The Intel Core 2 Extreme QX6700: The Quad Core Era Begins - Hardware Upgrade

Earlier this year, Intel unleashed their Conroe line of processors aimed towards the desktop PC market. After lagging behind AMD for quite some time, Conroe gave Intel the chance to pull ahead in terms of performance and cost. In the end, Conroe is a testament to how more efficient planning and design could impact the final performance of any product. To us 2006 is known as a year in which Intel has made one of their largest ever comebacks.

In recent years, Intel has greatly changed the focus of their processors. Just a little while back, one of the most popular trends for CPU manufacturers was increasing the operating frequency of their processors. As Intel blazed on faster and faster in terms of clock speed, they reached a point where they could not increase the performance of their most recent architecture of that time (Netburst) by the traditional means of increasing the operating frequency. Because of high energy consumption, Netburst based CPUs emitted large amounts of thermal heat and thus required additional cooling in order to keep them running properly. Although Intel kept the architecture alive as long as they could, they finally ended it with the release of the last Netburst based processor, the 3.73 GHz Pentium EE 965.

Intel seems to have learned a couple of lessons from their Netburst failure. For one, they changed their processor branding method in the later portion of Netburst’s life, showing that Intel didn’t place as much focus on operating frequency as they used to. In addition, for their successive architecture AMD began focusing on reducing two major characteristics: thermal output and power consumption. Both of these qualities go hand in hand. Increased power consumption results in increased thermal dissipation, while decreased power consumption results in lower thermal dissipation. With large corporations and home users beginning to see the need for more power efficient components, energy efficiency has become a key issue for both Intel and AMD. This is where the fork in the road came for Intel and AMD. How would they increase performance while decreasing power efficiency? The answer was Dual-Core processors.

The first Dual-Core solutions began appearing for retail back in the Spring of 2005. Intel launched off their Dual-Core solutions with the Pentium D 800 series while AMD started off with their Athlon 64 X2 processors. AMD and Intel also launched off Dual-Core server series. Intel had their Xeon DP server processors while AMD had their Opteron server CPUs. Dual-Cores have allowed users to greatly benefit in programs that are multi-threaded. It has also allowed users to run multiple CPU intensive programs at the same time.

The market has come a long way from where it used to be. We are now at a point at which Dual-Core CPUs are available to a large amount of consumers. Currently, the most economic Dual-Core processor is the AMD Athlon 64 X2 3800+. It is available at a price of 152 U.S.D. With low figures like this, Dual-Core CPUs are much more cost effective than what they used to be.

As we all know, technology moves fast. As such, Intel and AMD have now set their eyes on Multi-Core processors. In the course of the next six months, we will be seeing Quad-Core solutions appear on the market. The first of these solutions is the Quad-Core Intel Core 2 Extreme QX6700, codenamed Kentsfield. Kentsfield is a solution that has two Dual-Core Conroe die in the same processor, so basically, it can be considered a Dual-Dual-Core processor.

The “Extreme” at the end of the processor name indicates that Kentsfield is aimed towards the high-end sector of the processor market. As can be expected, since this is a processor that utilizes a very new technology, the price tag shows it. Starting at an MSRP of 999 USD, the Quad-Core QX6700 is not for the faint of heart. It is mostly aimed towards hardcore enthusiasts and for professionals that use CPU intensive applications that might be able to benefit from multiple cores.

Kentsfield is based on Intel’s Core 2 architecture. Some of the architecture’s characteristics are that that all processors are based on a 65nm process, the FSB is at 1066 MHz, and the processors have a shared secondary memory cache of 2 or 4 Mbytes on a single die. An L2 cache of 2Mbytes is used for Intel’s lower end models while the higher end processors usually come with 4 Mbytes.

Externally, Kentsfield looks exactly the same as other Core 2 Duo CPUs. It also uses the same LGA 775 socket. As we have said before, it is important to remember that the Core 2 Extreme QX6700 has 2 separate die on the same processor. Each die contains 2 cores and has 4Mbytes of L2 cache each adding up to a total of 8 Mbytes. The large cache, no doubt, will help capture the heart of enthusiasts.

The only main difference that can be seen between the two processors is the filter caps. These can only be seen when looking at the bottom side of the chip.

In the above screenshot of CPU-Z, we can see that the FSB is 1066.7 MHz. This is identical to all other Core 2 solutions from Intel. L1 data and instruction cache are independent and designated for each core.

Intel has certified their 975X and 965 chipset for use with the Intel QX6700. NVIDIA and ATI have also certified a couple of their chipsets to be compatible with the QX6700. Even though certain solutions have been certified, it is still essential to check whether your specific motherboard is compatible with Quad-core processors. We advise to always check on the manufacturer’s website whether the motherboard is certified or not.

