Logo   Banner   TopRight
AMD Athlon 64 FX-60 - Finally, An Enthusiast's Dual Core
Date: Jan 09, 2006
Author: Dave Altavilla
Intro and Specs


Intel's launch of their next generation Pentium Extreme Edition 955 dual-core processor last month, marked a solid improvement in overall performance for what is to be the last core iteration utilizing the Pentium 4 Netburst architecture.  While this Intel dual-core kicker, essentially beefing up a pair or Prescott dies in a single package, was a welcomed addition in performance and an impressive feat in 65nm process technology, thermal characteristics of the new Presler core Pentium were still pretty toasty and performance wasn't enough to catch AMD's high end Athlon 64 FX-57 or their Athlon 64 X2 4800+ dual core chip in many of our real world benchmark tests.

And today AMD lets fly with yet another volley of the dual-core persuasion, this time targeted squarely at the Enthusiast and Power User.  While Intel's branding of their Pentium Extreme Edition dual-core CPU as "extreme" conjures up thoughts of best-in-class performance for the enthusiast, whether running multimedia apps, gaming or otherwise, it is painfully obvious that AMD's dual-core Athlon 64 X2 line-up is significantly more capable and well rounded in terms of performance.  Regardless, AMD held off branding their high-end Athlon 64 X2 4800+ a true enthusiast dual core chip, leaving the "FX" moniker for the 2.8GHz single core Athlon 64 FX-57.  With today's launch of the Athlon 64 FX-60 though, AMD is clearly positioning the new dual core as an all around high-end enthusiast class processor and with a pair of 2.6GHz Athlon 64 cores under the hood, essentially a pair of Athlon 64 FX-55 chips if you will, you could definitely speculate that their positioning is fair and accurate.  But why speculate?  That's why we're here.  Read on, as we've put AMD's new Athlon 64 FX-60 to the test in the pages ahead.

AMD Athlon 64 FX-60 - A 2.6GHz Dual Core
Specifications & Features
Dual Athlon 64 FX Cores Clocked at 2.6GHz

AMD64 - When utilizing the AMD64 Instruction Set Architecture, 64-bit mode is designed to offer:

__Support for 64-bit operating systems to provide full, transparent, and simultaneous 32-bit and 64-bit platform application multitasking.

__A physical address space that can support systems with up to one terabyte of installed RAM, shattering the 4 gigabyte RAM barrier present on all current x86 implementations.

__Sixteen 64-bit general-purpose integer registers (per core) that quadruple the general purpose register space available to applications and device drivers.

__Sixteen 128-bit XMM registers (per core) for enhanced multimedia performance to double the register space of any current SSE/SSE2 implementation.

Integrated DDR memory controller:
__Allows for a reduction in memory latency, thereby increasing overall system performance.

An advanced HyperTransport link:
__This feature dramatically improves the I/O bandwidth, enabling much faster access to peripherals such as hard drives, USB 2.0, and Gigabit Ethernet cards.

__HyperTransport technology enables higher performance due to a reduced I/O interface throttle.

Large level one (L1) and level 2 (L2) on-die cache:
__With 128 Kbytes of L1 cache and 1MB of L2 cache per core, the AMD Athlon 64 FX-60 and Athlon 64 X2 processors are able to excel at performing matrix calculations on arrays.

__Programs that use intensive large matrix calculations will benefit from fitting the entire matrix in the L2 cache.

64-bit processing:
__A 64-bit address and data set enables the processor to process in the terabyte space.

__Many applications improve performance due to the removal of the 32-bit limitations.
Processor core clock-for-clock improvements:
__Including larger TLB (Translation Look-Aside Buffers) with reduced latencies and improved branch prediction through four times the number of bimodal counters in the global history counter, per core, as compared to seventh-generation processors.

__These features drive improvements to the IPC, by delivering a more efficient pipeline for CPU-intensive applications.

__CPU-intensive games benefit from these core improvements.

__Introduction of the SSE2 instruction set, and now SSE3 (Rev. E and Athlon 64 X2) which along with support of 3DNow! Professional, (SSE and 3DNow! Enhanced) completes support for all industry standards.

