The world’s first quad-core desktop processor from Intel promises to be something pretty spectacular. Being the next Extreme Edition processor, it has some big shoes to fill. How does it fare against its predecessor?
When reviews of the Core 2 family of processors from Intel hit the web exactly 111 days ago, enthusiasts everywhere were pretty impressed with the performance. Here was a processor that, on its lower end, was vastly superior to anything the competition had to offer. A $300 Intel Core 2 Duo E6600 wiped the board with AMD’s top model, the FX-62, while its big brothers, the E6700 and X6800, made sure nobody had any doubts. In the last 111 days, countless more reviews have been published that further defined the end result: Intel had once again taken the processor performance crown.
But something else happened in the last 111 days as well. AMD loyalists around the world were left wondering what their favorite company had up their sleeves to combat this new Intel powerhouse. The answer came with the announcement, just 12 days after the Core 2 launch on the 26th of July, of something they were calling “4×4”, a radical new platform design that takes advantage of AMD’s historically strong communications bus and is expected to provide some impressive performance. Four processing cores and four graphics cores? Sounds like a gamer’s dream. The whole thing was strangely reminiscent of the launch of Intel’s Extreme Edition processors back in 2003, which attempted to trump AMD’s Athlon64 launch that happened earlier in the same week. It stands to reason that this announcement was strategically planned and poised to steal some of the Core 2 thunder.
However, Intel saw this one coming and, in true Intel fashion, made a nice little counter-announcement. You see, Intel was working on something themselves. Two Core 2 Duos on one chip, a “Core 2 Quad” as it were. Codenamed Kentsfield, or Clovertown for the server segment, this first-ever Quad-Core consumer CPU was slated for launch in early 2007. We presume that after hearing about this whole AMD “4×4” thing, Intel decided they would push that launch date up a bit, a good two or three months to be exact. Before AMD even announced their new technology to the public, rumors hit the web that Kentsfield would be hitting shelves sooner than expected. Shortly after, Intel confirmed these rumors at their fall Intel Developer Forum (IDF).
Sure, Intel’s first Quad-chip would not be the monolithic native quad-core beast many enthusiasts were expecting, but two unlocked E6700’s packaged on the same chip was not something to scoff at. Previews of Intel’s new Core 2 Extreme QX6700 dropped at the end of September after reviewers got their paws on the chip at the fall IDF. Since then, everyone’s been begging for a comprehensive review of the chip. Seeing as the NDA expires today, we thought we’d oblige.
{mospagebreak heading=Introduction&title=ICM C2Q}
ICM: Core 2 Quad – Same old same old
In terms of chip architecture, there is nothing new to report with the QX6700. It uses the exact same ICM technology that is found on the other members of the Core 2 family, except it has a twin sitting right next to it on the processor package. To save us the time and you the sedative, here is our explanation of the features and advantages of Intel Core Microarchitecture, as related, verbatim, in our initial Core 2 Duo review.
… Intel Core Microarchitecture is, as previously related, sort of a combination of the P6 marchitecture used in the Pentium III, and the Intel NetBurst marchitecture used in the Pentium 4. It combines existing features from P6 that lend to its more energy and time efficient design, as well as features from NetBurst that allow higher levels of performance. Before touching on each new innovation, here is a list of features that will be discussed:
- Intel Wide Dynamic Execution
- Intel Intelligent Power Capability
- Intel Advanced Smart Cache
- Intel Smart Memory Access
- Intel Advanced Digital Media Boost
Wide Dynamic Execution is Intel’s name for the system of decoders and execution cores used by ICM. While all processors have their Execution system, P6 had primitive dynamic execution and NetBurst had the Advanced Dynamic Execution Engine, Wide Dynamic Execution is a large step in right direction in terms of power efficiency and overall performance. Essentially what Intel has done is, as the name suggest, widened each execution core. This allows more instructions to be handled simultaneously, four rather than the three found in P6 and NetBurst. More specifically, Intel’s improvements focus around the implementation of more accurate branch prediction and deeper instruction buffers (where instructions sit before entering the execution core).
The most significant development here is something called “macrofusion”. Intel uses what they call macrofusion to combine common instruction pairs into something called a micro-op to be sent through the decoder to the execution core together at the same time. Previously, each instruction was sent through the decoder to the execution core by itself, even if a similar instruction was waiting right behind it. So by using macrofusion, Intel’s Wide Dynamic Execution engine increases the amount of instructions that can be sent to the execution core in any given period of time. This increases efficiency, which in turn increases performance and decreases power usage. Combine this with an upgraded algorithmic logic unit (ALU), and ICM has quite a powerful branch prediction/execution engine at its disposal.
