http://www.blizzard.com/press/070724.shtml

World of Warcraft's Subscriber Definition
World of Warcraft subscribers include individuals who have paid a subscription fee or have an active prepaid card to play World of Warcraft, as well as those who have purchased the game and are within their free month of access. Internet Game Room players who have accessed the game over the last thirty days are also counted as subscribers. The above definition excludes all players under free promotional subscriptions, expired or cancelled subscriptions, and expired prepaid cards. Subscribers in licensees' territories are defined along the same rules.

http://www.azcentral.com/sports/suns/articles/0515suns-ON.html?&wired

It becomes more interesting.... 

http://www.youtube.com/watch?v=sc4yz__akIU&NR=1

To be honest, when I was watching the game, I didn't notice that happened. 

These numbers are highly questionable. The r600's technical specs point to a performance at least alot better then the x1950xtx so there has got to be something wierd going on. I guraentee you they will be alot better when real reviews come out in two-three weeks.TOAO_Cyrus1

Someone posted this at Dailytech:

"The R600 has 64x5 Vec5D units which maces each unit handle a maximum of 320 stream ops per second, but makes it's worst case scenario a lot worse with 64 stream ops per second (for the lack of a better term). You can think of that in the same manner as you see that SIMD units in our current processors can deliver huge amounts of processing power if, and only if, used correctly and optimized accordingly, otherwise we see no gains.



In my opinion AMD/ATI made a design compromise, they used this approach as it could prove to be way better in the dx10 world, and in a much more interesting way, in the GPGPU world.



Think about it, if you open up your architecture with CTM and give the people the power of 64x5 vec5d units you end up with an amazing amount of processing power. That's where I think they are focusing.



Nvidia has a much more favorable place in the gaming world. If you have 128 scalar units, in a worst case scenario you'd still issue 128 stream ops (all else constant, and given you have the bandwidth). But your best case scenario isn't that good.



I believe they delayed it because they were expecting dx10 games (of course, this is just speculation on my part). And I hope, for their sake, that it performs a lot better in that world.



Still, if I am somewhat right, drivers could provide better optimization for shader programs that aren't made with a simd architecture in mind, but then again, I could be entirely wrong.

http://forums.anandtech.com/messageview.aspx?catid=31&threadid=2039097&frmKeyword=&STARTPAGE=8&FTVAR_FORUMVIEWTMP=Linear