Athlon 64 Vs. Pentium 4

AMD's Athlon 64 and Athlon 64 FX Vs. Intel's Pentium 4 Extreme Edition. We showcase the latest CPUs from AMD and Intel and take a look at what they will deliver.

Introduction

There's been so much information made public about AMD's new desktop parts over the last year and a half, that we intend to make our reiteration of the specs short and sweet and cut right to the info you really want to see. Please do review the specifications, as they will show you the similarities as well as the differences between all of the CPUs.

All things considered, AMD and Intel are sharing a couple of strategies today with their flagship CPUs. For one, they are both "simply" remarking a server grade CPU and calling it a desktop unit. They are also asking for many hundreds of dollars for the privilege of owning one.

AMD's Athlon 64 CPUs

AMD has been promoting their K8 CPU, better known as the Hammer series, for well over a year now. These desktop parts were postponed for about ten months while AMD's strategy changed a bit and began to focus on their Opteron server-class CPUs. Today, AMD is announcing their Athlon 64 3200+ CPU and their Athlon 64 FX 51. From this point on, we will refer to them as the Athlon64 and the AthlonFX to avoid confusion.

The "64" in their name comes from the ability of these new CPUs to run the AMD64 instruction set that will be supported by the upcoming Windows XP 64-Bit Edition operating system. Moving to a 64-bit operating system with a 64-bit CPU will have some obvious advantages, although it's clear that you will not be seeing many of them right away. A 64-bit computing experience means that our PCs will no longer be limited to addressing only 4GB of RAM. In a 64-bit environment, it's now possible to address 8 terabytes of memory. In order to make the Athlon64 CPUs 64-bit compliant, the appropriate registers have to be added to the CPU. This has only increased the die size of the processor by 1 to 2%. Going to 64-bit has actually not been that costly from the standpoint of physical implementation.

The real beauty here lies in the Athlon64's ability to run current 32-bit applications and operating systems right now, and with "no" penalty. When 64-bit applications and OSes come to market, the Athlon64 will be ready to take advantage of them and still be able to run native 32-bit apps right alongside the 64-bit ones. Of course, we're focusing on Windows operating systems, as they are by far the majority of the desktop market. From what we've seen, current 32-bit applications run just as well on the new 64-bit OS from Microsoft. Some run a bit slower, some run a bit faster, but overall it's pretty much the same. Most likely no tangible performance differences will ever be noticed. These new processors include a wide range of specialized instruction sets besides the 64-bit expansion, including AMD64, MMX, 3DNow!, SSE, and now SSE2.

Both of the new Athlon64 CPUs will utilize HyperTransport technology. They will have available one HyperTransport link that is 16-bits wide and can move 3.2GB/s of data in each direction simultaneously. This link is used to communicate with the northbridge controllers. Obviously this gives a large amount of bandwidth to the CPU to work with other devices on the mainboard. Currently the AMD Athlon 64 FX and Athlon 64 can communicate with the chipset at speeds of up to 1600MHz. Keep in mind that "up to" does not mean "always", as we will show later.

The memory controller is no longer on the northbridge chipset. Instead, AMD has taken control of this aspect of the system and has integrated the controller onto the CPU die itself. Latency has been reduced significantly due to this, so this certainly comes as a very welcome change. Because the memory controller is located on the processor die, the memory subsystem traffic no longer has to go through the chipset for CPU-to-memory calls. Hence, the old term "front-side bus" isn't really applicable any longer. With AMD64 processors, the speed at which the CPU and memory controller interface is at full processor frequency.

AMD Athlon 64 FX 51

This is currently AMD's flagship processor. Gone is the performance rating, though we do get a little more marketing. The CPU we tested was supplied by AMD. The "FX" and "51" are ambiguous at best, meaning nothing beyond the overused FX cliché. 51 is the model number, which will scale with newer processors. The next CPU will likely be a model 53, with the 53 obviously being faster than the 51.