__32-bit instruction set extensions.

Fab location:
__AMD's Fab 30 wafer fabrication facility in Dresden, Germany

Process Technology:
__.09 micron SOI (silicon-on-insulator)

Die Size:
__Toldeo - 199mm2

Transistor count:
__Toldeo Core - Approximately 233.3 million

Nominal Voltage:

Beyond the obvious clock speed boost, not much has changed with respect to the inner workings or architecture of the Athlon 64 FX-60, versus AMD's next fastest dual core CPU, the Athlon 64 X2 4800+.  The only real difference between the two processors, other than their clock speeds, are the fact that the FX-60 has the ability to run at higher and lower core multipliers for added flexibility in overclocking.  The Athlon 64 X2 4800+ can only have its multiplier lowered but not increased beyond its standard 12X setting.  The Athlon 64 FX-60, however, comes with a standard core clock multiple of 13X and we were able to run it higher as well.  We'll show you more details in this area shortly but the primary benefit for a higher core multiplier is the ability to leave system clock speeds and the memory controller clock speed, at their stock 200MHz setting, which allows memory timings to remain within specification while overclocking the CPU.

Core Details And Vital Signs


Same socket 939 micro-PGA package, only with two 2.6GHz Athlon 64s under the hood....


The Athlon 64 FX-60 Dual Core Processor
High-Level System Block Diagram

Below is a block diagram of the Athlon 64 FX-60 dual core processor.  We've covered the technology and system architecture in some detail, in previous articles, so we won't re-hash much here.  However, we did want to highlight a couple of key architectural differences between AMD's dual core architecture and Intel's.

Integrated Memory Controller and HyperTransport Links:
AMD's integrated memory controller on the Athlon 64 has afforded the architecture a considerable latency advantage for many generations of the processor.   Conversely, this is a functional block that Intel breaks out into the Northbridge of their core logic chipsets.  This does afford Intel the flexibility of being able to migrate to next generation System Memory technologies more easily but at again, at the cost of latency.   In addition, AMD's bi-directional serial HyperTransport link that interfaces the CPU to the system core logic chipset, PCI Express ports and I/O expansion, is also a higher speed lower latency dedicated point-to-point architecture, versus Intel's QDR bus architecture, that is shared by Memory, PCI Express and I/O resource calls.  Frankly, it's somewhat surprising to us that Intel hasn't made the move to serial interfaces on their core CPU architecture, as the benefits of latency and pin-count reduction are obvious.

System Request Interface and Crossbar Switch:
Finally, the Athlon 64 FX-60 employs the same System Request Interface (or queue manager if you will) and Crossbar Switch architecture as the Athlon 64 X2 series of dual-core chips, allowing more efficient core-to-core communication, than the current Intel dual-core architecture, which require the cores to message each other over the front side bus.  There are inherent latency trade-offs within a Crossbar Switch architecture as well but without question this is a more elegant and forward looking solution than a shared bus architecture, no matter how you slice it.  With the Athlon 64 X2 and Athlon 64 FX-60 core-to-core communications reside on chip.

Initial Tests and Vital Signs
A True Dual-Core FX


Above are a few screen captures from Windows Task Manager's "Performance" tab which shows our Athlon 64 FX-60 at 50% CPU utilization driven by running one instance of Prime 95 for this simple test.  Here you can see the load-balancing that is going on between the CPU cores.  In addition, are screen shots from a recent version of CPU-Z which shows the Athlon 64 FX-60's dual 2.6GHz cores and their per-core dedicated 1MB of L2 cache and the 128-bit System Request Interface, which is indicated in the "bus-width" section under the cache tab.

Test System & System Power Characterization


How we configured our test systems: When configuring the test systems for this review, we first entered their respective system BIOSes and set each board to its "Optimized" or "High-Performance Defaults."  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.

