<!–#set var="article_header" value="Clash Of The OC'd Titans:
Athlon XP 2300+ vs. Pentium 4/3000″ –>
Speed Euphoria: AMD vs. Intel – Who’s Ahead?
It might sound like a standard topic, but in reality it’s quite dramatic: the AMD Athlon XP 2300+ challenges the Intel Pentium 4/3000. We’ve overclocked these two processors to the extreme, and in the end, they bedazzle with performances that will not be available on the market for another nine months. After all, with their latest offerings, the two manufacturers tout their respective processors as representatives of maximum performance levels. The task of coaxing that maximum performance out of the two competitors presented us with a particular challenge, but it was all done under conditions for absolutely stable operation.
The top products AMD Athlon XP 2000+ and Intel Pentium 4/2200 provide the foundation for extreme overclocking.
AMD Athlon XP 2000+ at the limit: 1912 MHz with the help of watercooling.
Testing station with the Athlon XP, which has attained the level of 2300+.
Speed Euphoria: AMD vs. Intel – Who’s Ahead?, Continued
The more reasonable of us might claim that the performance potential of the top processors from the AMD Athlon and Intel Pentium 4 series is completely sufficient. However, the urge to reach greater heights lies within all ambitious users who do not regard the PC as merely a simple tool. And lucky for us, for even the strongest and most expensive CPUs, there are methods of enhancing performance. (One can never get enough!)
This exorbitant play of numbers continues with our latest efforts: we equipped the test platforms with the best of the currently available watercooling systems. In addition, more than 50 of the latest boards were tested for their overclocking ability, and the best out of the group were used for this test. In the preparatory stages, we removed the multiplier lock from three of the Athlon XP 2000+ series CPUs.
A view of the water-cooled CPU element in the AMD platform.
In comparison, today the average system is equipped with a 900 MHz processor as well as an MX2000 graphics card. The test candidates reach performance levels of which most of us can only dream: in the Quake III test, we measured a top rate of 325 fps, while the 900 MHz system only barely manages to reach 141 fps. The performance in video encoding via Xmpeg is also impressive at 39 fps – that is far more than 25 fps real time, not to mention 18 fps from an Athlon 900.
The watercooling system from Innovatek enables extreme overclocking. Not much will run if you use a conventional cooler with an AMD system.
Tuning AMD & Intel CPUs In Practice
Both the Intel and AMD platforms were operated with the help of a watercooling system, otherwise it would have been impossible to overclock the Athlon XP and the Pentium 4. Those that are currently looking for the fastest CPU can choose between an AMD Athlon XP 2000+ with 1666 MHz or an Intel Pentium 4 with 2200 MHz. However, there is a great difference in price: a Pentium 4/2200 usually costs twice as much as an Athlon XP 2200+. Users who are still unsatisfied with the performance offered by these two top CPUs can enhance it even further, but only by overclocking. AMD and Intel have each integrated an overclocking protection in their processors, though, so overclocking isn’t all that easy a task. Both manufacturers have implemented a fixed clock multiplier to limit the actions of inexperienced users.
Depending on the external temperature, the enormous heat exchanger can cool the water to almost 0 degrees Celsius. This is how efficient cooling of AMD and Intel CPUs is made possible.
The “small” heat exchanger is only appropriate for marginal overclocking. For extreme overclocking, the larger version is required.
At the same time, the fact of the matter is that a high performance processor alone does not a high performance system make. An important component to successful overclocking is the motherboard, which must allow you to set the Front Side Bus and the memory clock. The most important criterium in checking for appropriate boards was the CPU core voltage – the performance of a processor cannot be continually increased without substantially increasing the processor voltage! At the moment, there are only a few boards that will allow voltages which do not adhere to the specifications of the CPU manufacturer. In thoroughly examining 50 boards for overclocking suitability, we found that Gigabyte offered the best conditions for both platforms. So for our AMD platform, we used the Gigabyte GA-7VTHX, which, however, does not offer the best performance among boards with the VIA KT266A chipset. Ultimately, what’s more important is that the board provides a good basis for an extreme increase in clock speed, and that it guarantees stable operation, as well.
