|Introduction, Specification and Related Info|
For the past few years, Intel's desktop processors have been playing second fiddle to AMD's Athlon 64 in a number of ways. Both single and dual-core variants of the Athlon 64 have been at the top of most benchmark charts essentially since their introduction, and AMD's processors have been more energy efficient as well. Conversely, since the introduction of the Prescott core, Intel's processors have earned a somewhat notorious reputation for running hot and consuming more than their fare share of power.
This scenario has played out since September of 2003, when AMD first released the Athlon 64 FX-51. But it all ends today. A while back Intel announced that they would be abandoning the Netburst microarchitecture, on which all of their current Pentium 4 and Pentium D processors are based, in favor of a new microarchitecture that incorporated the best of Netburst but borrowed heavily from their low-power, high-performance Pentium-M.
The Core microarchitecture as it is now known will be the basis for a whole line of mobile (Merom), desktop (Conroe), and server (Woodcrest) processors. Over the last few months, we've posted some information and performance data regarding Conroe, but up until now we haven't been able to fully evaluate one of these processors ourselves in our own labs. Today we're going to serve up a detailed look at Conroe and hopefully explain how Intel reclaimed the performance crown from their rivals.
We have posted a myriad of information related to Intel's Core microarchitecture and Core 2 Duo and Extreme processors over the last few months here at HotHardware.com. For some more background on the technologies employed by the Core microarchitecture and Intel's platform as a whole, we suggest taking a look at few of these related articles. They contain detailed explanations of some of the features common to Intel's legacy products, compatible chipsets, and the new Core 2 Duo and Core 2 Extreme processors:
We cover some specifics regarding Intel's 65nm manufacturing process in our 955XE / i975X evaluation and outline Intel's AMT (Active Management Technology) and IVT (Intel Virtualization Technology), among other things, in our Pentium D 820 and i945G/P evaluation. We also cover more of the features integrated into Intel's dual-core processors in our Pentium Extreme Edition 840 preview. The Conroe benchmarks from this year's IDF and our June performance update will give you some background as to how the Core 2 Duo's performance has changed over the last few months, leading up to today's official launch.
|The Core Microarchitecture|
The Intel Core microarchitecture at the heart of the Core 2 Duo and Core 2 Extreme processors is a significant departure from the Netburst microarchitecture used in the Pentium 4 and the Pentium D. There are some brand new technologies employed in the Core microarchitecture that we'll break down on the pages ahead. For now though, here is a condensed list of the features and benefits provided in this new CPU architecture.
Shorter Staged Pipeline: Processors based on the Core microarchitecture, like the Core 2 Extreme X6800 pictured here, look just like the Pentium Ds they'll eventually replace in Intel's desktop processor line-up. The new Core 2 Duo and Core 2 Extreme processors are built using the same LGA775 packaging and will work with many existing chipsets and motherboards. Looks can be deceiving, however. The die that lurks beneath that nondescript heat spreader is very different from the Pentium D. For one, the Core 2 Duo and Core 2 Extreme feature much short 14-stage pipelines, as opposed to the 31 stages used in Netburst-based processors.
Wide Dynamic Execution: The new Core microarchitecture also incorporates Intel's new "Wide Dynamic Execution" technology which enables delivery of more instructions per clock cycle to improve execution time and energy efficiency. Wide Dynamic Execution gives Core 2 Duo and Core 2 Extreme processors the ability to complete up to four instructions per clock cycle, versus Netburst's three instructions. Other enhancements in the execution cores include more accurate branch predictors and deeper instructions buffers.
Another important feature of Intel's Wide Dynamic Execution technology is dubbed Macro-Op Fusion. In most current processors, each incoming instruction is individually decoded and executed. But with Macro-Op Fusion, some common instruction pairs can be combined into a single micro-op during decoding (Intel defines a micro-op as a single internal instruction). The two instructions can then be executed as a single micro-op which increases the number of instructions that can be executed in any given length of time. According to Intel, with Macro-Op fusion, instruction loads and micro-ops can be reduced by approximately 15% and 10%, respectively.
|The Core Microarchitecture (Cont.)|
Intel Advanced Smart Cache: Unlike Netburst-based, dual-core Pentium D processors that feature a certain amount of distinct L2 cache per processor core, Core microarchitecture based processors, like the Core 2 Duo E6700 and Core 2 Extreme X6800 we'll be evaluating in this article, are equipped with a single, shared cache that can be accessed entirely by either core. With current dual-core processors, each core has to store data in its own L2 cache, and that data cannot be shared between cores. With the Core microarchitecture, however, each execution core has access to the whole L2 cache (4MB or 2MB depending on the model), improving overall efficiency. Intel's Advanced Smart Cache technology also allows each core to dynamically utilize up to 100% of available L2 cache. So, if one core doesn't require large amounts of L2 cache and another does, the second core can increase its allocation of L2 cache on the fly, which will reduce the number of cache misses, and in turn increase performance.
