Over the last 18 months, graphics adapters have evolved probably more than in the last 5 years. DirectX 10, regardless of whether it makes a difference or not in gaming has pushed both nVidia and AMD to come out with a whole slew of new graphics processors and implement them in the latest lineup of graphics cards. SLI and Crossfire capabilities are not even worth talking about anymore since their level of acceptance has made both of them essentially household items by now, at least in the high-end gaming community. Let’s take a quick step back … about a year to remember the introduction of nVidia’s GeForce 8800 Ultra featuring as much as 768 MB. 6 months ago came the possible 3-way SLI configuration  and AMD’s dual GPU Radeon 3870 X2, essentially combining two separate graphics cards on a single board and connecting them internally through a PCIe link.

Finally, with a little help from Anand's article, we were able to shed some light on some of the erratic performance numbers that have made our Phenom benchmarking life miserable from the beginning. We might actually have figured out some of what is going on... and share it!

The latest re-spin of AMD's multicore die fixes the bug known from Erratum 298 and opens up a bit of overclocking headroom. Gone is also the requirement for a BIOS fix causing the known performance hit. To make things even more appealing, running in native performance mode actually unleashes better performance in memory-intensive applications than what AMD Overdrive ever brought back after the patch. All in all, AMD seems to be on the right track with the latest -50 series

AMD's release of the Phenom processor along with the Spider platform was not exactly a head turner. More precisely, if heads actually turned, they probably went the other way. Ever since, there has been a lot of confusion, benchmarks chasing each other, whether they were relevant or not and by the end of the day, the entire platform is still a mystery. We took the current flagship edition, that is, the Phenom 9900 and pitted it against the Phenom 9600 Black Edition with the lower speed NorthBridge. We ran AOD and different memory frequencies and then compared the results with whatever competition is out there. We also did some more research into Erratum 298 and the bug it describes along with possible scenarios where it can occur and the workaround solution a.k.a. patch that could be worse than the bug itself - and not just in terms of performance.