<!–#set var="article_header" value="At The Last Second:
AMD’s Trump Card – The Athlon XP 2600+” –>
Overhauled: AMD Athlon XP 2600+
Suddenly, the results have come in: Intel, for all its ambition, has failed. On the other hand, AMD, despite being less well-off, has all the more reason to be pleased, because it has succeeded in passing the test with flying colors, without even trying that hard. That’s reason enough for the company to be proud of itself.
Never been seen before: an AMD Athlon XP with 2880 MHz – this corresponds to a model number of over 3700.
What in the world has happened? After the launch of the new AMD Athlon with the Thoroughbred core, the fans were rather disappointed because the new processor only provided a nominally higher clock rate (Athlon XP 2200+ with 1800 MHz), while having an identical thermal loss. There was no trace of overclockability, which was a big turn-off for tuning freaks. This quickly widened the performance gap between it and arch-enemy Intel with its Pentium 4 by a significant amount.
Intel turned up the clock rate repeatedly and introduced the P4/2533, and here, the Athlon XP 2200+ could not keep up with it in any of the disciplines. Not even when the theoretical and older benchmarks were used. AMD had a big problem, and it went all out in search of an explanation that could serve as a sales pitch at the same time.
The result of this is that some of the disappointed AMD fans wandered over to the Intel camp and got themselves a P4. Others comforted themselves with the thought that the “Hammer,” slated for launch at the end of the year, would take care of the performance deficit, and be cheaper to boot.
The top two models compared: Thoroughbred “B” (Athlon XP 2600+) vs. Thoroughbred “A” (Athlon XP 2200+). The die size has increased by 4 mm?, which is visually unnoticeable.
Now, however, it’s all different. Even AMD employees were surprised by the announcement from Dresden: various optimizations have made it possible to turn up the clock speed significantly for the Athlon XP with the Thoroughbred core. This is a first: AMD makes a 333 MHz jump in clock speed and launches the Athlon XP 2600+. Up till now, the clock rate was only increased by 66 MHz. And that’s not all – the test sample of the Athlon XP shows that it’s possible to overclock this puppy to 2800 MHz, which would correspond to an Athlon XP 3600+.
Overhauled: AMD Athlon XP 2600+, Continued
The new flagship model: AMD Athlon XP 2600+ works with 2133 MHz.
Nowadays, even as the lines at the employment office continue to grow, engineering skills are in greater demand than ever. After the Thoroughbred “A” core did not bring about the success that was hoped for, despite the shrink in die size to 0.13-micron, AMD wanted to quickly introduce the Athlon XP with the Barton core. The concept behind this was that, in the short term, a larger cache would make up for the performance gap between it and Intel.
However, at the last minute, AMD developers in Dresden (called “Fab 30”) made a breakthrough: suddenly, introducing the Thoroughbred “B” would make it possible to achieve significantly higher clock speeds, and so now an early launch of the Barton is no longer necessary. Moreover, this new development presents completely new possibilities, so the Athlon XP should continue to be available on the market for a while. It’s all thanks to the former Robotron engineers.
The Athlon XP 2600+ provides performance similar to the Intel Pentium 4/2533.
The latest BIOS version from KT333 and KT400 motherboards do not support the Athlon XP 2600+ yet – so we can bring you the exact clock frequencies.
Power With the Athlon XP 3400+ and 2666 MHz
The Athlon XP 2600+ runs stably at 2666 MHz – this corresponds to an Athlon XP 3400+.
In our last test of the previous Athlon XP 2200+ with Thoroughbred “A,” we concluded the following:
“…our attempt to overclock the processor’s FSB to 148 MHz, similar to the Athlon XP 2100+ (Palomino), failed. Here, it should be mentioned that both systems were operated with a high-quality watercooling setup. The result: the system remained stable only at 140 MHz FSB, meaning that the CPU clock was increased to 1890 MHz. This is an increase of 5%, which is clearly a weaker result than with the older Palomino, which ran with perfect stability at 1918 MHz (11.2% increase).”
Speed, thanks to 166 MHz FSB – the Athlon XP runs with 2666 MHz.
Good cooling allows the CPU to remain quite cold.
The situation has now changed entirely: the THG lab in Munich succeeded in overclocking the Athlon XP 2600+ (2133 MHz) with the Thoroughbred “B” core to the extreme. Together with the best watercooling system available on the market (see The Third THG Video: Silent and Ice Cold), the new Athlon XP, with 150 MHz FSB (instead of the factory-set 133 MHz) and an increased core voltage of 1.90 Volt (instead of the factory-set 1.625 Volt) could be overclocked to a true 2400 MHz. This clock rate corresponds to the model number (or “P” rating) of an Athlon XP 3000+.