Tuning AMD & Intel CPUs In Practice, Continued
We equipped the Northbridge of the chipset with a cooling fan to ensure its safety.
The CPU element of the Intel Pentium 4 based on Socket 478.
A view of the test system with the Intel Pentium 4/3000.
A similar problem was found with the Intel platform: there are mainboards for the Pentium 4 based on socket 478, with DDR SDRAM as well as RDRAM support. Using a slow SDRAM (PC133) was out of the question. A motherboard with Rambus memory is unsuited to extreme increases in clock speed, for the simple reason that there is currently no board on the market that allows a high enough CPU core voltage (non-specification). Additionally, there’s the problem of limited bus clock speed on boards with the Intel 850 chipset: when overclocking, the AGP bus speed increases in direct proportion to the Front Side Bus. The result: with an FSB of 137 MHz (required for a CPU clock of 3000 MHz), for example, the AGP bus runs at 90.5 MHz. At this clock speed, graphics cards with the GeForce 3 chip cannot be operated, which was the result of our tests. Cards with the GeForce 2 chip proved to be the exception – sometimes, they were even able to reach 100 MHz. Nevertheless, in the 3D category, the performance values of the older GeForce 2 cards are too low for the benchmark records. Because of this, we used an Intel 845D mainboard to reach the 3000 MHz record. In contrast to the Intel 850, the Intel 845D at 137 MHz FSB offers an almost standard AGP clock speed of 67.9 MHz.
The following table lists the FSB and AGP clock speeds.
Processor | PCI Clock | AGP Clock | Bus Clock |
Pentium 4/3000A | 33,9 MHz | 67,9 MHz | 137 MHz |
Pentium 4/2600A | 39,3 MHz | 78,5 MHz | 119 MHz |
Pentium 4/2400A | 36,3 MHz | 72,6 MHz | 110 MHz |
Pentium 4/2200A | 33 MHz | 66 MHz | 100 MHz |
Pentium 4/2000A | 33 MHz | 66 MHz | 100 MHz |
Pentium 4/2000 | 33 MHz | 66 MHz | 100 MHz |
Pentium 4/1900 | 33 MHz | 66 MHz | 100 MHz |
Pentium 4/1800 | 33 MHz | 66 MHz | 100 MHz |
Pentium 4/1700 | 33 MHz | 66 MHz | 100 MHz |
Pentium 4/1600 | 33 MHz | 66 MHz | 100 MHz |
Pentium 4/1500 | 33 MHz | 66 MHz | 100 MHz |
Pentium 4/1400 | 33 MHz | 66 MHz | 100 MHz |
Overclocking In Detail: Athlon XP With 1866 MHz
The fastest AMD processor to date is the Athlon XP 2000+ with 1666 MHz. The manufacturer delivered a CPU that was factory-fitted with a fixed multiplier, which we promptly unlocked according to our previous experience. You can read about this process in the article “Plastic Surgery: Releasing the Athlon XP To Hit 2000+” and see it in the accompanying video “Second THG Video: Unlocking The Athlon XP/MP.”
In order to attain the maximum increase in clock speed in an Athlon XP 2000+, it is necessary to lower the multiplier from 12.5 to 12.0. In the test, the Athlon ran stably up to a clock speed of 1866 MHz – at clock speeds even higher than this, it was no longer possible to conduct the benchmark tests. A clock speed of 1866 MHz corresponds to an Athlon XP 2300+.
At this point, we should explain why we lowered the multiplier to 12.0: on the one hand, a clock rate of 1866 MHz cannot be achieved with an Athlon XP 1900+ that is factory-equipped with a multiplier of 2.0. On the other hand, it is necessary to set the FSB clock and memory clock as high as possible, so that the memory performance required for such a high CPU clock is available. Take the Quake III benchmark, for example, which gives you a visual comparison between the two multiplier settings (12.0 and 12.5) at an identical CPU clock speed.