In most current processors, all of the logic subsystems on the die are turned on at all times, whether they are being used or not. In ICM, Intel implements advanced power gating that allows each logic subsystem to be turned on only when it needs to be used. This means that the rest of the logic subsystems are turned off, or in some kind of hibernation mode, while they are not being used. The result is a significant decrease in overall power consumption, which of course further increases efficiency. This means that ICM uses much less power than traditional processors, making it the ideal platform for mobile computing. Intel’s name for these features is Intelligent Power Capability.
Another improvement that Intel made with ICM is what they call Advanced Smart Cache. With dual-core NetBurst processors as in the current Pentium D line, each execution core had its own independent bank of L2 cache. The problem, well not really a problem but more of an underutilization, with this was that each execution core only had access to half of the processor’s total available L2 cache. As we have learned empirically, increased L2 cache pretty much always means greater performance. Advanced Smart Cache effectively doubles the utilization of L2 cache with ICM, allowing the processors to access and write information to the cache in larger amounts, decreasing cache misses and increasing performance. This technique is similar to what we already see in AMD’s Athlon 64 X2 offerings.
{mospagebreak title=Features (cont.)}
Intel’s Smart Memory Access is yet another way Intel has increased the overall performance of ICM. In Smart Memory Access, Intel implements something they call memory disambiguation. Memory Disambiguation is something that allows execution cores to speculate what data is needed from the memory before it actually executes. This allows the system to “pre-load” information while the previous instruction is being executed, drastically decreasing memory latency effects.
The final improvement we see in ICM is the Intel’s Advanced Digital Media Boost. This feature is basically the end result of the doubling of the speed at which SIMD instruction sets are executed. Previously to ICM, execution cores had to break down the 128-bit SSE,2,3 instructions into two 64-bit subsets in order to execute them, allowing one SIMD instruction set to be executed per every two clock cycles. With ICM, the entire 128-bit SIMD instruction set is executed in the same clock cycle, allowing one instruction set to be executed every clock cycle. As a result of Intel’s Advanced Digital Media Boost, the end-user experiences dramatically increased performance when processing multimedia instructions.
Specific to the Core 2 series of processors, and probably every processor hereafter, Intel has included several very interesting features. They are:
- Intel Virtualization (VT) Technology
- Intel Extended Memory 64 Technolgy (EM64T)
- Intel Execute Disable Bit
Virtualization Technology is something we’ve been hearing about for quite a while now. The main idea for VT technology is to allow multiple applications to be running simultaneously under different profiles. The way it does this it creates “Virtual” platforms, or environments, and allows processes to run in those environments when the computer is actively being used on a different platform. The application of this technology is likely far more promising and relevant in the enterprise/business world as opposed to the average home user, but the potential for overall productivity increases looks very attractive.
EM64T, while previously implemented on the 600 series of the Pentium 4 line and later processors, will again be used in Core 2 processors from Intel. EM64T is essentially the means by which Intel processors operate on x64 instructions. It also allows Intel processors to utilize greater amounts of virtual and physical memory. There is nothing ground-breaking to report here.
Intel’s Execute Disable Bit is basically a further enhanced attempt at hardware-based virus protection. The Execute Disable Bit marks memory banks as executable or non-executable, depending on whether or not it is infected by a virus or other malicious agent. If the operating system attempts to read data from a non-executable piece of memory, this will be recognized and avoided, disallowing the spread of the virus.
…
The only difference between the Core 2 Duo and the Core 2 Quad architecture is that Intel’s Advanced Smart Cache feature does not work on it to the extent it does on Core 2 Duo. This is due to the fact that the cache on the processors is not pooled and used by all four cores. Each set of two cores, dies, has it’s own cache bank which it pulls data from, and the two dies then have to communicate through the FSB link. This is far from ideal, and would be the main advantage of a "true" quad core processor.
Another difference between the Core 2 Quad and Core 2 Duo is power consumption. Where the Core 2 Duo carries a supposed thermal design power (TDP) of 65-75W, the Core 2 Quad, expectedly since it is two Core 2 Duos on one chip, carries a TDP of 120-130W. This is up there with the power consumption of the previous generation Extreme Edition chips, the 965 and 955.
