Battle of the 64-bit Budget CPU - Sempron 64 vs. Celeron D 64

Which budget 64-bit CPU platform offers the best bang for buck? Read on as we investigate Sempron 64 vs. Celeron D 64.

Manufacturer: none
11 minutes & 30 seconds read time


While we have covered the high-end crowd of the 64-bit arena, we haven't touched on what might be the 64-bit revolution for the value and mainstream users. Both AMD and Intel now produce 64-bit budget chips.

AMD's first budget CPU came in the form of the Duron series back in the Socket A days. This CPU was not only fast, but cheap too. The price in its peak was around $100 AUD for a Duron 750MHz which would power most games you wanted to play. Not only that, but would overclock as well due to the lesser cache memory not getting in the way. Next came the Sempron in Socket A which were basically re-badged AMD Athlon XP Thoroughbred-A cores with less cache.

Now that Socket A has been phased out, and no more Socket A CPU's are hitting the market, to keep budget at its best, AMD has gone to its original Socket 754 K8 architecture to produce the Sempron 64 CPU.

Meanwhile, Intel is also shipping its value based Celeron D now with 64-bit instructions. The first Celeron appeared back when Slot 1 made its debut, back when Pentium II was the unaffordable CPU of many. To keep the masses happy, Celeron was introduced initially without a Level 2 cache, this caused a lot of controversy as the CPU was extremely slow at executions. Celeron-A came out shortly after which included 128K of L2 cache rather than the 512K that the Pentium II had. The one thing Celeron had over Pentium II was the fact that its L2 cache was built onto the CPU die, the first CPU from Intel to do this. This led the way for Socket 370 as well as the later Pentium 3 CPU's on this same socket.

Celeron moved over to the Socket 478 platform based on the Northwood core Pentium 4 design, with less cache, as always. Today Celeron-D is being produced on the LGA775 format and including 64-bit instructions.

Today we are pitting the AMD Sempron 3000+ against the Intel Celeron-D 336. How will they perform? Who will win? Let's find out.

Architecture vs. Architecture

The Architectures

First off let's take a look at the two competing architectures.

The Sempron 64 derives from the original Hammer architecture that AMD introduced back in 2003 at Computex. The Hammer architecture has slowly progressed with changes in each new core revision but the basics remain the same.

64-bit Instruction Set: We'll start with the most important part of a modern CPU, its support for 64-bit OS's and applications. AMD was the first to introduce a 64-bit CPU using the x86 architecture for both desktop and workstations. While supporting 64-bit registers, AMD also keeps legacy support alive by including 32-bit ISA registers. What this means to users is you can run either Windows XP 32-bit or Windows XP 64-bit, there is no limitation on this part. When running Windows XP 64-bit, you have the ability to address memory ranges above 4GB, in fact up to and beyond 16GB is possible.

Integrated Memory Controller: AMD K8 architecture is the first desktop CPU to integrate a memory controller directly to the die. AMD uses a crossbar to connect a single DDR memory channel directly to the CPU die, allowing almost instantaneous access between the L2 cache and the system memory. This results in lower latencies than a CPU accessing the system memory along a traditional FSB arrangement.

SSE3 Instructions: SSE3 was introduced by Intel in order to increase the multimedia encode and decode of the Pentium 4 based CPU's. Since then it has now been added to AMD's latest CPU's to give AMD that extra advantage in applications that make use of these new instructions such as media encoding tasks.

90nm Fab process: The Sempron uses AMD's latest 90nm Silicon on Insulator process or SOI as its known. This process has been the success for AMD in the battle of performance per watt. AMD has been able to keep temperatures down using the same 90mn die size compared to that of the 90nm Intel CPU's.

Cache Cut: AMD Sempron 64 reduces the amount of L2 cache that is used on the CPU die. Athlon 64 based CPU's use either 512KB or 1MB of on die L2 cache. Sempron only supports 256K L2 cache onboard. This results in a slightly slower CPU clock for clock.