We will be benchmarking the Core 2 Extreme QX6700 against Intel and AMD’s most recent Dual-Core solutions. Intel’s line of processors will include their Core 2 LGA 775 processors while for AMD we will have their Athlon 64 X2 and Athlon 64 FX AM2 processors. All processors we are benchmarking today use DDR-2 memory.

This table displays the characteristics of the CPUs we will be benchmarking with:

AMD Athlon 64 X2 processors are available in a number of different frequncies. Currently, they range from 2 to 2.6 GHz. For most models, each core has a separate cache of 512 KB, however, for the Athlon 64 X2 5200+ and Athlon 64 FX-62, 1 MB of cache is integrated into each core. All of the processors have a declared TDP of 89 watts except for the Athlon 64 FX-62 which has a TDP of 125 watts.

The L2 cache, TDP, and clock frequency are in fact the only architectural differences between the AMD processors we are benchmarking today. They are all derived from the same dual-core architecture and are all constructed under a 90nm manufacturing process with SOI technology (Silicon on Insulator).

Moving on to the Intel solutions, we find that they all use the Intel Core 2 architecture. This is the same architecture which Intel introduced back in July of this year. All of these processors have a FSB of 1066 MHz. It is important to remember that AMD internally connects their components using HyperTransport technology in place of Front Side Bus. HypterTransport has a speed of 1,000 MHz (2,000 effective since there are two clocks). The base frequency for HyperTransport is 200MHz.

Intel’s Core 2 Extreme processors have a TDP, or Thermal Design Power, that is larger than the Core 2 Duo processors. The Core 2 Extreme X6800 has a TDP of 75W while the Core 2 Extreme QX6700 has a TDP of 130W. The hike in TDP that we see with the QX6700 is pretty much expected taking into consideration the fact that the processor has two separate Dual-Core die.

As can be expected, Intel has bundled their own certified cooling solution with the Core 2 Extreme QX6700 in order to ensure correct processor operation. The cooling solution that Intel provides is able to efficiently dissipate 130 TDP, however, it has one major drawback; the cooling fan can get pretty loud.

Above are two photographs of the cooling solutions Intel provides with three of their processors. From left to right, the heatsink for the Core 2 Extreme QX6700, the heatsink Intel for Core 2 Duo processors, and last the heatsink provided with the Pentium D 800 and 900 series. The heatsink for the Pentium D 800/900 series has the largest amount of fins. The first two have similar CPU contact bases.

The Core 2 Extreme QX6700’s fan runs at a default speed of 5,100 RPM. As can be expected, the fan generates a large amount of noise. Going into the bios, the most we could reduce the fan speed was 40%. Although operating temperatures were increased with the reduced RPM, the noise emission was much more bearable.

Speed of fan	velocità di rotazione	dB	CPU Temperatures in Celcius Core temp beta 0.93
Bios at 60%	2850 rpm	60 dB	71-68-69-69
Bios at 100%	5100 rpm	76 dB	65-60-61-62

To give you an idea of the amount of noise emitted by the fan, we ran a couple of tests. At 100% operating capacity, the noise level from 10cm away was 76dB. When forcing the fan to run at 60%, the noise level went down to 60dB. The above temperatures were taken with Povray running; therefore the CPUs were at 100% load.

In the analysis of the Core 2 Extreme QX6700 we used two platforms. For Intel processors, we used the Intel D975XBX2KR motherboard and for AMD Socket AM2 processors we used an ASUS M2R32-MVP. To keep the benchmarks as consistent as possible, we tried to use the same hardware between the two platforms whenever we could.

SYSmark 2004 is a benchmarking program that supports multithreaded applications and multitasking components. It benchmarks the system by simulating various PC tasks and applications. In our testing, we began with an overall system rating, and then moved on to Internet Content Creation and Office Productivity

Intel processors lead the pack in all of the benchmarks we ran using SYSmark 2004. The QX6700, though, wasn’t the top performing CPU from Intel in every benchmark. The Core 2 Extreme X6800, a processor which has a higher clock of 2.93 GHz, was able to best the QX6700 in all but one benchmark. The results the QX6700 had in the above benchmarks was a result of SYSmark 2004 not being able to take full advantage of all four cores.

For the 3D gaming tests, all platforms used the ATI Radeon X1950 XTX. Anti-aliasing was set at 4x while anisotropic filtering was set at 16x. We would have chosen more CPU dependant graphics settings; however, we opted upon simulating a more realistic test environment.

3D Mark contains a benchmark that is tailored specifically testing CPU power. For this reason we see the Core 2 Extreme QX6700 blowing away all other processors in terms of performance in the second benchmark.

In F.E.A.R., nearly all processors scored similar to each other in minimum FPS. The only exception was E6300 which scored a minimum of 20 FPS. We believe that the lower frame rate we see for the E6300 is less of a true performance issue and more of a specific compatibility issue F.E.A.R. has with the E6300.