The final, and one of the biggest differences between the Core 2 Duo and Core 2 Quad chips is that the latter will not work in all 965 and 975 chipsets. In fact, the Core 2 Quad apparently will not work in Intel’s own X975XBX Bad Axe board. For this review, we had to use a brand new X975XBX2 Bad Axe 2 motherboard supplied by Intel.
By now we all know what ICM can do. We know that its implementation on the Core 2 family of processors has been nothing short of a resounding success. We know that the Core 2 Extreme X6800 is the fastest processor on the market. What we don’t know is what kind of performance differences we will see with the Core 2 Quad. To get a definitive idea of what 4 physical processing cores are capable of, lots of testing has to be done. So, lots of testing we shall do.
{mospagebreak title=Test Setup and Methods}
Test Setup and Methods
Hardware Configuration
- Case: Vigor Force
- Power Supply: PCP&C Silencer 750 EPS12V
- Motherboard: Intel Desktop Board D975XBX (Bad Axe)
- Processor: Intel Core 2 Extreme X6800 (Varying MHz) | Intel Core 2 Extreme QX6700 (Varying MHz)
- Hard Drive: Western Digital 2500KS 7200RPM, 250GB with 16MB Cache Buffer
- Video: eVGA NVIDIA GeForce 7900GT KO Superclocked (580/1580)
- Memory: 1024MB (2×512MB) Corsair XMS2 PC2 6400 (TWIN2X1024A-6400)
- Optical Drive: Lite-ON SHW160P6S05
- Cooling: Vigor Monsoon II
- Audio: Creative X-Fi XtremeMusic
Software Configuration
- Motherboard BIOS: D975XBX2 Express
- Operating System: Windows XP Professional with Service Pack 2
- Video Driver: NVIDIA ForceWare Version 91.47 WHQL Certified (September 14 release)
Methods
Testing methodology for this processor is slightly different than for the processors we’ve tested before, or any processor for that matter. The simple reason for this is that it has 4 cores; not 2, not 1, 4. In a perfect world, we could just run all of our normal benchmarks and see a representative 100% increase over dual core processors. Unfortunately, there are quite a few things stopping this from happening. One such thing is that just because there are now twice as many cores, there is not going to be twice the performance. There are several reasons for this in itself, but the main reasons are bottlenecks in and around the chip that prevent it from performing twice as good as a dual core. Another reason we won’t see the full potential of this chip is because most applications are single-threaded, meaning they are designed to run on one core. Without running multiple instances of a single-threaded application, the point of which is highly questionable, or running one of the relatively few multi-threaded applications, we aren’t going to see huge performance increases while adding more cores.
Therefore, our main focus in this review will be the multi-tasking performance of the processor. To clarify what we mean by this, we’re going to get our usual performance numbers in addition to numbers achieved when multitasking. We will be testing the processor in games and synthetic benchmarks, under normal and multi-tasking conditions, to get an idea of performance increases and decreases. Specifically, in synthetic benchmarks, our multi-tasking of choice will be running three communications programs and one instance of Prime95. In gaming benchmarks, we will be loading the processor with streaming audio, two communications programs, steam downloading games, and one lovely instance of Prime95. This will allow us to record performance hits with each processor to see how good they are at multi-tasking.
Also, before we start, it should be noted that the Core 2 Extreme QX6700 ships with a clock speed of 2667MHz. We will be testing the performance at this setting first, and then overclocking it to 2933MHz in an effort to go apples-to-apples with its predecessor, the Core 2 Extreme X6800 (2933MHz). Oh, and speaking over overclocking, we will also be overclocking these processors to the maximum stable speed. For reference, it just so happened that both processors ended up at the same overclocked speed, but that is more a result of the motherboard than the processors themselves. We will touch more on that in the overclocking section. Now let’s take a look at the applications we will be using to test the Core 2 Extreme QX6700.
{mospagebreak title=Test Suite, 3DMark06}
Test Suite
When it comes to choosing the products that will be included in our benchmarking suite, we often have a hard time making the decisions. We don’t simply use the benchmarks that we have used in the past, because they may not be as applicable for the specific component we are trying to test. We have chosen the following tools based on the way the tests are run, the kind of results they generate, and the current interest/demand for them in the community:
- Kanada Lab - Super PI (mod 1.5 XS)
- SiSoftware - Sandra Lite 2007
- ScienceMark - ScienceMark 2.0 Benchmark Suite
- Futuremark - PCMark05 CPU Test Suite
- Futuremark - 3DMark06
- Id Software - Quake IV
- Monolith - F.E.A.R
- Valve - Half-Life 2: Episode 1
- Infinity Ward - Call of Duty 2
- Bethesda - The Elder Scrolls IV: Oblivion
It’s a shorter list than we wanted, but we had to make do with the time we had. In the same fashion as they did with the Core 2 Duo embargo date, Intel pulled the NDA on the QX6700 in from November 14th to November 1st. When it really comes down to it, we wouldn’t have added too many other applications to this list. What we would have done, however, is combined several of them in an attempt to paint a better picture of the multi-tasking performance of this processor. As is, this benchmarking suite is very solid and represents a good sample of what is out there, what people want to see, and what we think will yield the most meaningful results.