Intel Smart Memory Access: The Smart Memory Access technology employed in the Core microarchitecture, is designed to improve overall system performance by optimizing the use of available memory bandwidth and hiding the latency associated with accessing system memory over the front side bus. Smart Memory Access incorporates a new feature called memory disambiguation that increases the efficiency of out-of-order processing through the use of built-in logic that pre-fetches and loads data for instructions before all previous load instructions are executed. Smart Memory Access uses advanced algorithms to evaluate whether or not a load can be executed ahead of a preceding store. If it can, then the load instructions can be scheduled before the store instructions to enable increased instruction-level parallelism. Intel Smart Memory Access also incorporates new memory pre-fetchers that grab data and store it in the processor's cache before it's specifically requested, so that data can be readily accessed from fast on-die cache when needed. The Core microarchitecture uses two prefetchers per L1 cache and two prefetchers per L2 cache.
Intel Advanced Digital Media Boost: With Intel's previous generation of processors, 128-bit SSE, SSE2, and SSE3 instructions were typically executed at a rate of one complete instruction every two clock-cycles. With the Core microarchitecture based processors that feature Intel's Advanced Digital Media Boost feature though, 128-bit Streaming SIMD Extension (SSE) instructions can typically be completely executed in a single clock cycle, which effectively doubles the speed at which instructions of this type can be executed. Advanced Digital Media Boost will come into play when executing multimedia operations that involve graphics, video and audio that use SSE, SSE2, or SSE3 instructions.
Intel Intelligent Power Capability: As its name implies, Intelligent Power Capability is a set of features designed to reduce overall power consumption. This feature is in control of the power consumption of all of a Core-based processor's execution cores. Intelligent Power Capability incorporates fine-grained logic control that enables individual components within the processor core only when they are needed. In addition to this capability, many arrays within the CPU are split so that data required in some modes of operation can be put into a low-power state when it's not in use. Reductions in power consumption also come by way of enhancements made to Intel's 65nm manufacturing process node, like the use of Low-K dielectrics and strained silicon, and through the use of low-leakage and "sleep" transistors and Enhanced SpeedStep.
|The Core 2 Duo E6700 and Core 2 Extreme X6800|
For the past few weeks, we've been testing a pair of new processors from Intel, the Core 2 Duo E6700 and the new flagship Core 2 Extreme X6800. Before we got down and dirty with the benchmarks, we spent some quality time with these two processors to get a feel for their performance in day-to-day use, and also spent some time overclocking. For now, let's get some particulars out of the way...
First up, we have some information courtesy of CPU-Z. We fired up CPU-Z to get a closer look at the internal workings of the Core 2 Duo E6700 and Core 2 Extreme X6800 processors. Please note, that the 1600MHz clock speed reported by CPU-Z for both processors was a result of Intel's SpeedStep technology cranking down the CPU frequency while idling to conserve power. Also note that a full 4MB of cache is reported for both CPU cores. These processors do not sport 8MB of total L2 cache, but because Intel's Advanced Smart Cache technology can dynamically allocate some or all of the cache to each core, CPU-Z is "tricked" into reporting double the amount of actual L2 cache per core.
The first pair of screenshots on the left contain details regarding the Core 2 Duo E6700 processor. Although a lower speed is reported here due to SpeedStep, the Core 2 Duo E6700 actually runs at a default frequency of 2.66GHz (10 x 266MHz) with a 1066MHz FSB (266MHz Quad-Pumped). The Core 2 Extreme E6800 is similar, but it has a default multiplier of 11, for a final clock speed of 2.93GHz. The Core 2 Extreme E6800 is also "unlocked", so enthusiasts can alter the multiplier manually with some advanced motherboards for more flexible overclocking. Voltages vary depending on load, but in their idle state, CPU-Z reported a scant 1.2v supplying each CPU.
Both processors are manufactured on Intel's 65nm process node, and unlike the CPU's we tested last month in a closed door session in NYC, both of the processors seen here are based on the same stepping and revision, F and B1, respectively.