Test System Specifications
"Intel & AMD Inside!"
Pentium Extreme Edition 955
Pentium Extreme Edition 840
3.73GHz Pentium Extreme Edition

Intel "BadAxe" Motherboard
(i975x Chipset)

2x512MB Corsair DDR2-667
CL 3-2-2-8

GeForce 7800 GTX
On-board Ethernet
On-board Audio

WD "Raptor" HD
10,000 RPM SATA

Windows XP Pro SP2
Intel INF

NVIDIA Forceware v81.95
DirectX 9.0c
AMD Athlon 64 FX-60 (2.6GHz)
AMD Athlon 64 X2 4800+
AMD Athlon 64 FX-57

Asus AN832-SLI
(NVIDIA nForce 4 SLI X16)

2x512MB Corsair PC3200
CL 2-2-2-5

GeForce 7800 GTX
On-board Ethernet
On-board Audio

WD "Raptor" HD
10,000 RPM SATA

Windows XP Pro SP2
nForce 4 Drivers v6.82
NVIDIA Forceware v81.95
DirectX 9.0c
System Power Characterization
Affects of CPU power consumption in total system power requirements

Before we got down to performance testing of AMD's new high-end dual-core CPU, we wanted to do a bit of power characterization of the product.  Power consumption and heat dissipation continues to be a growing concern in modern computing, as chip complexity and die sizes grow.  Process technology enhancements can only stem the tide of power consumption so much, if along with those enhancements, additional resources are added to the architecture again increasing die area.  And so, below we'll show you a total system power consumption characterization with AMD's latest high-end processor cores based on their 90nm SOI process technology and we have compared them to a similarly configured Intel architecture-based system and their new 65nm Presler dual-core Pentium Extreme Edition 955 processor, Extreme Edition 840 dual-core and Pentium 4 Extreme Edition single core CPUs based on 90nm Intel process technology. 

Here we've listed two measurements for the Athlon 64 FX-60 driven system, one with AMD's Cool-n-Quiet power savings technology activated and one without.  Cool-n-Quiet technology must be switched on in a BIOS menu option and also it's not advisable to try overclocking the CPU with Cool-n-Quiet enabled, due to the voltage fluctuations and automatic adjustments the core makes on its own.  Regardless, we're not sure many performance enthusiasts will be running in Cool-n-Quiet mode but here you can see the affects of it on total system power consumption.

Which brings us to an important data point with the above test measurements.  The power consumption listed above is indicative of total system power requirements, configured with as many identical components as we could possibly put together.  The graphics card used in our test systems was an NVIDIA GeForce 7800 GTX, other than that motherboards, system memory, even system on-board audio solutions were all different.  For example, the Intel-based systems were all configured with the required DDR2 DRAM, versus standard DDR that is required on the AMD based systems.  Again, please check our test system setup specs on the top of this page to see how each complete test system was configured, before assessing the data in the chart above.

The quick-take on the data we've show here, is that, at full load, the latest dual-core Athlon 64 FX-60 and Athlon 64 X2 4800+ consume approximately 15 - 17% less power than Intel's latest 65nm dual core CPU, the Pentium Extreme Edition 955.  At idle, unless AMD's Cool-n-Quiet is enabled, the latest Athlon 64 FX-60 consumes about 5% more power than the Intel Pentium Extreme Edition 955.  On the other hand obviously, with Cool-n-Quiet enabled, the Athlon 64 FX-60 shows a 20% lower power consumption advantage versus Intel's new Presler core CPU.

In this analysis we've given you a snapshot of power consumption over time, from an idle state running Windows XP's desktop GUI, to a fully loaded state with Prime 95 and Folding At Home clients loaded.  As you can see, the dual-core Athlons show significantly more total power consumption versus the FX-57, and with twice the processing resources under the hood, it's not surprising.

SANDRA, Overclocking And Stress-Testing


Below are synthetic benchmark tests with SiSoftware's SANDRA, the System ANalyzer, Diagnostic and Reporting Assistant. SANDRA consists of a set of information and diagnostic utilities that can provide a host of useful information about your hardware and operating system. We ran four of the built-in subsystem tests that partially comprise the SANDRA 2005 suite (CPU, Multimedia, Cache, and Memory) with the Athlon 64 FX-60 installed into our test rig.  All of the scores reported below were taken with the processor running at its default clock speeds of 2.6GHz.