Tests
Please note that every result you see from this point forward is representative of the calculated average of the results of at least three tests and often times many more. This is to eliminate performance fluctuation anomalies and other random occurrences that may have the potential to throw off results.
3DMark06
As mentioned before, multitasking is going to be a major focus during this review. However, when not multitasking, it would be nice to have your processor able to use all of its cores on the one thing you are doing. This is what we call multi-threading, and unfortunately, it is something that must be implemented in the code of the specific application you are running. More unfortunately still is the fact that there are not too many multi-threaded applications out there. That said, Futuremark’s benchmarking suite has several tests that can make use of more than one core, and 3DMark06 is one of them. We tested the processors at stock settings, overclcoked settings, and while running an instance of Prime95 for the multi-tasking aspect.
As you can see, the QX6700 performs exceptionally in this test. It performs so well, in fact, that at stock settings it beats out its predecessor, which is running nearly 1GHz faster, by a solid 100 3DMarks. When a processor beats out another, faster processor, you would think something is wrong. However, as we mentioned, 3DMark06 IS a multi-threaded application. This means that the performance will increase substantially by adding more cores. The difference here might not seem like much, but 3DMark06 is very taxing on any system, and squeezing out an extra 100 3DMarks is often easier said than done. Clearly, we can see from this test that there is indeed some potential with the QX6700.
{mospagebreak title=PCMark05 and SuperPI}
PCMark05
We ran 3DMark06 because it is part of the Futuremark suite that is multi-threaded. This is also one of the reasons we chose PCMark05. Not only is PCMark05 multi-threaded, it includes two CPU tests that are exclusively multi-tasking performance tests. Running these will give us a good idea of the multi-tasking performance of these processors with respect to each other, as well as the overall performance of the processors. The CPU tests in PCMark05 are a very good way of getting a grasp on the overall computing power of your processor in the real world, as every test is based off of real world applications.
SuperPI (mod 1.5 XS)
The next test we chose to use was the venerable Super PI. Developed in 1994 by the Kanada Lab at the University of Tokyo, Super PI has been and remains to this day one of the simplest and clearest ways to test your computer’s sheer processing power. As has become common practice, the specific test we are reporting is the ‘1M’ test, or a test to see how fast your processor can calculate the value of pi out to one million decimal digits. This version of SuperPI has been modified by the gurus over at XtremeSystems.org to produce time figures that are accurate to the thousandths of a second. This allows us to pick up more subtle differences in computation times.
SuperPI is the first of our tests not to be multi-threaded. As a result, we wouldn’t expect the increased number of cores to have an effect, positive or negative, on the scores we achieved. However, you can see that the scores of the QX6700 at stock and at 2933MHz are significantly slower than those of the X6800 at stock. We don’t really have an explanation for this, other than that there could have been some sort of RAM influence going on. We will explain the whole RAM situation a bit later. It is easy to tell that when both processors were overclocked, the scores we achieved were pretty much the same, with no discernable advantage in either direction. You can also see that the addition of an instance of Prime95 slowed down the calculations pretty significantly as well, though much more so on the X6800 than the QX6700.
{mospagebreak title=ScienceMark 2.0}
ScienceMark 2.0
Since computers are computers, and they were originally developed to handle increasingly complex scientific calculations, it is only appropriate that we benchmark the very latest and greatest in computing with a tool called ScienceMark. Really though, ScienceMark 2.0 is probably the best collection of CPU processing tests that there is. The benchmark suite has your processor do everything from calculating 1536×1536 matrices to cryptographic calculations and simulating molecular dynamics.
We were very happy with the comprehensiveness and clarity of the results we achieved with ScienceMark in our Core 2 Duo review. So much so, that we decided to utilize it again. This time, however, we have added a memory/cache performance test to get an idea of the throughput rates and cache latency of the processor/memory interface. First, the overall test results.