Overclocking: Due to the fact that the new Core 2 micro-architecture features a shorter pipeline than current Netburst-based Pentium D processors (14 stages vs. 31 stages), some analysts were concerned that Intel wouldn't be able to hit sufficiently high clock speeds at launch to push the Core 2 family of processor's performance much higher than currently available processors. Looking back at our Conroe performance "sneak peak" articles, you already know that didn't turn out to be the case and the Core 2 Duo is performing just fine, even though the highest clocked Core 2 processor has a frequency almost 900MHz lower than Intel's fastest Pentium D.
Of course, with a shorter pipeline, the higher clock speeds of current Pentium D processors will not be possible with a Core 2 until more advanced manufacturing processes are used. But there is always some headroom left under the hood, so we overclocked the E6700 and X6800 to see just how fast these early samples would go. In the end, we had trouble hitting a stable 3.3GHz with the E6700. We're unsure if the motherboard or the CPU was hitting a wall, however, and will have to experiment further. Our Core 2 Extreme E6800 though, turned out to be a demon of an overclocker. With a simple bump in voltage up to 1.425v, and FSB manipulation, we were able to hit a completely stable 3.55GHz using a Thermaltake Mini-Typhoon air cooler. We've got some screenshots with Cinebench and SANDRA performance data above to give you an idea as to just how fast the Core 2 Extreme E6800 was at this high clock speed.
We also have some information regarding temperatures with both processors running at their stock and overclocked speeds. The Core 2 Extreme E6800 typically idled at about 42oC and peaked at about 56oC under load when at its default 2.93GHz. While overclocked though, the X6800 idled at about 53oC and peaked at a toasty 66oC. The E6700 ran somewhat cooler with stock idle and load temps of 41oC and 49oC, and overclocked temperatures of 45oC and 56oC. Of course, these are "only" the temperatures as they're interpreted by Intel's D975XBX motherboard. A different motherboard would likely report different temperatures.
|Chipsets and Motherboards: 975X and P965|
Intel DX975XBX Rev. 2: We used a number of motherboards to test Intel's new Core 2 processors, a couple of which you've probably seen before. The first motherboard we used was Intel's own D975XBX, which is based on the 975X Express chipset. We've already shown you this motherboard in our coverage of the Pentium Extreme Edition 955, but the latest revision has been outfitted with a new voltage regulator module and a new BIOS to fully support the lower voltages and features inherent to Intel's new micro-architecture.
The D975XBX's features and specifications read like some of the enthusiast-class boards that come from partners like Asus or Abit. The D975XBX features three physical PCI Express X16 slots (with varying electrical connections) and supports ATI's CrossFire multi-GPU technology. NVIDIA's SLI technology isn't supported, however. By simply inserting a second video card, the first and second PEG slots' PCI Express lane configuration is automatically changed to an 8x8 configuration.
The D975XBX has a 5-phase power array and each FET in the array is adorned with a relatively large aluminum "flame" heatsink. The Northbridge and Southbridge are also equipped with their own aluminum heatsinks as well. Four of the board's SATA ports are controlled by the ICH7R Southbridge and the other four by a Silicon Image controller. Its got 4 DDR2 DIMM slots, Gigabit Ethernet, and HD Audio support courtesy of a Sigmatel CODEC.
Intel recently released the new P965 chipset with Core 2 support as well. Although you'd think the "P" in P965 and this chipset's coincidental release just prior to the introduction of the Core 2 would mean this is Intel's performance chipset, it's not. The P965 is being marketed as a mainstream chipset, and as such it'll command a more palatable mainstream price than the 975X Express. Unlike the very similar i875 and i865 chipsets from days past, the P965 and 975X Express are actually quite different.
With the P965, Intel is introducing a new I/O controller hub, the ICH8, that features 6 SATA 3Gb/s channels, along with 6 PCI Express lanes, Matrix Storage Technology, 10 USB ports, and of course HD audio and GbE LAN support. Noticeably absent from the ICH8's specifications is PATA support. With this chipset, Intel is going all SATA, which means we'll likely see an influx of SATA optical drives in the not so distant future. Another difference between the P965 and 975X Express is the chipset's PCI Express lane configuration for graphics. Whereas the 975X Express has the flexibility to be set up in an 8x8 or a 16x1 lane configuration when dual graphics cards are used, the P965 will offer only 16x4 and 16x2 configurations.
