As we've seen here at VL, the Athlon 64 (A64) is one speedy CPU. Although there is no official 64-bit version of Windows, the chip can run 32-bit code natively. For those who are anti-Microsoft, there are some 64-bit versions of Linux, but if you're feeling adventurous, you can always pick up a beta of Windows XP 64-Bit Edition.

Though the Socket-940 and Socket-939 A64 and FX CPUs are still incredibly pricey, the Socket-754 A64 comes in at a much more reasonable price. Being a poor student on a budget, that doesn't mean I want a substandard CPU for my gaming needs. I picked up an A64 3000+ for a great deal, and today I'll be comparing it with the A64 3200+. Why? I'll explain that shortly.

The A64 has already been covered in detail here, but I'll cover a few highlights of the retail version of the 3000+ and get right into testing.

The Athlon 64 3000+

The box packaging is the same cellophane as we've seen with previous AMD CPUs. Inside are the instruction manual, warranty papers, case sticker, the CPU and a copper based heatsink. The heatsink has a thermal pad pre-applied, which was removed since after a couple days of testing as I wasn't terribly satisfied with the performance.

The A64 3000+ is physically identical to every other Socket-754 A64. A heatspreader protects the core by distributing the heat evenly, and to protect the CPU during heatsink installation.

Above we see the markings for our particular CPU. Though our test CPU is a 3000+, the clock speed is 2GHz (10x200). In case you're wondering, the answer is yes, it is exactly the same clock speed as the A64 3200+. All Socket-754 A64s share the same 0.13 micron silicon-on-insulator technology (SOI), 128KB L1 Cache, a transistor count of about 105.9 million and 193mm2 die size.

How the 3000+ (codenamed NewCastle) differs from the other Socket-754 ClawHammer based CPUs is the L2 cache is 512KB, which is half of its more expensive brothers. The difference in Performance Rating between the 3000+ and 3200+ (ClawHammer) can be attributed to the 512KB of L2 Cache (3200+ has 1MB).

Test Setup

MSI K8T Neo-FIS2R: Athlon 64 3000+ (10x200: 2GHz), 2 x 512MB Kingston HyperX PC4000, HIS Radeon 9600 XT, 80GB Western Digital SE 8MB Cache, Windows XP SP1, VIA Hyperion 4in1 drivers 4.51, ATI Catalyst 4.7.

MSI K8T Neo-FIS2R: Athlon 64 3200+ (10x200: 2GHz), 2 x 512MB Kingston HyperX PC4000, AIW Radeon 9600 XT, 120GB Western Digital SE 8MB Cache, Windows XP SP1, VIA Hyperion 4in1 drivers 4.51, ATI Catalyst 4.7.

The A64 3200+ was chosen because the clock speed matches the 3000+, and it costs over 50$ more on average. Although 50$ may not seem like all that much, I can think of better uses for starving students, and that's beer and shooters. In all seriousness, I'm pretty frugal by nature, so hey, 50$ saved is 50$ in your pocket. I was going to include some Athlon XP numbers, but I think Hubert covered it enough in his 3200+ review.

We've tried our best to match the two platforms evenly, with the only difference being an 80GB Western Digital SE and 120GB Western Digital SE and the 9600 XTs. For the hard drives, we set the primary partition (where all benchmarks were done as well) to 60GB, and both video cards run at the same default core/memory speed. The ram settings were configured to 2-3-3-5, and the video card drivers set to "Performance" in the control panel for the application benchmarks.

We selected the ATI 9600 XT VPU as the video card for a couple reasons. It's the best video card I currently have (:P), but also because the price point works into our theme of a reasonably priced setup for serious gamers.

Note that for our video game tests, we'll be dropping the typical 640x480 tests in favour of 1024x768 through 1600x1200 at maximum detail, but no AA/AF enabled. We feel these resolutions and IQ settings are what most gamers will use. Unlike the application benchmarks, sound will be enabled throughout game testing.

All our benchmarks were run on a 32-bit version of Windows XP, five times total with the average displayed in this review, except in the case of SiSoft Sandra where we took the score that resulted the most often (three times in this case). I understand the review title hints that the focus will be on gaming, and we'll cover that on the next page, but I'm sure some people would like to see standard application performance numbers. We'll start with those first...

SiSoft Sandra 2004

Although a synthetic benchmark, it's a popular one, freely available if you wish to make comparison benchmarks. We will be testing the CPU, MMX, and memory speeds, using the 32-bit 2004 version.

CPU	MMX	Memory

CPU Arithmetic Benchmark

	Whetstone FPU	Dhrystone ALU
Athlon 64 3200+	8672	3168
Athlon 64 3000+	8382	3167

The extra L2 cache is probably the difference maker in the floating point test, but the arithmetic performance is almost exactly the same.

CPU Multimedia Benchmark

	Integer aEMMX/aSSE	Floating-Point iSSE2
Athlon 64 3200+	18783	20570
Athlon 64 3000+	14974	19774

The 3200+ is clearly the winner here, though unlike the CPU tests, the scores are almost reversed for the 3000+ where the MMX floating point does better than the arithmetic.

Memory Benchmark

	Int Buffered iSSE2	Float Buffered iSSE2
Athlon 64 3200+	3057	3058
Athlon 64 3000+	3036	3036

Although the 3200+ holds the lead again, the 3000+ scores very well in comparison. Enough so, that I'm ready to say the difference here is negligible.

ZD Content Creation 2004

The ZD Content Creation suite is a script that runs a series of actions and calculates a final score that measures a PC's overall performance.

	Score
Athlon 64 3200+	29.8
Athlon 64 3000+	27.7

Another close call for the 3200+, as it's barely keeping ahead of the 3000+. Just over two points separates the CPUs, and in real-world use, I doubt many will see the difference.

PiFast

A good indicator of CPU/Motherboard performance is PiFast version 4.2, by Xavier Gourdon. We used a computation of 10000000 digits of Pi, Chudnovsky method, 1024 K FFT, and no disk memory. Note that lower scores are better, and times are in seconds.

	Time in Seconds (lower is better)
Athlon 64 3200+	48.24
Athlon 64 3000+	49.34

The 3200+ is just under one second faster. For a "budget" CPU, things are looking great for the 3000+ so far.

CDex Audio Conversion Wav to MP3

CDex was used to convert a 414MB Wav file to a 320kbs MP3.

	Time in Minutes:Seconds (lower is better)
Athlon 64 3200+	1:33
Athlon 64 3000+	1:37

The 3200+ is the clear winner here, being about four seconds faster than the 3000+. I'd hardly call it a wipeout, but use a larger file, and those seconds will become minutes. Of course, we're talking about a HUGE WAV file in that case.

TMPGEnc 2.521

We used an Animatrix file, titled The Second Renaissance Part 1, and a WAV created from VirtualDub. The movie was then converted it into a DVD compliant MPEG-2 file with a bitrate of 5000. Times are in minutes:seconds, and lower is better.

	Time in Minutes:Seconds (lower is better)
Athlon 64 3200+	3:58
Athlon 64 3000+	4:01

Another victory here for the 3200+, leading the 3000+ by two seconds in video encoding. Like our WAV conversion, use a larger file, do the math and the end result will paint a different picture. Still, not bad for a much cheaper processor.