Logo   Banner   TopRight
TopUnder
Transparent
Pentium 4 Extreme Edition 3.46GHz 1066MHz FSB
Transparent
Date: Oct 31, 2004
Section:Processors
Author: Dave Altavilla
Transparent
Introduction

 

The folks out in Sunnyvale have been releasing new additions to their Athlon 64 processor lineup with the kind of fevered pace we've come to expect from major-league semiconductor manufacturers.  So too has the ever-dominant Intel.  With 78,000 employees worldwide and almost 10 times the revenue of its nearest competitor at $30 billion annually, its this kind of critical mass that allows a company to pour endless resources into a program and drive higher reaching product line goals.

Earlier this month, we brought you a view of AMD's new Athlon FX-55 and Athlon 64 4000+ processors and their relative performance versus Intel's flagship CPUs.  Today we've got Intel's latest weapon locked and loaded into its "Enthusiast Class" silo.  The Pentium 4 Extreme Edition 3.46GHz with 1066MHz FSB is on our test bench here at HotHardware.  Intel had to further validate the 925X chipset for this higher system bus speed this time around as well, so we're also looking at the release of the kicker chipset enhancement on this launch, dubbed the i925XE. We'll dig in and crank it all up for you in the pages ahead.


Intel Pentium 4 In LGA Packaging And i925X Chipset

Intel Pentium 4 Extreme Edition 3.46GHz With 1066 Frontside Bus
Cranking up the FSB and dropping the multiplier
  • Pentium 4 Extreme Edition Processor
  • Clock speed: 3.46GHz
  • 1066MHz "Quad Pumped" frontside bus
  • 0.13-micron manufacturing process
  • Hyper-Threading Technology
  • 512kB on chip, full Speed L2 Cache
  • 2MB on chip, full speed L3 Cache
  • 8KB L1 Data Cache
  • Streaming SIMD Extensions - SSE2
  • 20-stage "Hyper Pipelined" Technology
  • Rapid Execution Engine - ALU clocked at 2X frequency of core
  • 128-bit Floating Point/Multimedia unit
  • Intel "NetBurst" micro-architecture
  • Supported by the Intel i925XE chipset, with Hyper-Threading support
  • Intel MMX media enhancement technology
  • Memory cacheability up to 4GB of addressable memory space and system memory scalability up to 64GB of physical memory
  • 1.525 - 1.6V operating voltage range
  • LGA775 Packaging - Land Grid Array
  • 110.7 Watts TPD (Thermal Design Power) 



Architecturally, not a thing has changed for the Pentium 4 Extreme Edition, with the exception of its multiplier ratio.  The only significant change with this Intel CPU is that Intel validated it with a lower bus multiplier along with a significantly higher frontside bus at 1066MHz, versus the previous 800MHz system bus.  In fact, enthusiast overclockers have probably already seen some of the benefits of higher frontside bus speeds, while taking multiplier-locked retail chips to their outer limits. Also note that this new chip has a higher TPD (Thermal Power Dissipation) number at 110 Watts versus a standard P4 EE 3.4GHz chip at 89 Watts.

The other notable is that this CPU is being targeted strictly for Intel's new i925XE chipset, which is essentially identical to the previous i925X, only it too has been validated for the new 1066MHz FSB.  Motherboard manufacturers no doubt will likely migrate to this new chipset for their Alderwood (i925)-based offerings moving forward.  We should also mention that most likely motherboard partners will not be able to just release a BIOS revision that will enable 1066MHz FSB operation without an associated board level hardware revision, as well.  This short-term limitation is most likely due to a new system clock PLL divide ratio that needs to be accommodated in the new design for support of 1066MHz FSB.  Regardless, if you took the plunge into the i925X, we're sure you'll be more than disappointed that in fact you'll need an entirely new motherboard to support the next generation P4 FSB.

Transparent
Intel D925XECV2 Motherboard and i925XE Chipset

As we noted earlier, architecturally, the i925XE chipset is identical to the i925X.  Intel has simply gone through the life cycle and burn-in processes required to validate the chipset at the new 1066MHz rated bus speed.  As a refresher we'll cover the high-level functional blocks of the i925X and i925XE chipset here for you.  If you would like a full run-down on the platform, please refer back to our June 19th article from earlier this year, where we cover the architecture in full detail.  