The scene doesn’t change going to other games. As resolutions and filter levels increase, the margins of difference between the CPUs decrease. At 1600x1200, we see performance become very similar between the processors. In the gaming tests, it is important to remember that the main limiting factor is not the processor but the video card. Also, as can be seen, none of the game titles are able to truly take advantage of the four cores of the QX6700.

In AutoGK, a benchmark utility that tests by converting a media file to DivX, Intel processors yet again lead in terms of performance. The Core 2 Extreme comes in first place at 106.7 FPS. The QX6700’s lead compared to the X6800 isn’t very large. The X6800 comes in at 99.8 FPS. Compared to the Core 2 Duo, the E6700 performs 15% better. The Athlon 64 FX-62, AMD’s current best offering, isn’t able to go past 73.5 FPS. Compared to the FX-62, the QX6700 holds an advantage of 45%.

Intel’s dominance in the above test is a result of some of the innovations Intel has introduced with the Core architecture. Primarily, optimizations to SSE 128bit instructions are a strong reason for the performance displayed by the above Intel processors.

In the above 7-Zip test, it is necessary for us to assume the same thing that we did in Sysmark 2004 benchmarks; that the program has poor exploitation of 4 cores.
For this benchmark, it seems that higher clock frequencies is what impacted performance the most.

Lame MT is a command line interface tool that converts Wav to MP3. In the results, Intel processors for the most part outperform their AMD counterparts. The only exception is the Intel Core 2 Duo E6300 which falls behind the AMD Athlon FX-62 by 8 seconds. Considering the large difference in price between the two cards, kudos should still be given to the E6300 for holding up a pretty good fight.

A general trend which is beginning to appear in programs that have poor exploitation of four cores is that the QX6700 turns out to have very similar performance to the Core 2 Duo E6700. This is a result of both processors having equal clocks indicating that only two cores of the QX6700 are actually being taken advantage of.

In this benchmark, we have used iTunes to convert a Wav audio file to 192kbps MP3. Again we see poor use of the QX6700’s four cores. The Core 2 Extreme QX6700 ends up tying for second place with the Core 2 Duo E6300.

Unlike most of the benchmarks we have run so far, Sony Vegas is a program that is able to take decent advantage of the QX6700’s four cores. As is evident in our benchmarks, it is able to properly use multiple cores. The QX6700 was able to complete all of its conversion in 241 seconds. The fastest competitor was the 2.93 Ghz X6800, which was able to complete its conversion in 325 seconds. The fastest AMD solution was the Athlon 64 FX-62, which came it at 436 seconds.

Entering the second portion of our multimedia tests, the QX6700 is finally able to take first place in every test. This is a result of better multi-threaded software. Unfortunately, many of the programs we have benchmarked with earlier have had poor exploitation of the extra cores the QX6700 has to offer, however, stepping into conversion utilities, the scene significantly changes.

3D rendering isn’t an everyday task for a majority of PC users. It is important to remember, however, that the Core 2 Extreme QX6700 isn’t just meant for everyday users. The processor can also prove to be invaluable for professionals who run CPU intensive programs and need as much product efficiency as they can get.

Keeping that in mind, the Core 2 Extreme QX6700 had very strong results in rendering. In 3DS Max the QX6700 dominated with its four cores by impressive margins and maintained its performance throughout the remainder of the rendering tests.

The QX6700 also had sweeping victories in PovRay 3.7 and Cinebench 9.5. In PovRay, the QX6700 nearly had double the performance of the Core 2 Duo E6700 indicating that all four cores were actually being used. Looking at Cinebench’s multithreading tests, we see a huge jump for the QX6700’s score compared to what it had in single thread testing.

Viewperf is a benchmark from Spec.org that was created by IBM. The utility provides a complete analysis of system performance when executing OpenGL applications. The program is comprised of 9 scenes and benchmarks system performance in each scene. As a heads up, some of the scenes aren’t very CPU dependant and, since this is a graphical benchmark, focus more on the video card.
Thus, the video card will probably be a main limiting factor in the benchmarks.

The results are all around the place. The Athlon FX-62 leads the pack for AMD’s solutions while the Core 2 Extreme QX6700 is unsuccessful in gaining first place. The best it reaches is third.

Power consumption, as we all know, has become a very essential part of a CPU’s performance. Lower power consumption doesn’t just reduce power bills but also helps lower the cooling requirements for a CPU. A processor that consumes more power will require a stronger cooling system therefore potentially raising operational noise.

