After about a year since the Intel Company started mass manufacturing of x86 processors using the 0.13-micron technology, the long-time Intel's rival, AMD, also finished the transition to the new technology. It should be mentioned that AMD managed to compete quite successfully with Intel without the most advanced technology for a long time. The progressive K7 architecture used in the Palomino based processors allowed Athlon XP to perform as fast as high-end Pentium 4 CPUs that work at higher clock-rates. But recently the situation has greatly changed. The Pentium 4 working frequencies have sprung up and AMD has been left in a kind of dead end with its 0.18-micron technology limiting the working frequencies of Palomino based processors to 1.73GHz.

That's exactly how the things were by the beginning of the summer when after numerous delays AMD could at last launch its first desktop CPUs on the new Thoroughbred core made with 0.13micron manufacturing technology. Everything mentioned above is exactly the reason why AMD fans pin great hopes on the new core. Thoroughbred, although it differs slightly from its predecessor in the architecture, can allow to continue raising the clock frequencies of the Athlon XP family thus giving the opportunity for high-performing AMD processors to catch up with their Intel's competitors.

So I've got the first Thoroughbred processor that works at 1.8GHz clock frequency, that is, only 66MHz higher than the top-end Palomino model. But this CPU is interesting not only for its clock-rate but for its new core, made with the more advanced technology. That's why I will talk not so much about the new Athlon XP performance, but mostly about some other quite important things. They are:

• The differences between Thoroughbred and Palomino cores
• The thermal limitations of the new Athlon XP processors
• The further AMD plans for Athlon XP family promotion
• The overclocking of Thoroughbred based processors

What is Thoroughbred?

First of all let's point out the differences between the new Thoroughbred core and its predecessor, the Palomino core, used in the Athlon XP processor family. In brief, the main difference between the two cores is the technological process used for their manufacturing. In more words, we should mention that the transition to the 0.13-micron process for Thoroughbred production allows AMD to kill a few birds with one stone.

The "birds" are mostly of a technological nature. As the perfection of the technological process leads to a smaller processor die size it also means that they will be able to produce more dies of a single wafer. According to AMD, a wafer 200mm in diameter (that's exactly the wafers used at Fab30 in Dresden) yields 322 Thoroughbred dies while earlier such a wafer could only give out 201 Palomino cores. The result is smaller production cost, which gives us, the end-users, hope that AMD's current pricing policy will keep pleasing us.

We shouldn't also miss out the fact that the change of the technological process will allow raising Athlon XP working frequencies higher up. As it has been already said, the 1.73GHz frequency is virtually the limit for the 0.18-micron Palomino core. The move from the Palomino core to Thoroughbred is going to offer AMD the possibility of speeding up its high-performance CPU family. For example, preliminary estimates suggest that Thoroughbred based Athlon XP processors will at least reach the 2GHz bar by the end of the year. By this time, AMD is also planning to replace the Palomino core with Thoroughbred in the low-end Athlon XP models with ratings of 1700+ and higher, though.

It wouldn't be correct to consider the Thoroughbred core as a remake of Palomino with some new norms imposed by the finer manufacturing technology. Thoroughbred considerably differs from Palomino in its inner structure; the following snapshots bear the evidence:

Palomino core		Thoroughbred core

By the way, it's exactly the changes mentioned that delayed the launch of Thoroughbred to the market. The original variant of the core didn't satisfy the developers and that's why we didn't see Thoroughbred in the beginning of the year as had been previously planned. AMD had to work on the design of the 0.13-micron core and the results of the work are evident, but unfortunately the time was lost and after the launch of the Palomino based Athlon XP 2100+ the regular introduction of new processor family has been hitched a little.

Notwithstanding the outer dissimilarities, the two cores are nearly identical in their characteristics. The difference only lies in the technology and, as a result, in the size and geometry of the core. The main characteristics of Athlon XP cores in comparison with Northwood core of Pentium 4 are listed below:

The fact that we don't see any architectural innovations by Thoroughbred as compared with Palomino is quite easy to explain. Right now AMD is focusing on polishing up its 8th-generation Hammer processor, which is to be launched at the beginning of 2003. So, AMD has neither desire nor available resources to work on improvements in the old K7 architecture. The further development of Athlon XP is going to be as usual - by simply increasing the core clock frequency.

