Logo   Banner   TopRight
TopUnder
Transparent
AMD QuadFX Platform & FX-70 Series Processors
Transparent
Date: Nov 29, 2006
Section:Processors
Author: Marco Chiappetta
Transparent
Introduction, Specs, and Related Info

By now you're all familiar with Intel's Core micro-architecture, as well as the dual and quad-core Core 2 Duo, Extreme, and Quad processors based on it. With the launch of their Core 2 processors, Intel no longer played second fiddle to AMD in terms of overall performance or power efficiency. And for the last few months since their release, Intel has been steadily re-gaining fans in the enthusiast community that were put off by years of relatively hot-running, and under-performing Netburst micro-architecture based processors.

Today, AMD aims to re-capture the hearts of those hardcore enthusiasts with the official launch of their QuadFX platform. News about QuadFX has been trickling out for months under the "4x4" banner. But today we can show you exactly what QuadFX is all about. At its most basic level, QuadFX is AMD's vision of a high-end quad-core desktop computing platform. There's a lot more going on under the virtual hood, however. We'll start with some quick specifications and move on to some of the more juicy details on the pages ahead.

AMD QuadFX Platform & FX-70 Series Processors
Specifications & Features

As you can see in the specifications above, AMD's new FX-70 series of processors, which are the heart of the QuadFX platform, are very similar to current socket AM2-based FX processors. They differ only in their packaging, the number of HyperTransport links incorporated into the CPU, and of course by their clock speeds.


We have posted a wealth of information regarding AMD's Athlon 64 processors and related core-logic chipsets over the last few months here at HotHardware.com. For some more background on the technologies employed by AMD's Athlon 64 processors and the AMD platform as a whole, we suggest taking a look at few of the related articles listed below. They contain detailed explanations of some of the features common to AMD's legacy products, compatible chipsets, and the enhancements make to various dual-core processors:

We cover some specifics regarding AMD's dual-core processors in our FX-62 and 5000+ evaluation, and cover many of the details regarding the nForce 600 family of chipsets in our evaluation of the nForce 680i SLI. And our Quad-Father update has some early details regarding QuadFX. We suggest perusing these articles if you're unfamiliar with AMD's technology. They'll lay the groundwork for the tech covered on the next few pages. 

Transparent
Platform Specifics

Today's launch is as much about AMD's new FX-70 series processors as it is about the rest of QuadFX platform as a whole. In addition to a pair of FX-70 series processors, QuadFX requires a dual-socket motherboard to accommodate the two processors.  Although AMD recently acquired ATI and all of their chipset assets and IP, they have chosen to launch the QuadFX platform with a chipset built by NVIDIA. All initial QuadFX systems will be powered by the nForce 680a SLI chipset. The high-level block diagram below outlines the interconnects and main features of QuadFX.

As you can see, each dual-core FX processor is linked to its own banks of DDR2 RAM, and the processors themselves are linked with a coherent HyperTransport bus. Only one of the CPU sockets is directly connected to the nForce 680a chipset via a pair of HyperTransport links, however. This was done to make it possible to use QuadFX motherboards in single CPU configurations. We should note that all of the HyperTransport links on the board are full 16-bit links, operating at 1GHz (2GHz DDR).

As of today, the QuadFX platform allows for two, dual-core processors to be installed, for a total of four execution cores in a single system. In the future though, when AMD has native quad-core processors in their product line-up, QuadFX will scale to an 8-core platform.  And technically, since AMD opened up their Opteron architecture with their Torenzza initiative, that second socket doesn't have to contain a traditional x86 CPU at all. Instead, specialized co-processors can be installed in the second socket that interface with the system via HyperTransport. It's going to be interesting to see what the future brings to QuadFX.