However, we still wanted to know what the absolute limits of the processor were. To this end, we used a special cooling technique to keep the CPU die temperature to -41 degrees Celsius.
To sum up: the Athlon XP ran completely stably with a 166 MHz FSB clock and a core voltage of 2.20 Volt. The resulting clock speed is 2666 MHz, which corresponds to the model number of an Athlon XP 3400+. Further details can be found in the benchmark results. We could only attain the absolute record for a short time: the Athlon X with Thoroughbred “B” ran at an FSB clock of 175 MHz and a clock speed of 2800 MHz – which would correspond to an Athlon XP 3600+…
Data on the overclocked Athlon XP 3400+.
And Now, More Speed: Copper Interconnects at Nine Levels
Architecture of the new Thoroughbred “B” core.
A modern CPU, such as the AMD Athlon XP or the Intel Pentium 4, consists of a transistor layer (silicon) and multiple layers of interconnects. Ever since the launch of the first Athlon with the Thunderbird core with its six layers of interconnects, AMD has been increasing the number of layers. The Palomino core followed with seven layers of interconnects, which were required for the SSE logic as well as for the optimization of the cache lines.
The next step came with the Thoroughbred “A,” which added another layer. By shortening the circuit paths and distances by 0.05-micron to 0.13-micron, as well as re-organizing all of the components, additional new circuits were necessary. The current record-holder is the Thoroughbred “B” with a total of nine interconnect layers made of copper.
The increased number of circuit paths results in better conductivity for the connections within the processor, which, in turn, automatically decreases the amount of thermal loss. Also, AMD has added further condensers in order to significantly decrease the interference of transistors at high clock speeds in this combination of interconnect layers and capacities.
View of the interconnect layers in a modern processor. Pictured here is the P4 Northwood.
To sum up: the resistances between the transistors are increased (crosstalk). To differentiate between the two Thoroughbred types, AMD’s new CPU with nine layers was given the designation “B.” In the future, AMD plans to use the SOI technique (nearly pure copper interconnects) for the first time with its Barton core. CPUs produced with SOI technology are supposed to process signals faster and with less power consumption. This enables a 35% increase in clock speed with the same power consumption.
Heat Dissipation and Clock Increase
Compared to the Thoroughbred “A,” the attainable clock increase of the Thoroughbred “B” is immense because the internal heat dissipation can be decreased. The top model of the Thoroughbred “A” CPUs (the Athlon XP 2200+ with 1800 MHz) has a maximum heat dissipation of 67.9 Watt, which is just behind its predecessor, the Athlon XP 2100+ (1733 MHz) with the Palomino core, at 72 Watt.
The Thoroughbred “B” has a heat dissipation of 68.3 Watt at 2133 MHz – which, if you consider the higher clock speed, is a smaller increase. Future versions of the Athlon XP with 2400 MHz (XP 3000+) will have a heat dissipation of about 77 Watt.
Correction in Model Numbering: Now at 133 MHz Increments
With the launch of the Athlon XP with the Palomino core, AMD also introduced its new model number system, called P-rating. The idea behind this was to create a direct comparison to arch-rival Intel, which markets its CPUs using “real” megahertz figures. This is a language that the less tech-savvy customers can understand: megahertz and gigabyte. As we’ve already determined in numerous articles, a processor’s performance capabilities cannot be analyzed in terms of pure clock rate numbers.
In any case, our recent test comparisons made it clear that the model numbering for the top AMD CPUs did not correspond to the performance of comparable P4s. Because of this, AMD adjusted its virtual values to become lower: previously, a clock increase of 66 MHz meant the equivalent of 100 additional points; now, the 100 points are related to the absolute values. This is AMD’s admission that the previous performance scale was set too high, especially when it came to the higher clock speeds.
The Chipset Bottleneck: Memory Interface Puts the Brakes On Athlon
A thorough analysis of the memory-dependent benchmarks clearly reveals that, when used with the top AMD and Intel models (Athlon XP 2600+ and Pentium 4/2533) under the same conditions (DDR333 with CL2.0), the P4 can show a significant advantage in applications.
Here, the Intel 845G and the VIA KT333 platforms act as the basis. In extreme cases, the difference in memory access speed can be up to 30%. However, in this regard, AMD is not the guilty one, since the chipsets come from other manufacturers. VIA continues to have problems designing a high-performance memory interface – a fact that is supported by our benchmarks.
The Same Ol’ Topic: Protection Against an AMD Thermal Death
Approximately nine months ago, THG stirred up quite a commotion with its article Горячо! Как современные процессоры защищены от перегрева?. Accompanying the article was our very first downloadable video, which shows the dramatic behavior of an AMD Athlon with a Palomino core, equipped with a thermal diode, when the CPU cooler fails during operation.