One change in Athlon XP will happen, though. The next core, code-named Barton, which is to be used in the processors of this family up to its end, will have a larger L2 cache - 512 KB. By the way, that'll be the only difference between Barton and Thoroughbred. AMD decided to make this change to its new product because the 0.13-micron core required certain redesign, so that its production could be moved to UMC's fabs.

As there's about half a year before Barton appears in the market, let's return to the hero of our story, the Thoroughbred core.

To sum up everything mentioned above, we can only state that there're really few points of difference between Thoroughbred and Palomino. As for the "outer manifestations" of the new core, that is, the features which allow distinguishing between the processors on Palomino and Thoroughbred cores, we should mention the following:

• The CPUID of Thoroughbred based processors is changed into 68xh. Let me remind that the CPUID of Palomino based processors is 66xh.
• The passive components, which were previously situated at the bottom of the processor are now moved to its top
• Processor's marking, which was previously made on the processor die itself, now is moved to an additional plastic sticker on the CPU's package.
• The marking itself has been changed slightly: now the line begins with letters "AXD". Overall, the Thoroughbred-core Athlon XP marking should be read as follows:

As for compatibility, AMD claims the new Thoroughbred based processors are fully compatible with Socket A infrastructure and the only thing required from mainboards to support new Athlon XPs is an updated BIOS version that can "recognize" the new processors. Nevertheless, there have been claims from users that some mainboards do not support Thoroughbred yet. So when buying the new CPU don't forget to check the platform maker's web-site to make sure that your specific mainboard does support Thoroughbred based processors (check the mainboard revision first).

Is Athlon XP As "Hot" As We Think?

One more advantage of the shift to finer manufacturing technology is lower heat dissipation. The new Athlon XP on the Thoroughbred core is no exception. As the new processors use lower core voltage they do produce less heat. Here are the official AMD data on heat dissipation of Athlon XP processors built on Thoroughbred and Palomino cores:

As you can see from the table, the move to 0.13micron technology allowed reducing the heat dissipation of slower Athlon XP CPUs by 22% (1.5V Vcore) and by 12% for faster models (1.65V Vcore). It is interesting that according to the table drawn from the official data, future Thoroughbred based processors will dissipate no more heat than the today's fastest Palomino based Athlon XP 2200+.

It should be mentioned that the 12% reduction of heat dissipation is no impressive result. Though, as the voltage of the higher Thoroughbred models has only been lowered by 6% there's no reason to hope for more.

On the other hand, is it that necessary to reduce Athlon XP's heat dissipation? For example, in the following diagram I compared heat dissipation of top-end Thoroughbred based Athlon XP with that of Northwood based Intel Pentium 4:

Yes, the typical Athlon XP processor does dissipate a little more heat than Pentium 4. But there's nothing to be so worried about. The situation isn't so dramatic as most users think. In my opinion, no one would ever notice the allegedly awful heat dissipation of Athlon XP if it were not for three facts. These facts, acquiring with the time numerous details and exaggerations, push most users to consider Athlon XP an extremely warm CPU. So, here are the real facts:

• Processors from Athlon XP family, especially the ones based on the new Thoroughbred core, need a high-quality cooler with a copper base while Pentium 4 doesn't put any strict requirements on the cooling system.
• Athlon XP processors fail due to overheating more often than Pentium 4.
• Athlon XP processors working in normal conditiond have higher core temperature than Pentium 4.

Once facing these facts, the unsophisticated reader would instantly infer that the Athlon XP's temperature is too high, even dangerous for the CPU. But the facts need a few comments making clear the point of the matter. On the whole, the comparison to Pentium 4 is not correct as the operation mechanisms, thermal conditions and monitoring of the Intel's processor differ greatly from the functions implemented in Athlon XP.

So, let's try to work it out. As for the necessity of using copper base coolers with Athlon XP processors on Thoroughbred core, it's quite simple to explain: the Thoroughbred core is very small, i.e. the surface dissipating heat is small. That is why the density of the heat stream to be taken off the Thoroughbred processor is rather high and the dissipated heat should get quickly spread upon the heatsink footing to avoid local hot-spotting. That's why the footing should be made of a material with high thermal conductivity.