This "quad-core" setup is obviously much different than Intel's.  Whereas Kentsfield-based processors are a multi-chip package with two, dual-core dies on a single chip, QuadFX is a pair of dual-core processors linked via serial interface. Intel's current solution is much more cost efficient, and more elegant in our opinion, but the AMD architecture has a major advantage, namely the coherent HT link between the processors. For any communication via the individual cores, Kentsfield has to send data off one die, over the FSB, and back to the other die. A pair of Athlon 64 FX-70 series processors can communicate directly over a dedicated HT link without contention over a shared front side bus.

Having two independent processors though, each with their own memory controller, means QuadFX has a non-uniform memory access, or NUMA. The system can be configured as two independent nodes or with node interleaving, but the operating system being used has to have proper support for NUMA because the way the OS accesses system memory can have a major impact on performance.  For example, Windows XP doesn't properly support NUMA due to the way its scheduler loads execution cores in a multi-processor system. With XP a single thread could bounce from core to core.  Windows Vista does properly support NUMA, however, and Vista's scheduler shouldn't shift single tasks between individual cores. As you'll see later (on our SANDRA and PCMark05 pages specifically), there is a large difference in available memory bandwidth when the OS has native NUMA support.

Also note the impressive list of features brought forth by the nForce 680a SLI chipset. The 680a is a multi-chip solutions that is derived from NVIDIA's nForce Professional 3000 family of products. As such, the chipset has support for up to 56 PCI Express lanes, 4 Ethernet MACs, 12 SATA ports, HD Audio, Firewire, and 20 USB 2.0 ports. The block diagram above shows one configuration with four PCI Express x16 slots, which would support quad graphics cards, but the PCI Express lane configuration is flexible should other motherboard partners eventually decided to outfit their upcoming QuadFX boards in a different manner.  The 680a SLI has a slew of other unique features as well, like FirstPacket and DualNet, for example. We go into detail regarding the nForce 600 chipset family's features in this article.

Transparent
Inspecting the Hardware

The first, and for now only, motherboard available for the QuadFX platform is the Asus L1N64-SLI WS. AMD has informed us that other motherboard partners are looking to introduce QuadFX motherboards some time before the end of this year, or more likely in the first half of 2007.

As its name implies, the Asus L1N64-SLI WS is a workstation class motherboard, but unlike many other boards in this class, it is outfitted with a host of features and options targeted squarely at hardcore power users and enthusiasts. Of course the board has dual CPU sockets and four DIMM slots; two per CPU. It is also equipped with four physical PCI Express X16 slots (two with x16 and two with x8 electrical connections), single standard PCI and PCI Express x1 slots, 12x SATA ports, ADI SoundMax HD audio, and dual gigabit LAN ports. Although the nForce 680a SLI chipset has four Ethernet MACs, only two are used on this board.

   

The Asus L1N64-SLI WS' BIOS is also relatively well equipped. Unlike most other dual-socket, workstation class motherboards, the L1N64-SLI WS has a full compliment of overclocking and tweaking tools. All of the board's integrated peripherals can be enabled or disabled from within the BIOS, and a number of voltages and frequencies can be tweaked as well.  All of the HyperTransport link frequencies can be altered independently, as can the PCI Express frequency, and the HT reference clock that determines the ultimate CPU clock speed. Memory, chipset, and processor voltages can also be manipulated, and memory timings can be tweaked on a per CPU basis.

 

      

     

With a motherboard this complex, you'd expect it to have a relatively cramped layout, and you'd be right. The heatsinks and heatpipes mounted to the voltage regulation modules and chipset are positioned in such a way that they won't interfere with any expansion cards, but because the board has two CPU sockets, and two banks of DIMM slots, in addition to a number of other integrated peripherals, it's a pretty tight fit once cables are connected and expansion cards are installed.


    
AMD Athlon 64 FX-74
(
Middle: Athlon 64 FX-74 / Core 2 Duo Comparison)

If you flipped through the pictures of the motherboard above, you probably noticed that the QuadFX platform utilizes a different CPU socket than previous desktop Athlon 64 processors. The FX-70 series of processors for the QuadFX platform use AMD's 1207-pin Socket F platform, similar to the one used on current high-end Opterons. The move the Socket F for QuadFX was necessary because the standard AM2 desktop platform doesn't have the pin-count accommodations for the multiple HyperTransport links that reside between the processors and chipset.

