Not denouncing their benchmarks entirely - I believe Intel's benchmarks of their own chip are valid, and I think they have a winner on their hands.
But someone evaluating them raises some real questions about the comparison: http://voodoopc.blogspot.com/2006/03/if-only-they-had-time-machine.html You'll notice that the image I am referring to on Anandtech's website (the bios image) states that the AMD processor is "unknown" which makes me believe that the bios they are running is outdated. So, I did a bit of digging and low and behold, the DFI bios version "D49C-32" they are running is from 10/11/05. There has been 1 major revision with major fixes that include: Set Cool 'n' Quiet Default to Disabled - With Cool & Quiet enabled, AMD processors will throttle in order to save power and bring their thermal load down. This means the processor could be running as low as 800MHz in certain programs - no matter what the program is. In theory Cool & Quiet is supposed to throttle up to maximum in games but this is not always the case. No enthusiast PC goes out with Cool & Quiet enabled unless it's a fanless machine or media center. Add Support for AMD Athlon 64 FX60 CPU - According to DFI the FX-60 will not operate correctly without this bios update. Without official support for the FX-60 CPU I'm not sure what we're comparing against here. Fix Memory Timings 2-1-1-1-1 and 4-1-1 Mode Wrong & Fix Read Preamble Table Error. - Memory latency can make a massive difference in performance. If the latency was not running at the correct latency we can see a pretty big difference in all kinds of performance. Anandtech stated "The AMD system used 1GB of DDR400 running at 2-2-2/1T timings." Apparently this isn't the case, but they would not be able to tell without having the platform in house. Fix Fill 3114 SVID&SSID under Cross fire mode. - More apparent performance issues under Crossfire mode. Next, when you take a future Intel chipset and compare it to a chipset that no enthusiast supports (RD480) with an outdated bios it's like taking a Ferrari and putting it on Bias-Ply tires. It's just not a good way to show off a "new" technology. If we go and check out the numbers on Anandtech we'll see the Unreal Tournament 2004 benchmark showing 160fps on the unknown AMD X2 processor while the Intel Conroe at 2.66GHz came in significantly higher at 191fps. Though this isn't exactly conclusive, if you go back and re-read some old FX-57 reviews on Tom's Hardware you'll see a benchmark for the same game set at the same resolution (and the same color depth), the FX-57 running at 2.8GHz scored 183.4fps. The thing is it's using an Nvidia Geforce 6800 GT which seems to me that there are many variables here when it comes to benchmarking. Perhaps it's somewhere locked in the settings, but I won't know until I sit down and compare our own benchmarks with consistant settings. Note that a single core Athlon 64 4000 achieved a better score in the benchmark run by Tom (160.5fps) than the one provided by Intel (160.4) at IDF. Like I said, I don't view this as conclusive, but it shows that there are variances depending on how the benchmark is setup. Here is a link to Tom's review.