Processor | Memory Clock | CAS Latency | Type of Memory |
Athlon XP 2300+ | 155 MHz | 2.5 | DDR-SDRAM (DDR333, PC2700) |
Athlon XP 2200+ | 150 MHz | 2.5 | DDR-SDRAM (DDR333, PC2700) |
Athlon XP 2100+ | 144 MHz | 2.0 | DDR-SDRAM (DDR333, PC2700) |
Athlon XP 2000+ | 133 MHz | 2.0 | DDR-SDRAM (DDR266, PC2100) |
Athlon XP 1900+ | 133 MHz | 2.0 | DDR-SDRAM (DDR266, PC2100) |
Athlon XP 1800+ | 133 MHz | 2.0 | DDR-SDRAM (DDR266, PC2100) |
Athlon XP 1700+ | 133 MHz | 2.0 | DDR-SDRAM (DDR266, PC2100) |
Athlon XP 1600+ | 133 MHz | 2.0 | DDR-SDRAM (DDR266, PC2100) |
Athlon XP 1500+ | 133 MHz | 2.0 | DDR-SDRAM (DDR266, PC2100) |
Athlon 1400 | 133 MHz | 2.0 | DDR-SDRAM (DDR266, PC2100) |
Athlon 900 | 100 MHz | 2.0 | DDR-SDRAM (DDR266, PC2100) |
In the latest Athlon XP processors with the Palomino core, the maximum multiplier value is 12.5, which is technically limited. Every Socket 462 board has only four registers available for coding the CPU multiplier. Things will look different though, once AMD’s Athlon XP 2200+ (Thoroughbred core) makes its debut at the CeBIT 2002 in March. With an FSB of 133 MHz, a board would have to offer a multiplier of 13.5 – AMD solves this problem with a translation table in the CPU. A further possibility is that the multiplier, as delivered with the board, would be fully ignored and that the CPU would dictate the settings.
CPU Clock (MHz) | Multipl. | L1 | |||||
FSB100 | FSB133 | X | 1 | 2 | 3 | 4 | 5 |
500 | 667 | 5.0 | open | open | open | open | open |
550 | 733 | 5.5 | open | open | open | open | open |
600 | 800 | 6.0 | open | open | open | open | open |
650 | 867 | 6.5 | open | open | open | open | open |
700 | 933 | 7.0 | open | open | open | open | open |
750 | 1000 | 7.5 | open | open | open | open | open |
800 | 1067 | 8.0 | open | open | open | open | open |
850 | 1133 | 8.5 | open | open | open | open | open |
900 | 1200 | 9.0 | open | open | open | open | open |
950 | 1267 | 9.5 | open | open | open | open | open |
1000 | 1333 | 10.0 | open | open | open | open | open |
1050 | 1400 | 10.5 | open | open | open | open | open |
1100 | 1467 | 11.0 | open | open | open | open | open |
1150 | 1533 | 11.5 | open | open | open | open | open |
1200 | 1600 | 12.0 | open | open | open | open | open |
1250 | 1667 | 12.5 | open | open | open | open | open |
Free | closed | closed | closed | closed | closed |
CPU Clock (MHz) | Multipl. | L3 | ||||
FSB100 | FSB133 | X | 1 | 2 | 3 | 4 |
500 | 667 | 5.0 | closed | closed | open | open |
550 | 733 | 5.5 | open | closed | closed | open |
600 | 800 | 6.0 | closed | open | open | closed |
650 | 867 | 6.5 | open | open | closed | closed |
700 | 933 | 7.0 | closed | closed | open | open |
750 | 1000 | 7.5 | open | closed | closed | open |
800 | 1067 | 8.0 | closed | open | open | closed |
850 | 1133 | 8.5 | open | open | closed | closed |
900 | 1200 | 9.0 | closed | closed | open | open |
950 | 1267 | 9.5 | open | closed | closed | open |