As for Pentium 4, this processor is equipped with IHS (Integrated Heat Spreader) that plays the same role as the copper cooler base in Athlon XP. In Pentium 4 the heat produced by the core first spreads along the IHS, which is a 2-millimeter copper plate with nickel overcoat, and then is taken to the main heatsink through the large surface of the heat-spreader. As a result, the cooler for Pentium 4 can be made of aluminum as well, as IHS actually none other but the copper cooler footing integrated directly into the processor.

The use of IHS in Pentium 4 has its shortcomings, though. As the IHS is in contact with the cooler, the temperature lag of the Pentium 4 cooling system is a little bit higher than it could be if the heatsink were in direct contact with the core. Anyway, IHS helped Intel to solve quite successfully the problem of the processor core hot-spotting.

Unfortunately, AMD has not come to use IHS in Thoroughbred based processors. CPUs on Barton core won't have IHS either. But all 8-generation AMD processors, which are going to be rolled out in the end of 2002 - beginning of 2003, will be equipped with a heat-spreader.

As for burnt Athlon XPs, we cannot deny these sad facts, unfortunately. These CPUs do sometimes fail due to overheating while the rivalry Pentium 4 family is equipped with special heat-protection that stops the processor when the critical temperature level is exceeded. Athlon XP has no protection like that, so, if the user does something wrong when installing the cooler or uses low-quality or not very tight sitting cooling systems, the processors on Palomino core as well as on Thoroughbred may just burn up.

It's not fair to claim that AMD didn't care to protect its CPUs from overheating. The company just wanted the mainboard to bear part of the responsibility for overheating and, as we see, miscalculated greatly. AMD equipped its Palomino and Thoroughbred based processors with a built-in thermal diode. The data taken by this diode was originally supposed to be then processed by the mainboard. So, AMD was left to the mercy of mainboard makers, who didn't prove eager to support the new initiative. The implementation of thermal control on the mainboard requires additional costs. Although the costs are not so big (the required chips cost less than $1), no one appeared willing to pay them. So, nowadays there are very few mainboards in the market that have a correct heat-control system. Anyway, there's still some hope that the situation will improve: AMD decided to force mainboard makers to support heat-protection and won't certify products that don't have it.

It'd be fair to mention that 8th generation processors, unlike their predecessors, will have a built-in protection against core overheating, the same as by Pentium 4.

Now let's talk about the higher working temperature of Athlon XP. However paradoxical it may seem it's mainboard makers again, and not the processors, that are to blame here. The point is that when processing the HALT command, which means processor's stopping as there're no instructions to run, the CPU usually switches to the "waiting" mode (Halt and Stop Grant). The processors of the Pentium 4 family automatically turn into the power saving mode. In case of Athlon XP it's necessary to disconnect from the system bus (Bus Disconnect) if you want to switch to the power saving mode, which is supposed to be done via the chipset and mainboard BIOS.

Unfortunately, the same mishap happened to the system bus disconnect task, just like in case of thermal diode. As practice shows, BIOSes of nearly all mainboards is configured in such a way that Athlon XP never switches to the power saving mode. So, while Pentium 4 processors are at rest, Athlon XPs go on working. Considered that computers usually stay idle in office tasks waiting about 95% of the time for the data to be processed, it's natural that in normal conditions Athlon XP is warmer than Pentium 4.

Right now, I can only think of one mainboard maker that implemented the possibility to switch Athlon to power saving mode in the BIOS. It is Fujitsu-Siemens, which mainboards are hard to find in the market.

Thoroughbred Temperature: Practical Test

I conducted a few experiments measuring the working temperature of Athlon XPs on the Palomino and Thoroughbred cores as well as of Northwood based Pentium 4. The goal of the experiments was to prove the above-expounded theory as well as to study the temperature of new Athlon XP CPUs in practice.

