processor

 

Core™2 Extreme Processor Lineup:
 
The Core™ 2 Extreme processor family just grew by one member as the new Core™ 2 Extreme QX6700 Quad-Core was added last month. The new QX6700 Quad-Core operates 300MHz less than the X6800 Dual-Core Extreme, but since the clock multiplier isn't locked in this family you can very easily overclock the 2.66GHz QX6700 by 10% to 2.93GHz without any problems to match the raw speed of the X6800. Besides being able to run 4 concurrent software threads the QX6700 Quad processor differentiates itself by having an 8MB L2 cache (4MB x 2) and a 128KB L1 cache (64KB x 2). It still uses Intel's 65nm manufacturing technology and the stellar Core™ micro-architecture, which we covered in the September ASI Technical Newsletter.

The QX6700 is packed with 582 million transistors into a die size of approximately 286mm², double that of the Core 2 Duo E6000 "Conroe" series. Like the previous Core™ 2 Extreme x6800 processor, the QX6700 utilizes a 1066 MHz front-side bus, comes in the LGA775 package and supports DDR2-800 memory. The QX6700 processor voltage draw ranges from 1.100V to 1.372V, is rated at 65W thermal design power (TDP), and has a thermal specification of 130W, or the amount of heat that needs to be dissipated by the cooling system.

ASI SKU	Processor Number	Clock Speed	Cache Size L2	Front Side Bus	Quad Core	Intel® VT	Enhanced Intel SpeedStep® Technology	Intel® 64	Execute Disable Bit	sSpec#
54454	QX6700	2.66 GHz	8MB	1066 MHz						SL9UL
50918	X6800	2.93 GHz	4MB	1066 MHz						SL9S5

 
Multimedia Performance:

The Core™ 2 Extreme QX6700 Quad-Core processor makes significant improvements to overall system performance by offloading certain tasks or threads to one of the four specific cores to help get more done in less time. Today's multimedia applications such as video & audio editing, graphics rendering, and 3D modeling take advantage of multi-threading performance and some even demonstrate significant scalability with these Quad-Core processors. If you're in the business of professional content creation, purchasing one of these processors is almost a no-brainer. In the testing results show below and reported at other review web sites, the primary difference between two and four cores is in the sheer amount of work that got done. 3D renderers like Autodesk's 3ds Max, absolutely love more processing cores, as do popular applications such as Photoshop CS2 and Lightwave 9.

The Cinebench 9.5 test is also a multi-threaded 3D rendering benchmark that takes advantage of any and all available processing cores and performance numbers posted by Anandtech show performance gains of over 60% when moving from two to four cores. But one of the more interesting findings reported is with the QX6700 being a more efficient overall CPU with higher performance per watt numbers as compared to any "Conroe" processor, even though Kentsfield consumes twice the amount of power to operate. 

Courtesy of AnandTech, the graph below illustrates how having more cores does increase efficiency if the software is designed to take advantage of those additional cores. The point of diminishing returns hasn't been reached with adding more cores, but the two downward trending curves for Quicktime H.264 encoding (purple line) & iTunes MP3 encoding (yellow line) show the current problem with scaling from two to four cores. Very few desktop applications can actually take advantage of a dual-core CPU, even fewer are geared for Quad-Core processors, and in those applications these Quad-Core processors actually take a step backwards in terms of overall efficiency; that's not the fault of the processor, but rather of the software not being optimized to support multiple threads.
 

 

Gaming Performance:
 
It seems all the new multi-core game consoles have made a big impact on the way PC game developers are programming for multi-core desktop processors.  Numerous gaming companies are now working on completely new gaming engines, which can take advantage of four processor cores and potentially any number of cores down the road (possible desktop Octa-Core in the 2008 timeframe with Intel 45nm "Nehalem" microarchitecture). From looking at numerous reviews detailing multi-cores and gaming, it seems the results clearly didn't show any advantage using multi-core processors right now, beyond the moderate speed gains from a couple games that can take advantage of 2 software threads. Even the multi-threaded Quake 4 benchmark doesn't show a performance increase when going from two to four cores, and it's one of very few games that actually takes advantage of multiple cores. Without significant software re-writes of today's games, you just won't see the sort of benchmark improvements you need in order to drive gaming performance forward.

2007 will mark the beginning of multi-threaded games making their impact on the PC gaming market segment. Upcoming game titles that will support Quad-core (or better) support will be Supreme Commander (Gas Powered Games / THQ), Epic's Unreal Tournament 2007 (and all Unreal Engine 3 titles), Valve's Half-Life 2: Episode 2 and Ubisoft's just-released Splinter Cell: Double Agent. At the Fall 2006 Intel Developer Forum, Intel showed off Remedy's Alan Wake, which will support a staggering 5 independent execution threads ->  one each for rendering, audio, streaming, physics and terrain bit-mapping. In this game apparently just the rendering and physics threads alone are enough to max processor utilization for a dual-core CPU, but the additional 3 threads are what may improve the gaming experience on a Quad-Core CPU.

Below are some screenshots from the upcoming PC game Alan Wake:

From Tim Sweeney, Founder and President of Epic Games about Unreal Tournament 2007's thread usage: "Currently Unreal Engine 3 runs two heavyweight threads all the time: one for gameplay and one for rendering. In addition, there are several helper threads to which we offload all of the physics (using Ageia's multithreaded PhysX library), streaming, and several other tasks. We plan to extend the threading support further in time for the release of Unreal Tournament 2007 next year, to further exploit multi-core PC CPUs. Major opportunities for multithreaded optimization include particle systems, animation, and terrain. Also, since UT2007 uses very extensive vehicle and ragdoll physics, we expect that at peak times during gameplay that we'll have no trouble fully exploiting 4 threads at the maximum detail settings."

