Clash of the Titans: Dothan vs Turion
August 30, 2005 at 06:10:00 PM, by
The Intel Pentium M 760 2.0GHz and AMD Turion 64 ML-37 2.0GHz compared!
Introduction
UPDATE: THE BATTERY PERFORMANCE TESTS HAVE BEEN ENTIRELY UPDATED. THE ORIGINAL UN-TOUCHED RESULTS ARE INCLUDED, AS WELL AS THE UPDATED AND NEW (MOBILEMARK 2005) RESULTS - [9/27/2005]
For many years, Intel's Pentium M processor and the the Centrino platform, has reigned as king. Offering high performance at a fraction of the power consumption of its competitors, the Pentium M was the obvious, no-brainer choice for most mobile users.
However, things are becoming more complicated. With AMD's recent release of its Turion processor, Intel's dominance of the mobile market faces its first real challenge since it introduced the Centrino platform. Based on AMD's powerful Athlon 64 line, the Turion 64 mobile processor claims to offer comparable battery life and even better performance than Intel's Pentium M champion.
Unfortunately, since so little is known about mobile CPUs, as opposed to the maelstrom of information available for their desktop counterparts, many misconceptions have been lodged in people's minds.
After the release of Asus' adapter to allow the Pentium M to be used on a desktop motherboard, Pentium M ( Dothan) CPUs became a prime target for overclockers. Clock for clock, the Dothan CPU trounced both the Pentium 4 and the Athlon 64 at most activities. Therefore, enthusiasts believe that Dothan should easily outperform Turion in laptops as well.
On the other side of the camp, many believe that Turion based notebooks will surpass the Pentium M in gaming performance since the Athlon 64 is generally the fastest in gaming performance.
This article fully covers the architecture of both the Pentium M and Turion CPU, along with a brief history outlining the various tweaks and optimizations that were performed during their development. To appeal to a broader audience, this article will first discuss the effect of the respective designs rather than delving into any engineering details. Then, the pros and cons of each CPU architecture are discussed, followed by a brief evaluation of price. Finally, the bulk of the article compares the Turion ML-37 against the Pentium M 760 in a myriad of benchmarks, starting from simple synthetic benchmarks to the four main categories of computing: office/business, multimedia, gaming, and scientific use. Battery life is also compared, and the conclusion will provide a brief summary.
Intel DothanYears ago, due to the relatively tiny laptop market compared to the desktop market , CPU companies were not willing to waste millions of dollars on R&D in order to develop a low heat, low power consumption processor. Therefore, laptop CPUs have traditionally been merely lower TDP (Thermal Design Power) versions of desktop CPUs. However, as desktop CPUs rose in power consumption and the once-miniscule laptop market swelled, it became apparent that a true mobile CPU was needed. As a prime example, the current fastest Pentium 4 Mobile CPU runs at 3.46GHz with a sky-high TDP of 88W. This CPU would chew up any battery thrown at it and could easily start a fire in your lap. It soon became obvious to Intel that a real laptop-optimized CPU was required, clearly prohibiting such a high performance chip from being used in a thin and light notebook.
Intel's Pentium M line was designed from the ground up, optimized specifically for the highest possible performance per watt. Aside from creating a revolutionary processor, Intel unveiled a marketing stroke of genius: the Centrino marketing campaign. Instead of advertising a CPU, Intel heavily advertised an entire computer platform; consisting of its Pentium M processor, i855 mobile chipset, and 2100B wireless mini-PCI card all under the name "Centrino." Only laptops utilizing all three of these components, all manufactured by Intel, could be considered a "Centrino certified laptop." The benefit was that the manufacturer could cash in on Intel's massive marketing campaign, while Intel would benefit from the simple fact that they were guaranteed to sell three products (CPU, chipset, and wireless) instead of merely one.
Centrino was a complete success. Intel dominated the mobile market, creating the CPU, chipset, and wireless solution. Intel's massive marketing campaign was so successful that many consumers thought Centrino was actually the name of the CPU, as opposed to the name of the platform.