The alarms sounded at AMD, and the THG crew met several times with engineers from the manufacturers in question (AMD and Siemens). A few weeks later, AMD introduced a new circuitry logic to a small group of people. This was supposed to turn off the power supply as soon as the die temperature surpassed 85 degrees Celsius.
In order to guarantee faultless protection, the CPU temperature is measured by the thermal diode in very short time intervals, which ensures that the power can be switched off in time. Currently, all motherboard makers have a copy of the new “Thermal Guide” from AMD. Still, only a few manufacturers integrate thermal protection into the older motherboard designs, because this involves huge additional costs.
There are some examples of this to be found in the THG lab. The QDI Kudoz 7X with the VIA KT400 chipset (reviewed here), which arrived recently, implements overheating protection. In this technique, the data for temperature control (thermal diode) is read by the BIOS. The problem with this is that booting a PC without a CPU cooler takes too much time to ensure effective protection, with the result that the Athlon XP/MP CPU dies a thermal death.
Asus currently offers an interesting solution: here, the CPU temperature is monitored entirely with external circuitry on the motherboard. This results in more immediate reaction to temperature changes, nearly free of delay. Booting up without a CPU cooler and without having the CPU burn up is not a problem.
Comparison: T-Bred “B” vs. T-Bred “A” vs. Palomino
| Processor (Thoroughbred “A”) | Clock frequency | Model Number |
| AMD Athlon XP 1500+ | 1333 MHz | 1500 |
| AMD Athlon XP 1600+ | 1400 MHz | 1600 |
| AMD Athlon XP 1700+ | 1466 MHz | 1700 |
| AMD Athlon XP 1800+ | 1533 MHz | 1800 |
| AMD Athlon XP 1900+ | 1600 MHz | 1900 |
| AMD Athlon XP 2000+ | 1666 MHz | 2000 |
| AMD Athlon XP 2100+ | 1733 MHz | 2100 |
| AMD Athlon XP 2200+ | 1800 MHz | 2200 |
| Processor (Thoroughbred “B”) | Clock frequency | Model Number |
| AMD Athlon XP 2400+ | 2000 MHz | 2400 |
| AMD Athlon XP 2600+ | 2133 MHz | 2600 |
| AMD Athlon XP 2800+ | 2266 MHz | 2800 |
| AMD Athlon XP 3000+ | 2400 MHz | 3000 |
| AMD Athlon XP 3200+ | 2533 MHz | 3200 |
| AMD Athlon XP 3400+ | 2666 MHz | 3400 |
| AMD Athlon XP 3600+ | 2800 MHz | 3600 |
| Processor core | Number of layers | CPU classes |
| AMD Thunderbird | 6 | Athlon 650 MHz to 1400 MHz |
| AMD Palomino | 7 | Athlon XP 1500+ to XP 2100+ |
| AMD Thoroughbred “A” | 8 | Athlon XP 1700+ to XP 2200+ |
| AMD Thoroughbred “B” | 9 | Athlon XP 2400+ to XP 3400+ |
Comparison: T-Bred “B” vs. T-Bred “A” vs. Palomino, Continued
With the launch of the Thoroughbred core in the “B” version, AMD Athlon reaches the fourth level in its evolution. The first core, the Thunderbird, was introduced about two years ago. AMD made a further step in October 2001 by introducing the Palomino core, which essentially consisted of the SSE command set expansion and an optimization of the cache line. The Thoroughbred core has been around since June 2002, and this one involves a rearrangement of the CPU’s internal components, as well as smaller structures (0.13-micron). In order to give you a complete picture, we’ve included the data for the Barton core, which is expected to be out in Q3. The Barton represents the fifth and last level of development of the Athlon program, based on Socket 462.