Preliminary Benchmarks with SiSoft SANDRA 2005
Synthetic Testing

CPU Arithmetic Benchmark
Athlon 64 FX-60

Multimedia Benchmark
Athlon 64 FX-60

CPU Cache Benchmark
Athlon 64 FX-60
(DDR2 400)

Memory Benchmark
Athlon 64 FX-60

(DDR2 400)

SANDRA's CPU test reports that the Athlon 64 FX-60 is the single fastest CPU in terms of Dhrystone ALU performance and second only to the Intel Extreme Edition in terms of Whetstone FPU calculations. This test also showcases the FX-60's 200MHz advantage, with scores significantly faster in both cases, versus the Athlon 64 X2 4800+ reference system. The Multimedia tests show a very similar scaling with the FX-60's faster Integer performance and the Pentium Extreme Edition showing better FPU throughput. Cache & Memory benchmark show the Athlon 64 FX-60 performs a bit better in terms of available bandwidth versus the reference Athlon 64 FX-53 but not in the same league as any of the Intel dual core Pentiums in the database.  The memory bandwidth benchmark reported peak bandwidth scores just under 6GB/second, which is some of the fastest standard DDR400-based performance this test has shown us to date with an Athlon 64 X2 processor.

AMD Athlon 64 FX-60 Overclocking
3GHz and stress tested

In our next series of tests, we decided to stress the chip a bit more with some overclocking at full CPU load.  The goal here was to determine what the Athlon 64 FX-60's threshold of pain was, while utilizing standard air cooling practices.  To be fair, we should note that we used a very leading-edge and capable CPU cooler throughout our testing, Zalman's CPNS9500 LED all copper integrated HSF cooler.  This Zalman cooler has brought down temperatures on high-end CPUs in our test labs, by an average of 5oC or more.  In fact, under full load with Prime 95 and Folding@Home clients running at an overclock of 3GHz, the Athlon 64 FX-60 only reached a 50oC core temperature in an open air test bench environment.  Core temps in a closed case configuration would undoubtedly be higher but regardless, with the excellent performance of this particular Zalman cooler, along with the Athlon 64 FX-60's moderate thermal characteristics, our results overall were rather impressive.

To achieve the 3GHz mark we only had to change the CPU core multiplier to 15X in the BIOS and leave the system clock speed at a standard 200MHz which allowed us to keep system memory and HyperTransport speeds within specifications.  We did increase the core voltage to 1.3V + .2V boost for 1.5V total as well in the CPU voltage settings of our Asus A8N32 SLI test system motherboard. The new AMD core proved itself to be very stable at this configuration, remaining glitch-free after several extensive 100% CPU load tests.  The obligatory "no guarantees" rule of overclocking applies here but we're fairly confident that anyone interested in taking the Athlon 64 FX-60 higher will realize very positive results, at the very least completely stable 2.8GHz operation and that's with perhaps a more modest CPU cooler than we used as well.


Athlon 64 FX-60 @ 3GHz - SANDRA CPU, Memory and Multimedia Tests

Faster than an Athlon 64 X2 4800+ dual core CPU by 600MHz and faster than AMD's Athlon 64 FX-57 single core CPU by 200MHz, an overclocked 3GHz dual core Athlon 64 FX-60 is a thing of beauty.  AMD clearly has more room to grow with this dual core architecture, even within the confines of 90nm SOI process technology, which almost sounds silly to say, since 90nm fabrication is very much state-of-the-art.  Although 65nm is on the horizon for AMD as well.



Again from more of a synthetic testing perspective, we have FutureMark's PCMark05 and specifically have run the suite's CPU and Memory test modules.  Our philosophy with respect to "synthetic" benchmarks, like PCMark05, is that while the end user doesn't have complete access to the applications or measurements that are run and tabulated in these various test scenarios, it does give a perspective of potential performance within a computing architecture and it's capabilities within a group of specific tasks.  Furthermore, synthetic testing sometimes affords the ability to provide a "look-ahead", if you will, of new usage models that perhaps do not exist in current applications available on the market.  All told, we feel synthetic benchmarks are a component of a complete product performance profile, targeted at complimenting critical real-world application testing.