Moving on to the memory tests, we recorded the values of latency of 256 byte strides through the memory and L2 Cache, as well as the total realized memory bandwidth. These numbers should be interesting, as they will give us insight about the performance hits that may or may not have been incurred while mounting two dies on one processor package.
It is pretty clear from these results that there was indeed a latency hit involved with putting the two processors on the same package. This is similar to what we saw with the Pentium-D line of processors, which were essentially two Prescott dies bolted together, much the same way the QX6700 is two Conroe dies bolted together. There was also a significant difference in memory throughput between the two processors, which was pretty surprising considering the RAM settings were constant throughout stock and 2933MHz testing. This could be a possible reason for some of the weird numbers we have been getting previously. Whatever the case, these performance hits seem to vanish when the processor and memory are overclocked significantly.
{mospagebreak title=Sandra 2007}
SiSoft Sandra 2007 Lite
We used the Sandra 2007 benchmarking tool to get some raw statistics of our processors. Since these numbers are not calculated in any kind of special scoring system, there is nothing much to explain here, so we will let the pictures do the talking.
 |
 |
 |
 |
 |
As you can see, the total arithmetic and multimedia processing capacity of the QX6700 is very close to double in every single test. This is expected, considering this benchmark is measuring the total processing power of all the cores and is not dependent on any other variables. If you were reading this review hoping to see the Kentsfield performing twice as good as the Conroe, here you go. However, it is rare that CPU utilization is ever at 100% in normal work modes, even with several applications open and running. It is possible, however, and it certainly does happen. We will talk more about that in the multi-tasking section to come.
{mospagebreak title=Games, Quake 4}
Games
This is what everyone has been waiting for. Sure, the QX6700 has roughly twice the theoretical processing power of the X6800; we already knew that. But does this new Extreme Edition exceed at the task that Extreme Editions were originally designed for? While seeing theoretical performance in synthetic benchmarks is a good indication of the capabilities of your processor, the real worthwhile performance numbers are generated in games, where results are entirely indicative of real world performance. Here is how the QX6700 performs in the latest single-threaded and multi-threaded games.
Quake 4
Quake 4 is a game that can utilize symmetric multi-processing, or dual core CPUs. We investigated the performance benefits of using dual core and hyperthreading CPUs with Quake 4 in our previous article entitled “The Effect of Dual Core Patches”. We will be running this game on both processors with SMP turned on and OFF to see if there is indeed a performance benefit when the number of cores increases to four. We don’t expect there will be, but we can dream. Also, we have shown time and time again that Quake 4 is extremely CPU-limited on the low end, while being rather GPU-limited on the high-end. As such, we would expect the results on low settings to be more telling in terms of processor performance.
Here you can see that the QX6700 performs consistently worse than the X6800. This is undoubtedly due to the extra 266MHz on the X6800. It is also clear from these results that the multi-threaded capabilities of Quake 4 are limited to two cores, which we already kind of knew. Since Quake 4 is historically the game that benefits the most from symmetric multi-processing, we fear that this result may be a foreshadowing of things to come. Let’s find out.
{mospagebreak title=F.E.A.R.}
F.E.A.R.
We love F.E.A.R. Not only is it easily the most visually stunning First Person Shooter currently out there, it is excessively easy to benchmark. Thanks to the built-in “Test Settings” feature, users can benchmark the game very easily to see how well their system does. While being majestically beautiful at the highest possible settings, the game looks pretty darn ugly when everything is turned off. As such, we would expect it to be very CPU-limited on the low end. Another great thing about F.E.A.R is that it is extremely sensitive to changes in your system. Increasing the clock speed of your processor almost always results in a representative increase in frame rates. Increasing the load on your processor almost always results in a representative decrease in frame rates. Unfortunately, however, F.E.A.R is still a single-threaded game, meaning we probably won’t be seeing any performance increases with the QX6700.
As you can see, the X6800 consistently outperforms the QX6700, even when the latter is overclocked to even the playing field. This is almost certainly due to the memory throughput and latency issues we uncovered earlier. There may be several readers out there that are so incredibly on top of things that they realized that these numbers are not very consistent with the performance numbers we presented when testing the X6800 in F.E.A.R as part of our initial Core 2 Duo review. You are absolutely correct if you noticed this; not only is there a difference, but there is a pretty big difference. Since the hardware being used here is not radically different — in fact it is identical with the exception of a slightly updated motherboard — we can only attribute this difference to software differences. Possible software differences include video driver updates and different game versions.