| Manufacturer | AMD | AMD | AMD | AMD | Intel |
| Processor | Athlon XP Thoroughbred “B” Core |
Athlon XP Thoroughbred Core “A” Core |
Athlon XP Palomino Core |
Athlon Thunderbird Core |
Pentium 4 Northwood Core |
| Introduction | Q3, 2002 | Q2, 2002 | October 9, 2001 | October 9, 2001 | January 7, 2002 |
| Clock Frequencies | 1.86 – 2.66 GHz | 1.46 – 1.80 GHz | 1.2 – 1.80 GHz | 0.65 – 1.40 GHz | 1.6 – 3.06 GHz |
| Manufacturing Process | 0,13 µm | 0,13 µm | 0,18 µm | 0,18 µm | 0,13 µm |
| Die size | 84 mm2 | 80 mm2 | 128 mm2 | 128 mm2 | 146 mm2 |
| Number of Gates | 37,5 Million | 37,5 Million | 37,5 Million | 37,5 Million | 55 Million |
| Platform | Socket462 | Socket462 | Socket462 | Socket462 | Socket478 |
| CPU Bus Clock (Front Side Bus) | 133 MHz 266 MHz DDR |
133 MHz 266 MHz DDR |
133 MHz 266 MHz DDR |
100/133 MHz 200/266 MHz DDR |
100/133 MHz 400/533 MHz QDR |
| L1 Execution Cache Size | 64 KB | 64 KB | 64 KB | 64 KB | 12.000 µ-Ops (Trace Cache) |
| Execution Pre Decode? | no | no | no | no | yes |
| L1 Data Cache Size | 64 KB | 64 KB | 64 KB | 64 KB | 8 KB (unconfirmed) |
| Hardware Data Prefetch | yes | yes | yes | yes | yes |
| L1 Cache Clock | core clock | core clock | core clock | core clock | core clock |
| L1 Data Cache Bus Width | 64-bit | 64-bit | 64-bit | 64-bit | 256-bit |
| L2 Cache Size | 256 KB | 256 KB | 256 KB | 256 KB | 512 KB |
| L2 Cache Clock | core clock | core clock | core clock | core clock | core clock |
| L2 Cache Addressable Range | 64 GB | 64 GB | 64 GB | 64 GB | 64 GB |
| Processor Data Bus Width | 64-bit | 64-bit | 64-bit | 64-bit | 64-bit |
| Platform Support | |||||
| Chipsets | VIA KT333 to KT400 SiS 735 and SiS 745 Nvidia nForce, nForce 2 AMD 750 and 760 |
VIA KT133A to KT400 SiS 735 and SiS 745 ALi Magik 1 Nvidia nForce, nForce 2 AMD 750 and 760 |
VIA KT133 to KT400 SiS 735 and SiS 745 ALi Magik 1 Nvidia nForce, nForce 2 AMD 750 and 760 |
VIA KT133 to KT400 SiS 735 and SiS 745 ALi Magik 1 Nvidia nForce, nForce 2 AMD 750 and 760 |
Intel i845, Intel i845D, Intel i845E, i845G, i845EG, i845EP Intel i850 and Intel i850E VIA P4X266, VIA P4X266A VIA P4X333, VIA P4X400 SiS 645, SiS 645DX, SiS 648 |
| Type of Memory | SDRAM DDR-SDRAM |
SDRAM DDR-SDRAM |
SDRAM DDR-SDRAM |
SDRAM DDR-SDRAM |
SDRAM DDR-SDRAM, RDRAM |
| Memory Clock | 100/133/166/200 MHz | 100/133/166/200 MHz | 100/133/166 MHz | 100/133/166 MHz | 100/133/166/300/400/533 MHz |
| Instruction Extensions | |||||
| MMX | yes | yes | yes | yes | yes |
| Enhanced 3DNow! | yes | yes | yes | yes | no |
| 3DNow! Professional | yes | yes | yes | yes | no |
| SSE | yes | yes | yes | yes | yes |
| SSE2 | no | no | no | no | yes |
| Electrical Specifications | |||||
| SMP-Support | no (‘not officially endorsed or supported’) | no (‘not officially endorsed or supported’) | no (‘not officially endorsed or supported’) | no (‘not officially endorsed or supported’) | no |
| Core Voltage | 1,65 Volt | 1,65 Volt | 1,75 Volt | 1,75 Volt | 1,5 Volt |
| Thermal Protection (Thermal Diode) | yes | yes | yes | yes | yes |
| Integrated Thermal Protection Logic | no, requires logic on motherboard | no, requires logic on motherboard | no, requires logic on motherboard | no, requires logic on motherboard | yes |
Heat Dissipation: All AMD CPUs Compared
Yield For The Thoroughbred “B”: 306 CPUs Per Wafer
| CPU core | Thoroughbred “B” | Thoroughbred “A” | Palomino |
| Wafer size | 31416 mm² | 31416 mm² | 31416 mm² |
| Die-size | 84 mm² | 80 mm² | 128 mm² |
| Process | 0,13 µm | 0,13 µm | 0,18 µm |
| Cut loss | 18 Prozent | 18 Prozent | 18 Prozent |
| Yield | 306 parts/Wafer | 322 parts/Wafer | 201 parts/Wafer |
AMD works exclusively with wafers with a diameter of 200 mm, which constitutes a total surface area of 31415 mm? of silicon. The quotient from wafer surface and CPU die size gives you the theoretical yield, without waste. In the production process, an average of 18% clippings results from the 200 mm wafers.