Futuremark PCMark05
More Synthetic CPU and Memory Benchmarks

"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.


PCMark05's CPU performance module is a multi-threaded test that benefits from not only the second core on the Pentium Extreme Edition 955 but also Intel's HyperThreading technology, running more than two threads at time but potentially up to four, 2 logically per core. This makes the EE 955 appear as four virtual 3.46GHz processors to a benchmark like this one, hence the high score. According to PCMark05 the EE 955 outscores AMD's Athlon 64 FX-60 by 656 points, or approximately 12%.

"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.


The scores reported by PCMark05's memory performance module are relatively mixed for various reasons. The single-core Athlon 64-FX 57 and 3.73GHz Pentium 4 Extreme Edition post the highest scores because they don't share a memory controller with a second hardware level execution core. Beyond that, it's all about CPU system bus speed and bandwidth, as well as system memory interface speed, bandwidth, and related latencies.  From strictly a dual-core performance perspective the two competing architectures are fairly well matched here, with a virtual dead heat between Intel's latest Pentium EE 955 and the AMD Athlon 64 FX-60.

Office XP SP2 And Photoshop


PC World Magazine's WorldBench 5.0 is a new breed of Business and Professional application benchmarks. We've made the jump from our aging Winstone tests, to this more updated suite of benchmarks with much more up-to-date applications incorporated within it.  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 to completion of run-time workloads.

PC World's World Bench 5.0: Office XP SP2 & Photoshop 7 Modules
More Real-World Application Performance


Worldbench 5.0's Office XP SP2 performance module reports all of the competitive CPUs and test systems here as finishing the workload within a few seconds of each other.  Regardless there are a couple of observations to be made.  Clearly Microsoft Office XP doesn't benefit much in this test from dual core CPUs and multi-threading.  The Athlon 64s  put up some of the fastest times in the field but not by much, and amongst each other, the faster the clocks speed, the better the score.  The same can be said for the Intel processors as well, with the 3.73GHz single core Pentium 4 Extreme Edition putting up the best time for the Intel camp.

The Photoshop 7 test is somewhat more taxing, but it too isn't completely multi-threaded, save for a few of the filters applied at various points during the benchmark. The Athlon 64 FX-57 is the best performer here by a hair but the Athlon 64 FX-60 clips right in behind it literally 1 second later, which is virtually an identical score considering the error variance of the test.  The fastest Intel-based time, put forth by the Pentium Extreme Edition 955, is smoked by 47 - 48 seconds, or about 17%, by both the Athlon 64 FX-60 dual core CPU and the single core FX-57.

3ds Max and Windows Media Encoder


We continued testing of the new AMD Athlon 64 FX-60 processor with a few more tests that are part of the Worldbench 5.0 suite. Up next we have some performance results of WB 5.0's 3Ds Max test (3D Studio Max) in Direct 3D mode. 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 score here indicates better performance and faster completion times.

PC World's World Bench 5.0: 3ds Max
More Real-World Application Performance

This test draws a nearly identical parallel to the Photoshop test in World Bench 5.  Everyone of the Pentiums are completely out-performed by the group of Athlon 64 processors, with spreads ranging up to a full minute slower between the fastest time put up by the Athlon 64 FX-57 and the slowest put up by the Pentium EE 840.  Also of note is that this test and application does benefit from dual-core CPUs since it is a multi-threaded application.  In fact the Athlon 64 FX-60 once again drops in right on top of the Athlon 64 FX-57, although the FX-60's cores are running a full 200MHz slower at 2.6GHz.  Here the FX-60 is the best of both worlds, a balance of very fast CPU clock speeds, with two cores at the ready for twice the compute resources.

Windows Media Encoder 9
Digital Video Encoding

For our next text, we moved onto a benchmark based on Windows Media Encoder 9.  PC WorldBench 5's Windows Media Encoding test reports encoding times in seconds, and like the tests in 3DStudio Max above, lower times indicate better performance.