Asus P5B Deluxe WiFi-AP Edition: The P965 based motherboard we used for testing is a brand new member of Asus' AI Lifestyle line of products, the P5B Deluxe WiFi-AP Edition. In typical Asus fashion, this motherboard is chock full of integrated peripherals, including an 802.11g wireless NIC. Other features of this board include dual-PEG slots, power LEDs, an 8-Phase power array, Firewire support, dual Gigabit LAN, and Asus proprietary features like AI NOS (automatic overclocking), AI Gear (overclocking / power profiles), and AI Nap (low-power standby).
As you can see in the pictures above, although the board is equipped with the ICH8. The Asus P5B Deluxe does have a PATA port. PATA support comes by way of a JMicron JMB363 PATA / SATA controller, that also happens to power this board's eSATA port.
The Asus P5B Deluxe WiFi-AP Edition also features a very complete system BIOS, loaded with overclocking and performance related features. Although not officially supported by Intel, this board gives users the ability to dial in RAM frequencies of 1066MHz and higher without overclocking the processor, and there are a variety of tweakable voltages as well. Using this board and a combination of multiplier, voltage, and FSB tweaks, we had no trouble taking our Core 2 Extreme X6800 processor to well over 3.5GHz. It may be based on a mainstream chipset, but this motherboard has "power user" written all over it.
|Chipsets and Motherboard: NF4 SLIX16|
The third motherboard we used to test Intel's new Core 2 processors was based on the NVIDIA nForce 4 SLIX16 Intel Edition chipset. The Asus P5N32-SLI SE is a revision to a motherboard we've already reviewed here at HotHarware, the P5N32-SLI Deluxe. However, the P5N32-SLI SE is essentially revision 2.0G of the P5N32-SLI Deluxe. The majority of the motherboard remained unchanged, but like Intel did with the D975XBX, the P5N32-SLI SE has an updated BIOS and a revamped VRM to fully support the Core 2 Duo and Extreme line of processors.
We detailed many of the features of the nForce 4 in our coverage of the first nForce 4 SLI Intel Edition chipset last year. In addition to what's mentioned in that article, the nForce 4 SLIX16 sports a number of additional features including true dual-PCI Express x16 graphics slots allowing full bandwidth SLI and support for all Intel dual core processors.
The P5N32-SLI SE also sports Asus' "two-slot thermal design" which consists of a pair of heatsink / heatpipe assemblies that cool the VRM, SPP, and MCP. It's equipped with an 8-Phase power array, eSATA (aka SATA on the Go), and it supports some of the Asus' proprietary features we mention on the previous page, like AI NOS and AI Quiet. Speaking of quiet, the P5N32-SLI will not contribute to the noise level of a system because it boasts a completely fanless design. Of course, in typical Asus fashion the P5N32-SLI SE's BIOS is loaded with overclocking and tweaking options too.
We ran our entire suite of benchmarks on this motherboard using a couple of Corsair DDR2 memory kits, and can wholeheartedly recommend this motherboard to anyone who wants to build a Core 2 based system with SLI support. This motherboard, in conjunction with Corsair's TWIN2X2048-6400C3 memory kit, a pair of GeForce 7900 GTX cards, and a Core 2 Extreme X6800 processor posted some of the highest benchmark scores we have ever seen.
|Our Test System and SiSoft SANDRA 2007|
How we configured our test systems: When configuring our test systems for the following set of benchmarks, we first entered their respective system BIOSes and set each board to its "Optimized" or "High-Performance Defaults." We then saved the settings, re-entered the BIOS and set memory timings for DDR2-800 at 4,4,4,12 1T latency. The hard drives were then formatted, and Windows XP Professional (SP2) was installed. When the Windows installation was complete, we installed the drivers necessary for our components, and removed Windows Messenger from the system. Auto-Updating and System Restore were then disabled, and we set up a 768MB permanent page file on the same partition as the Windows installation. Lastly, we set Windows XP's Visual Effects to "best performance," installed all of our benchmarking software, defragged the hard drives, and ran all of the tests.
We began our testing with SiSoftware's SANDRA, the System ANalyzer, Diagnostic and Reporting Assistant. We ran five of the built-in subsystem tests that partially comprise the SANDRA 2007 suite (CPU, Multimedia, Memory, Cache, and Latency) with the new Core 2 Duo E6700 and Core 2 Extreme X6800 processors. All of the scores reported below were taken with the processors running at their default clock speeds of 2.66GHz and 2.93GHz, respectively.