1000 | 1333 | 10.0 | closed | open | open | closed |
1050 | 1400 | 10.5 | open | open | closed | closed |
1100 | 1467 | 11.0 | closed | closed | open | open |
1150 | 1533 | 11.5 | open | closed | closed | open |
1200 | 1600 | 12.0 | closed | open | open | closed |
1250 | 1667 | 12.5 | open | open | closed | closed |
Free | X | X | X | X |
CPU Clock (MHz) | Multipl. | L4 | ||||
FSB100 | FSB133 | X | 1 | 2 | 3 | 4 |
500 | 667 | 5.0 | open | closed | closed | open |
550 | 733 | 5.5 | open | closed | closed | open |
600 | 800 | 6.0 | open | closed | closed | open |
650 | 867 | 6.5 | open | closed | closed | open |
700 | 933 | 7.0 | closed | open | open | closed |
750 | 1000 | 7.5 | closed | open | open | closed |
800 | 1067 | 8.0 | closed | open | open | closed |
850 | 1133 | 8.5 | closed | open | open | closed |
900 | 1200 | 9.0 | open | open | closed | closed |
950 | 1267 | 9.5 | open | open | closed | closed |
1000 | 1333 | 10.0 | open | open | closed | closed |
1050 | 1400 | 10.5 | open | open | closed | closed |
1100 | 1467 | 11.0 | closed | closed | open | open |
1150 | 1533 | 11.5 | closed | closed | open | open |
1200 | 1600 | 12.0 | closed | closed | open | open |
1250 | 1667 | 12.5 | closed | closed | open | open |
Free | X | X | X | X |
CPU Voltage: 1.92 Volt for the Athlon
Processor | CPU Clock | Clock Multiplier | CPU Core Voltage |
Athlon XP 2300+ | 1866 MHz | 12 | 1,920 V |
Athlon XP 2200+ | 1800 MHz | 12 | 1,920 V |
Athlon XP 2100+ | 1733 MHz | 12 | 1,920 V |
Athlon XP 2000+ | 1666 MHz | 12.5 | 1,750 V |
Athlon XP 1900+ | 1600 MHz | 12 | 1,750 V |
Athlon XP 1800+ | 1533 MHz | 11.5 | 1,750 V |
Athlon XP 1700+ | 1466 MHz | 11 | 1,750 V |
Athlon XP 1600+ | 1400 MHz | 10.5 | 1,750 V |
Athlon XP 1500+ | 1333 MHz | 11 | 1,750 V |
Athlon 1400 | 1400 MHz | 10.5 | 1,750 V |
Athlon 900 | 900 MHz | 9 | 1,750 V |
An important condition for successful overclocking is the increase in CPU core voltage. AMD Athlon XP 2200+ processors from the factory work with 1.75 Volt. At such a setting, it is nearly impossible to increase the clock speed by a substantial amount. Extreme clock speeds of up to 1900 MHz are possible only if the CPU core voltage is increased to 1.90 Volt. Also, an increase in core voltage automatically causes an increase in the switching speed of the transistors, as well as a higher thermal power. The large amount of heat generated can be controlled by an efficient watercooling system. Otherwise, the processor will die a thermal death (see our video: Hot Spot – How Modern Processors Cope With Heat Emergencies). Only a few boards with the VIA KT266A chipset offer a core voltage of more than 1.85 Volt. In the test, we use the Gigabyte GA-7VTXH, which allows a maximum of 1.92 Volt. Epox EP-8KHA+, a well-known and frequently used board, only allows a maximum of 1.85 Volt, which is not optimal for extreme overclocking. The only way to remedy this with the Epox is to modify the voltage regulator.