Article Image Article Image Article Image Article Image

The packaging is a 940-pin lidded ceramic micro PGA. The lidded portion refers to the big metal heatspreader on top of the CPU, which is certainly a welcomed addition. This part runs at a 2.2GHz (11 x 200 CPU Clock) clock speed and supports Dual 64-bit DDR memory channels, which provides a 128-bit wide memory bus and 6.4GB/s of bandwidth working at DDR400 specifications. Fallback to DDR200 is supported. This AthlonFX will require the use of registered DIMMs. Remember that the memory controllers on the K8 are now on-die and fully in the control of AMD, not the chipset builders.

Many of you might be looking at this CPU and asking, "Isn't that an Opteron?" and you would be close to right. The AthlonFX is very much an Opteron in desktop clothing, right down to the pin count. The differences are that the Opteron is not specified to run DDR400, only DDR333, and officially two of the HyperTransport bus interfaces have been removed from the AthlonFX.

AMD Athlon 64 3200+

This will pan out to be AMD's mid-level CPU. It still carries a performance rating and picks up where the AthlonXP left off with the "3200+" moniker. It should be noted that AMD would not supply a sample of this CPU to be tested. Our connections in Taiwan supplied samples for testing.

Article Image Article Image Article Image

The packaging is a 754-pin lidded organic micro PGA. Again, we see the heatspreader. This part runs at a 2.0GHz (10 x 200 CPU Clock) clock speed and supports an integrated single 64-bit DDR memory channel providing 3.2GB/s of bandwidth working up to DDR400 specifications. It also has fallback to DDR200 as well.

AthlonFX and Athlon64 Similarities

Basically what you have seen listed above are the differences in our two new CPUs from AMD. Instead of writing everything twice, we've decided to give you a list of specifications that are shared between the two parts.

L1 Cache Size: 128KB L1 (64KB instruction + 64KB data)

L2 Cache Size: 1MB L2 (exclusive)

HyperTransport Links: Supports single link - 6.4GB/sec per link I/O bandwidth

Fab location: AMD's Fab 30 wafer fabrication facility in Dresden, Germany

Process Technology: 0.13 micron SOI (silicon-on-insulator) technology

Die Size: 193mm2

Transistor count: Approximately 105.9 million

Nominal Voltage: 1.50v

Certainly what stands out here is the massive L1 and L2 cache sizes that are packed onto the AthlonFX and Athlon64. Also, the die size might jump out at you as being massive as well. Certainly these are not inexpensive CPUs to build. We do see a reduction in voltage, which will be welcomed by the enthusiast.

There is an on-die thermal diode that will allow the die temperature to be monitored when used with the proper software. This diode will also be utilized to shut down the CPU clock if there is a point where the die becomes too hot. The Athlon64 CPUs are specified to work from -55°C to 85°C. So, in short, we shouldn't see any more burned up AMD CPUs.

To drive the point home even more on just how alike these two CPUs are, take a look at their block diagrams side by side. The memory interface is truly the only difference.

Article Image

AthlonFX / Athlon64 and AthlonXP Differences

This chart will lay out all of the changes from the current AthlonXP. Besides the obvious interface differences that have been discussed above, the transistor count stands out like a sore thumb. The transistor count, having nearly doubled, can easily be blamed on the on-die L1 and L2 cache increases. Those caches have basically doubled from the XP to the Athlon64 as well as the transistor count and size of the die.

Article Image

There has been a lot of talk about the K8 core, but AMD has spent very little time marketing the actual advancements of the CPU pipeline. Here is the official word from AMD:

Processor core clock-for-clock improvements, including larger TLB (Translation Look-Aside Buffers) with reduced latencies and improved branch prediction through four times the number of bimodal counters in the global history counter, as compared to seventh-generation processors. These features drive improvements to the IPC, by delivering a more efficient pipeline for CPU intensive applications. CPU-intensive games like Comanche 4 and Unreal Tournament benefit from these core improvements.

The benchmarks will surely show us the truth here.

Athlon64 & AthlonFX in Action

Below we have screenshots of how the CPUs look in action. Notice the expanded L2 cache. Do note that the AthlonFX is not running at stock clock in our screencap.

Article Image Article Image