The lite-duty SANDRA benchmarks tell an interesting story. The CPU arithmetic benchmark shows the Core 2 processors performing very well in the ALU portion of the test, but in the FPU test the Core 2 CPUs hang alongside an FX-62 and get thrashed by Netburst based Pentium Extreme Editions. According to the multi-media benchmark though, the Core 2 Duo is simply in a league of its own -- nothing comes close to the 6 digit scores put up by the E6700 and X6800.
SANDRA's memory benchmark isn't very interesting because nothing has really changed in regard to available bandwidth over current systems. Because the Core 2 is compatible with current chipsets, it's bound by the same memory controller used in today's Pentium D based systems. Their 5.6GB/s+ scores put the Core 2 processors we tested on-par with their Pentium powered counterparts. We should note, however, that we also ran a few tests on the nForce 4 SLIX16 based Asus board using Corsair's TWIN2X2048-6400C3 memory kit and the X6800 and bandwidth scores peaked at almost 5.8GB/s.
The Cache and Latency tests also produced some interesting results. Because these benchmarks can reside solely in each CPU's 4MB of cache, latency and cache bandwidth is much better on the Core 2 architecture versus the Athlon 64, even though the A64 features an on-die memory controller.
|Futuremark PCMark05: CPU and Memory|
For our next round of synthetic benchmarks, we ran the CPU and Memory performance modules built into Futuremark's PCMark05 suite.
"The CPU test suite is a collection of tests that are run to isolate the performance of the CPU. The CPU Test Suite also includes multithreading: two of the test scenarios are run multithreaded; the other including two simultaneous tests and the other running four tests simultaneously. The remaining six tests are run single threaded. Operations include, File Compression/Decompression, Encryption/Decryption, Image Decompression, and Audio Compression" - Courtesy FutureMark Corp.
From this point forward, we'll be comparing Intel's new Core 2 Duo E6700 and Core 2 Extreme X6800 processors to a Pentium Extreme Edition 965, an Athlon 64 FX-62, and an Athlon 64 5000+. Also note that we've benchmarked the Core 2 Extreme X6800 on a trio of different chipsets, the 975X Express, the new P965, and the nForce 4 SLIX16. Our goals were to give you all an idea as to how Intel's new processors perform versus their main competition and on a variety of platforms.
As you can see, the chipset used has little bearing on performance in PCMark05's CPU benchmark when coupled to a Core 2 Extreme processor. Both Intel chipsets and the nForce 4 SLIX16 perform similarly. The Core 2 micro-architecture, however, gives Intel's new processors a marked edge over its competition, besting the Athlons and the XE965 by considerable margins.
"The Memory test suite is a collection of tests that isolate the performance of the memory subsystem. The memory subsystem consists of various devices on the PC. This includes the main memory, the CPU internal cache (known as the L1 cache) and the external cache (known as the L2 cache). As it is difficult to find applications that only stress the memory, we explicitly developed a set of tests geared for this purpose. The tests are written in C++ and assembly. They include: Reading data blocks from memory, Writing data blocks to memory performing copy operations on data blocks, random access to data items and latency testing." - Courtesy FutureMark Corp.
PCMark05's memory performance module tells a similar story. Both the Core 2 Duo E6700 and Core 2 Extreme X6800 significantly outperform the Pentium Extreme Edition and both Athlons. The Core 2 processors' increased cache size, deeper buffers, and Smart Cache technology can be credited with the performance boost here, as the XE965 was tested using the same 975X Express chipset and Corsair DDR2 memory.
|WB5: Office XP SP2 and Photoshop 7.0|
PC World Magazine's WorldBench 5.0 is a new breed of Business and Professional application benchmark, that has replaced the aging and no-longer supported Content Creation and Business Winstone tests in our suite. WorldBench 5.0 consists of a number of performance modules that each utilize one, or a group of popular applications to gauge performance.
Below we have the results from WB 5's Office XP SP2 and Photoshop 7 modules, recorded in seconds. Lower times indicate better performance here, so the shorter the bar the better.
WorldBench 5.0's Office XP SP2 benchmark isn't quite as intensive as some of the suites' other tests, but Intel's new processors still manage to outpace their competition by about 10 to 20 seconds depending on the specific comparison. Obviously, any of the processors tested here are plenty powerful enough to run a typical office suite though.