Intel D925XECV2 Motherboard & i925XE chipset
Same chipset and features, new 1066MHz FSB

The i925X  and i925XE chipsets are Intel's high end/highest performance solution for the Pentium 4 desktop platform. These chipsets do have slightly better memory bandwidth efficiency and latency characteristics versus the i915.  Other than that, the general feature set is the same as Intel's mainstream-value i915 product, which is very well rounded, with all of the latest IO technologies any power user could need.


The kit we received from Intel included the Pentium 4 EE 3.46GHz chip and Intel's D925XECV2 motherboard that is built on the new i925Xe chipset. This motherboard is actually set up a bit like many of the "enthusiast" class products we've seen from the likes of ABIT, ASUS, and DFI. As an aside, we haven't received any i925XE boards from any of the Taiwanese OEMs as of yet, but we hear they are right around the corner.  As a result, we would caution anyone looking to make the jump to this new Pentium 4 LGA PCI Express-enabled architecture to wait until i925XE boards are available in the retail channel.  You'll have the ability to run standard 800MHz FSB CPUs today and the migration path to 1066MHz FSB-driven CPUs in the future.

Four channels of RAID 0, 1-capable SATA hard drive connections are available off the ICH6 Southbridge chip with this motherboard, and the list of creature comforts goes on from there. The D925XECV2 is capable of 8-channel audio but set up with a 5.1-jack combination with left and right front and rear along with combination center and channel/sub woofer ports, along with digital audio optical and coaxial ports.  It also has eight USB 2.0 ports available, four on the I/O plate and four internal with a backplate slot bracket.  Finally, as you can see in the shots above, there is one X16 PCI Express slot for standard PCI slots and two 1X PCI Express slots.

 

Transparent
Vital Signs, Temps, Overclocking & D925XECV2 BIOS

Intel has been refining and tuning the Pentium 4 LGA775 heatsink design since its initial release in June.  The result is a slightly taller sink with a shrouded fan that seems to spin a bit more quietly with its scooped blade design.

Pentium 4 Extreme Edition 3.46GHz Vital Signs
New higher speed FSB, same basic thermal characteristics

    

The retention mechanism for the stock Pentium 4 HSF is the same press fit cam loaded design that we saw in the first iteration for the LGA775 socket assembly, only this time out there seems to be significantly more insertion force required to get these clips to lock into place.  As a result, it "feels" as though the retention pins themselves now actually provide a bit more downward pressure on the socket once locked into place.

      

Pentium 4 Extreme Edition 3.46GHz - LGA775 at ideal and load

Upon inspection of the processor's chip ID taken with the CPUZ utility, you'll note that this is indeed an Intel  "Gallatin" core based processor, identical to previous versions of what is now known as the Pentium 4 Extreme Edition processor.  As you'll note, the FSB is set to 266MHz with a 13X multiplier on this chip, which when driven though the chipset's "Quad-Pumped" architecture, yields an effective 1066MHz system bus speed.

Thermal characteristics of the new 3.46GHz Extreme Edition are still in the same ballpark as previous 800MHz FSB versions in the LGA775 package.  On an open air test bench environment, we observed idle temps at 36ºC and full load temps at 55ºC.  To take this one step further we've observed full load temps running a distributed computing application like Folding At Home in a closed ATX chassis environment, reaching a stabilized 61ºC with a fair amount of ventilation in the case.  Make no mistake, this chip isn't any cooler than previous versions.  We've heard rumblings that Intel is refining its process technology in future iterations of upcoming Prescott core processors with a 2MB L2 cache and 1066MHz FSB that will mitigate these higher thermals to a certain extent.  We'll cross our fingers on that one because all of these chips are really cooking at the moment.

Overclocking The P4 EE 3.46GHz Processor
Intel burn-in mode

Though the new P4 EE 3.46GHz chip was humming along under full load with thermals hot enough to fry us a hearty breakfast sunny side up, we still thought it could handle a bit of overclocking, so we fired up the D925XECV2's "burn-in" mode in the BIOS menu.