The original Pentium M design, called Banias, is predominantly a tweaked Pentium III with Pentium 4 traits and optimized for low wattage. To increase clock speeds, the pipeline was lengthened from 10 to 12 stages. Built on the 130nm process and offering a 24.5W TDP, the chip featured a 32KB 4-way associative L1 cache and a 1MB L2 cache, built with low-voltage transistors to minimize static power consumption. Banias also inherited the Pentium 4's quad-pumped single channel 100MHz Front Side Bus (FSB) as well as its SSE2 implementation, albeit with a few more power optimizations. Aside from borrowing technology from the Pentium 4, the developers of the Banias also pioneered a few new technologies.
To increase processor efficiency (IPC), Banias utilizes a highly advanced branch prediction unit. Based off the Pentium 4's branch predictor, Banias' branch prediction unit is not only optimized further, but also adds two more specialized predictors: the Loop Detector, which improves general loop branching predictions, and the Indirect Branch Predictor, which improves indirect branch performance used to select data-dependent targets. A branch predictor determines whether a conditional branch in the programming flow is likely to be taken or not. A conditional branch is where the code might jump to an entirely different location in memory, causing the processor to have to read and load data from a different location. Branch predictors allow processors to fetch and execute instructions without waiting for the branch to be resolved. The Loop Dectector and the Indirect Branch Predictor account for about 30% of the net increase in branch prediction accuracy. The net result is a 20% lower branch misprediction rate, which translates to as much as a 7% increase in overall performance.
One of the key additions to Banias is the innovative micro-ops fusion engine. In modern mainstream processors, x86 instructions are broken down into small pieces, called micro-ops, before sent down the pipeline to be processed. To lower the amount of micro-ops sent down the pipeline and increase overall efficiency, the micro-ops fusion engine "fuses" dependent micro-ops instructions together. Studies found that the mechanism reduced the number of micro-ops handled by the out-of-order logic by more than 10%. The result is a performance increase of 5% for integer operations and an amazing 9% for floating point operations.
Another innovation in Banias is the appearance of the dedicated stack engine, which improves stack (instruction queue) management. The net result is an approximate 5% reduction of micro-ops, increasing its efficiency even more.
With an unprecedented power-centric design philosophy, Banias was truly revolutionary. However, Intel was not yet finished. Dothan, the follow-up to Banias released in 2004, continued to innovate. MS Office and other memory intensive applications were sped up dramatically due to a doubling of L2 cache to 2MB. The dramatic increase in transistor count that followed was alleviated by the shrink to a 90nm strained silicon process, which also helped improve clock speeds while reducing power consumption from 24.5W to 21W TDP. Furthering the quest for minimal power consumption, Intel's Advanced Tight Loop Execution takes every opportunity possible to shut down parts of the processor that are not being used. The branch prediction unit was optimized even further, especially for loops in branches. The micro-ops fusion engine was also upgraded, resulting in a net boost of 8% to integer operations and 10% to floating point operations (up from 5% and 9%). A "bug" was also fixed: the Enhanced Register Access Manager removes the stall condition that would otherwise impact the register renaming mechanism, regaining anywhere from 0 to 20% performance in legacy applications.
With the release of the Sonoma platform, Dothan got yet another upgrade. The memory bandwidth bottleneck was alleviated with the support of a new quad-pumped 133MHz FSB, as well as the adoption of DDR2 PC4200 (533MHz) SO-DIMM modules. Intel also followed AMD's footsteps by including support for the XD (Execute Disable) bit, identical to AMD's NX bit, which prevents buffer overflows thus providing a form of virus protection via hardware. Currently, 533MHz Dothan CPUs run at up to 2.26GHz. Unfortunately, these features came at a price: Dothan suffered almost a 30% increase in TDP, going from 21W to 27W TDP. However, due to its high-tech design Dothan still offers incredible battery life while increasing performance.