The Photoshop 7.0 benchmark shows a much larger delta separating the test systems. The Core 2 Duo E6700 and Core 2 Extreme X6800 processors finished this test about 70 to 105 seconds faster than the competition, regardless of which chipset was powering the system. That's a huge advantage in a very popular application. We're sure some of you image manipulation experts are drooling over these results.
|WB5: 3ds Max and Nero Burning ROM|
We continued our testing of the new Core 2 Duo E6700 and Core 2 Extreme X6800 processors and chipsets with a few more tests that are part of the WorldBench 5.0 suite. Up next we have some performance results from WB 5.0's 3Ds Max (Direct 3D) and Nero Burning ROM tests.
A number of different 3D objects are rendered and animated in this test, and the entire time needed to complete the tasks is recorded. As is the case with all of the individual WorldBench tests, a lower time here indicates better performance.
Once again, we see Intel's latest flagship processors trouncing their competition. The 975X Express and P965 chipsets performed similarly, with the nForce 4 SLIX16 coming in a few seconds faster, but all of the Core 2 based systems were faster than their nearest competitor by about 10% to 15%. We need to point out a special case here, however. The Athlon 64 FX-62 score was recorded with the processor installed on a reference ATI RD580 board, the 5000+ score was recorded using an nForce 590 SLI. We uncovered a glitch with the nForce 590 SLI that causes lower performance than expected in this test, so to put the FX-62's performance in a more accurate light, we used the RD580's score. We've been in contact with NVIDIA over this issue and suspect a future driver update will bring the nForce 590 SLI's performance up to match other AM2 compatible chipsets.
Here's one for the record books folks. The Core 2 Duo E6700 and Core 2 Extreme X6800 powered systems simply dominated the competition in WorldBench 5.0's Nero Burning ROM benchmark. We suspect the Core 2's dominant performance can be attributed to the new processor's Smart Cache and Smart Memory Access technologies. The E6700's and X6800's 4MB of Smart Cache and Smart Memory Access features lower the number of instances where the CPU must access main memory, which increases performance. If we run this same test with 2GB of RAM installed in a XE965 or Athlon system instead of 1GB like we did here, their scores drop significantly.
|WB5: WME 9 and Mozilla MT|
For our next test, we moved on to a benchmark based on Windows Media Encoder 9. PC WorldBench 5's Windows Media Encoder and Mozilla multi-tasking test reports encoding times in seconds, and like the tests on the previous page, lower times indicate better performance.
In this test, a video is encoded using Windows Media Encoder 9, while an instance of the Mozilla browser is running and navigating through various cached pages in the background. Because the system is multi-tasking with two different applications, this test is more taxing than just running one instance of WME.
The Core 2 Duo E6700 and Core 2 Extreme X6800 posted another set of dominant scores in this multi-tasking benchmark. Intel's new processors were almost twice as fast as the Pentium Extreme Edition 965 here, and between 53 and 149 seconds faster than either Athlon 64 we tested.
We also found a small anomaly with the P965 chipset here that resulted in lower performance. We suspect the relative immaturity of the P965 is what caused the lower than expected results in this benchmark, and believe that its performance will increase as motherboard manufactures and Intel have more time to tweak and tune the chipset drivers and motherboard BIOSes.
|LAME MT: MP3 Encoding|
In our custom LAME MT MP3 encoding test, we convert a large WAV file to the MP3 format, which is a very popular scenario that many end users work with on a day-to-day basis to provide portability and storage of their digital audio content.
In this test, we created our own 223MB WAV file (a never-ending Grateful Dead jam) and converted it to the MP3 format using the multi-thread capable LAME MT application in single and multi-thread modes. Processing times are recorded below. Once again, shorter times equate to better performance.
Once again, the new Core 2 Duo E6700 and Core 2 Extreme X6800 processors put up some impressive scores. Our custom LAME MT encoding test had the Core 2 Extreme X6800 finishing the encoding process between 10 and 30 seconds faster than the XE965 and either Athlon 64. And the Core 2 Duo E6700 was right behind the X6800, trailing Intel's new flagship by only 5 seconds in either test.
We also found that all of the chipsets we tested offered similar performance in this test. As an additional experiment, we installed Corsair's low-latency 6400C3 kit in the nForce 4 SLIX16 based system, but we didn't see any difference in performance. In fact, with 2GB of CAS2 RAM installed on the nForce 4 SLIX16 board, it finished in exactly the same amount of time.
For this next batch of tests, we ran Kribibench v1.1, a 3D rendering benchmark produced by the folks at Adept Development. Kribibench is an SSE aware software renderer where a 3D model is rendered and animated by the host CPU, and the average frame rate is reported.
We used two of the included models with this benchmark: a "Sponge Explode" model consisting of over 19.2 million polygons and the test suite's "Ultra" model that is comprised of over 16 billion polys.