DDR333 Memory For 166 MHz FSB
Processor | PCI Clock | AGP Clock | Bus Clock |
Athlon XP 2300+ | 38,5 MHz | 76,9 MHz | 155 MHz |
Athlon XP 2200+ | 37,2 MHz | 74,4 MHz | 150 MHz |
Athlon XP 2100+ | 34,5 MHz | 68,9 MHz | 139 MHz |
Athlon XP 2000+ | 33 MHz | 66 MHz | 133 MHz |
Athlon XP 1900+ | 33 MHz | 66 MHz | 133 MHz |
Athlon XP 1800+ | 33 MHz | 66 MHz | 133 MHz |
Athlon XP 1700+ | 33 MHz | 66 MHz | 133 MHz |
Athlon XP 1600+ | 33 MHz | 66 MHz | 133 MHz |
Athlon XP 1500+ | 33 MHz | 66 MHz | 133 MHz |
Athlon 1400 | 33 MHz | 66 MHz | 133 MHz |
Athlon 900 | 33 MHz | 66 MHz | 100 MHz |
In order to operate the Athlon XP at a maximum clock speed of 1866 MHz, we increased the Front Side Bus and memory speed to 155 MHz (12.0 x 155 MHz = 1866 MHz). Such a clock speed is not possible with normal DDR SDRAM modules (PC2100 or DDR266). For this reason, we used the fastest DDR333 module, which is suited to running at 166 MHz. In the near future, new memory modules that run up to 266 MHz (DDR400 and DDRII/533) will be available. According to the SiS roadmap, the corresponding chipsets should begin to make their appearances starting in July or December of this year.
Memory modules compared (from top to bottom): DDR333, DDR266 and RDRAM with PC800.
Overclocking in Detail: Pentium 4 With 3000 MHz
The limiting factor with the factory-delivered Intel Pentium 4/2200 is the fixed multiplier. This is how Intel prevents the processor from being easily overclocked. It is impossible to get around this hurdle because a special SRAM register on the CPU contains the multiplier, and this cannot be changed. So this leaves only one option: overclocking the Front Side Bus. In order to measure the performance of an Intel CPU of the future, we used two different platforms – one based on the Intel 845D chipset and the other based on the Intel 850 chipset. With the help of the Gigabyte GA-8IRXP (Intel 845D), we set a record with 3 GHz.
In our last test “Tom’s Hardware Speed Project: Pentium 4, Over 3 GHz,” the benchmarks have already shown that, in individual disciplines, the performance did not reach levels that one might expect at such a high clock speed. Here, we can deduce that a DDR SDRAM module clocked at 137 MHz cannot deliver the necessary bandwidth in combination with the Pentium 4, thereby accounting for performance that is somewhat lower than expected. By comparison, our Pentium 4 system with the Intel 850 chipset shows better results in a few of the benchmarks, despite a lower clock rate of 2600 (Pentium 4/2600).
RDRAM Memory: High Speed With 472 MHz
Our RDRAM system (Intel 850 chipset) was able to run at a maximum clock speed of 2600 MHz. The only limiting factor was the AGP bus, which ran at 77.9 MHz in this case. The GeForce 3 graphics card was unable to reach higher speeds. By contrast, the memory ran at 472 MHz, which did not present any problems. In the summer of this year, Intel will introduce the Pentium 4/2533 (19,0 x 133 MHz = 2533 MHz), which will operate with an FSB clock of 133 MHz. Likewise, the speed of the RDRAM memory will also be increased to 533 MHz (PC1066).
Up Close: All Test Components
The Gigabyte GA-7VTXH: this board is suitable for extreme overclocking an AMD Athlon XP 2000+. This is the only motherboard that offers 1.92 Volt. (We used a chipset fan for extra cooling from another motherboard.)
According to our test findings, the Gigabyte GA-8IRXP provides the optimal basis for overclocking an Intel Pentium 4 (Socket 478). It offers a core voltage of 1.85 Volt max. The majority of boards work with 1.65 Volt max.