To test the thermal characteristics of the processors I used an open stand working at 25oC room temperature. The temperature of the processor core was measured by the mainboard with the help of the thermal diode built into the processor. For a Socket A platform I used ASUS A7V333 based on the VIA KT333 chipset (one of the few that can use the thermal diode built into Athlon XP). i845G chipset based MSI 845GMax was used as a Pentium 4 platform. To conform to the "pure experiment" principle I utilized the same Thermaltake Volcano 7 cooler for both Socket A and Socket 478 platforms. I chose it for one simple reason: it can be installed on both sockets thanks to different fastening mechanisms shipped with the cooler. The cooler's fan rotation speed was set to 3,000rpm to get an acceptable noise level.

   

Before every test the processors and the cooler were left to cool down for 30 minutes, the temperature was measured once the testing participants were stably working for 30 minutes after the hardware boot-up and software loading. The operation system used was MS Windows XP Professional. On the whole, we tested four processors: Athlon XP 2100+ on Palomino core, Athlon XP 2200+ on Thoroughbred core, and Pentium 4 2.2 and Pentium 4 2.53 on Northwood core.

First of all I measured the processor temperature in Windows XP in the "passive mode". Half an hour after the operation system had been started, the following picture cleared up:

This is an obvious proof of all the above-said. When idle, Athlon XP processors turn out to be somewhat warmer than Pentium 4 Northwood. The problems of Athlon XP are rooted in the mainboard, which doesn't switch it to the power saving mode during the idle moments.

Now let's see how the processors do when working hard. To emulate an "extreme" CPU heating-up I used programs from the CPUBurn set.

Here we are! It turns out that if we give Pentium 4 a loadful of work, thus not letting it to perform the saving HALT instructions, its temperature will be no less than that of Athlon XP. Moreover, Pentium 4 2.53GHz is even generating more heat than Athlon XP 2100+ Palomino, notwithstanding the lower theoretical heat dissipation of the Pentium 4 CPU. By the way, the new Thoroughbred based Athlon XP 2200+ proves to be 4oC colder than its predecessor.

The next experiment implies the study of processor temperatures in real applications. For this purpose, the CPU temperature was measured every 5 seconds while running the SYSmark 2002 test. This benchmark is known to model the actions of an ordinary user, starting applications and working in them with the speed of an average "hypothetical" person. That's why we can state with a high degree of accuracy that the average temperature in SYSmark 2002 is to be considered the average temperature of a processor when working in office applications and digital content creation tasks.

As we see, the general situation is not in favour of Athlon XP. While under full workload Pentium 4 2.53GHz heated up more than its AMD's rivals, in our real model Athlon XP on both cores turned out to be warmer. The reasons are the same: AMD processors don't switch to the power saving mode when idle. It'd be fair to mention that in similar tests, performed on a Socket A mainboard by Fujitsu-Siemens, the temperature of Athlon XP is much lower and doesn't exceed the temperature of the top-end Pentium 4 model. The results of that test will hopefully be published later.

Let me also note that the 0.13micron Athlon XP 2200+ proves to be not much colder than its fellow made with 0.18micron technology, namely the Athlon XP 2100+ processor on Palomino core.

In the last and maybe the most picturesque diagram in this section (click here to see it) the protocol of the temperature change in SYSmark 2002 is depicted.

The picture is not actually funny but useful. First of all, it catches one's eye that the curve illustrating the changes in the temperature for Pentium 4 differs greatly from that for Athlon XP. While the temperature of AMD processor is changing smoothly, the gradient of Pentium 4 temperature has considerable ups and downs. I have no intention to expatiate upon the fact, as the difference is possibly the result of alternative functioning principles of temperature measurement design or different thermal diode location within the processor die.

Secondly, note that Athlon XP is nearly always warmer than Pentium 4. However, somewhere in the end of the first third of the graph there's a peak where Pentium 4 2.53 has a higher temperature than both Athlon XP CPUs. The peak occurs when processing a video-movie in Adobe Premiere, which evidently uses up all the processor recourses.

Thirdly, let's once again point out that in real-life tasks the temperature of Thoroughbred based Athlon XP 2200+ does not differ greatly from that of Palomino based Athlon XP 2100+. Well, as AMD is not planning to raise the heat dissipation of upcoming Athlon XP models on Thoroughbred core, the temperatures of Athlon XP 2200+ look quite acceptable.