HyperTransport Technology: AMD first brought out HyperTransport technology back when Athlon was Socket A. With the advent of an on-die memory controller and basic Northbridge, AMD has decided to use Hypertransport to connect the K8 architecture CPU's to external chips.

Intel Celeron-D goes a different way from the AMD side of things with its line of CPU.
Netburst Architecture: Netburst has been the base of Intel Pentium 4 CPU's since its introduction. Netburst uses an extremely long pipeline with double speed ALU's. While this allows Intel to increase the speed of its CPU on a MHz race, keeping the pipeline full is a hard job and results in less work being done per clock cycle. Celeron-D uses a 133MHz Quad Pumped bus (533MHz) that connects the CPU to its Northbridge.

EM64T: While not named AMD64, Extended Memory 64 Technology (EM64T) is a direct copy of the AMD64 instruction set. Intel introduced this into its Pentium 4 600 series CPU, and has now filtered down to the Celeron-D range for the budget line. Celeron-D CPU's with a 1 or a 6 at the end of its identification number indicate its support for 64-bit instructions (331 and 336 are 64-bit compatible).

SSE3: Celeron-D CPU's are based on the Prescott core design, with this comes the longer pipeline and the SSE3 instruction sets. These are introduced to give the Netburst architecture a leap when it comes to heavy media encoding

90nm Strained Silicon: Rather than follow AMD down the SOI road, Intel decided to go Strained Silicon. This technology for its 90nm has been somewhat controversial due to its high voltage leak capacity as well as the extreme heat 90nm Intel CPU's generate on a performance per watt basis.

Cache Rise: Since the Celeron introduction, all Celeron CPU's have had a 128K L2 cache. Celeron-D increases this to 256K L2 cache. This is the highest cache level any Celeron CPU has experienced.

As you can see, there is quite a difference in architectures.

Processors in Detail

Our Test Processors

Before we get into the specifics of the tests we thought it might be good to show the test candidates.

Here we show the top side of the CPU's. Both use an alloy heatspreader on the top which aids in two respects. First is to transfer heat evenly. The die itself on both CPU's is rather small and fragile. In order to increase the amount of surface area, a large alloy block is placed on the die of the CPU. This means the die has to heat up a greater area before it overheats, however, having an extra piece of metal between the die and heatsink does mean less heat transfer - it's a catch 22 situation.

The second reason is the die's strength or lack there of. As many know first hand (including me) the die of the CPU, when exposed, is very fragile. AMD Athlon XP owners know this all too well, as even a small chip when installing the heatsinks leads to CPU death. Adding a heatspreader prevents the heatsink damaging the die even when large pressure is on the CPU from the heatsink.

Flipping over the CPU's we see a contrast of differences. First off AMD uses a 754 pin layout with a brown PCB. No resistors or capacitors are exposed, just a small gap in the middle. Intel Celeron-D however is built on the Socket 775 format. This means there are no pins on the CPU at all to bend. When dealing with micro pins like AMD, Intel has the advantage of not having any pins on the CPU to damage, only on the motherboard. Overall we do prefer a pin-less setup on the CPU as it's more practical to replace a cheap motherboard rather than an expensive CPU if a pin is to be broken or bent.


For this article, we are going to test the CPU's both stock, and both overclocked to their highest level reached. Sempron has one advantage over Intel Celeron-D, with the Sempron you can change the multiplier from its default down; this means you can clock the bus up to high levels while still keeping the CPU close to a sane clock speed. Celeron-D however doesn't give this option, so you are stuck with your default multiplier.

With our Sempron sample we clocked it from 1.8GHz to 2.13GHz using a ratio of 9x and a FSB of 236MHz. For our Celeron we clocked it from 2.8GHz to 3.17GHz using the default 21x ratio and a FSB of 151 or 604 MHz.