Comparative Performance:

I think this quote from The Inquirer web site says it all: "...taking a look into media encoding shows that the AMD Quad FX (FX-74) just gets crushed by Kentsfield in raw performance, especially in MPEG-2 8Mbit reproduction. But the worst result for AMD is a look into power consumption and performance per watt. The AMD system consumes far more power than Intel, sometimes even double that of the Kentsfield setup."

All benchmarks found here => http://www.intel.com/performance/desktop/extreme/index.htm

Platform Support for Core™2 Extreme Quad-Core processors:
 
Motherboard Support:

Quite a few motherboard companies have announced support for the new Intel QX6700, but not all motherboards will support this quad-core CPU. For example the Intel® 975X Express Chipset supports the Intel Core™2 Quad-Core processor if you have a MB, which has support for the new input voltage range along with a very recent BIOS update that adds the new CPU Microcode signature for this processor. For example, Intel's own D975XBX board doesn't support the QX6700, so Intel is now shipping a new version, the D975XBX2 "Bad Axe 2" (ASI SKU: 53844) board with full Quad-Core support as well as changes to allow for DDR2-800 memory. Most of ASUS's current high-end desktop offerings based on the 975X, P965, and nVidia 680i / 650i chipsets will support the new Quad-Core processor => Click here to see full list.
 
Chassis Support:
Intel thermal specifications require the use of a Thermally Advantaged Chassis (TAC) version 1.1 when integrating an Intel® Core™2 Extreme processor into your system. A TAC version 1.1 chassis is defined by the presence of an 80mm side-panel air duct, a 92mm rear chassis fan and side-panel venting holes above the graphics and add-in card slots to provide additional cooling for high-end PCI Express graphics and other peripherals. Some chassis even have super-quiet 200mm fans as found in the new Antec "Nine Hundred" ultimate gamer chassis (ASI SKU: 52553).

     To view Intel's Thermally Advantaged Chassis list => Click Here
   
Power Supply Support:
Intel requires an ATX12V version 2.2 power supply for use with the Core™2 Extreme QX6700 processor. Please check www.intel.com/go/powersupplies for the appropriate support and validated power supplies. But please only use this chart as a guideline, since your particular system configuration will dictate the total wattage needed to run your system reliably. For a system with a discrete PCI Express x16 video card, Intel recommends a power supply in the 450-600W range, and if you employ two high-end nVidia 8800GTX in SLI with multiple hard drives, you might even consider going with a 700W to 1KW power supply.
 

With the intent of putting itself way out ahead of it's competitors, NVIDIA has come up with a brand new architecture which will entrench them in the hearts and minds of hardcore gamers and performance users worldwide. In just one generation, NVIDIA has introduced performance that, in many cases, doubles the competitions performance numbers on their top end product.

An Entirely New Design

NVIDIA has brought forth a completely new design for their video card offering. Instead of the normal process of simply increasing pixel and vertex shaders in order to improve quality or bumping up core clock and memory clock speeds to increase performance, the G80 relies on an entirely new unified architecture to do it's bidding. Created over the past 4 years, NVIDIA has been fairly quiet about the entire development cycle of the 8800 GPU and it's unified architecture model. Even as of July 2005, David Kirk (Chief Scientist at NVIDIA), was saying that they "will do a unified architecture in hardware when it makes sense." Well, it seems as though it definitely makes sense now!

The GeForce 8800GTX GPU consists of a massive array of 128 parallel Stream Processors (with a tidy 96 available on the 8800GTS), which are capable of performing the previously individualized tasks of vertex and pixel operations. Along with these tasks, the stream processors also handle geometry and even physics operations leaving room for NVIDIA to take a swipe at 3rd party physics add-on cards as well. Each of the stream processors can be dynamically allocated to whichever task makes the most sense for load balancing the GPU's resources for the current task at hand. In addition to the huge number of stream processors available, they also run on a separate clock than the GPU's core clock. The Shader clock runs at 1350MHz on the 8800GTX and a very respectable 1200MHz on the 8800GTS.

(Stream Processors shown as SP in the diagram below)

(near) Future Proof?

The NVIDIA 8800 Series is the first card to support Microsoft's DirectX 10 API and it's not hard to guess why. NVIDIA engineers designed the new product from the beginning with DirectX 10 and Windows Vista in mind. Also good to note is that the GeForce 8800GTX was the reference GPU for DirectX 10 API development and certification. DirectX was introduced back in 1995 to be launched in conjunction with the Windows 95 operating system. DirectX 10 represents the most significant enhancement of the technology since the introduction of Pixel Shading in DirectX 8.0. Shader Model 4.0 has been released as a part of the DirectX 10 specification and the 8800 Series is fully compliant with all of it's requirements.

In a nutshell, the big differences between Shader Model 3.0 and 4.0 boil down to three main changes; unification of shaders, geometry shaders, and resource virtualization. With the incoming Microsoft Windows Vista operating system, and it's exclusive use of DirectX 10, this card is primed and ready to fully support both gaming and and kind of creative production work in the foreseeable future.

Specifications