AMD TurionLike Intel, AMD has traditionally converted its desktop CPUs into laptop CPUs by merely lowering voltage and clock speeds. Instead of Intel's ground-up approach when designing the Dothan CPU, the design of the Turion CPU was much easier. Since the AMD64 architecture already ran cool with relatively low power consumption, AMD already had a Dothan competitor on its hands. Therefore, in order to analyze the Turion's CPU architecture, one must analyze its desktop equivalent.
The AMD64 originated from the incredibly successful Opteron server CPU line. Released in 2003, the original Opteron Sledgehammer design was essentially a 130nm process K7 Barton core using the new Socket 940, optimized for especially high multi-core performance. To increase clock speeds, the pipeline length was increased from 10 to 12 stages. Although it kept the 2-way associative 64KB L1 cache from the Barton core, the Sledgehammer offers a total of 1MB L2 cache, up from Barton's 512KB. The number and arrangement of the execution units in its execution core remained the same, but Opteron is able to use the existing units much more efficiently through a number of optimizations.
When a processor receives x86 instructions, in order to process them, the instructions need to be broken down into 1 to 3 pieces, called micro-ops. In the original Barton core design, complex (2-3 micro-op) x86 instructions are converted into micro-ops through a microcode decoder, while the simple (1 micro-op) instructions go through the hardware, or Fastpath decoder. The difference between the two decoders is that the hardware decoder can convert up to 3 instructions at a time, while the micro-code decoder can only process 1 at a time. Obviously, if the code consists of predominantly complex instructions, then the CPU would be facing a serious bottleneck. The Sledgehammer CPU design alleviates the problem by modifying the Fastpath decoder to accept a wider array of (mostly floating-point and SSE) instructions, thus increasing average decode bandwidth considerably.
Branch prediction had also been drastically improved. While the algorithm remains predominantly the same, both the branch selector table and the global history table were drastically increased in size. The result was a 5-10% improvement in accuracy. The Sledgehammer also added the branch target address calculator , which lowers the misprediction penalty in certain situations.
One of the most celebrated improvements in Sledgehammer is the addition of an on-die memory controller . With CPU speeds drastically outpacing RAM speeds, the system memory has become a major bottleneck to overall performance. Traditionally, RAM traffic is routed from the CPU to the memory controller on the northbridge, and then finally to the RAM. By moving the memory controller onto the CPU die itself, the intermediate step is removed. Thus, memory latency is drastically reduced by literally shortening the distance between the CPU and RAM. In fact, AMD attributes up to 20% of the performance increase from Barton to Sledgehammer to this controller. However, lowering memory latency is not the only benefit. In standard multi-CPU systems since there is only a single northbridge, with every additional CPU the portion of memory bandwidth available for each CPU becomes less and less, presenting a huge bottleneck. In AMD64 multi-CPU systems, each core has its own memory controller instead of sharing the single memory controller on the northbridge. This results in a much more efficient availability of total memory bandwidth with respect to the number of CPUs, thus raising performance by a considerable margin.
Traditionally, the front-side bus (FSB) has been a bottleneck in performance. The Sledgehammer solves this problem by replacing the aging EV6 bus with the high bandwidth, low latency HyperTransport (HTT) computer bus. The double-pumped 800MHz (1600MHz effective) HTT allows for up to 6.4GB/s of bandwidth, up from 3.2GB/s on the Barton core. Coupled with the Socket 939 platform's integrated memory controller (not found on Turion), PC3200 dual channel memory on the efficient HTT bus was easily able to keep up with a Pentium 4 paired with dual-channel PC5400.
The Sledgehammer's Translation Look-Aside Buffers (TLBs) have also been improved. When a CPU accesses its memory, it must first convert the address location given by the code into the physical address that tells the CPU where in the RAM the data is located by fetching it from the page table. Essentially, to perform a memory access the CPU must first perform the page table memory access, which adds a considerable amount of latency to the process. The TLB stores the results of previous conversions. Thus, if an address has already been cached, the CPU can directly fetch the address from the TLB, reducing latencies even further. Sledgehammer doubles the size of the TLB, in addition to lowering the latency involved with accessing the TLB. The TLB has also been modified to work better when running multiple processes. Unfortunately, the problem is that the TLB is only helpful when processing highly memory-intensive tasks, which mainly useful in servers and only rarely on the desktop. Since Turion also inherited this trait, the enlarged TLB wastes a decent amount of both transistors and power consumption.