The FX-70 series of processors utilizes LGA processor packaging, which moves the pins off the processor itself to the CPU socket, similar to Intel's LGA 775 platform - except with many more pins. The FX-74 pictured here is a dual-core processor with a default clock frequency of 3.0GHz (15x200MHz). The FX-72 runs at 2.8GHz, and the FX-70 at 2.6GHz. These processors will not be sold individually at first, but rather will be sold in pairs.

Transparent
Vital Signs and Overclocking

Before we began benchmarking the QuadFX platform, we fired up CPU-Z to see if there was anything interesting to report with regard to the default CPU, motherboard, and memory configurations.  We then took some notes regarding temperatures and overclocking.

      
CPU-Z: Athlon 64 FX-74 & Asus L1N64-SLI WS

Other than the Athlon 64 FX-74's clock speed, nothing much has changed since we last looked at the Socket AM2 Athlon 64 FX-62 a few months back. As you can see, the FX-74 processor is built using AMD's .09-micron manufacturing process and is clocked at 3.0GHz (15x200MHz) with 1MB of L2 cache per core. The other new members in the FX-70 series are all similarly configured, but with lower multipliers (14 for the FX-72, 13 for the FX-70).  You'll notice in the screenshots above that CPU-Z can't properly identify any of the components used in the QuadFX platform currently, but the actual specifications listed regarding clock speeds and caches are all correct.


Athlon 64 FX-74 Overclocked to 3.2GHz

We also spent some time overclocking the QuadFX FX-74 platform and had some very interesting results. To overclock the processors, we used the stock AMD copper / aluminum heatsink and fan combo pictured on the previous page. We bumped the CPU voltage up by .15v and set the memory voltage to 2.2v. Then we raised the processors' multipliers and HT speed until the test system was no longer stable. In the case of the FX-74, because the CPUs are unlocked, we raised the multiplier first before cranking up the HT speed.

Ultimately, we were able to take the new FX-74 up to 3.215GHz, a 215MHz (7.1%) increase, raising the multiplier to 16 and the HT link to 202MHz. It seems AMD is getting close to hitting the virtual clock speed wall with their 90nm SOI manufacturing process.

Through all of our preliminary tests, we also monitored processor temperatures via the system BIOS. While idling at stock speeds, we found that the FX-74 processors' core temperatures hovered around 46oC. And under load temperatures peaked at around 67oC. Overclocking didn't drastically effect temperatures, but we did see a max of about 70oC at one point, at least according to the Asus L1N64-SLI WS' BIOS.

Transparent
Our Test Systems & SiSoft SANDRA

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 1024MB 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.

HotHardware's Test Systems
AMD & Intel Inside!
System 3:
AMD Athlon 64 FX-74

AMD Athlon 64 FX-72
AMD Athlon 64 FX-70

(3.0GHz, 2.8GHz, 2.6GHz)

Asus L1N64-SLI WS
(NVIDIA nForce 680a SLI)

4x512MB Corsair PC-6400
CL 4-4-4-12 - DDR2-800

GeForce 7950 GX2

On-Board Ethernet
On-board Audio


WD740 "Raptor" HD
10,000 RPM SATA

Windows XP Pro SP2
nForce Drivers v9.53
NVIDIA Forceware v91.27
DirectX 9.0c
System 2:
Core 2 Extreme QX6700
(2.66Hz - Quad-Core)
Core 2 Extreme X6800 / E6700
(2.93GHz & 2.66GHz)

Asus P5W DH Deluxe
(975X Express)