Test Configuration and Details
Intel Hardware Socket 478 |
|
Processor | Intel Pentium 4/3000A MHz (~548 MHz QDR FSB) Intel Pentium 4/2600A MHz (~472 MHz QDR FSB) Intel Pentium 4/2400A MHz (~440 MHz QDR FSB) Intel Pentium 4/2200A MHz (400 MHz QDR FSB) Intel Pentium 4/2000A MHz (400 MHz QDR FSB) Intel Pentium 4/2000 MHz (400 MHz QDR FSB) Intel Pentium 4/1900 MHz (400 MHz QDR FSB) Intel Pentium 4/1800 MHz (400 MHz QDR FSB) Intel Pentium 4/1700 MHz (400 MHz QDR FSB) Intel Pentium 4/1600 MHz (400 MHz QDR FSB) Intel Pentium 4/1500 MHz (400 MHz QDR FSB) Intel Pentium 4/1400 MHz (400 MHz QDR FSB) |
Motherboard | ASUS P4T-E (I850) Revision: 1.00 |
Memory | 2 x 128 MB, RDRAM, 400 MHz, Viking |
Motherboard | GIGABYTE GA-8IRXP (P4 Titan DDR) Revision: 2.0 |
Memory | 256 MB, DDR-SDRAM, 166 MHz, Micron (DDR333) |
AMD Hardware Socket 462 |
|
Processor | AMD Athlon XP 2300+ (1866/~310 MHz DDR) AMD Athlon XP 2200+ (1800/~300 MHz DDR) AMD Athlon XP 2000+ (1666/266 MHz DDR) AMD Athlon XP 1900+ (1600/266 MHz DDR) AMD Athlon XP 1800+ (1533/266 MHz DDR) AMD Athlon XP 1700+ (1467/266 MHz DDR) AMD Athlon XP 1600+ (1400/266 MHz DDR) AMD Athlon XP 1500+ (1333/266 MHz DDR) AMD Athlon 1400 MHz (1400/266 MHz DDR) AMD Athlon 900 MHz (900/200 MHz DDR) |
Motherboard | EPOX EP-8KHA+ (VIA KT266A) Revision: 2.0. |
Memory | 256 MB DDR-SDRAM, CL2, PC2100, Micron |
Motherboard | GIGYBYTE (VIA KT266A) |
Memory | 256 MB DDR-SDRAM, CL2, PC2700 |
General Hardware | |
Graphics Card | GeForce 3 Memory: 64 MB DDR-SDRAM Memory clock: 400 MHz Chip clock: 250 MHz |
Hard Drive | 40 GB, 5T040H4, Maxtor UDMA100 7200 rpm 2 MB Cache |
Drivers & Software | |
Graphics Driver | Detonator 4 Series V21.88 |
DirectX Version | 8.1 |
OS | Windows XP, Build 2600 (English) |
Benchmarks & Settings | |
Quake III Arena | Retail Version 1.16 command line = +set cd_nocd 1 +set s_initsound 0 Graphics detail set to ‘Normal’ Benchmark using ‘Q3DEMO1’ |
3DMark2000 | Version 1.1 Build 340 – default Benchmark |
3DMark2001 | Build 200 – default Benchmark |
SiSoft Sandra 2001 | Professional Version 2001.3.7.50 |
Newtek Lightwave | Rendering Bench SKULL_HEAD_NEWEST.LWS |
mpeg4 encoding | Xmpeg 4.2a DivX 4.11 Compression: 100 Data Rate: 1500 Kbit Format: 720×576 Pixel@25 fps 150 MB VOB-Datei, no Audio |
Studio 7 | Version 7.02.7 (MPEG 2) |
Sysmark 2001 | Patch 3 |
Lame | Lame 3.89 MMX, SSE, SSE 2, 3DNow |
WinACE | 2.04, 178 MB Wave-Datei, Best Compression, Dictonary 4096 KB |
Cinema 4D XL R6 | CineBench 6.103 |
Suse Linux 7.3 | Kernel 2.4.13 Compiling |
Benchmarks Under Windows XP
OpenGL Performance | Quake 3 Arena “Demo 1” |
Direct3D Performance | 3D Mark 2000 and 3D Mark 2001 |
3D-Rendering | Cinema 4D XL R6 |
3D-Rendering | SPECviewperf “Lightscape” |
3D-Rendering | Lightwave 7 |
Audio-Encoding MP3 | Lame MP3 Encoder |
Video-Encoding MPEG-2 | Pinnacle Studio 7 |
Video-Encoding MPEG-4 | XMpeg 4.2a und Divx 4.2 |
Office Performance | Sysmark 2001 |
Archiving | WinACE 2.04 |
Linux Kernel Compiling | Suse Linux 7.3 (Kernel 2.4.13) |
SiSoft Sandra 2001 | CPU and Multimedia Bench |
We used a total of 19 different benchmark tests in order to obtain a well-rounded and balanced picture of the extremely overclocked AMD Athlon XP and Intel Pentium 4 processors. A quick glance at the benchmark results gives you an overview of 22 of the latest processors from AMD and Intel. Low-cost CPUs, such as the AMD Duron and Intel Celeron, were intentionally left out of the benchmark evaluation; such processors are meant for a different market segment in which high performance is not crucial. Here, the Intel Pentium 4 (overclocked to 3000 MHz) and the AMD Athlon XP 2300+ (overclocked to 1866 MHz) take center stage.
