AMD's New Athlon 64: Fast and Pricey

By Jason Cross  |  Posted 2004-10-18 Email Print this article Print
 
 
 
 
 
 
 

Review: AMD ships a new FX series processor, and renames the old one to the 4000+. Performance is terrific, but will you pay for one?

Despite all the recent talk about dual-core processors, it'll still be a year or more before they're widely available as desktop CPUs. The march of processor technology comes in many forms: Process size shrinking and materials changing, caches changing, major and minor architectural enhancements, and of course, clock speed increases. The most interesting new CPU launches contain several of these elements--perhaps a shrink from a .13 micron manufacturing process to .09 along with an increase in cache size and a tweak to the CPU's branch prediction logic. The least interesting are the mere clock speed increases, and that's what AMD has in store for us today.

Like the Highlander, there can only be one Athlon 64 FX processor on the market at a time. The Athlon 64 FX-53's introduction came with the halt in production of the FX-51. So, too, does today's release of the FX-55 mean the end of the line for the FX-53--but in name only: This time around, AMD has essentially renamed the FX-53 as a standard Athlon 64 chip, the 4000+.

The new 4000+ is made in exactly the same way as the FX-53, on exactly the same .13 micron manufacturing process, runs at the same 2.4GHz clock speed, has the same L2 cache size of 1MB, and is available only in socket 939 format. The one small difference between an Athlon 64 4000+ and FX-53; the FX line of chips allows the clock multiplier to be overclocked, while the normal line does not. Since we won't be overclocking our chips in this review, the two chips are interchangeable.

So what's new in the FX-55, then? Not much. AMD has managed to increase the clock speed another 200MHz, to 2.6GHz, keeping them a full 1GHz behind Intel's best. The FX-55 is made with the same .13 micron manufacturing process, and we don't expect to see FX chips made with the company's new .09 micron process until 2005.

Let's take a look at how our systems were set up for these tests. We tested two Athlon 64 CPUs, the new FX-55 and the 4000+, which is identical to the old FX-53 if you aren't overclocking. For comparison's sake we ran the numbers with Intel's latest CPU in one of the newest high-performance motherboards. As you can see, we tried to keep components similar between the two, but there are slight differences. The hard drive in our P4 rig supports Native Command Queuing, as does that motherboard chipset, so driver performance may be slightly higher there. Also, we used the best RAM we could find for that system, DDR2/533.

Component Athlon 64 system Pentium 4 system
Motherboard and Chipset Asus A8V Deluxe Intel D925XCVLK
Memory 2 x 512MB Kingston DDR400 2 x 512MB Kingston DDR2/533
Graphics Radeon X800XT (AGP) Radeon X800XT (PCIe)
Hard Drive Seagate Barracuda 160GB SATA Seagate 160GB SATA Drive with support for Native Command Queuing
Optical Drive ATAPI DVD-ROM Drive ATAPI DVD-ROM Drive
Audio Sound Blaster Audigy 2 Sound Blaster Audigy 2
Operating System Windows XP Professional with SP2 Windows XP Professional with SP2

Since this is just a small clock-speed bump from AMD's previous FX-53 chip, with no architectural improvements, we ran a somewhat abbreviated suite of benchmarks.

  • BAPCO SYSmark 2004, which replaces our now-obsolete Winstone suite;
  • PCMark04, to gauge memory and CPU performance;
  • 3DMark 2003 CPU Test as a memory and CPU check;
  • Windows Media Encoder 9 video encode, DivX 5.2 video encode, and WMA audio encode, to check throughput digital media encoding;
  • Four games: Unreal Tournament 2004, Painkiller, Doom 3, and Halo; and
  • SPEC Viewperf 8.01. This is a mostly graphics intensive test using recordings from a suite of professional 3D applications. But it hammers on the graphics interface and actually stresses both AGP and PCI Express a bit more than the game benchmarks do.

The new FX-55 chip shows a modest but appreciable difference in SYSmark performance, matching our Pentium 4 system very well. Pentium 4 CPUs tend to perform pretty well on SYSmark 2004, because the heavy multitasking environment makes good use of Hyper-threading technology. It's also a test in which hard drive performance can make a real difference, and we wouldn't be surprised at all if the Native Command Queuing feature of the SATA hard drive in our P4 system adds a few points to the scores. A 4 to 5% speed increase from AMD's former best CPU is nothing to sneeze at, though.

PCMark's synthetic tests tend to fit very well in a CPU's L2 cache, and so the higher clock speed of a Pentium 4 gives them a decided advantage. What's more interesting than the relative comparison between the Pentium 4 and FX-55 is the comparison between the FX-55 and FX-53. An 8% increase in clock speed has resulted in a 7% increase in overall PCMark score, and more than 8% increase in CPU score. Because Athlon 64 chips have an integrated memory controller that runs at the chip's core clock speed, the memory score has even gone up by about 4%. Those are impressive results, but it's improvements in real-world application performance that matter.