This is really where the rubber meets the road in our opinion.  Media encoding is very much a mainstream function for today's modern PC user.  Converting video and audio streams to different formats is something "even your Dad" is capable of and may have use for.  Beyond word processing, web surfing and gaming, media encoding applications are probably the most widely utilized software on the market today.  Windows Media Encoder 9 is a multi-threaded  application as well and it benefits greatly from not only dual physical cores under the hood but also multiple logical threads being run on a single core, as with Intel's HyperThreading technology.

That said, Intel is very much served up a beating once again by AMD's fastest dual core processor, the Athlon 64 FX-60.  The FX-60 processed the WorldBench 5 workload some 50 seconds faster than the fastest Pentium EE 955 score and even the 2.4GHz Athlon 64 X2 4800+ blows by Intel's current best, in this test.  The EE 955 does have enough gas to take on the Athlon 64 FX-57, but at nearly a full 1GHz clock speed deficit, the FX-57 in turn takes out the 3.73GHz Pentium 4 Extreme Edition, even with its HyperThreading technology at work.

WME9/Mozilla Multi-tasking & LAME MP3


Continuing on in our media encoding vein, we took a look at media encoding and conversion while multi-tasking with a web browser.  WorldBench 5's WME and Mozilla Multitasking module adds a bit more work to the mix, with a very common usage model, that of an end user taking time out to surf the web while a video file is being encoded.

Windows Media Encoder 9 & Mozilla Multi-Tasking
More Digital Video Encoding

There's no need to provide extensive analysis for you here, you realize a full-fledged beating when you see it.  Even the fastest Pentium Extreme Edition dual core CPU can't compete with the Athlon 64 FX-60, not by a long shot.  Even the single core FX-57 takes Intel's fastest chip to task here.

LAME MT MP3 Encoding Test
Converting a Large WAV To MP3

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.   The LAME MP3 DLL is one of the most widely used MP3 audio encoding engines in the market today, utilized by many third party software vendors.  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.

In our single threaded tests, the Pentium 4 EE and Athlon 64 FX-57 share a photo-finish but our multi-threaded Lame MP3 encoding tests show a more significant advantage on the Intel side of the fence. The Pentium EE 955 boasts a 6 second lead on the Athlon 64 FX-60, what equates to a roughly a 13% performance advantage and the slower 3.2GHz core Pentium Extreme Edition 840 was also able to edged past the FX-60.  Though not as decisive a victory as the Athlon 64s took our Windows Media Encoder tests with WB5, LAME MP3 audio conversion is pretty much all Intel here.

Kribibench 3D Rendering


Next we ran Kribibench, a 3D rendering benchmark produced by the folks at Adept Development.  Kribibench is an SSE aware software renderer.  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 very large "Ultra" model that is comprised of over 16 billion polygons.

Kribibench v1.1
Details: www.adeptdevelopment.com


Dual core processors rule the roost with Kribibench but there's an interesting contrast between our Sponge model rendering test and the Ultra model rendering test, which is significantly more taxing.  With the Sponge model, the Athlon 64 FX-60 and 4800+ hold approximately a 19% advantage over the fastest Pentium EE 955 score but the Ultra model shows a much tighter grouping of performance with only a 1.5% variance between the Athlon 64 FX-60 and the Pentium EE 955.  Perhaps we're seeing the Athlon 64 system architecture running out of overall memory bandwidth here, reining in its performance significantly versus the lighter workloads in the Sponge test model, which obviously allows all the CPUs tested to stretch their legs a bit more.

Cinebench Rendering & 3DMark05 CPU Test


Cinebench 2003 is an OpenGL 3D rendering performance test, based on the commercially available Cinema 4D from Maxon.  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). 

Cinebench 2003 Performance Tests
3D Modeling & Rendering Tests

In this test the Athlon 64 FX-57 put up the best single threaded score of the bunch and the fastest multi-threaded score went easily to the Athlon 64 FX-60.   As you'll note, we also tossed in an overclocked score of the FX-60 at 3GHz, to give you a taste of speeds bins to come perhaps.  The Pentium EE 955 put up a solid showing, neck and neck with the Athlon 64 X2 4800+ but was in turn then trumped by the stock speed Athlon 64 FX-60 with its 6% performance lead.