The following table shows the TDP (Thermal Design Power) of the processors we are benchmarking:

Processor	Clock	TDP	Socket
Athlon 64 FX62	2,8 GHz	125 Watt	AM2
Athlon 64 X2 5000+	2,6 GHz	89 Watt	AM2
Athlon 64 X2 4600+	2,4 GHz	89 Watt	AM2
Athlon 64 X2 4200+	2,2 GHz	89 Watt	AM2
Athlon 64 X2 3800+	2 GHz	89 Watt	AM2
Core 2 Extreme QX6700	2,67 GHz	130 Watt	775 LGA
Core 2 Extreme X6800	2,93 GHz	75 Watt	775 LGA
Core 2 Duo E6700	2,67 GHz	65 Watt	775 LGA
Core 2 Duo E6600	2,4 GHz	65 Watt	775 LGA
Core 2 Duo E6400	2,13 GHz	65 Watt	775 LGA
Core 2 Duo E6300	1,86 GHz	65 Watt	775 LGA

It is important to remember that the TDP values declared by Intel are not the true maximum power dissipation. Instead, they are the typical TDP the processor has
The AMD measurements, on the other hand, are the maximum TDP of the processor

In addition, AMD’s TDP might be slightly higher considering that the memory controller is integrated into the CPU. For Intel, the memory controller is external and instead integrated in the Northbridge.

To measure the consumption of the CPU we measured consumption of the whole system. In order to keep our results as accurate as possible, we tried to use exactly the same parts wherever possible, the exception being the motherboard difference between AMD and Intel CPUs. Here are the components:

The below measurements are for idle. For idle, we booted up Windows XP and tested the power consumption on an empty desktop.

The above test values are in watts. Also, keep in mind that it is the consumption of the whole system and not just of the CPU. Practically identical values came in for all AMD solutions while Intel solutions varied by each processor. For Intel solutions, the ones with larger L2 caches ending up consuming more power compared to those that had smaller cache sizes. AMD’s Cool’n’Quiet technology really helped reduce power consumption for AMD CPUs. By reducing the CPU’s clock down to 1GHz when idle, Cool’n’Quiet really helped increase power efficiency.

The above tests were taken with the system running at full load. Our tests were run while running a scene in Povray 3.7. As can be seen, the QX6700 came in with the most power consumption. This is expected because of the larger amount of cores. Unlike the other CPUs, the QX6700 has two physics die on the processor, whereas all other CPUs only have one.

Intel was first to the market with Dual-core CPUs when they released their Pentium D 800 series, and they have yet again beat AMD to the market with the first Quad-core solution, the Core 2 Extreme QX6700. Hot on the heels of the desktop PC version, Intel has already planned the release of a Quad core Zeon CPU for the server market. The Quad Xeon MP, Intel’s first Quad-core sever processor, will be released later this month. The approach Intel has used so far to release Quad-core CPUs is similar to how they began their release of Dual-core CPUs. Intel began by first releasing a desktop PC version and then releasing a version aimed towards the server market. AMD, on the other hand, released their first Dual-core processors as part of their Opteron server series, and then released one for the mainstream desktop PC market.

Intel has released the Core 2 Extreme QX6700 at 999 USD. Since Quad-core technology is still relatively new and is still not a mainstream technology, it will take a while for prices to recede. The price of the Core 2 Extreme raises the question of whether the CPU is actually worth its price. The answer is relative to what the intended use of the CPU will be.

As our gaming benchmarks have shown, current games do not take very good advantage of Quad cores. To be frank, many games don’t even take advantage of multiple cores yet. Keeping this in mind, it wouldn’t be the wisest of decisions to immediately purchase a Quad-core processor if your sole reason would be to use it for gaming. It is understandable that there aren’t many game engines that can truly take advantage of multiple cores due to the complexity of its programming, however, we would really like to see more games being published that actually benefit from multiple cores. For gamers, the hardware industry has gotten ahead of itself. Without proper support for multiple cores, their performance is nothing compared to what it should truly be. For gamers, the Core 2 Extreme QX6700 would be little more than a novelty.

The story is different for users who are oriented for a more professional use. As our tests with 3DS Max 9 proves, the QX6700 allows for obvious performance advantages in rendering applications. The reductions in time to perform these tasks can greatly increase productivity, thus ultimately justifying the cost of buying the Kentsfield.

Kenstfield also has a lot of headroom for overclocking. In our tests, we were able to reach a maximum stable clock of 3.3GHz. Although we were able to reap the obvious benefits that come with higher clock speeds, the overclock came with some problems of its own. Because of the higher operating frequency, we ended up also having higher operating temperatures meaning we had to meet additional cooling requirements. Moreover, total power consumption of the system was increased.

In around two weeks’ time, AMD will be introducing their own Quad-core platform known as 4x4. This new technology, instead of immediately jumping to four cores on one processor, will split up the job between two dual-core Athlon 64 FX processors, adding up to a total of four cores. Once 4x4 is released, it will be a good reference point to see the actual effectiveness of Intel’s Core 2 Extreme QX6700.