Our results were typical of previous-generation 3.4GHz Extreme Edition CPUs, clocking in at a default voltage of 3.8GHz.  This was set with the D925XECV2 board's "+ 10%" mode BIOS option, which is a limitation of the Intel motherboard more than anything else.  Perhaps a more interesting datapoint is the fact that our system bus speed is also now at 1173MHz, which is derived from a 293MHz base clock and equates to 586MHz DDR2 DRAM timings.

Intel D925XECV2 BIOS Menus
1006MHz FSB Ready

Just for giggles we thought it would be interesting to take you through a quick tour of a few of the D925XECV2's BIOS screens.  Again, Intel BIOS options aren't set up for the "overclocker" or enthusiast in mind, but they give the end user access to many settings that are required to tune the board for optimal performance.

Again, this Intel board offers "burn in mode" for both CPU and Memory timings but has no voltage adjustments for either.  You can also over-speed and tweak PCI Express links to a certain degree, but we're not sure there is any real potential upside benefit here to offset any sacrifices you'll make in stability.  All told, the D925XECV2's BIOS does just what it was intended to do, nothing more nothing less.  Enthusiasts looking to squeak out additional performance through overclocking and voltage settings will be left a bit flat.  You do however have the saving grace of being able to tweak memory timings within the BIOS, which will certainly help get some of the system's inherent DDR2 memory latency in line somewhat.

 

Transparent
Test System Details & SiSoft Sandra Testing

Our test system specs are detailed below for your edification.  Please note we've also migrated to Windows XP Service Pack 2 for this round of testing.

HotHardware Test Systems
Tough Competition
SYSTEM 1:
Intel Pentium 4 EE 3.46GHz - LGA 775


Intel D925XECV2

i925XE Motherboard

2x512MB Kingstonl DDR2 533 PC2-4300

CL 3, 3, 3, 8

NVIDIA GeForce 6800 GT


On-Board 10/100 Ethernet
On-Board Audio

WD "Raptor" 36GB Hard Drive
10,000 RPM SATA

Windows XP Pro SP2 (Fully Patched)
NVIDIA Forceware v66.81
DirectX 9.0c
SYSTEM 2:
Intel Pentium 4 EE 3.4GHz - LGA775

Intel Pentium 4 560 3.6GHz - LGA775

ASUS P5AD2

i925X Motherboard

2x512MB Kingston DDR2 533 PC2-4300

CL 3, 3, 3, 8

NVIDIA GeForce 6800 GT


On-Board 10/100 Ethernet
On-Board Audio

WD "Raptor" 36GB Hard Drive
10,000 RPM SATA

Windows XP Pro SP2 (Fully Patched)
NVIDIA Forceware v66.81
DirectX 9.0c
SYSTEM 3:
AMD Athlon 64 FX-55 (2.6GHz)

AMD Athlon 64 FX-53 (2.4GHz)
AMD Athlon 64 4000+ (2.4GHz)
AMD Athlon 64 3800+ (2.4GHz)

MSI K8N Neo2 Platinum
NVIDIA nForce 3 Ultra Chipset

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

NVIDIA GeForce 6800 GT


On-Board 10/100 Ethernet
On-Board Audio

WD "Raptor" 36GB Hard Drive
10,000 RPM SATA

Windows XP Pro SP2 (Fully Patched)
NVIDIA Forceware v66.81
DirectX 9.0c


Intel System Information Utility Detail

 

Just a quick note on our test system specs before we dig into benchmark specifics:  In all our test setups, we used an NVIDIA GeForce 6800 GT graphics card.  In the case of the PCI Express-capable Intel systems, we used the NV45 version of the GF 6800 GT, and for the AGP graphics-based systems, we used standard GeForce 6800 GT AGP cards.  In both cases, GPU core and memory clock speeds were identical.  For the Intel-based systems, we used the latest chipset drivers provided with the i925XE launch kit.  In all cases, we were making an effort to utilize identical components across the various Intel and AMD platforms wherever possible.