The Sledgehammer core also supported a new instruction set, x86-64 (later called AMD64, to the annoyance of Intel), which was essentially a straightforward extension of the original 32 bit x86 instruction set to 64 bit. This evolution of the x86 instruction set added many enhancements. The number of general purpose registers was increased from 8 to 16, and the register capacity was increased from 32 to 64 bits. This alone is responsible for much of the performance increase when running in native 64 bit mode. 64 bit support also allows the CPU to access up to 256 terabytes (1 TB = 1024GB) of RAM, a large step up from the 32 bit's 4GB limitation. This is especially useful for servers, which were becoming hampered by the 4GB limitation. x86-64 removed the inefficient segmentation support (a form of memory protection), replacing it with the faster new "SYSCALL" function. Since most programs are currently still 32 bit-based, and 64 bit operating systems are still immature, x86-64 support is not very valuable to the average computer user and is only a form of "future-proofing" the system. Fortunately, even when in 32 bit mode, x86-64 performs identically compared to the standard x86 instruction set, making x86-64 a useful feature to have.
The final addition is the NX bit , which stops certain viruses by preventing buffer overflows, if supported by the operating system. Windows XP SP2 added NX and XD (Intel's renamed version of NX) support, so if your OS is properly updated, you have some basic virus protection. Remember that NX does not stop all viruses, so having antivirus software is still highly recommended.
Sledgehammer was an instant hit, defeating its Intel counterpart by a considerable margin. AMD soon launched a whole lineup for desktops, under the blanket name AMD Athlon 64. There were two models: the single-channel desktop version, on Socket 754 , and the dual channel desktop version, on Socket 939. Every processor on Socket 939 , aside from the FX series, had its L2 cache halved to 512KB. The Socket 754 versions featured either 1MB or 512KB of L2, depending on the model. The soon-released CG stepping improved on the memory controller, offering wider RAM support, and removing some of the annoying restrictions, such as forcing the user to have identical DIMMs in the 2nd and 3rd memory slots. AMD also released DTR laptop versions of these chips, with Socket 754 as the socket of choice. These processors scaled up to 4000+ (2.6GHz), but had a high TDP of either 81.5W (DTR) or 62W (Mobile). The entire Athlon 64 lineup was immensely successful and was followed by the much-heralded die shrink, from 130nm to 90nm . Intel had just shrunk its Pentium 4 to 90nm as well, but it was facing many problems caused by the shrink, including leakage and overheating. AMD's new core, code-named Winchester and available only on Socket 939, was an instant success. AMD faced minimal problems associated with the die shrink. Winchester not only saved AMD lots of money due to its smaller die but also ran cooler and used less power, providing the backbone for AMD's first foray into low-wattage mobile computing with the AMD64 line. AMD merely ported Winchester to Socket 754, removed the IHS (integrated heat sink), and renamed it the Oakville core. Unfortunately, Oakville wasn't very popular at all, only getting adopted in a mere two laptops: the Acer Ferrari 3200 (Athlon 64 2800+ Low-Voltage Mobile 35W), and the Acer Ferrari 3400 (Athlon 64 3000+ Low-Voltage Mobile 35W). The latest desktop core change, codenamed Venice for the 512KB L2 versions and San Diego for the 1MB L2 versions, had several new improvements. Memory controller performance and compatibility continued to improve, and SSE3 support was added. This newest SIMD instruction advanced 11 new instructions to improve multimedia performance. The new cores were also able to scale incredibly high, easily overclocking to 2.8GHz and beyond. This was due to a new technology, Dual Stress Liner (DSL), developed by a joint operation with IBM. The technology stretches and squeezes silicon atoms in the CPU to improve response time (up to 24% higher), as well as lowering heat and power.