The OpenGL performances are measured through various Quake 3 tests – the Direct3D performance from the DirectX package is measured with 3D Mark 2000 (based on DirectX7) and 3D Mark 2001 (based on DirectX 8). A comprehensive test scenario is created by a variety of benchmarks for MPEG encoding: with the help of the Lame MP3 Encoder, a 178 MB WAV file is converted to the MPEG-1 Layer 3 format. One of the established standards is our MPEG-4 test, in which data from a commercial DVD-ROM is converted into MPEG-4 via Xmpeg and the Divx codec. In addition, an MPEG-2 file is created with the video editing software Pinnacle Studio 7. For a while, we have been using the professional Lightwave package version 7b from Newtek to evaluate rendering performance. Archiving is also important for practical applications; we use WinACE-Packer for that. Compiling the newest Linux Kernel 2.4.13 has long been part of our standard repertoire. In order to test office performance, we use the Sysmark 2001 benchmark.
A new addition is the synthetic benchmark suite SiSoft Sandra 2002. In comparison to its previous version, 2001, the resulting test values in this version can be reproduced. Nevertheless, a run-through of Sandra does not prove that a system is fully stable. As we’ve confirmed in our tests, the AMD Athlon XP could attain a benchmark value of 1912 MHz with Sandra 2002, while all other test programs caused a system crash.
OpenGL Performance: Quake 3 Arena
In both of the Quake 3 Arena timedemo runs, the Pentium 4/2600 and the Pentium 4/3000 beats the AMD Athlon XP 2300+. Also noteworthy is the frame rate of the Pentium 4/2600, which accesses the Rambus memory.
Direct3D – DirectX 7: 3D Mark 2000
3D Mark 2000 shows the Direct3D performance from DirectX 7 under Windows XP. The Pentium 4/3000 makes 10843 points, just a nose ahead of the Athlon XP 2300+.
Direct3D Performance – DirectX 8: 3D Mark 2001
3D Mark 2001 reveals the Direct3D performance from DirectX 8 under Windows XP. In this benchmark, the results of the Pentium 4/2600 (RDRAM) are identical to those of the Pentium 4/3000 (DDR SDRAM). The Athlon XP 2300+ reaches the same level.
MP3 Audio Encoding: Lame MP3
With the Lame MP3 Encoder, a 178 MB sound file in WAV format is converted to MPEG-1 Layer 3 format under Windows XP. The chart above clearly shows that the extremely overclocked Pentium 4 /3000, with its 512 KB L2 Cache, is superior to the Athlon XP 2300+.
Video-Encoding MPEG-4: Flask Mpeg and Divx
SiSoft Sandra Benchmarks: CPU and Multimedia
The Pentium 4/3000 is ahead of the Athlon XP 2300+ in all categories, except for Memory Performance. Only the RDRAM-based Pentium 4/2600 can top the Athlon XP 2300+ in Memory Performance.
3D Rendering: Newtek Lightwave 7b
In the Lightwave benchmark, the Intel optimization (SSE2) is evident. The result: for rendering, the Athlon XP 2300+ takes almost double the amount of time the Pentium 4/3000 requires.