Preliminary Benchmarks With SiSoft SANDRA 2004
Synthetic Testing

We began our testing with SiSoftware's SANDRA, the System ANalyzer, Diagnostic, and Reporting Assistant. SANDRA consists of a set of information and diagnostic utilities that test some basic metrics on major subsystem performance levels in a number of areas. We ran three of the built-in subsystem tests that comprise the SANDRA 2004 suite: CPU, Multimedia, and the Memory test, which is based on the "Stream" memory benchmark algorithm.


    
P4 EE 3.46GHz 1066MHz FSB - Intel i925XE Chipset

 

      
P4 EE 3.4GHz 800MHz FSB - Intel i925X Chipset

We've taken scores from an 800MHz FSB P4 EE at 3.4GHz and compared them for you here versus the new 3.46GHz P4 EE with 1066MHz FSB. We should also point out again that both test systems are running system memory at CAS 3, 3, 3, 8 timings. The scores at this setting are actually slightly better than stardard DDR400 RAM at CAS2. As you'll note, the only major difference that is reported in these Sandra test modules are the memory bandwidth scores.  There's about a 14% advantage in overall memory bandwidth for the i925XE system.  Beyond that, the P4 EE 3.46GHz chip scores only marginally better than a standard P4 EE 3.4GHz 800MHz FSB chip, in floating point and integer performance.

Transparent
PCMark04 Tests

 

PCMark04 makes use of some fairly common desktop computer tasks for its benchmark metrics, although with its weighted scoring scale, it is still to be considered more of a synthetic benchmark.

Futuremark PCMark04
More Synthetic CPU and Memory Benchmarks

Source - FutureMark White Paper - "The CPU test suite is a collection of tests that are run to isolate the performance of the CPU. There are nine tests in all. Two pairs of tests are run multithreaded - each test in the pair is run in its own thread.  The remaining five tests are run single threaded. These tests include such functions as file encryption, decryption, compression and decompression, grammar check, audio conversion, WMV and DivX video compression."

The P4 560 takes the lead here with its overall clock speed advantage in the field.  Additionally, it's a complete sweep of this test by the P4 lineup versus the Athlon 64 lineup. 


Source - FutureMark White Paper - "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."

In the PCMark04 Memory test, the Athlon 64's integrated memory controller allows it to surge ahead of some of the fastest Pentium 4s.  However, the P4 EE 3.46GHz CPU, with its 1066MHz FSB, shows a clear advantage over all others, even the top-end Athlon 64 FX-55.  PCMark04's memory read/write operations obviously favors the higher-speed system bus of this new Extreme Edition P4, as expected.

Transparent
3DMark05 CPU Test

 

We've included Futuremark's 3DMark05 in our testing suite for CPUs only at this point in time.  The CPU tests of this benchmark were the only subroutines we enabled for our test run.  Additionally, we've taken a snippet from a section of the Futuremark 3DMark05 white paper that explains exactly what the CPU test does specifically.

3DMark05 CPU Test
DirectX Gaming Performance - CPU Test

Source - 3DMark 0 5 Whitepaper-

As in the previous 3DMark version, the CPU test runs game tests in low resolution using software vertex processing and disabled post-processing. This decreases the graphics card workload, and makes the test result reflect above all the CPU's performance rendering 3D scenes and performing other 3D game related tasks like performing matrix calculations. The CPU test also uses fixed frame rendering to further ensure the workload stays the same for all systems.

In 3DMark05, an additional workload typical for the CPU in 3D games has been added. The CPU not only calculates the vertex shaders; it also continuously calculates the flight path of the air ship. The air ship actually flies the same path every time, in order to keep the workload the same between different systems, but the calculations are performed as if it would intelligently steer according to the canyon shape and other obstacles like the sea monster jumping up from the water. The path finding algorithm used is D*Lite (http://www.cc.gatech.edu/fac/Sven.Koenig/).

For the Pentium 4 lineup in this test, it looks like overall clock speed dictates the results.  However, this is definitely not the case for the AMD Athlon scores, where a 2.6GHz Athlon FX-55 beats out the 3.6GHz Pentium 4 560 Prescott core CPU.  That's a full 1GHz differential in clock speed, and AMD's flaghship desktop core still wins by a small margin.  The new 3.46GHz Extreme Edition P4 drops in about where we expected in this benchmark with a slight edge over the standard 3.4GHz P4 EE.