Futuremark 3DMark05 - CPU Test
Simulated DirectX Gaming Performance

Easing gently into our gaming benchmark line-up, we have 3DMark05's built-in CPU test.  This is a multi-threaded synthetic DirectX benchmark 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 dependant on the host CPU's performance.  Calculations normally reserved for the system's 3D graphics GPU, are instead sent to the central processor via software.  The number of frames rendered per second in each test are used to determine the final weighted score.

The CPU test built-into 3DMark05 is another multi-threaded test that shows the benefits of a dual-core processor. In this test, the Pentium Extreme Edition 955 posted the highest score of the bunch, nudging past the AMD Athlon 64 FX-60 by 100 points or so and past the X2 4800+ by 180 points, or roughly 2.7%.  Beyond that, the scores are self explanatory but in general, the average score between the Athlon 64s in this test was slightly higher overall than the Pentium Extreme Editions we tested.

Half Life 2 and Quake 4 - Low Res Gaming


To start our in-game testing, we did some low-resolution benchmark loops with the latest version of Half Life 2.  Like some of the other in-game tests in this performance showcase, we used low-quality graphical settings and a low screen resolution to isolate CPU and memory performance and take as much of the load off the graphics pipeline as possible.

Benchmarks with Half Life 2: Low-Res / Low Quality
DirectX 9 Gaming Performance

Clearly, Half Life 2 runs faster on AMD's Athlons, as is evident by the 30-40+ frame per second advantages for the FX-57 and X2 4800+. The new Pentium Extreme Edition 955 did manage to jump ahead of the other Intel processors in this test though, besting the higher-clocked, single-core 3.73GHz Extreme Edition by about 7 FPS, and the 840XE by about 25 FPS.  The Athlon 64 FX-60 falls right in between the single core FX-57 and the 2.4GHz dual core X2 4800+, an easy performance trade-off, if you ask us, in return for exponentially smoother multi-tasking and multi-threaded performance that will undoubtedly become more prevalent in future game titles.

Benchmarks with Quake 4 v1.05: Low Quality
OpenGL Gaming Performance

We then benchmarked all of the test systems again using our custom single-player Quake 4 timedemo file. Here, we installed the new v1.05 patch which is SMP capable, cranked the resolution down to 640 x 480, and configured the game to run at its "Low-Quality" graphics setting. Although Quake 4 bears a fairly heavy burden today's high-end GPUs, when it's configured at low detail and resolutions settings such as these, it too is much more CPU and memory-bound.

Quake 4 clearly now benefits from dual-core CPUs and multi-threading with this new patch.  In fact the Pentium EE 955 is able to outscore its brethren single core 3.73GHz P4, even though it runs at nearly a 300MHz slower clock speed.  However, none of the Intel-based scores were even in the same league as the Athlon 64 brigade.  The Athlon 64 FX-60 took the top post with a 29%  advantage over the fastest Intel Pentium Extreme Edition 955 frame rate. 


Half Life 2 And Quake 4 - High Res Gaming


We took a different approach for this next batch of in-game benchmarks. In a recent poll of our readers, we found that 1280x1024 is the most popular resolution that folks use to play their games. So, for this next set of tests we configured Half Life 2 and Quake 4 to run at that resolution with anti-aliasing and anisotropic filtering enabled, and re-ran the same benchmark demos as we ran in our low-res "CPU bound" tests on the previous page.

Benchmarks with Half Life 2: Mainstream Settings
DirectX 9 Gaming Performance

Although the framerates are lower, this Half Life 2 benchmark tells basically the same story as the one on the previous page. The Athlon 64 FX-57 was clearly the fastest stock speed CPU we tested, followed by the Athlon 64 FX-60, and then the Athlon 64 X2 4800+.  At an overclocked 3GHz, the Athlon 64 FX-60 dropped in another 13% better frame rate number versus its stock 2.6GHz speed and it was about 8% faster than the single core, 2.8GHz FX-57.  Apples to apples however, AMD's fastest dual-core CPU took out Intel's fastest by a solid 16+ percent.