With the production of the Thoroughbred core in Dresden, AMD calculates a yield of approximately 306 CPUs – based on 0.13-micron technology and a die surface of 84 mm?. Compared to its predecessor, the Thoroughbred “A,” the theoretical yield decreases to 16 pieces per wafer. Note that the count of 306 processors assumes an error of 0%.
What Does 166 MHz FSB Bring Us?
Although there are many rumors going around that the FSB clock will be increased, AMD has kept to 133 MHz FSB with the Thoroughbred “B” as well. Among the chipsets for the Socket 462 platform, both the VIA KT333 and the KT400 support 166 MHz FSB.
The latest motherboards with the VIA KT333 and KT400 chipsets already offer an FSB clock of 166 MHz and 200 MHz.
The First Final Version Board With VIA KT400: Gigabyte GA-7VAXP
A view of the Gigabyte GA-7VAXP with the VIA KT400 chipset.
Accessories included with the Gigabyte GA-7VAXP.
All aboard: FireWire, Network, Sound and USB 2.0, including an FSB clock ranging from 133 MHz to 200 MHz.
Retail package of the Gigabyte board with the KT400 chipset.
New Guidelines for CPU Coolers: No Cheap Models
A new CPU cooler for Athlon XP 2200+ with high contact pressure, approved by AMD.
Effective cooling for the processor is an important topic when you make the switch from Athlon XP with the Palomino core to the Athlon XP with the Thoroughbred core (“A” and “B” versions). Although the new Athlon XP has a marginally lower power intake, a higher performance CPU cooler is required nevertheless.
This can be explained by the following: because the CPU surface is 40% smaller, there is automatically less surface area available for heat dissipation. This begs the question of why AMD still doesn’t use a heat spreader, as do the Intel Pentium 4 and the AMD “Hammer.”
For OEM manufacturers and those who buy new Athlon XP processors, the updated guidelines from AMD (valid since June 10, 2002) are:
Only CPU coolers equipped with a copper plate (or adequate heat transfer capabilities) as the contact surface to the CPU die may be used. Or, to put it another way: if you want to upgrade your PC with an Athlon XP based on the T-bred, you can safely throw out your old aluminum cooler. Otherwise, if your Athlon XP (the new model) dies a thermal death when used together with a cheap cooler, AMD is not giving you any guarantees.
A countersunk copper contact on a CPU cooler.
A CPU cooler from Taisol.
Price Comparison: AMD Athlon XP vs. Intel Pentium 4
| Processor | Price per 1000 |
| Intel Pentium 4/2533A | US$ 243 |
| Intel Pentium 4/2400A | US$ 201 |
| AMD Athlon XP 2600+ | US$ 297 |
| AMD Athlon XP 2400+ | US$ 193 |
| AMD Athlon XP 2200+ | US$ 183 |
| AMD Athlon XP 2100+ | US$ 174 |
Retail packaging of an AMD Athlon XP 2200+ with an old Thoroughbred “A” core.
Who’s buying what? On the left, a passer-by eyes an AMD, while the second one seems to ogle a P4.
The biggest computer mile in Germany – special offers at Schillerstraße in Munich.
Test Setup and Details
| AMD Hardware (Socket 462) | |
| Processors 133 MHz FSB 166 MHz Memory Clock |
Athlon XP 2600+ (2133 MHz) Athlon XP 2200+ (1800 MHz) Athlon XP 2100+ (1733 MHz) Athlon XP 2000+ (1666 MHz) Athlon XP 1900+ (1600 MHz) Athlon XP 1800+ (1533 MHz) Athlon XP 1700+ (1466 MHz) Athlon XP 1600+ (1400 MHz) Athlon XP 1500+ (1333 MHz) Athlon 1400 (1400 MHz) |
| Processor 100 MHz FSB 133 MHz Memory Clock |
Athlon 850 (850/100/133 MHz) |
| Motherboard and Memory | Epox EP-8K3A+ (VIA Apollo KT333 Chipset) Revision: 1.0 Bios: 8k3a2328 (19.06.2002) 512 MB DDR-SDRAM, CL2.0, 166 MHz, PC2700, Winbond |
| Drivers | VIA KT333 Driver4 in 1 Version: 4.40(a)P3 INF File Version: 1.60a AGP Driver Version: 4.20a IDE Filter driver Version: 1.20 |
| Intel Hardware (Socket 478) | |
| Processors 133 MHz FSB 400/533 MHz Memory Clock |