The Core 2 Duo E6700 and Core 2 Extreme X6800 processors posted the highest scores we have seen to date in the two Kribibench tests we ran. We hesitate to say they "tore" through this benchmark with scores roughly in the 2 to 6 frame per second range, but Intel's new processors clearly outperformed the Pentium Extreme Edition 965 and both Athlon 64s.
This was one of the few tests where the chipset had a measurable impact on overall performance. In both tests, the nForce 4 SLIX16 based system was fastest, followed by the P965 and then the 975X Express.
|Cinebench and 3DMark06: CPU|
The Cinebench 2003 benchmark is an OpenGL 3D rendering performance test, based on the commercially available Cinema 4D application.
This is a multi-threaded, multi-processor aware benchmark that renders a single 3D scene and tracks the length of the entire process. The time it took each test system to render the entire scene is represented in the graph below (listed in seconds).
Chalk up another victory for the Core 2 micro-architecture. Whether in single or multi-thread mode, regardless of which chipset was at the heart of the system, the new Core 2 Duo E6700 and Core 2 Extreme X6800 processors significantly outpaced the Pentium Extreme Edition 965 and both Athlon 64s. The X6800 was between 7 and 14 seconds faster than a FX-62, and the E6700 was roughly 5 to 8 seconds faster as well.
This was another application we tested using Corsair's TWIN2X2048-6400C3 memory kit on the nForce 4 SLIX16. Using higher capacity, lower-latency memory improved the X6800's performance to 51.6 and 27.9 seconds in the single and multi-threaded tests, respectively.
3DMark06's built-in CPU test is a multi-threaded "gaming related" DirectX metric that's useful for comparing relative performance between similarly equipped systems. This test consists of two different 3D scenes that are generated with a software renderer, which is dependent on the host CPU's performance. This means that the calculations normally reserved for your 3D accelerator are instead sent to the central processor. The number of frames generated per second in each test are used to determine the final score.
Intel's new processors also performed quite well in 3DMark06's built-in CPU benchmark. In this test, the E6700 was roughly 10% faster than anything from AMD and the X6800 was faster still, coming in about 19% ahead of the Athlon 64 FX-62.
|Low-Res Gaming: F.E.A.R. and Quake 4|
To start our in-game testing, we did some low-resolution benchmarking with F.E.A.R. When testing processors with F.E.A.R, we drop the resolution to 640x480, and drop all of the in-game graphical options to their minimum values to isolate CPU and memory performance as much as possible.
In the Conroe performance "sneak peek" articles we've posted over the last few months, Intel used F.E.A.R. as a showcase for their Core 2 line of processors. The reason? Intel's new processors handily outpace anything currently available from AMD here, and absolutely crush the Pentium Extreme Edition 965.
For our next game test, we benchmarked all of the test systems using a custom single-player Quake 4 timedemo. Here, we installed the game's official v1.2 patch which is SMP capable, tuned the resolution down to 640x480, and configured the game to run at its "Low-Quality" graphics setting. Although Quake 4 typically taxes today's high-end GPUs, when it's configured at these minimal settings, it too is more CPU and memory-bound than anything else.
The new Core 2 Duo E6700 and Core 2 Extreme X6800 also performed extremely well in our custom Quake 4 benchmark. In these tests, the E6700 was about 19 FPS faster than an FX-62 and the X6800 was 30 to 40 frames per second faster depending on the chipset used. And of course, the Pentium Extreme Edition 965 and Athlon 64 5000+ trail the Core 2 processors by even larger margins. Interestingly, the P965 / X6800 combo put up the best scores, followed by the 975X Express and then the nForce 4 SLIX16.
|High-Res Gaming: F.E.A.R. and Quake 4|
To see how Intel's new Core 2 processors would fare in a typical high-end gaming scenario, we also tested the CPUs with some popular games at high-resolution settings that taxed the graphics sub-system of each of the platforms.
For these tests, we installed a second GeForce 7900 GTX into our nForce based tests systems and enabled SLI multi-GPU rendering. As you can see, when the graphics sub-system becomes the bottleneck, performance between the different platforms and processors levels out somewhat. But because the F.E.A.R. benchmark also incorporates physics and AI unlike a typical timedemo, faster processors also boost performance a bit, as is evident by the X6800's and E6700's first and second place finishes.