2x1GB Corsair PC-6400
CL 4-4-4-12 - DDR2-800

GeForce 7950 GX2
On-Board Ethernet
On-board Audio

WD740 "Raptor" HD
10,000 RPM SATA

Windows XP Pro SP2
Intel INF 8.0.1.1002
NVIDIA Forceware v91.27
DirectX 9.0c

System 3:
AMD Athlon FX-62

(2.8GHz)

Asus M2N32-SLI Deluxe
(NVIDIA nForce 590 SLI)

2x1GB Corsair PC-6400
CL 4-4-4-12 - DDR2-800

GeForce 7950 GX2

On-Board Ethernet
On-board Audio


WD740 "Raptor" HD
10,000 RPM SATA

Windows XP Pro SP2
nForce 4 Drivers v6.86
NVIDIA Forceware v91.27
DirectX 9.0c
Preliminary Testing with SiSoft SANDRA XI
Synthetic Benchmarks

We began our testing with SiSoftware's SANDRA XI, the System ANalyzer, Diagnostic and Reporting Assistant. We ran six of the built-in subsystem tests that partially comprise the SANDRA XI suite with the new QuadFX platform equipped with a pair of Athlon 64 FX-74 processors (CPU, Multimedia, Multi-Core Efficiency, Memory, Cache, and Memory Latency). All of the scores reported below were taken with the processors running at their default clock speeds of 3.0GHz.


AMD QuadFX
Athlon 64 FX-74 @ 3.0GHz
CPU Arithmetic



AMD QuadFX
Athlon 64 FX-74 @ 3.0GHz
MultiMedia



AMD QuadFX
Athlon 64 FX-74 @ 3.0GHz
Multi-Core Efficiency



AMD QuadFX
Athlon 64 FX-74 @ 3.0GHz
Memory Bandwidth


AMD QuadFX
Athlon 64 FX-74 @ 3.0GHz
Cache and Memory

AMD QuadFX
Athlon 64 FX-74 @ 3.0GHz
Memory Latency

In the CPU Arithmetic and Multimedia benchmarks, the QuadFX platform performed as expected; it was faster than a slightly lower clocked dual, dual-core Opteron system, but it couldn't quite compete with the quad-core Intel powered setups. According to the new multi-core efficiency benchmark, however, the QuadFX platform fares very well. Until block sizes hit a certain point, the QuadFX platform offered significantly more inter-core bandwidth than anything else. The QuadFX system also fared well in the Cache & Memory bandwidth, and in the standard memory bandwidth tests, but memory latency was actually higher than all of the other reference systems in SANDRA's database.


AMD QuadFX
Athlon 64 FX-74 @ 3.0GHz
Memory Bandwidth

With Windows Vista
Node Interleaving
Enabled

AMD QuadFX
Athlon 64 FX-74 @ 3.0GHz
Memory Bandwidth

With Windows Vista
Node Interleaving
Disabled

The six tests above were run with Windows XP Professional, but as we mentioned earlier, Windows Vista has native support for NUMA which can drastically affect performance. We ran a couple of memory bandwidth tests using the final build of Windows Vista with node interleaving enabled, and again with it disabled. With it enabled via the system BIOS, the QuadFX platform manages about 6.6GB/s of memory bandwidth. With node interleaving disabled in the BIOS though, SADNRA reports over 12GB/s of available memory bandwidth under Windows Vista.

Transparent
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. 

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.

The QuadFX FX-74 system put up the third highest score in PCMark05's CPU performance module, falling in between the Core 2 Duo E6700 and Core 2 Extreme X6800, and the other members of the FX-70 series follow suit right behind the FX-74.

What's interesting to note is how each quad-core platform scales at like clock speeds. In all of our graphs from this point forward, pay attention to the scaling between the Intel Core 2 Extreme QX6700 (2.66GHz) and Core 2 Duo E6700 (2.66GHz), and between the QuadFX FX-72 (2.8GHz) and Athlon 64 FX-62 (2.8GHz).  In this test, going from two to four cores on the Intel platform resulted in a 25.2% increase in performance; on the AMD platform the increase was "only" 23.4%.