On Thoroughbred's Future

Although the Thoroughbred launch marked a new era for AMD, the use of 0.13micron manufacturing technology, the core seems unlikely to last for long. On the whole, there's a strikingly curious tendency in AMD: every new core has a shorter life cycle. The first Socket A Thunderbird core was present in Athlon XPs for about 16 months. Within that time its working core frequencies jumped from 700MHz to 1400MHz. The Palomino core, which came to replace Thunderbird, managed to last 8 months only. Its core clock frequencies ranged from 1.2GHz to 1.73GHz. The new Thoroughbred core seems likely to stay in the market for even shorter time: up to the end of the year only.

As for the working frequencies of 0.13micron Thoroughbred based processors, AMD plans show that the company is going to roll out Athlon XP (Thoroughbred) with 2300+ and 2400+ ratings, and core clocks of 1.87GHz and 1.93GHz accordingly, in the third quarter. Then, in the fourth quarter, Athlon XP 2500+ and 2600+ (2.0GHz and 2.07GHz respectively) will be launched. And then …that's all. Athlon XP on Thoroughbred core will not be manufactured anymore.

There's an easy explanation to this paradox. AMD now lays special emphasis on the promotion of 8th generation processor family, so that the K7 architecture is going to get less attention. During 2003 AMD wants to give up manufacturing Socket A processors at all, turning to the production of ClawHammer and SledgeHammer, at least in its Fab30. As is known, Fab25 is going to stop producing CPUs and switch to flash-memory production.

As a result, the Socket A CPU production will be assigned to the Taiwanese UMC Company. But UMC will take the baton without the help of Thoroughbred. It's going to start producing Athlon XP processors on the new core, code-named Barton. That's why Thoroughbred is now viewed as a transitional step to Barton.

Barton, as well as Thoroughbred, will be manufactured with 0.13micron technology without SOI. This core will boast a twice as large L2 cache. As a result, the size of the Barton core will be 115sq.mm and the number of transistors - 53.9 million.

As Barton is a much more complex processor core than Thoroughbred, AMD is unlikely to raise its core frequencies far beyond those of Thoroughbred. So it's quite possible that Barton based Athlon XP by UMC, coming into market in the beginning of 2003, will work at the frequencies not exceeding 2GHz. Anyway, the larger L2 cache is going to contribute to the overall processor performance, thus giving AMD the opportunity to rate 2GHz Barton higher, for example, 2800+.

So, the preliminary schedule of the new Socket A processors launch looks as follows:

That's where the further increase of Socket A CPUs core frequencies is supposed to stop, as otherwise these processors will be competing with perspective 8th generation ClawHammer CPUs. Nevertheless, the manufacturing of Socket A processors will be stopped much later. At least, AMD promises to support the socket throughout all 2003.

Thoroughbred Overclocking Experience

A lot of our readers must be interested in the overclocking potential of the new Athlon XP processors on Thoroughbred core. It seems that the 0.13micron technology used for their production should bring in good overclocking results. Added the fact that AMD plans to raise Thoroughbred core clock above 2GHz, new Athlon XPs were supposed to have significant overclocking potential.

Unfortunately the reality turned out to be much worse. Our Athlon XP 2200+, working at 1.8GHz and using 133MHz FSB, could function with the FSB being no higher than 142MHz. We could by no means improve the result. Better cooling and higher core voltage didn't help. So, the maximum frequency we got was 1917MHz. On the one hand, I managed to raise the core clock by only 6.8% above the nominal. But on the other hand I nearly reached the speed of Athlon XP 2400+, which core clock will be 1933MHz.

At the same time we should consider the following. First processors from AMD built on the new core are traditionally not very good for overclocking. We have already witnessed the same situation with Thunderbird, then with Palomino and now with Thoroughbred. The point is that the new manufacturing technology, 0.13micron Athlon XPs are produced with, is not perfect and the yields of processors dies supporting higher core clock frequencies is expected to improve with the time. So, it won't be a surprise, if in a few months Thoroughbred based Athlon XP processors will prove capable of overclocking above the 2GHz limit. For now, we are left with the modest results we get today.

Testbed and Methods

On the whole we should not expect much from the new Athlon XP. As these CPUs don't have any architectural innovations compared to their Palomino based predecessors, the performance growth by Athlon XP 2200+ on Thoroughbred core compared with the predecessor is achieved only due to the 66MHz increase in the core clock.