Transparent
Business & Content Creation Winstone 2004 Tests

 

Our Business and Content Creation Winstone testing will showcase overall desktop performance using standard business, desktop publishing, multimedia creation and 3D rendering applications.

Business & Content Creation Winstones
Real-World Application Performance

The PC Magazine Business Winstone 2004 test utilizes the following applications in its benchmark:

  • Microsoft Access 2002
  • Microsoft Excel 2002
  • Microsoft FrontPage 2002
  • Microsoft Outlook 2002
  • Microsoft PowerPoint 2002
  • Microsoft Project 2002
  • Microsoft Word 2002
  • Norton Antivirus Professional Edition 2003
  • WinZip 8.1

The Pentium 4s posted scores all within this test's margin of error versus each other.  However, the Athlon 64s averaged 13 – 16% lead in the Business Winstone test suite.

 


The PC Magazine Content Creation Winstone 2004 test utilizes the following applications in its benchmark:

  • Adobe Photoshop 7.0.1
  • Adobe Premiere 6.50
  • Macromedia Director MX 9.0
  • Macromedia Dreamweaver MX 6.1
  • Microsoft Windows Media Encoder 9 Version 9.00.00.2980
  • NewTek's LightWave 3D 7.5b
  • Steinberg WaveLab 4.0f

The Content Creation Winstone test shows a nearly identical performance spread with the Pentium 4s all bunched up at the 33-point mark and the Athlon 64s showing a 15 – 17% performance advantage.  Somewhat interesting to note is that the higher-bus bandwidth of the new 3.46GHz Pentium 4 EE allows it to keep pace with the 3.6GHz P4 Prescott core CPU, which has a full 200MHz clock speed advantage.  Regardless, the differential between those scores and the standard 3.4GHz P4 EE is negligible. This test also underscores what we've all known for quite some time now and what Intel has had to admit wholeheartedly over the past few years: Higher clock speeds do not always equate to higher performance.

 

 

Transparent
Windows Media Encoder 9 and Lame MP3 Encoding Tests

 

Windows Media Encoder 9
More Digital Video Encoding

We continued our testing with a video encoding benchmark using Windows Media Encoder 9.  In this test, we took a 416MB Digital Video file and encoded to WMV9 format.  Run times were recorded in Minutes:Seconds, with lower times again indicating better performance.

It seems that media encoding has always been Intel's strong suit, and this benchmark data point is no exception.  WME9 is a heavily SSE optimized and multithreaded application, which does take advantage of the P4's Hyper-Threading features.  As a result, we see the chips fall here with a solid advantage for Intel, where the P4 560 captures a 10 – 18% lead, depending on which Athlon we compare it to.  However, once again we don't see any advantage for the 3.46GHz P4 EE versus its 800MHz FSB driven counterpart in this test.  In terms of the overall performance scale here, relatively speaking within each processor architecture, clock speed is the single most limiting factor with Windows Media Encoder 9.

LAME MP3 Encoding Tests
Converting A Large WAV To MP3

In our Lame MP3 tests we convert audio files to MP3 format, which is a very popular scenario that many end users work with on a regular basis, to provide portability and storage of their digital audio content.  In this test, we chose a large 223MB WAV file (a never-ending Grateful Dead jam) and converted it to the MP3 format.  Processing times are recorded below, and shorter times equate to better performance.

For MP3 audio encoding, memory bandwidth is probably the most critical aspect of overall system performance.  Here the new 3.46GHz P4 Extreme Edition takes a first place tie with the Athlon 64 FX-55, and with its slightly higher clock speed, just edged out the standard 3.4GHz P4 EE.  Although P4s are strong again in this test, capturing three of the top four slots performance-wise, a 1066MHz frontside bus with the 3.46GHz P4 EE doesn't provide any additional performance enhancement.