Pentium 4 2.53 GHz (2533 MHz) Pentium 4 2.4 GHz (2400 MHz) Pentium 4 2.26 GHz (2266 MHz) |
| Processors 100 MHz FSB 400 MHz Memory Clock |
Pentium 4 2.4 GHz (2400 MHz) Pentium 4 2.2 GHz (2200 MHz) Pentium 4A 2.0 GHz (2000 MHz) Pentium 4A 1.6 GHz (1600 MHz) Pentium 4 2.0 GHz (2000 MHz) Pentium 4 1.7 GHz (1700 MHz) Pentium 4 1.4 GHz (1400 MHz) |
| Motherboard and Memory | Asus P4T533-C (Intel 850E Chipset) Revision: 1.01 Bios: 1006 BETA 002 (07.08.2002) 2x 256 MB RDRAM, PC800, 533 MHz, 40ns, Infinion 2x 256 MB RDRAM, PC1066, 533 MHz, 32ns, Kingstone |
| Drivers | Intel 850E Driver V 4.00.1013 (7.06.2002) Intel IAA Driver V 2.2.0.2126 |
| Common Hardware | |
| Graphics Card | GeForce 4 Ti 4600 (MSI MS-8872) Version: 2.00 Memory: 128 MB DDR-SDRAM Memory Clock: 650 MHz Chip Clock: 300 MHz |
| Hard Drive | 40 GB, 6L040J2, Maxtor UDMA100, 7200 U/min, 2 MB Cache |
| Network | D-Link DFE-530TX (10/100 Mbit) |
| CDROM | Asus 52x |
| Drivers and Software | |
| Graphics Driver | Detonator 4 Serie Version: 29.42 |
| DirectX | Version: 8.1 |
| OS | Windows XP, Build 2600 (English) |
| Benchmarks and Settings | |
| Bapco Sysmark 2002 | Version 1.0 |
| Quake III Arena | Patch V1.16 640×480 – 16 bit / 1024 x 768 – 32 bit Timedemo1 / demo demo001 / nv15demo command line = +set cd_nocd 1 +set s_initsound 0 Graphics detail = Normal |
| 3DMark 2000 Pro | Version 1.1 – Build 340 1024 x 786 – 16 bit Default Benchmark |
| 3DMark 2001 SE | Version 1.1 – Build 340 – Patch Build 330 1024 x 786 – 32 bit Default Benchmark |
| PCMark 2002 Pro Pack | Build 100 CPU and Memory Tests |
| SiSoftware Sandra Standard 2002 SP1 | Version 2002.6.8.97 CPU Multi-Media- / CPU Arithmetic- / Memory Bandwidth- Benchmark |
| Newtek Lightwave | Version 7.5 – Build 572 Render First Frame = 1 Render Last Frame = 60 Render Frame Step = 1 Rendering Bench “SKULL_HEAD_NEWEST.LWS” Show Rendering in Progress = 320×240 Ray Trace Shadows, Reflection, Refraction, Transparency = on Multithreading = 8 Threads |
| DivX Pro | Version 5.0.02 Xmpeg – Version 4.5 150 MB VOB Datei, 3780 Frames, 720 x 480 pixel, 25 fps Audio = off, Crop = on, Video Format = YV12 Performance/quality = Slowest, Encoding bitrate = 780 kbps |
| Pinnacle Studio 7 | Version 7.13.5 Rendering – DVD Compatible no Audio |
| Lame | Version 3.92 32 bit Dos Promt, 178 MB Wave Datei, 44100 Hz 32 – 320 Kbit sampling |
| e-merge Winace | Version 2.2 178 MB Wave-Datei, Kompression = Best, Dictonary = 4096 KB |
| Maxon Computer Cinema 4D XL 7 | Version 7.303 Rendering in 1024 x 786, “Radiosity-Stairs.c4d” |
| magix mp3 maker platinum | Version 3.04 D 178 MB Wave Datei, 44100 Hz, VBR = on and Quality |
| SPEC Viewperf | Version 7.0 1280 x 1024 / 32 bit / 85 Hz, Vsync = off |
| Comanche 4 | 1024 x 768 / 32 bit / Audio = off |
| discreet | Characters “Dragon_Charater_rig” Rendering Single, 1024×768 |
Benchmarks under Windows XP
| OpenGL Performance | Quake 3 Arena “Demo 1” and “NV15 Demo” |
| 3D Rendering | SPECviewperf 7 (new) Lightwave 7.5 Build 572 Cinema 4D XL 7.303 3D Studio Max 4.2 SP1 |
| DirectX7 Games | 3D Mark 2000 Pro (ver. 1.1) |
| DirectX8 Games | 3D Mark 2001 SE (ver. 1.1) Comanche 4 (new) |
| MP3 Audio Encoding | Lame MP3 Encoder 3.92 mp3 Maker Platinum 3.04 |
| MPEG-2 Video Encoding | Pinnacle Studio 7.3.15 |
| MPEG-4 Video Encoding | XFlask 4.5 and Divx 5.02 Pro |
| Office Performance | Sysmark 2002 |
| Archiving | WinACE 2.2 |
| CPU und Multimedia Bench | PC Mark 2002 SiSoft Sandra 2002 Pro SP1 |
We used a total of 28 different benchmark tests in order to give you a complete and balanced picture of what the new Athlon XP 2600+ can do. The benchmark results, from a total of 31 different CPUs, provide a general overview. As a comparison, all AMD Athlon XP processors, the classic Athlon with the Thunderbird core, and the slowest Athlon 850 are also included.