Benchmarks with Quake 4: Mainstream settings
OpenGL Gaming Performance

When we re-ran our custom Quake 4 benchmark at a higher resolution with anti-aliasing enabled, the test systems were much more severely limited by the 256MB GeForce 7800 GTX graphics card we used. The Intel powered systems all posted the same framerate, as did the AMD powered systems. What may be surprising is that this time around, however, the Intel based systems were technically the "fastest", although less than a frame per second separated them from the AMD based systems. In GPU-bound circumstances, the type of CPU used in the system has a minimal impact on performance in most circumstances.  

However, what we may be witnessed to here is a specific case where AMD's obvious architectural advantage in the CPU core, is negated significantly by some of the inherent system level challenges, for example taking a native HyperTransport serial domain to PCI Express in the MCP (or Southbridge) of the nForce 4 chipset within the system.  Here a translation of HT calls to the CPU coming in from the PCI Express links in the graphics subsystem could in fact be causing a slight performance degradation not linear with the same degradation that the Intel based system sees at higher graphics workloads.  Intel's architecture is largely PCI Express-based and is on chip in the Northbridge, where the memory controller also resides.  There could be any one of a myriad of system level performance reasons for the scores we see here however.

Benchmark Analysis & Conclusion


Benchmark Summary:
AMD's Athlon 64 FX-60 was clearly THE fastest all around processor in our barrage of tests, save perhaps for a few of the synthetic benchmarks and our high-res Quake 4 test, where for all intents and purposes all of our high end systems performed similarly due to fill-rate limitations of the graphics subsystem.  Furthermore, it seems as though some of the traditional Media Encoding tests which have historically fallen to Intel, are now clearly showing an AMD advantage as the Athlon 64 dual-core architecture continues to scale in clock speed.  What's perhaps almost as important to some of you as well, were the excellent thermal and power characteristics that the new Athlon 64 FX-60 put forth.  At full load, the Athlon 64 FX-60 based system consumed nearly 40 watts less power versus our test system built around the 65nm Pentium Extreme Edition 955, and it did so running significantly cooler, at least according to the readings we were getting from Intel's motherboard environmental monitoring software.

AMD's latest "FX" enthusiast class dual-core CPU, the Athlon 64 FX-60,  is sure to be a hit for those looking for the utmost performance with little compromise and not much regard for the budget.  With a Processors In A Box (P.I.B) MSRP of a mighty steep $1031, you'll have to get serious clearance from the wife or significant other, if you want to slide this upgrade past and keep the peace.  While there are clearly better total cost values in the market (take the Athlon 64 4400+ for example at half the price), those who must have the best all around computing experience in virtually all desktop applications, while not sacrificing gaming performance, will love this new high-end CPU from AMD.  Not to mention, this new speed bin continues to show great promise for overclocking, where 2.8GHz should be a walk in the park for most and 3GHz a reasonable goal with a capable cooling solution.

So, while Intel seems to be coming to the end of the line with their Netburst architecture even in .065 micron, AMD seems to be clipping right along at .09 in their dual-core architecture, with more headroom available, even before they make the jump to 65nm and a new socket.  The question is when will that transition happen and with that transition, what other AMD architectural enhancements will come along?  We know DDR2 and a new socket is in the Athlon 64's future, but AMD could make other architectural enhancements as well.  Intel has made it relatively clear that in the second half of this year, their new CPU core architectures will give AMD a serious run for their money.  But for now we'll just have to wait and see.  AMD obviously still has a significant performance lead in the high-end desktop CPU space with the introduction of yet another fastest dual-core processor in the market.

_Fastest Dual-Core, period
_64-Bit support
_Great gaming performance
_Great Overclocking potential
_Runs cooler and uses less power than competing dual-core CPUs

HotHardware's PC Hardware Forum
Get Tweaked, Overclocked and Ready To Rock...

Content Property of HotHardware.com