Benchmarks - Test Systems and Sandra

AMD Sempron Test System

Processor: AMD Sempron 3000+ (1.8GHz)
Memory: 2x 512MB Corsair DDR-533
Hard Disk: 2x Seagate 7200.9 (Supplied by Seagate)
Motherboard: EPoX 8NPA SLI
Graphics Card: ATI Radeon X800XT Platinum (Supplied by ASUS)
Operating System: Microsoft Windows XP SP2

Intel Celeron-D Test System

Processor: Intel Celeron D 336 (533MHz FSB) (Supplied by Intel)
Memory: 2x 512MB Corsair DDR2-800
Hard Disk: 2x Seagate7200.9 (Supplied by Seagate)
Motherboard: ASUS P5AD2-E Premium (Supplied by ASUS)
Graphics Card: ATI Radeon X800XT Platinum (Supplied by ASUS)
Operating System: Microsoft Windows XP SP2

We choose these particular setups as they were somewhat in the higher than average range of the performance specs for both platforms. Intel's Celeron-D was populated with Dual Channel DDR-2 memory, as this is a feature of the provided chipset.

While we did use two modules in the AMD system, they aren't running Dual Channel, since Socket 754 doesn't provide for this feature.

Let's begin our benchmarks and find out which platform offers the best performance.

SiSoft Sandra

Version and / or Patch Used: 2005 SR3a
Developer Homepage:
Product Homepage:
Buy It Here

SiSoft Sandra (System ANalyser, Diagnostic and Reporting Assistant) is a synthetic Windows benchmark that features different tests used to evaluate different PC subsystems.

Here we can see that the Sempron CPU lacks in the synthetic tests compared to the Celeron-D and its Prescott core, however, this isn't a measure of real world power.

Benchmarks - PCMark


Version and / or Patch Used: 2005 Build 1.1.0
Developer Homepage:
Product Homepage:
Buy It Here

PCMark is a multipurpose benchmark, suited for benchmarking all kinds of PCs, from laptops to workstations, as well as across multiple Windows operating systems. This easy-to-use benchmark makes professional strength benchmarking software available even to novice users. PCMark consists of a series of tests that represent common tasks in home and office programs. PCMark also covers many additional areas outside the scope of other benchmarks.

While the Celeron-D comes out ahead in both CPU and memory scores, the end results between the two aren't that far apart. This shows Sempron does have some hidden power.

Benchmarks - Super PI

Super PI

Version and / or Patch Used: 1.1e
Developer Homepage:
Product Homepage:

Developed by some folks from the University of Tokyo (yes, Japan), Super PI is a small utility that does just as the name implies. It figures PI to a set number of decimal places. Since PI is an infinite number to the right of the decimal point, the utility measures the time it takes to figure a set number of places. It runs the calculations a set number of times and gives a time for the completion of the task. This is a simple and effective way to measure the raw number crunching power of the processor being used to compile the results.

Super Pi shows us that the Sempron despite its lower clocks is able to crunch the Pi numbers quicker than Celeron-D with its longer pipeline.

Benchmarks - Media Encoding

High Quality MPEG-4 Encoding

For this test we used Cyberlink Power Producer in conjunction with DIVX 6.0Pro to encode "The Matrix" MPEG2 down to a MPEG-4 High resolution video file.

The Cyberlink software comes with its own time clock to indicate the time taken to encode the movie as well as average CPU usage and average FPS during the encode. We used DIVX 6.0 due to its support for Dual Core CPUs as well as Intel SSE3 and Hyper Threading.

Media Encoding is still best handled by an Intel CPU, but the difference isn't much at all. In fact we wouldn't really recommend any of the low-end CPU's for daily encoders.

Benchmarks - 3DMark03


Version and / or Patch Used: Build 360
Developer Homepage:
Product Homepage:
Buy It Here

By combining full DX8 and partial DX9 support with completely new tests and graphics over the previous version, 3DMark03 continues the legacy of being the industry standard 3D benchmark.

Please Note: Due to recent events with the 3DMark03 series, we are adding results purely for those who are still in favor of 3DMark03. These results should not be taken too seriously and are only added for interest sakes.