The OpenGL performances are evaluated using 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 relative newcomer is the first game with DirectX 8 support: Comanche 4. 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 4.5 and the codec DivX 5.02 Pro. In addition, an MPEG-2 file is created with the video editing software Pinnacle Studio 7. Part of our standard repertoire is the professional Lightwave package version 7b from Newtek, used to evaluate rendering performance. Also important for practical applications is data-packing, for which we use WinACE-Packer 2.11. In order to test office performance, the Sysmark 2001 benchmark is used. A comprehensive 3D benchmark suite is provided by the new SPEC Viewperf 7. And last but not least, SiSoft Sandra 2002 Pro should be mentioned. The new versions of 3D Studio Max 5.0 and Pinnacle Studio 8.1 arrived at our lab too late, so we were unable to include them in the test.
OpenGL-Performance: Quake 3 Arena
The four time-demo runs from Quake 3 Arena consistently show that the Athlon X 2600+ is unable to keep up with the P4/2533. The reason for this is that the Athlon XP lacks memory throughput, causing the AMD platform to land in the lower ranks. Note: all overclocked processors are indicated by the grey bars and serve as reference only.
DirectX 7 Games: 3D Mark 2000
3D Mark 2000 shows the Direct3D performance of DirectX7 under Windows XP. According to old tradition, the AMD Athlon XP 2600+ soars above the Intel Pentium 4/2533 with Rambus memory (PC1200) by almost 700 points. The absolute leader of the pack is the overclocked Athlon XP 3400+.
DirectX 8 Games: 3D Mark 2001 SE
3D Mark 2001 shows the Direct 3D performance from DirectX 8 under Windows XP. This test shows that the Athlon XP 2600+ lies just 180 points behind the Pentium 4/2400.
DirectX8-Hardcore-Spiel: Comanche 4
Comanche 4 is one of the first games on the market to support DirectX8. With its high clock speed, the Athlon XP 2600+ can basically gain ground and reach the performance of a Pentium 4/2533 with PC800. When older processors are used, such as the Athlon 850, the game can hardly be played.
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 shows that the AMD Athlon XP 2600+, with 96 seconds, clearly surpasses the P4 with 2533 MHz with about 105 seconds. And so the tables have turned. Overclocking to 2666 MHz (XP 3400+) provides even more speed – here, the CPU only requires 78 seconds.
MP3 Maker Platinium
Video-Encoding MPEG-4: Xmpeg 4.5 und Divx 5.02 Pro
Memory performance is an important factor for MPEG-4 encoding. In this category, the AMD Athlon XP 2600+, combined with DDR333 (CL 2 – max. timing), is not quite able to keep up with the Pentium 4/2400. However, its performance, compared the previous Thoroughbred, has increased by five frames. The Athlon 850 simply shows its age with about 16 fps, as does the Athlon 1400 with 26 fps, because it lacks the SSE codes.
Video-Encoding MPEG-2: Pinnacle Studio 7
In encoding an MPEG-2 film with Pinnacle Studio 7, the AMD Athlon XP 2600+ does the job in 70.3 seconds, which is almost as fast as the P4/2400. When extremely overclocked, the Intel Pentium 4/2800, despite its Rambus speed (PC1200), must admit defeat to the Athlon XP 3400+. The AMD CPU was able to render the video in 58.8 seconds. The slowest candidate is the AMD Athlon 850, which we’ve included as a comparison – it is slower than the Athlon XP 3400+ by more than three minutes, which is enough time for you to have a cup of coffee!
SiSoft Sandra 2002 Benchmarks: CPU und Multimedia
The SiSoft Sandra Pro Benchmark 2002 shows that the Athlon XP 2600+ comes pretty close to the P4/2533 in various tests. It’s clear to see that the benchmark does not represent realistic performance values. Many users get their hopes up too high when comparing their own results.
Multimedia-Performance: PC Mark 2002
In both of the benchmasrk tests, the AMD Athlon XP 2600+ is slightly ahead of its arch-rival, the Intel Pentium 4/2533. The reason for this positioning is that this benchmark is based on old programming code – without optimizations for AMD and Intel.