So the aim of the performance test was to unveil the performance growth by the new Athlon XP 2200+ compared to its predecessor Athlon XP 2100+ and Intel Pentium 4 competitors.

To put the processors in equal conditions, they were tested on platforms with DDR333 SDRAM. The platform for Athlon XP was based on the today's fastest VIA KT333 chipset, and Pentium 4 worked on the i845G based mainboard, as this chipset proved to be the fastest of all DDR333 chipsets for Socket 478 processors, according to our previous tests.

The general system configuration looked as follows:

All the tests were done in MS Windows XP Professional operation system; the mainboards were set to the maximum performance (with the minimal memory timings).

I decided to complement the test results with the showings of an overclocked Athlon XP on Thoroughbred core. As during overclocking the CPU reached 1917MHz core frequency, which is close to 1933MHz, we denote the overclocked Athlon XP 2200+ in the diagrams as "Athlon XP ~2400+". Don't forget it to avoid misunderstandings.

As the number of tests we performed for the new processors is considerably bigger than usual, we split the results into several groups, according to the test tasks.

Performance

Office Applications and Data Encoding

According to SYSmark 2002, the processors of the Pentium 4 family look better than their AMD rivals. Moreover, the superiority of Intel CPU is so evident that even an overclocked Thoroughbred based Athlon XP cannot catch up with Pentium 4 2.2GHz. Obviously, it's explained by the fact that the new versions of SYSmark applications began to use SSE2 instructions set, which is not supported by the Athlon XP family.

Audio compression into the mp3 format is performed quite fast on Athlon XP. The new model with 2200+ rating is nearly as efficient as Pentium 4 2.4B GHz.

Video compression with the DivX codec gives the lead to the Pentium 4 family again. The new, fifth version of the codec we used for the test supports SSE2 instructions. By the way, video processing is an algorithm very sensitive to data transfer rate between the processor and memory and this is the point where Pentium 4 platforms surpass Athlon XP, thanks to their high-speed processor bus.

When archiving information with WinRAR, Athlon XP 2200+ performs close to Pentium 4 2.2GHz.

Although PCMark2002 is a synthetic test, I put its results in this section as the algorithms it uses to measure CPU performance include JPEG decompression, LZ77 compression and decompression, text search and audiostream modulation. The results in this benchmark show that Athlon XP 2200+ is slightly slower than Pentium 4 2.2GHz.

As for the memory subsystem performance, Pentium 4 leaves Athlon XP behind because of a faster processor bus, notwithstanding the same memory used in the platforms. Although in the Athlon XP system the peak memory bandwidth is 2.7GB/sec, the memory bus bandwidth is only 2.1GB/sec thus the processor cannot utilize the memory bus to the full extent. Pentium 4 processors have higher linear memory bus bandwidth: 3.2GB/sec or 4.2GB/sec depending on the FSB frequency (100MHz or 133MHz), so the practical memory work speed turns out higher.

3D Games

This test shows that new Athlon XP 2200+ performs better than Pentium 4 2.2GHz, but worse than Pentium 4 2.4B GHz.

In "Return to Castle Wolfenstein" Pentium 4 performs a little better so Pentium 4 2.2GHz leaves Athlon XP 2200+ behind. But the game is based on the Quake 3 engine, which is known for its loyalty to Intel processors.

A similar situation can be observed in another popular game - Comanche 4 helicopter simulator.

But in the new RPG "Dungeon Siege" Athlon XP 2200+ beats Pentium 4 2.2GHz and is close to Pentium 4 2.4B GHz.

The second version of the shooter Serious Sam: The Second Encounter, on the contrary, is especially "Athlon-friendly". New Athlon XP 2200+ working at 1.8GHz is even faster than the top-end Pentium 4 that has 40% higher core clock - 2.53GHz. It looks as if the game developers didn't optimize their engine for SSE2 so Pentium 4 didn't have an advantage over Athlon XP, which showed everything its FPU was capable of.