Office Performance: Sysmark 2001
The benchmark result of the Pentium 4/3000 gives us food for thought – we suspect an error in the counting algorithm. Most probably, the Sysmark 2001 is not suited for such high clock speeds. Otherwise, the Athlon XP 2300+ and the Pentium 4/2600 show noteworthy performance scores.
Compiling Linux: Suse Linux 7.3 / Kernel 2.4.13
Astonishing results are revealed in compiling the Linux kernel: AMD Athlon XP 2300+ and Intel Pentium 4/3000 are neck and neck, at about 175 seconds. The Athlon 900 seems almost lamed, requiring about double the time.
Archiving: WinACE 2.04
Archiving is a very practical application. With the help of WinACE 2.04 Packers under Windows XP, a 178 MB WAV file is packed, reporting the time to complete the task. Again, the advantages from RDRAM can be seen: the Pentium 4/2600 (Intel 850 platform) beats the Pentium 4/3000 as well as the AMD Athlon XP 2300+. Still, the Athlon XP 2300+ is able to outperform the Pentium 4/3000 by about half a minute – a sensational result!
3D-Rendering Performance: SPECviewperf “Lightscape”
AMD Athlon 2300+ tops the record, beating the Intel Pentium 4/3000 with DDR SDRAM. Only the Rambus system, together with the Pentium 4/2600, can achieve better scores.
MPEG-2 Video Encoding MPEG-2: Pinnacle Studio 7
The Intel Pentium 4/3000 is significantly faster than the Athlon XP 2300+ at creating an MPEG 2 film with Pinnacle Studio 7. Thanks to RDRAM, the Pentium 4/2600 can almost keep up with the Pentium 4/3000.
3D-Rendering Performance: Cinema 4D XL R6
In 3D-Rendering with Cinema 4D, the Pentium 4/3000 has a clear lead over the Athlon XP 2300+. Still, AMD manages to come relatively close to the performance of a Rambus system with Pentium 4/2600.
Conclusion: Pentium 4/3000 Hindered By DDR Memory – Athlon XP 2300+ Is The Stronger Performer
Our test of the world’s fastest CPUs shows that a Pentium 4/3000 based on DDR SDRAM cannot utilize its full potential. This fact is reflected in our benchmarks, where we use a Pentium 4/2600 based on RDRAM (Intel 850) as a basis for comparison. In a few of the benchmark categories, even the Pentium 4/2600 system, in combination with RDRAM, was able to outperform the Pentium 4/3000 with DDR memory. The limiting factor with the Pentium 4 at high clock speeds clearly lies in the limited bandwidth of DDR SDRAM. This influence was not seen in the RDRAM system.
In the end, it is worth waiting for DDR400 and DDR II/533, as well as RDRAM with 533 MHz. The test perpetuates the impression that Intel’s previous strategy regarding RDRAM made sense – at least with higher clock speeds of 2.5 GHz, which the manufacturer will introduce this year. At this point, the question arises: will Intel return to RDRAM and change its strategy once more? Otherwise, the chipmaker may still opt to go for DDR memory, such as DDR333, DDR400 and DDR II/533.
We broke a further record in speed with the Athlon XP 2300+, which is based on an overclocked Athlon XP 2000+. The limit was reached at 1866 MHz, at which completely stable operation was possible. However, to cool the CPU, we used an efficient watercooling system, as we had already done with the Pentium 4/3000. Otherwise, such a high speed for the Athlon processor would not be possible.
In the benchmark results, the Athlon XP 2300+ cannot quite keep pace with the Intel Pentium 4/3000, but the values that we measured are very impressive. In order to outperform the Intel Pentium 4/3000, AMD must switch over to 0.13 Micron in order to enable higher clock speeds. According to our laboratory results, with a speed of 1933 MHz, AMD can reach the performance of a Pentium 4/3000 based on DDR SDRAM. This clock speed can only be achieved by the new Athlon XP with the Thoroughbred core, which is expected soon.