Office-/Internet-Performance: Sysmark 2002
In all three parts, the Athlon XP 2600+ lands only in mid-field, achieving the performance level of a Pentium 4/2000 with the Northwood core. Even a high clock speed for the Athlon XP does not help. It’s the old dilemma again: there’s a lack of throughput for the memory interface of the VIA KT333 chipset. A further note about all AMD Athlon XP CPUs: optimizations are lacking, so the AMD processors fall a bit behind when compared to the Intel models.
Archiving: WinACE 2.2
Archiving data is a very practical application. With the help of the new WinACE 2.11 packer under Windows XP, a 178 MB WAV file was packed while the clock was running. The Athlon 2600+ ends up just slightly behind the Pentium 4/2400. This benchmark is heavily influenced by the memory interface – therefore the test scores for the overclocked Athlon CPUs (XP 3000+ and XP 3400+) are quite low. The reason is that, at high memory clock rates, the motherboard switches over to the slower CL3 mode. Here, it is not possible to manually set it to CL2.0.
3D-Rendering: Newtek Lightwave 7.5
In the Lightwave benchmark, the optimizations for the Pentium 4 processors become particularly noticeable – the Athlon XP 2600+ lands mid-field. Even high clock speeds of up to 2666 MHz don’t help.
3D-Rendering: Cinema 4D XL 7.303
The AMD Athlon gets agressive in the Cinema benchmark: here, the AMD Athlon XP 2600+ takes first place and puts the P4/2533 in its place. In any case, there’s a difference of 28 seconds!
3D-Rendering: 3D Studio Max 4.26
In this benchmark, ten frames from the “Rabbit” scene are calculated at 800 x 600. Ever since the software manufacturer came out with Service Pack 1, which they optimized in cooperation with Intel, it doesn’t look very good for the Athlon XP 2600+. The P4/2533 has a nine-second advantage. It’s only thanks to the extremely high clock speed of 2666 MHz that the Athlon XP 3400+ is able to achieve the same results as the Pentium 4/2800.
3D-Rendering Performance: SPEC Viewperf 7
The SPEC benchmark reveals varying results: due to the load caused by the code, there were breakdowns during the DRV8, DX07 and PROE01 tests. Pentium 4/2533 and Athlon XP 2600+ alternate the lead position.
Conclusion: An Upset In The Power League
With the launch of its Athlon XP 2600+, AMD has provided a big surprise: at the last minute, and without warning, the new CPU with the Thoroughbred “B” core landed at the THG lab in Munich. Compared to its predecessor, the Thoroughbred “A,” this one shows above all that a significantly higher clock rate (this top model now runs at 2133 MHz) automatically means greater speed. In the benchmark tests, the Athlon XP 2600+ manages to surpass the Intel Pentium 4/2533 once more, but not in all disciplines.
This brings AMD back in the running. These excellent results were achieved by modifying the CPU core, where the manufacturer has added an additional layer of interconnects. There have been no real changes to the CPU design at its transistor layer. The new Thoroughbred core is distinguished from the previous one by the “B” at the end of its name.
There’s also some good news for friends of overclocking: the Athlon XP 2400+, which is also well-suited for overclocking, is to be had for a more reasonable price. In the THG lab, we took the 2133 MHz model, combined it with a watercooling system, and were able to overclock it to 2400 MHz, which corresponds to an Athlon XP 3000+. An extreme increase in clock speed is only possible with ice-cooling, for a CPU die temperature of -41 degrees Celsius, at which even 2800 MHz (corresponds to Athlon XP 3600+) can be attained.
Intel P4’s seemingly permanent performance lead has been broken. AMD’s CPU has succeeded in becoming a powerful rival to Intel’s top P4 model. At the same time, the power consumption has been lowered, allowing more room for higher clock speeds. Above all, Intel should brace itself for the near future, because successor of the Thoroughbred “B,” the Barton core, will work with a double L2-Cache. To top it off, the identical interconnect technique will be used as it is with the latest T-bred. The only way out is to go for the higher megahertz regions, and those beyond the 3 GHz limit are soon to be exploited. Here, at the very least, the two manufacturers are sure to pull out their secret weapons.
And, to look at things another way, the system performance for each platform is strongly affected by the combination of chipset and processor. Despite the use of the fastest DDR333 memory (CL2.0), the memory performance lags quite far behind the scores made by the Intel 845G chipset for the P4 platform. It is most likely because of this that AMD has adjusted the model number to be lower.
[Update on Aug 23, 2002 for clarification: Rambus platforms with PC1066 memory offer more memory bandwidth compared to DDR333 platforms. Therefore, memory-intensive benchmarks show that Rambus has a performance advantage over DDR333. The overall performance is also affected by the bandwidth of the Front Side Bus. The major disadvantage the Athlon has compared to the Pentium 4 is its lower FSB bandwidth.]