AMD used the Venice and San Diego core as the basis for its Turion lineup. Due to its design, both cores already ran cool and used little power. The latest desktop AMD64 3500+ processor running at 2.4GHz only used about 55W at full load. To create a high performance, low power consumption CPU, AMD merely converted the Venice/San Diego to single-channel Socket 754 , used power-optimized transistors (as opposed to transistors optimized for speed on desktop CPUs), and added an additional C3 Sleep state . Turion processors come in 25W (MT) and 35W (ML) flavors. High end Turion CPUs all have 1MB L2 cache , while the midrange and low-end Turion CPUs is either 512KB or 1MB L2 cache, based on the model number. Currently, the highest rated model runs at 2.2GHz (ML-40). A 2.4GHz model is planned for Q4 2005, with the 25W version available Q1 2006. Even with relatively few mobile-specific optimizations, Turion is still extremely competitive with its Intel counterpart.
ComparisonIn order to correctly interpret the benchmarks, it is imperative to understand the strengths and weaknesses of the Dothan and Turion processor architectures.
Featuring brand new technology developed specifically for a high performance to power ratio, such as the accurate branch prediction, micro-ops fusion engine, and dedicated stack manager, Dothan obviously has the superior CPU architecture. In contrast, Turion is merely a slightly modified desktop/server CPU. In fact, it would make sense to call Turion a low-voltage, single-channel Athlon 64. Unfortunately, this means that many transistors are wasted on server and desktop-specific features. Due to its architecture, it is fair to assume that Turion will lose to Dothan in power consumption and heat, but we must keep a rather large detail in mind: Intel and AMD do not calculate thermal design power the same way. Intel measures TDP as the maximum power dissipated at 75% of the maximum power for a given frequency. AMD measures TDP as the maximum power dissipated when the CPU is drawing the maximum current under worst-case conditions. Thus while Intel's numbers may be lower, and they have a lot of power optimizations, the Pentium M is not as economical as its 27W TDP suggests. Suddenly the Turion ML's 35W TDP isn't looking so bad, and the Turion MT's 25W TDP is looking mighty good. Therefore, we can expect Turion to be, for the most part, competitive with Dothan. The question is where the strengths and weaknesses of each architecture lie.
One of the easiest comparisons is the L1 and L2 cache. Dothan has 32KB of 4-way associative L1 cache, coupled with 2MB of L2 cache. Turion has 64KB of 2-way associative L1 cache, coupled with 1MB of L2 cache. The 4-way associativity of Dothan's L1 cache basically negates the 32KB size advantage Turion has, making the two processor's L1 cache essentially equal in performance. Cache associativity means that there are "N" possible places that a given memory location may be in the cache. In this case, N is 4 for Dothan and 2 for Turion. Larger N increases L1 performance drastically. Since Dothan has twice the L2 of that of Turion, it has an advantage in memory intensive applications, especially gaming and business applications where large amounts of data need to be processed.
Although Turion and Dothan have 12 pipeline stages, determining each CPU's IPC (Instructions Per Cycle) is still difficult. Dothan features many IPC-enhancing features such as accurate branch prediction, micro-ops fusion engine, and dedicated stack manager. However, Dothan inherits all of the Pentium III's execution units, which unfortunately are only limited to 2 ALU and a mere 1 FPU unit. In contrast, Turion has an amazing 3 ALU and 3 FPU units. The number of ALU and FPU units determine the maximum number of arithmetic and floating-point operations that can be performed every clock cycle. Therefore, Dothan is extremely weak at FPU calculations, while competitive in ALU calculations since its other features make up for its lack of computational power.