{mospagebreak title=Half-Life 2: Episode 1, Call of Duty 2}
Half-Life 2: Episode 1
When Valve dropped the first of their series of episodic sequels to Half-Life 2, critics across the internet gave it pretty rave reviews. Gotfrag EXE’s very own Jason Baker rated the game as an A- and gave the graphics an A. The Source engine, though nearly 2 years old already, is still capable of producing some amazing visuals. To test this game, we recorded a custom timedemo that spanned three “chapters” of the game. The timedemo then had to be played in three separate parts. In order to get a good representation of the frame rates we were getting, we took the total number of frames rendered in all three playbacks, and divided that by the total time it took to render those frames. The demos we used can be downloaded here (.zip, 14.3MB).
Call of Duty 2
Still widely considered one of the most visually stunning games around, Infinity Ward/Activision’s Call of Duty 2 is one of the most ideal benchmarking platforms we could think of. While the volumetric smoke effects and complex textures in the game keep your graphics card sweating, the physics engine and 3D environments can cause your CPU to have a lot of trouble as well. We tested Call of Duty 2 on the highest and lowest graphics setting sets and four different resolutions. We have also revised our previous testing procedure for this game due to complaints that the multiplayer demo we were using was not a good representation of the game. After playing through the single player campaign, we couldn’t agree more. Our new test is based off of three different single player demos; one at the Point Du Hoc landing, one in Stalingrad, and one in some city in Africa that we can’t remember the name of. There is a very, very good sampling of all the different effects in this game in these three demos, and we think they are here to stay. We hope to make these demos available to you sometime in the future, but as of now they are over 800MB combined.
CoD2 utilizes symmetric multi-processing, as we explained in our previous article entitled “The Effect of Dual Core Patches”. From the results, it would stand to reason that the implementation here does indeed make use of four cores. We’re not entirely convinced that it does, however. One possible explanation for this is that while the results are different from what we saw in Quake 4, they are consistent with the fact that CoD2 is a much more taxing game than Quake 4. This means that although only two cores are being used in CoD2 gameplay, they are being entirely used, and any outside CPU load will negatively affect performance on dual core systems. Again, this is where quad cores shine, as performance hits are minimal with the QX6700.
{mospagebreak title=Oblivion}
Oblivion
We said previously that if Oblivion had some sort of built-in benchmarking method, or even a demo playback feature, it would, beyond a shadow of a doubt, be the best benchmarking tool available. We stand by this statement to this day. The sheer amount of graphic settings adjustments you can make is reason enough to think this, but throw in the fact that on the highest possible settings the game is incredibly beautiful just hammers the point home. Sadly, however, there is no built-in playback feature, and benchmarking in the game can be a logistical nightmare. Even sadder still is the fact that the game will soon be considered outdated, thanks to the impending arrival of DirectX 10 and the games that come with it. Our tests consist of running in a straight line in one direction for 60 seconds and recording the frame rates with FRAPS. We repeat this test AT LEAST 10 times in order to assure as accurate of a result as possible.
Oblivion is not a multi-threaded game and, as such, does not benefit from the two additional processing cores. The results are very representative of the 266MHz clock speed advantage of the X6800. There is nothing unusual to report here; the X6800 is still the world’s best processor for gaming.
{mospagebreak title=Multitasking}
Multitasking
Here is where things get interesting. The multi-tasking performance of the QX6700 is pretty much guaranteed to be better than that of any other consumer-level processor due to the simple fact that it has four physical processing cores. Previous Extreme Edition chips, like the 965 and 955, touted four processing cores, but of those, 2 were physical and two were logical. Logical cores (Hyper-Threading) are good, but the performance has been proven to be not nearly on the same level as a true dual-core or multi-core configuration. When you run an application, it is assigned a core affinity, meaning core 0 or 1 if you have a dual-core CPU. If it is single-threaded, it will run exclusively on that core. Run something else, and it might be assigned to core 1. You can sometimes set the affinity to run by adding a –A1 tag to target field in the shortcut properties, but that is not always guaranteed to work. If the CPU is loaded 100%, then any further programs you start will be budgeted into the CPU time and result in decreased performance across the board. We have already seen this with Benchmark + Prime95 tests we ran above. Now with a quad-core processor, you have twice the capacity. Let’s illustrate this graphically. The following test was performed with several programs/applications running in the background. They are: Streaming audio from WMP11, Ventrilo (VoIP), Downloading three games on Steam (Defcon, Dark Messiah, HL2:LC), mIRC, one instance of Prime95, and Adobe Photoshop was open but not running anything.