Our custom Quake benchmark basically told the same story as F.E.A.R. Here, all of the test systems were tightly grouped with the Core 2 Extreme X6800, once again finishing at the top of the charts. The FX-62 pulled off a second place finish followed by the E6700 and then the Athlon 64 5000+.
We have one final data point we'd like to cover before bringing this article to a close. Our goal was to give you all an idea as to how much power each configuration used while idling and running under load.
Please keep in mind that we were testing total system power consumption here at the outlet, not just the power being drawn by the processors alone. In this test, we're showing you a ramp-up of power from idle on the desktop to full CPU load and then a ramp to full CPU and Graphics load. We tested with a combination of Prime 95 and Folding@Home on the CPU.
Intel's new Core 2 processors are significantly more power-friendly than the Netburst-based Pentium Extreme Edition 965 and even a bit more green than AMD's Athlon 64 FX-62 when coupled to an Intel-built chipset. You may be surprised to see power consumption numbers that are relatively in-line with other architectures, but you must remember that performance is drastically increased with the new Core 2 processors.
If we used the closest to apples-to-apples comparison, the X6800 / NF4 versus the FX-62 / NF 590 SLI there is a 36 watt delta separating the X6800 from the FX-62 under load. So, while not only does the X6800 use less power when running at a full load, it can complete most jobs faster than the FX-62. This is where the Core 2's power efficiency will be far superior to most other architectures. It's not necessarily the processor's peak power consumption that'll be lower, but its total power consumption over time.
While we're talking about power, we should also point out the differences between the various Core 2 compatible chipsets. In our tests, the 975X based systems used the least power, followed by the P965 and then the nForce 4 SLIX16. However, please be aware that the P965 based Asus P5B motherboard we used for testing was loaded with integrated peripherals. It's the board's additional components that resulted in the high power usage and not anything inherent to the chipset. The nForce 4 SLIX16, however, is contenting with a couple of issues. The nForce 4 SLI SPP is manufactured on TSMC's .09 micron line, but the MCP is still manufactured at .13 micron. This combination of older and newer process technology results in higher power consumption as you can see in the comparison versus the RD580. We also suspect the nForce 4 SLIX16 wasn't properly utilizing the C1E halt state, which is why the Core 2 / nForce 4 SLIX16 combo used so much more power while idling. This is an issue that will likely be resolved in a future BIOS update, however.
|Our Summary & Conclusion|
Benchmark And Performance Summary:
Core 2 compatible chipset performance is also relatively clear. All of the chipsets we tested, the 975X Express, the P965, and the nForce 4 SLIX16, performed similarly throughout our entire array of benchmarks. There was some variation in a few tests likely due to drivers or relative immaturity, but there is no one chipset we tested that clearly outpaced another.
Intel's Core 2 Duo and Core 2 Extreme are almost here, and they seem to be everything Intel promised and then some. The Core 2's 4-issue core, shorter pipeline, Smart Memory Access and Advanced Smart Cache technologies make Intel the performance leader once again. It was a tough few years for Intel, as AMD's Athlon 64, FX, and X2 processors outperformed their products almost across the board. But Intel isn't playing second fiddle to AMD anymore. Make no mistake, the Core 2 Duo and Core 2 Extreme are very real, and their performance is undeniable.
There are some questions that remain, however. The chart above illustrates Intel's future desktop product line-up, with MSRPs for each. On July 27th, enthusiast-class Core 2 processor based systems should be available from some vendors, and we suspect some processors will show up at a few e-tailers as well, but volume won't truly ramp up until the August / September timeframe. Between now and then, we're unsure what pricing will actually be like for the end user. Considering the Core 2's dominant performance and the likelihood of limited availability, street prices will probably be quite high for some time.
We're also unsure how the lower-end E6400 and E6300 will perform with "only" 2MB of cache. Representatives tell us performance is very good and that we'll have samples sometime soon, but we won't know for sure until we test one for ourselves.
The last question is AMD's response. The company has been quiet as of late, but if Core 2 processors aren't widely available for some time and street prices are excessively high, AMD may not feel the need to be vocal about upcoming products just yet. We know AMD's 4x4 platform that'll feature a pair of dual-core FX processors is on the way, the FX-64 is in the works as well, the company's got low-power X2s available, they will eventually move to 65nm, and the K8L could arrive anytime between Q2 2007 and Q2 2008 if current rumors are to be believed. AMD will surely have an answer for the Core 2 Duo in time. Whether or not it's enough in the eyes of consumers is another question, however. For now though, Intel is riding high. "Intel - Leap Ahead". That's exactly what they just did.