Dothan and Turion are evenly matched in memory bandwidth. Even though Dothan uses RAM that runs at 33% higher frequencies (266MHz DDR2 vs 200MHz DDR1) and has dual-channel memory support (compared to Turion's single-channel support), the extremely high DDR2 RAM timings (4-4-4-12 standard) reduce memory performance significantly. Turion's integrated memory controller, on the other hand, reduces latencies even further to maximize memory performance. The result is that even though Dothan has much more POTENTIAL bandwidth, but since the latencies associated with the RAM are much higher Dothan doesn't even come close to its theoretical maximum bandwidth. Turion on the other hand, due to the combined power of its relatively low latency DDR1 (2.5-3-3-7) and integrated memory controller, comes extremely close to its theoretical maximum bandwidth. The result is similar average memory bandwidth. However, in programs with constant memory accesses, expect Turion to have more memory bandwidth. With programs that have high sustained memory access, expect Dothan to gain the edge.
One obvious advantage that Turion has over Dothan is SSE3 support. First introduced in Intel's Prescott core, SSE3 adds 13 new instructions (only 11 were implemented, since the other two concerned HyperThreading) designed to improve multimedia performance. Video encoding and gaming all gain a slight boost in performance, but only in programs that are optimized for SSE3.
A main reason many people consider purchasing a Turion-based laptop is for x86-64 support. With the release of the 64 bit-capable Windows XP X64 and with Windows Vista coming in 2006, a performance increase by merely updating the OS seems tempting to many. However, one must realize that the shift from 32-bit to 64-bit will take a much longer time than one would expect. Unless you are keeping the laptop for more than a couple years, converting to a 64-bit OS like Vista as soon as it's released is just asking for trouble, due to the countless bugs and lack of softwar e/driver support that are bound to follow. Of course, if you love tinkering with new technology, then x86-64 support is a good reason to choose Turion over Dothan.
Price: The Hidden FactorWe realize that price is usually one of the largest factors in choosing a laptop. Therefore, it is essential to provide an adequate price analysis to properly compare the Turion to the Dothan processor.
In order to successfully compete with Intel, AMD priced its processors at roughly half a speed-grade lower than that of its competition. Listed below are the official MSRP for both companies.
Model | AMD Turion | Model | Intel Sonoma | Price Diff. |
| ML-40 2.2GHz 1MB | $354 | 770 2.13GHz | $423 | $69 |
ML-37 2.0GHz 1MB | $263 | 760 2.0GHz | $294 | $31 |
ML-34 1.8GHz 1MB | $220 | 750 1.86GHz | $241 | $21 |
ML-32 1.8GHz 512kB | $184 | 740 1.73GHz | $209 | $25 |
ML-30 1.6GHz 1MB | $154 | 730 1.6GHz | $209 | $55 |
ML-28 1.6GHz 512kB | $154 | n/a | n/a | n/a |
Price checked 8/16/2005
Considering that the difference in price between the 2.2GHz ML-40 and the 2.13GHz 770 is only $69, it would not seem that Turion has much of an advantage in this department. Just to be sure however, we compared the Compaq V2000 with the Compaq V2000Z. These two notebooks, aside from their CPU, are essentially identical. To aid in comparison, advantages that one laptop may have over the other are in bold.
System | Compaq V2000 | Compaq V2000Z |
OS | Windows XP Home SP2 | Windows XP Home SP2 |
CPU | Pentium M 725 1.60GHz 2MB L2 | Turion ML-30 1.6GHz 1MB L2 |
Bus | 400 MHz | 1600MHz |
RAM | 2x256MB DDR | 2x256MB DDR |
CD/DVD | 8X DVD | 8DVD Drive |
Networking | Intel PRO/Wireless 2200BG WLAN | 54g 802.11b/g WLAN w/SpeedBooster |
Hard Drive | 60GB 5400RPM | 60GB 5400RPM |
Video | Intel Extreme Graphics 2 | ATI RADEON XPRESS 200M |
Battery | 6-Cell 4,400 mAh | 6-Cell 4,400 mAh |
Price | $1,049* | $899* |
*Price checked 8/28/2005
Screenshots from HPShopping:
V2000
V2000Z
Not only is the Turion-based V2000Z $150 cheaper, it also comes with a better graphics solution. Although the ATI Radeon Xpress 200M is also integrated, since it's based off the desktop X300 graphics card, it offers much higher performance, as well as broader compatibility compared to Intel's GMA900 offering. We do realize that the V2000 is based on the i855 Centrino platform, using the 400MHz FSB Dothan instead of the newer 533MHz FSB Sonoma platform. We believe that this comparison is accurate enough to make a point, since if HP were to offer the Sonoma platform on the V2000 the price would still be similar or even more expensive. It is also a less fair comparison to use the HP dv1000, a machine with the same chassis and more features, as HP charges a bit of a premium for the dv1000's multimedia features.