As stated in our previous mention of F.E.A.R, the game is extremely sensitive to minor fluctuations in hardware settings and system load. That makes it a very good platform to test multi-tasking with, since all of these programs will be adding their own bit of load to the CPU. You can see that the QX6700 is on an entirely different level than the X6800 in terms of multi-tasking. Where the X6800 suffered a 174 frame loss, the QX6700 took a mere 12 frame hit. This is a HUGE difference. This is a 37% performance hit with the X6800 that was pretty much avoided on the QX6700. Granted, gamers are most likely not going to be running an instance of Prime95 in the background while gaming. However, they could very well be running an instance of F@H, and the other applications listed are fairly commonly used. While the QX6700 was inferior to the X6800 in basically all of our gaming tests (with the exception of CoD2, but this was most likely a multi-tasking issue as well), it seems to make up for it in the multi-tasking department.
Not necessarily totally relevant to our audience, but important nonetheless, are the benefits this could potentially provide to digitally-oriented professionals. Running 3D Studio MAX, Adobe Premiere, After Effects, Combustion and watching your favorite episode of The Office has never seemed so easily do-able before. This processor, in the right hands, could increase workplace productivity enormously. Spend a cool grand on this processor, earn millions in return? Potentially.
{mospagebreak title=Overclocking}
Overclocking
Our experiences overclocking the Core 2 chips have pretty good in general so far. The X6800 was running along quite nicely at 3.68GHz stable as a rock before we started this Kentsfield review, and the E6700 overclocks very well also. We have read numerous accounts of these chips hitting around 4GHz pretty stably, and the SuperPI world record was recently broken by an infamous chap named “Coolaler” who overclocked his E6700 to 4603MHz. However, the thing that people sometimes don’t consider when talking about how well a chip overclocks is the motherboard that is facilitating the overclocking. Unfortunately for us, the two main motherboards we have overclocked the Core 2 chips with have been the Intel 975XBX, versions 1 and 2. While version 1 could hold its own in terms of maximum overclocking potential, version 2, the one that is apparently the only one that will work with the QX6700, has a ton more options and a much better BIOS layout that version 1. However, there is an FSB wall on the motherboard that will not allow processors to boot with FSB speeds above 279MHz. This is EXTREMELY frustrating, as right off the bat it was clear that the QX6700 had some serious potential. How did we discover this potential, you ask?
Well, we raised the multiplier to 11 right after we installed it to see if it would boot fine at X6800 speeds. It booted perfectly and was just as stable as at stock. Then, we bumped up the multiplier to 12, for a clock speed of 3.2GHz. It booted perfectly fine here as well, and was very stable. Then, we bumped it to 13. Very rarely does an Extreme Edition processor take multiplier bumps this well. It booted at 13, but had some stability issues. Up to this point we had not raised the voltage or changed any other settings on the board. We had to push the voltage up to 1.4875V to get the system very stable at 3.47GHz. Then, we reverted the multiplier down to 8, set the RAM frequency to go 1:1, and got ready for what we expected to be some serious FSB overclocking fun. It didn’t turn out that way. The max we were able to get to was 279MHz. The system would not successfully boot at 280MHz; it wouldn’t even try. There was no other possible reason the system would not boot at this frequency, considering at 8×280 the processor would be running far below its stock speed and the RAM would be running at less than ¾ its rated ability. Messing with the voltages, fiddling with RAM timings — nothing would make this system work with an FSB of 280MHz. These results were repeated on the X6800, which had been to well over 300MHz FSB in the past on the Bad Axe revision 1.
 |
 |
While it is a shame that the Intel X975XBX2 has this deficiency, the maximum overclock of 3630MHz we were able to achieve on the QX6700 is extremely impressive. That is nearly a full 1GHz overclock, and for those number freaks out there, is just above a 36% overclock. Theoretically, when the QX6700 is running at 3630MHz (and it does this stably, mind you), it has just over 14.5GHz of processing power. Yeah, 14.5GHz. Folding on 4 cores, anyone?