"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 - when run under Windows XP - reported scores for the QuadFX platform that were significantly lower than anything else we tested. We suspect these results were related to QuadFX's non-unified memory architecture, so we also ran this test under Windows Vista. Under Vista, the QuadFX FX-75 platform put up a score of 4668, an increase of almost 1200 points, but that was still lower than the competition.

Transparent
Worldbench 5: Office XP and Photoshop 7

PC World Magazine's Worldbench 5.0 is a Business and Professional application benchmark.  The tests consist of a number of performance modules that each utilize one, or a group of popular applications to gauge performance. 

Worldbench 5.0: Office XP SP2 and Photoshop 7 Modules
Real-World Application Performance

Below we have the results from WB 5.0's Office XP SP2 and Photoshop 7 performance modules, recorded in seconds.  Lower times indicate better performance here, so the shorter the bar the better.

Worldbench 5.0's Office XP SP2 performance module reported similar scores for all of the platforms we tested. Obviously, high-end systems like these are more than capable of running any single-threaded "office-type" application, hence the fractional performance deltas separating all of the platforms here. 

The Photoshop 7.0 benchmark module had a much more pronounced spread. Because this test isn't multi-threaded though, there aren't any major differences in performance between the similarly clocked dual and quad-core systems. It's clear that the Intel-powered rigs had a marked advantage, however, thanks the Core 2's increased IPC and unified SmartCache technology.

Transparent
LAME MT and Sony Vegas 7.0b

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. 

LAME MT MP3 Encoding Test
Converting a Large WAV To MP3

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.

LAME MT only utilizes two concurrent threads, hence the similar performance between the dual and quad-core systems tested here. Once again though, the QuadFX systems fall victim to the Intel Core 2-powered rigs, regardless of the test configuration. Interestingly, the QuadFX FX-72 system was marginally slower than the Athlon 64 FX-62, probably due to the QuadFX's NUMA and differences in the motherboard chipsets.

Sony Vegas Digital Video Rendering Test
Video Rendering Performance

Sony's Vegas DV editing software is heavily multithreaded as it processes and mixes both audio and video streams. This is a new breed of digital video editing software that takes full advantage of current dual and multi-core processor architectures.

The AMD QuadFX platform scaled very well in our Sony Vegas benchmark. The only system faster than any of the QuadFX rigs was based on Intel's quad-core Core 2 Extreme QX6700, which finished about a minute ahead of the FX-74. Scaling between the competing dual and quad-core architectures also favored Intel slightly. Going from two to four cores increased performance by about 42% on the Intel platform; it increased performance on the AMD platform by roughly 40%.

Transparent
Kribibench v1.1

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.

Kribibench v1.1

Details: www.adeptdevelopment.com

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 new QuadFX platform scaled relatively well in the two 3D models we used for testing with Kribibench, but Intel's quad-core Core 2 Extreme QX6700 had a clear advantage in both tests. The Intel platform also scaled better here. In the "Sponge Explode" test, the QX6700 was 45.5% faster than the E6700; the QuadFX FX-72 was 21.1% faster than the Athlon 64 FX-62. In the more taxing "Ultra" test, the QX6700 outperformed the E6700 by 76.5%, and the FX-72 beat the FX-62 by about 64%.

Transparent
POV-Ray Rendering

POV-Ray, or the Persistence of Vision Ray-Tracer, is a top-notch open source tool for creating realistically lit 3D graphics artwork. We tested with POV-Ray's standard included benchmarking model on all of our test machines and recorded the scores reported for each.   Results are measured in pixels-per-second throughput.

POV-RAY Performance

Details: www.povray.org

 

 

 

As you can see, the QuadFX platform scaled very well in the POV Ray benchmark in comparison to any of the dual-core systems listed here, but Intel's Core 2 Extreme QX6700 put up a monster score that couldn't be touched. Moving from two to four cores with the QuadFX platform (FX-72 vs. FX-62) resulted in an 81% increase in performance. Somehow, Intel's QX6700 managed to outperform the E6700 by over 200%.