When it comes to digital media encoding, AMD's new CPU delivers excellent results. The clock speed of the FX-55 is 8% higher than the FX-53 or 4000+. Bearing in mind that a real-world performance increase that's 70% of the increase in clock speed would be considered good, this is absolutely fantastic. Better than that: It's what you could call "ideal." There's an 8% reduction in decoding times, which means that the change in clock speed delivers a 100%-efficient change in performance. You can't ask for more than that.

Our Viewperf 8.01 benchmark figure is the geometric mean of the eight benchmarks in the Viewperf suite. The new Athlon 64 shows about a 5% performance improvement. The speed of the video card and OpenGL driver optimizations can greatly affect these benchmarks, so 5% from a CPU with an 8% faster clock speed alone is actually a pretty good figure.

Since the introduction of the Athlon 64, its strong suit has been games. Obviously, most game benchmarks depend greatly on video card speed. We used one of the fastest cards on the market, a Radeon X800 XT, and ran our games at 640x480 to minimize the video card's impact. The results are encouraging: The FX-55 is anywhere from 5% to 7% faster than the FX-53 / 4000+.

The gap between the best Athlon 64 and the best Pentium 4 in games is huge. As resolution is increased, these numbers might fall off due to a greater reliance on video card performance. On the other hand, actual gameplay tends to be more demanding on the CPU than benchmarks. The long and short of it is that you can expect the previous fastest gaming chip on the block, the FX-53, to be beaten by the FX-55 with a margin of around 5% to 8% in CPU-heavy games.

CPU launches that are merely a boost in clock speed sure aren't sexy, but they're the bread and butter of the processor industry. The huge investments made in new CPU architectures have to pay off with excellent CPU scalability, allowing chips to achieve not only higher clock speeds, but ensuring that those clock speeds translate into real performance and don't come with major engineering overhead.

In a way, it's impressive that AMD has managed to get the Athlon 64 architecture up to 2.6GHz without moving to a 90 nanometer manufacturing process. Maybe that's an impressive feat, but we would still prefer to see the crown jewel FX series processors made with that process. Since the FX series doesn't have locks on the CPU multiplier value, overclockers would probably see the most benefit from the cooler-running 90 nanometer chips.

However, this increase in clock rate comes at a price. The new CPU has a maximum thermal rating of 104 watts, almost 17% greater than the FX-53. AMD shipped the CPU with a beefier CPU cooler than we've seen in the past, to account for the higher thermal output. This may seem like a huge thermal increase, and it is: AMD needs to move to 90nm if it wants to continue pushing the Athlon 64's clock speed. We should note that the Pentium 4 560, already made using a 90nm process, carries a maximum thermal rating of 115W.

Ultimately, it's all about performance increases. In that respect, the Athlon 64 architecture continues to shine. All other things being equal, these chips have demonstrated nearly 100% clock scaling efficiency in many cases. Increase the clock speed 8%, performance goes up 8%. It sounds intuitive, but in the past it has not usually been the case.

The FX-55 chip is very expensive, taking the over-$800 spot previously occupied by the FX-53. That chip, now the Athlon 64 4000+ with a multiplier lock, drop to $729 in volume, still quite expensive. We're not big advocates of spending well over $500 for a CPU--that's just not where you get the most bang for your buck--but in that very high end, extreme, no-holds barred price range, AMD's Athlon 64 FX series continues to impress. If you're thinking about spending that much money on a CPU, this one comes highly recommended.

Product Athlon 64 FX-55
Web Site: www.amd.com
Pros: Fantastic game performance; performance increase in CPU-limited tasks is almost 1:1 with increase in clock speed.
Cons: Expensive; still made with a .13 micron process.
Summary: AMD's impressive CPU architecture shows its strengths by achieving very real performance boosts even with a simple increase in clock speed.
Price: $827 (in volume)
Score:

Product Athlon 64 4000+
Web Site: www.amd.com
Pros: Great performance, it's like an FX-53 only a little cheaper
Cons: Still costs over $700, multiplier lock
Summary: If you wanted an FX-53 but weren't planning to overclock it, now you can get it for about $100 less.
Price: $729 (in volume)
Score:
 
 
 
 
Jason Cross Jason was a certified computer geek at an early age, playing with his family's Apple II when he was still barely able to write. It didn't take long for him to start playing with the hardware, adding in 80-column cards and additional RAM as his family moved up through Apple II+, IIe, IIgs, and eventually the Macintosh. He was sucked into Intel based side of the PC world by his friend's 8088 (at the time, the height of sophisticated technology), and this kicked off a never-ending string of PC purchases and upgrades.

Through college, where he bounced among several different majors before earning a degree in Asian Studies, Jason started to pull down freelance assignments writing about his favorite hobby—,video and computer games. It was shortly after graduation that he found himself, a thin-blooded Floridian, freezing his face off at Computer Games Magazine in Vermont, where he founded the hardware and technology section and built it up over five years before joining the ranks at ExtremeTech and moving out to beautiful northern California. When not scraping up his hands on the inside of a PC case, you can invariably find Jason knee-deep in a PC game, engrossed in the latest console title, or at the movie theater.

 
 
 
 
 
























 
 
 
 
 
 

Submit a Comment

Loading Comments...
























 
 
 
 
 
 
 
 
 
Thanks for your registration, follow us on our social networks to keep up-to-date