{mospagebreak title=Final Thoughts and Conclusions}
Final Thoughts and Conclusion
So what have we learned from the QX6700? Well, one thing we have learned is that four cores are better than one when it comes to multi-tasking/productivity. In a professional environment, these chips could run circles around any other chip out there. There are computers using four and more cores out there today for machines that do serious multi-tasking. However, what’s the use of four processing cores when they are all, individually, weak? Never before have four processing cores this powerful been packaged, in any way, shape, or configuration- into a single consumer- or business-level computer. This is not to say that a QX6700-based computer is faster than the supercomputers you hear beating Chess grandmasters and cracking codes, etc. No way. The processing power of those supercomputers is many thousands of times that of the QX6700. Then again, you can’t play games on Blue Gene.
Speaking of gaming, it is clear from our tests that the QX6700 at stock speeds is slower than its predecessor. This would make sense, considering the QX6700 is clocked 266MHz lower than the X6800. However, when we overclocked the QX6700 to 2933MHz to bring it up to X6800 speed, it was still slower. We found out through alternate testing that this was probably the result of inferior memory performance due to deficiencies in the memory interface, and probably the communication between the two dies on the package. However, if you are like many gamers out there that like to do other things while gaming, then the QX6700 becomes slightly more appealing. If you are one of the select few that is an EXTREMELY HEAVY multi-tasker, and we mean running massively intensive applications and gaming at the same time while eating popcorn, then the QX6700 becomes ideal. Multi-tasking is the saving grace of the QX6700, no question about it. But what about those of us who don’t really multi-task that much?
Well, unfortunately, said people will have to wait until applications come out that can actually make use of four processing cores. This is not to say that there are not programs out there that already do this. Autodesk’s 3D Studio MAX 8 and 9 are capable of working with four processors; Adobe’s Photoshop CS2, Apple’s Quicktime Pro 7.1, and several media encoding applications are as well. However, we are only just starting to see a lot of mainstream applications, namely games, that make use of two processing cores, so the timeframe for four cores in games is largely uncertain. However, we have received information from Intel about several games that are slated for release soon or are just now coming out that make use of four cores. They include: Epic’s Unreal 3 Engine-based games, Ubisoft’s Splinter Cell: Double Agent, Remedy’s Alan Wake, THQ’s Supreme Commander, and last and certainly not least, Valve’s Half-Life 2: Episode 2. That’s right, HL2:EP2 will use a multi-threaded version of the Source engine that can use four processing cores. Assuming we have any idea what we’re talking about in terms of programming, it shouldn’t then be too hard to make all source games work with four cores. The performance benefits we might see here could be tremendous. In that list of games- there are quite a few that we were already looking forward to, and now hearing about their multithreaded-ness makes us extremely excited to give them a test.
So our final thoughts about the Core 2 Extreme QX6700? Well, it is irrefutably the most powerful consumer-level processor in existence. That much is certain. However, its use as the base for a gamer’s dream machine is something that depends on the gamer and what kind of things they do besides gaming. For someone like a professional graphic artist, animator, movie editor, the QX6700 could be an invaluable resource. For someone like a guy that likes to listen to music while gaming, it is more than likely a huge waste of money. Would we recommend it to everyone? Of course not. Would we recommend it to gamers with virtually unlimited budgets? Not necessarily. At this point it probably sounds like a broken record, but the only people that we can whole-heartedly recommend this processor to are those who do HEAVY MULTI-TASKING on their computers on a daily basis. Virtually everyone else, if they have the money of course, would be better off with an X6800, as its overall performance in gaming and single-threaded applications is in most cases superior to the QX6700. Does this mean the QX6700 will never be as good as the X6800 at gaming? Absolutely not, and we look forward to the day that it is. Because that will be nuts.
Popularity: 5% [?]
You Should Also Check Out These Post:
- Valve Joins the Cloud Computing Trend
- NVIDIA Updates
- Sapphire's Radeon HD4850X2
- Get out and vote
- AMD's Deneb to Sample Late November
More Active Posts:
- Some Updates (8)
- NVIDIA has an overcomplicated lineup (8)
- What happened... (8)
- Welcome To FPSLabs V3 With Added Awesome! (7)
- NVIDIA GeForce 9800GTX on April 1st (7)
- Razer Piranha Review (7)
- Microsoft to sell XP until 2010 (6)
- Weekly Hardware Recap (6)
- The Weekly Hardware Recap (5)
- Newbie's Guide to Vista: The Look (5)
Loading ...
FramesPerSecond Labs consists of a small team of dedicated, hardware enthusiast, PC gamers who want to bring the best there is to offer in hardware reviews and advice for gamers. The team at FPSLabs strives to achieve this through integrity, passion, and a love for the one thing that matters most... the current and future FPSLabs community...
Commenting For This Post Was Disabled