Transparent
Cinebench 9.5 and 3DMark06: CPU

The Cinebench 2003 benchmark is an OpenGL 3D rendering performance test, based on the commercially available Cinema 4D application. Cinema 4D from Maxon is a 3D Rendering and Animation tools suite used by many 3D Animation houses and producers like Sony Animation and many others. And of course it's very demanding of system processor resources.

Cinebench 9.5 Performance Tests
3D Modeling & Rendering Tests

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).

The QuadFX systems performed very well in the Cinebench 9.5 benchmark. In the single-threaded test, the FX-74s hung right alongside the QX6700 and E6700; only the Core 2 Extreme X6800 was faster. In the multi-threaded test only 1 second separated the QX6700 and FX-74 rigs. Once again, this is due to better scaling on the Intel platform. Whereas the QX6700 was 42.3% faster than the E6700 here in the MT test, the FX-72 was 39.3% faster than the FX-62.

Futuremark 3DMark06 - CPU Test
Simulated DirectX Gaming Performance

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 that 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.

3DMark06's multi-threaded CPU performance test had all of the QuadFX-based systems finishing well ahead of any dual-core system, but once again the Intel Core 2 Extreme QX6700 took the top spot outpacing the QuadFX FX-74 by 368 points. The Intel platform also scaled better - a 69% increase for the QX6700, 56% increase for the QuadFX FX-72.

Transparent
Quake 4: Low and High Resolutions

For our next set of tests, we benchmarked all of the test systems using a custom single-player Quake 4 timedemo. Here, we installed the game's official v1.3 point release which is SMP capable, turned 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 is much more CPU and memory bandwidth-bound than anything else.

Benchmarks with Quake 4 v1.3
OpenGL Gaming Performance

 

Low-Resolution, Low-Quality testing with Quake 4 showed a marginal performance decrease on the QuadFX platform, likely due to the non-unified memory architecture and the increased latency that comes with it. 

The same holds true on the High-Resolution, High-Quality tests. Here we raised the resolution to 1600x1200 and enabled anti-aliasing and anisotropic filtering, and yet again the QuadFX platform failed to outpace its dual-core counterpart, the Athlon 64 FX-62, let alone any of the Core 2-based machines.

Transparent
F.E.A.R.: Low and High Resolutions

For our last set of game tests, we moved on to more in-game benchmarking with F.E.A.R. When testing all 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.  However, the in-game "effects" and "advanced computer options" settings, which control the level of detail for F.E.A.R.'s physics engine and particle system, were all set to maximum values, since these actually do place some load on the CPU rather than GPU.

Benchmarks with F.E.A.R. v1.08
DirectX 9 Gaming Performance

Like the Quake 4 results on the previous page, low-resolution benchmarking with F.E.A.R. showed a performance decrease on the QuadFX platform, compared to the Athlon 64 FX-62.

Things balanced back out once we put a heavier load on the graphics sub-system and raised the resolution and level of pixel processing, however. At 1600x1200 with 4X anti-aliasing and 16X anisotropic filtering enabled, the QuadFX and Core 2 Extreme QX6700 rigs all finished within 1 frame per second of each other. The Core 2 Extreme X6800 took the top spot by a couple of frames per second though.

Transparent
Power Consumption

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.

Power Characteristics
Processors and Platforms

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.  We tested with a combination of Cinebench 9.5 and SANDRA XI on the CPU.


* PLEASE NOTE: QuadFX FX-74 Tested with and without Cool n' Quiet

Before you look at this graph and gasp at the overall power draw of the QuadFX systems, we have some explaining to do. The QuadFX rigs listed here were assembled in a full tower case that incorporated multiple intake and exhaust fans. The QuadFX systems also had three additional cooling fans mounted at various locations on the motherboard, dual-optical drives, a floppy drive, and a 1 Kilowatt PC Power & Cooling PSU. Due to the dual-socket nature of QuadFX, the systems were also equipped with four memory DIMMs versus two on the other systems, and they of course had two CPU coolers as well (In case you're wondering, yes, the QuadFX rig was incredibly loud with all these fans).