To see where the true competition lies, we configured the V2000Z to closer match the V2000 in pricing by upgrading to a faster Turion processor.
System | Compaq V2000 | Compaq V2000Z |
OS | Windows XP Home SP2 | Windows XP Home SP2 |
CPU | Pentium M 725 1.60GHz 2MB L2 | Turion ML-37 2.0GHz 1MB L2 |
Bus | 400 MHz | 1600MHz |
RAM | 2x256MB DDR | 2x256MB DDR |
CD/DVD | 8X DVD | 8DVD Drive |
Networking | Intel PRO/Wireless 2200BG WLAN | 54g 802.11b/g WLAN w/SpeedBooster |
Hard Drive | 60GB 5400RPM | 60GB 5400RPM |
Video | Intel Extreme Graphics 2 | ATI RADEON XPRESS 200M |
Battery | 6-Cell 4,400 mAh | 6-Cell 4,400 mAh |
Price | $1,049* | $999* |
*Price checked 8/28/2005
After configuring the V2000Z to have a closer price point as the V2000, the V2000Z has a powerful high-end CPU clocked at 2.0GHz making the Turion-based notebook a true bargain in this situation. You can add $25 for the 8-cell battery and another $25 to upgrade the 8X DVD drive for a DVD/CD-RW combo drive to match the V2000 in price, yet with better components. The features configured in the V2000 are a large step down from the V2000Z. For entertainment's sake, you can feel free to downgrade the V2000's hard drive and increase its CPU speed or RAM; but the result is the same and Turion proves itself to be a better value.
This example is not an anomaly; due to many hidden price premiums associated with using Dothan processors, Turion has a significant advantage in price. In order to benefit from being able to market the laptop as Centrino,' manufacturers are required to use Intel chipsets and wireless cards. Since Intel has a complete monopoly over Dothan chipsets, Intel is free to charge a premium for its products. AMD takes the opposite route and allows third party manufacturers to produce chipsets and wireless cards for its mobile platform. The downside to this approach is that most mobile chipset manufacturers (SiS, VIA, ATI) have not put the time and money into designing an exceptionally power and performance optimized chipset like Intel's. The upshot to this is a generally lower price point and a variety of chipset features.
However, even extremely low prices will not help Turion if it's not up to par in the most important category: performance. Will Turion flourish or flop? Read on to find out!
Setup PreperationIn order to perform logical, unbiased comparison, excruciating care was exacted to obtain the most accurate results possible. Since the use of desktop chipsets/motherboards would skew test results, two laptop platforms were used for the benchmark process: the Acer Ferrari 4005 and the Acer TravelMate 8104.