 

Transparent
Cinebench 3D Rendering Tests

 

Cinebench 2003 Performance Tests
3D Modeling & Rendering Tests

The Cinebench 2003 benchmark is an OpenGL 3D rendering performance test, based on the commercially available Cinema 4D application.  This is a multithreaded, multiprocessor 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).  We ran two sets of numbers, one in single-thread mode and another in the benchmark's multithread mode for our Hyper-Threading enabled P4 test systems.  The Athlons are only capable of running the single thread test, hence their lack of those data points in the graph below.

In this 3D Rendering benchmark we see similar results to what we saw in the MP3 encoding test runs of our Lame MP3 benchmarks, at least as far as single-threaded scores go. There is one small exception, and that is the P4 560 Prescott core CPU bringing up the rear of the pack by a 6 – 9% margin.  Beyond that, our two P4 Extreme Edition processors are in a virtual dead heat with each other, regardless of FSB speeds, and the Athlon 64 FX-55 takes our fastest single-threaded P4 score by about 5%.  That said, there's no point to running this application in single-threaded mode when you can realize a hearty 13% performance lead over the fastest Athlon 64 score with "multi-CPU" mode enabled in Cinema 4D and Cinebench.

 

Transparent
Kribibench 3D Modeling

Kribibench v1.1
Details: www.adeptdevelopment.com

We then fired up the "KribiBench" benchmark produced by Adept Development.  KribiBench is an SSE aware software renderer.  A 3D model is rendered and animated by the host CPU, and the average frames per second are reported.  We used two of the included models with this benchmark: an Office model consisting of 42,000 polygons and an enormous "Ultra" model that comprises over 16 million polygons.

Intel's SSE optimizations once again allow the three Pentium 4 processors to claim to poll positions.  Also, once again we see little advantage (only about 1%) for the 1066MHz FSB P4 EE versus the 800MHz FSB version.  The fastest P4 score here is about 9% faster than the fastest Athlon 64 FX-55 score.

 

In the significantly more strenuous "Ultra" model test, we see the Pentium 4 claim a 6 – 9% advantage over the Athlon 64 CPUs.  And again, the i925XE platform with the 3.46GHz P4 EE, doesn't offer much if any more performance than its 800MHz FSB i925X-derived sibling.

Transparent
Wolfenstein Enemy Territory Benchmarks

 

 

ids Quake engines have always been memory- and system-bandwidth sensitive, so we stepped into Wolfenstein Enemy Territory for some frag-worthy frames-per-second testing in a low-res drag race.

Wolfenstein: Enemy Territory
OpenGL Quake Engine Gaming

Wolfenstein: ET is a free, standalone multiplayer game that is based on the original Return to Castle Wolfenstein that was released a few years back. It uses a heavily modified version of the Quake III engine, which makes it a very easy-to-use benchmarking tool.  We ran the test using the "Fastest" setting at a low resolution of 640 X 480, using 16-bit color and textures.  Running this test with a high-end graphics card at these minimal settings isolates processor and memory performance, without being limited by the graphics subsystem.

No question about it, the Athlon 64's got game, and in a big way, over the P4.  Even the significantly less expensive 3800+ manages to best the 3.46GHz P4EE by a couple of FPS.  However, we do see a more significant advantage of about 3% or so for the 1066MHz FSB P4 EE. We were expecting more, frankly, so here's hoping that might pick up a bit more in UT2004 and Doom 3 testing.

Transparent
Unreal Tournament 2004 Benchmarks

 

When testing with UT 2004, we use a specific set of game engine initialization settings that ensure all of the systems are being benchmarked with the exact same in-game settings and graphical options.  Like the other in-game tests, we used a "Low-Quality" setting with UT2004 that isolated CPU performance. 

Unreal Tournament 2004
DirectX Gaming Performance

The scores below were taken from a custom demo recording of a multiplayer match that is relatively CPU-limited in terms of performance.

There's not much more to say here than what our benchmark numbers reveal: total dominance by the Athlon 64 across the board, along with little if any advantage for the 3.46GHz P4 EE and i925XE chipset.  To add insult to injury, the 3.6GHz Prescott P4 falls behind the slowest P4 3.4GHz Extreme Edition score by about 11%.  It seems that the ultra low latency of the Athlon 64's integrated memory controller really allows it to shine in some of these gaming benchmarks.