3D Rendering

We investigated the 3D rendering speed in three popular testing packages: 3ds max 4.26, Maya 4.0.1 and new version of Lightwave - Lightwave v.7.5. In all the tests we timed the scenes rendering speed that is why smaller values on the diagrams correspond to better performance. To test in 3ds max we used the islands scene, for Maya 4.0.1 we used Maya-Testcenter rendertest methodology and in Lightwave we used sunset and raytrace scenes.

Whatever Athlon XP fans may say, the last 3ds max 4.26 version has been well optimized for the SSE2 instruction set. That's why the final rendering on Pentium 4 processors is performed faster than on Athlon XP, despite the remarkable computing capacities of the latter.

Maya has no special optimization for any processors, so the results obtained in it show that the performance of new Athlon XP 2200+ is about the same as that of Pentium 4 2.4B GHz.

The new version of Lightwave software as well as 3dx max 4.26 has been optimized for SSE2 instructions. So there's no need to be surprised at the impressive results of the Pentium 4 family. More to that, the raytrace scene we used here is the benchmark recommended by Intel. Of course, I couldn't miss out the possibility to test the rendering speed of the scenes Intel does not recommend. The result, to put it mildly, was just the opposite.

So we can't definitely declare the superiority of any architecture in Lightwave 7.5. All depends upon the specific scene.

Scientific and Professional OpenGL Applications

It's quite natural that the Athlon XP family does well when it comes to math1ematical modeling tasks. In that sort of applications the most important thing is FPU performance and it is one of the strong sides of the Athlon architecture. Pentium 4 that works at a much higher core frequency is losing in this test to Athlon XP equipped with three-pipeline high-performance FPU.

In SPECviewperf 7.0 benchmarks Athlon XP processors are quite strong. The reasons are the same: the algorithms used in the tests are rather out-dated and don't use SSE2 instructions. And there's still no rival to Athlon XP as it comes to intensive calculations.

We also tested the CPUs in one more benchmark - SPECapc for 3dx max 4.26. This new test allows estimating the performance in viewports in the latest version of 3dx max 4.26. The test uses four different scenes and animates them within the program environment. For more information about the test, look here.

The computational rating of Athlon XP is high, but during visualization higher data transfer rate between the processor and memory, as well as the support of new SSE2 instructions hand the laurels over to Pentium 4. The overall Athlon XP 2200+ performance is a little better than that of Pentium 4 2.2GHz, but worse than that of Pentium 4 2.4B GHz.

Conclusion

It's time to sum it all up. The transition to the 0.13micron technology, undertaken by AMD, didn't bring in anything new. Thoroughbred core proved to be similar to Palomino, with all the highs and lows of the latter. As a result, we shouldn't expect too much from the new Athlon XP. AMD has now got the opportunity to raise slowly the core clock frequencies of its processors for another while, which will be going on for about the next half a year. Then Thoroughbred will be replaced with Barton, boasting a larger L2 cache, and 8th generation ClawHammer architecture.

Meanwhile, although new Thoroughbred based Athlon XP processors produce less heat and can potentially work at higher frequencies, their overclockability might be better. So, there's no reason to buy an Athlon XP on Thoroughbred core if you are pursuing overclocking goals.

By launching 0.13micron CPU into mass production, AMD proved once again that there shouldn't be any problems with the production of the budding hit - Hammer. Moreover, the use of more advanced technology for the production of Athlon XP family will automatically leaf to the reduction of the production costs, and thus to lower prices for AMD processors, which will definitely make them more attractive to the users.

As for performance, we can only state that AMD has fallen behind Intel and can't now offer processors performing close enough to Pentium 4 2.53GHz. New Athlon XP 2200+ in most cases works just a little bit faster than Pentium 4 2.2GHz. Moreover, as the SSE2 instructions set is getting popular, Pentium 4 processors show even better results. If we add to that the recent transition of Pentium 4 processor family to 533MHz bus, which also contributed in a way to its performance, then it's quite possible that Athlon XP rated adequately to the Intel processors core clock will be lagging behind them in a lot of applications.

On the other hand, it should be mentioned that though AMD lost the leading position in performance, there are no equals to Athlon XP processors in price-to-performance terms in the market and there will hardly be any in the future. And it should be also noted that there is an area where Athlon XP is an indisputable leader: the tasks involving FPU very actively.