| Specifications | Ferrari 4005 | TravelMate 8104 |
| Processor | AMD Turion 64 Mobile ML-37 | Intel Pentium M Processor 760 |
| FSB/HTT | 1600MHz | 533 MHz |
| Chipset | ATI Radeon Xpress 200M | Intel 915 PM Express |
| Wireless LAN | Broadcom 802.11b/g with SpeedBooster | Intel PRO/Wireless 2915ABG (802.11a/b/g) Bluetooth Wireless IrDA |
| LCD | 15.4' WSXGA+ TFT LCD (1680x1050) | 15.4' WSXGA+ TFT LCD (1680x1050) |
| Hard Drive | 100GB Seagate Momentus 5400RPM 8MB Cache (ST9100823A) | 100GB Seagate Momentus 5400RPM 8MB Cache (ST9100823A) |
| Memory | 1GB DDR333 PC2700 SDRAM | 1GB DDR2-533 SDRAM (2 x 512MB) on Dual-Channel Mode CL=4 |
| Graphics | ATI Mobility Radeon X700 | ATI Mobility Radeon X700 128MB of DDR Video RAM on PCI Express |
| Graphics Interface | S-Video/TV-out/DVI-D | S-Video/TV-out/DVI-D |
| Optical Drive | Slot-Load DVD-RW Super-Multi Double Layer | Tray-Load DVD-RW Super-Multi Double Layer |
| Audio | Realtek AC' 97 | Realtek High Definition |
| Audio Interface | Microphone, two stereo speakers, headphone/line-out with SPDIF support | Microphone, two stereo speakers, headphone/line-out with SPDIF support |
| Weight | 6.3 lbs. with 8-cell battery | 6.3 lbs. with 8-cell battery |
| Size (W x D x H) | 14.3' x 10.5' x 1.2'-1.4' | 14.3' x 10.5' x 1.2'-1.4' |
| Operating System | Windows XP Professional w/SP2 | Windows XP Professional w/SP2 |
| Battery | 4,800 mAh | 4,800 mAh |
Since the Ferrari 4005 only uses DDR333, in order to obtain a fair comparison, we enlisted the help of OCZ with DDR400 PC3200:
DDR400 is supposed to be standard for Turion laptops, although many manufacturers prefer to use DDR333 RAM instead. At default, the Ferrari recognizedour OCZ DDR400 RAM as DDR333. This was easily changed by changing the MEMCLK Frequency to 200 using A64 Tweaker:
With the aid of the excellent freeware program AMD64Tweaker, written by CodeRed, we were able to clock the OCZ RAM to run at DDR400 with the Acer Ferrari 4005WLMi.
The Ferrari recognized our OCZ DDR400 RAM at DDR333(167 MHz).
Ferrari clocked to DDR400 (400MHz) with AMD 64Tweaker.
The OCZ RAM uses Samsung UCCC, which has default settings of 200MHz 3-3-3-x timings. Samsung UCCC is built off the 90nm process, and is said to overclock well, up to 250MHz (albeit at higher timings of 2.5-3-3 or 3-3-3). Unfortunately, no overclock was able to be performed to test out the theory.
A hidden factor turned out to be the ATI Mobility Radeon X700 128MB frequencies. The X700 in the Ferrari 4005 is clocked higher than the TravelMateat 358/345 (core/memory), while the X700 in the TravelMate 8104 is clocked at 358/300. The problem was solved with the aid of ATI Tool by overclocking the graphics card in the Travelmate 8104to match the speeds of the graphics card in the Ferrari.
Default X700 clock speeds from the TravelMate 8104.
ATI Mobility Radeon X700 from TravelMate 8104 overclocked to match the Ferrari X700 speeds.
Another problem was that the two laptops used different graphics card drivers. The Ferrari 4005 used the 6.14.10.6546 driver set, dated at 6/13/05, while the TravelMate 8104 used the Page:1/1
Stay up-to-date with the latest mobile technology news and laptop reviews with our free newsletter! Get free tips, news and information from Laptop Logic and find the best laptop deals online.
Comments
SkinnerBETSY29 at 02:45, March 28, 2010
Lots of specialists claim that business loans help a lot of people to live their own way, because they are able to feel free to buy needed goods. Moreover, some banks present college loan for young and old people.
de Bill at 03:17, April 29, 2009
Hello, I am now trying to overclock exactly same Ferrari 4005 and exactly same 2 OCZs DDR400 modules at 200 MHz with the same A64Tweaker, but the computer still crashes. What version of the BIOS did you have on yours Ferrari? Anything else to set up to reach 200 MHz please? I can reach just 183 MHz with no crash .. coz I think it is impossible to do it with OCZ modules, they seems to need more strain. Thx.