 

Transparent
Doom 3 Benchmarks - Multiplayer Match Demo Benchmarks

 

 

Doom 3 Benchmarks
OpenGL Gaming Performance

For our last game test, we benchmarked all seven of the test systems using our custom multiplayer Doom 3 timdemo. We cranked the resolution down to 640 x 480 and configured the game to run at its "Low-Quality" graphics setting. Without question, Doom 3 is a graphics card killer of a benchmark, but when we turn down the resolution and all that graphical detail and glitz, the game is significantly more CPU-dependant.

Firstly, the Prescott core Pentium 4 is a bit happier in Doom 3 versus UT2004 and the 3.6GHz chip only drops behind the 3.4GHz P4 EE by a couple of frames.  Additionally, once again we netted only about a 3% gain going from an 800MHz FSB to a 1066MHz FSB with the P4 Extreme Edition.  Lastly, the Athlon 64s swept in a big "Boston Red Sox" kind of way, with the "Big Papi" Athlon 64 FX-55 blowing by the fastest P4 EE by about 15%.  Folks, if you are a gamer, there's no faster desktop processor architecture, at this point in time, than the Athlon 64.

 

Transparent
Performance Analysis & Conclusion

 

 

Benchmark Summary:
When we look at the entire spread of benchmark readings we took here for you today, we see the P4 pulling down six wins to the Athlon 64's five wins, with two virtual draws in the 3DMark05 and Lame MP3 tests.  These metrics should be contrasted by the fact that the benchmarks that were won by the Athlons were more mainstream desktop and gaming applications that many of our readership finds most important.  The P4 definitely excels in audio and video media encoding, as well as mutlithreaded applications like Cinema 4D in our Cinebench benchmarks.  The Athlons were the clear and distinct victors everywhere else.  Finally, we noted a smallish 1 – 3% advantage for the new 3.46GHz P4 Extreme Edition and i925XE chipset, versus the current standard P4 EE and the i925X.

 

We are consistently reminded by you, our readers, to "tell it like it is."  Our role at HotHardware.com, as tech journalists, is fairly straightforward and simple:  test, re-test, evaluate, summarize, and pass some level of judgment.  We'll sum things up for you here by saying that Intel's high-end desktop enthusiast platform has fallen behind, dollar for dollar, versus AMD's.  If you are the type who is heavily into media encoding, then there is no better choice than the Pentium 4 Extreme Edition processor.  If you're a gamer and standard business app user, like the apps we've shown you in our Business and Content Creation Winstone tests, then there's no better CPU right now than the Athlon 64.  Additionally, the latter group of end users we described in these two scenarios is arguably significantly larger than the former.  We can't tell the story any more plainly than that.

The new 3.46GHz Pentium 4 Extreme Edition coupled with the i925XE chipset offers a small performance gain for the enthusiast in certain gaming scenarios and the occasional Media Encoding or 3D Rendering workload.  Again, the performance gains we observed were in the 1 – 3% range, hardly what we expected frankly.  In fact, equal performance gains can be realized just by making the move to lower latency DDR2 DRAM set at CAS 3, 3, 3, 8 timings.  With identical system memory timings in our i925X and i925XE systems, the net gain from a high-speed 1066MHz FSB is observable but frankly nothing to write home about, at least in the applications we tested it in. 

The new 3.46GHz Pentium 4 EE is going to list in 1,000 piece pricing at $999, and we expect all motherboard OEMs to transition to the i925XE in their high-end lineup.  Our last price check shows Athlon FX-55 chips at around $900 in retail.  If you are the type to belly up with this kind of coin, then it certainly makes sense to make your move to the new 1066MHz-enabled P4 EE, but the Athlon FX-55 for all intents and purposes is a much better value for the masses, with significantly less expensive platform components like motherboards and RAM.  Intel is expected to release a new version of the Prescott core Pentium 4, with an integrated 2MB of L2 cache, a tweaked branch predictor in its pipeline and targeted improved power efficiencies, as well.  At a rumored 3.73GHz, Intel needs to get this chip out the door soon, but we're hearing that it is being pushed out to at least until sometime in early 2005.  We'll be here with the details and benchmarks when the time comes, rest assured.



Content Property of HotHardware.com