Our other tests systems were equipped with different PSUs, no additional cooling fans, only two DIMMs, and only a single optical drive. These differences in the configurations, along with the over-the-top nature of the QuadFX platform itself account for the huge power consumption deltas you see here. No matter how you slice it though, QuadFX is going to be a monster in terms of power consumption.  Enabling Cool 'n' Quiet brought idle power consumption way down from over 400W to a much more palatable 260W, but relatively speaking the QuadFX platform is much more power hungry than Intel's Quad-Core QX6700 desktop platform.

It's interesting to note the differences in consumption on the different platforms under idle and load conditions, however. The Core 2 Extreme QX6700 for example, consumed 96 more watts under load than at idle. The QuadFX FX-74 based rig consumed 100 more watts, the FX-72 91 more watts, and the FX-70 86 more. This shows the processors themselves to be on a similar level in terms of power consumption, but the complexity of the QuadFX platform as a whole requires significantly larger amounts of power to achieve similar, albeit somewhat lower performance, than Intel's current quad-core platform.

Transparent
Our Summary and Conclusion

Performance Summary: Throughout our entire suite of benchmarks, a system powered by Intel's quad-core Core 2 Extreme QX6700 processor outpaced all of the QuadFX-based systems. The Intel-powered system also generally scaled better moving from two to four cores in the multi-threaded benchmarks. Our in-game tests and PCMark05's memory performance module put the QuadFX platform at a disadvantage versus AMD's own socket AM2 dual-core platform, but in all of the multi-threaded application and rendering tests, QuadFX was significantly faster than a similarly equipped system powered by an Athlon 64 FX-62.

AMD is clearly laying the foundation for the future with the QuadFX platform. In its current state, a QuadFX machine powered by a pair of AMD's fastest FX-74 processors can't quite keep pace with the QX6700 strictly from a performance standpoint. The overall feature-set of the QuadFX is impressive, however, thanks in part to NVIDIA's nForce 680a SLI chipset and the dual-socket nature of the platform. Twelve SATA ports, quad PCI Express X16 graphics slots, the potential for quad gigabit Ethernet, and SLI support put the QuadFX in a league of its own from a feature standpoint. And that doesn't account for potential future innovations that could make use of the second CPU socket.


* FX-70 Series Processors Sold In Pairs

AMD will be selling Athlon 64 FX-70, FX-72, and FX-74 processors in pairs with heatsinks for prices of $599, $799, and $999 respectively. That's two 3.0GHz dual-core processors for just under a grand. The Asus QuadFX L1N64-SLI WS motherboard will sell for upwards of $300. And for maximum performance, QuadFX will also require four DIMMs, which will be marginally more expensive than two DIMMs of a similar total capacity.  Overall, the pricing structure makes configuring the fastest QuadFX system more expensive than the fastest quad-core Intel-based system, but QuadFX does offer more features, so pricing isn't out of line in our opinion.

Unfortunately for AMD, even with the launch of QuadFX, Intel is still in possession of the overall performance crown for now - the gap just got smaller, much smaller in some cases. With QuadFX, AMD could have a solid platform in place for a strong 2007, however. AMD's next-gen, native quad-core processors due out next year could significantly close or even erase the performance gap between AMD and Intel due to core architectural enhancements. Install two of them in a QuadFX motherboard, and you'll have eight cores at your disposal; an interesting prospect to say the least.

  • Good Performance
  • Feature Rich Platform
  • Tweakable Motherboard
  • Competitive Pricing Structure
  • Some Overclocking Headroom
  • Didn't Scale As Well As Intel
  • Power Consumption


Content Property of HotHardware.com