Here we start to see that the Sempron with its K8 architecture is starting to win the gamers crown.

Benchmarks - 3DMark05


Version and / or Patch Used: Build 120
Developer Homepage:
Product Homepage:
Buy It Here

3DMark05 is now the second latest version in the popular 3DMark "Gamers Benchmark" series. It includes a complete set of DX9 benchmarks which tests Shader Model 2.0 and higher.

For more information on the 3DMark05 benchmark, we recommend you read our preview here.

Again Sempron takes the lead.

Benchmarks - 3DMark06


Version and / or Patch Used: Build 102
Developer Homepage:
Product Homepage:
Buy It Here

3DMark06 is the very latest version of the "Gamers Benchmark" from FutureMark. The newest version of 3DMark expands on the tests in 3DMark05 by adding graphical effects using Shader Model 3.0 and HDR (High Dynamic Range lighting) which will push even the best DX9 graphics cards to the extremes.

3DMark06 also focuses on not just the GPU but the CPU using the AGEIA PhysX software physics library to effectively test single and Dual Core processors.

In 3DMark06 Sempron extends its lead over the Celeron-D.

Benchmarks - Doom 3

Doom 3

Version and / or Patch Used: Unpatched
Timedemo or Level Used: Custom Timedemo
Developer Homepage:
Product Homepage:
Buy It Here

Doom 3 is the latest game to hit our test lab and is one of the most intensive games to dates. With our own custom time demo we are able to give a realistic rating on what kind of FPS you will be achieving.

For more information on benchmarking Doom 3 we recommend you check out our extensive article regarding it here.

Sempron again holds the leaders position.

Benchmarks - Quake 4

Quake 4

Version and / or Patch Used: Unpatched
Timedemo or Level Used: Custom Timedemo
Developer Homepage:
Product Homepage:
Buy It Here

Quake 4 is one of the latest new games to be added to our benchmark suite. It is based off the popular Doom 3 engine and as a result uses many of the features seen in Doom. However, Quake 4 graphics are more intensive than Doom 3 and should put more strain on different parts of the system.

Sempron still continues to dominate the gaming sections.

Benchmarks - F.E.A.R.


Version and / or Patch Used: Unpatched
Timedemo or Level Used: Custom Timedemo
Developer Homepage:
Product Homepage:
Buy It Here

F.E.A.R. (First Encounter Assault Recon) is an intense combat experience with rich atmosphere and a deeply intense paranormal storyline presented entirely in first person. Be the hero in your own spine-tingling epic of action, tension, and terror...and discover the true meaning of F.E.A.R.

In our last gaming benchmark, we see that Sempron based system leads once again.

Final Thoughts

It is clear that the battle between AMD and Intel is at a hot point. Intel is still hurting with its losses to AMD over the Athlon vs. Pentium battles and not only does it have to fight this one, but now it has a competitor for the value segment, the AMD Sempron.

Intel's value market was unchallenged for some time with the demise of Duron, but now it's starting to show again that AMD can put its muscle into both high-end and value segments, and still give the users what they want.

Overall the Sempron proved itself to be the better CPU, while lagging behind in the encoding and synthetic segments, when it starts to put its K8 architecture behind games, it's clearly the winner overall.

Today we have shown what budget can do and what both companies have on offer. Hopefully we will see the Sempron evolve to include some of the higher functions like DDR-2 later on this year and give Intel something more to worry about.

PRICING: You can find products similar to this one for sale below.

USUnited States: Find other tech and computer products like this over at

UKUnited Kingdom: Find other tech and computer products like this over at

AUAustralia: Find other tech and computer products like this over at

CACanada: Find other tech and computer products like this over at

DEDeutschland: Finde andere Technik- und Computerprodukte wie dieses auf

Newsletter Subscription
We openly invite the companies who provide us with review samples / who are mentioned or discussed to express their opinion. If any company representative wishes to respond, we will publish the response here. Please contact us if you wish to respond.