The latest round of high-def video games for PCs – namely Call of Duty: Modern Warfare 3, Battlefield 3, and the impending launch of Star Wars: The Old Republic – have saved graphics chip maker Nvidia from being slammed by a slowdown in PC sales. In fact, explained Nvidia in a conference with Wall Street analysts call going over …
BIGGER does not necessarily mean BETTER
Servers like fast threads sometimes, and server applications these days usually need more than the 4GB of main memory
Do they? Let's not mistake Vmware and Xen needs with applications. If the 64 bit arm retains the power envelope of its 32 bit predecessor we will most likely see more physical servers or physical nodes in a multi-server chassis like the SeaMicro AtomSmasher. An arm server node in its entirety is likely to cost a fraction of the cost of the licenses for running the same workload on a Vmware partition.
So it depends whose market will arm/nvidia decide to eat. They may go for Intel's jugular and I honestly wish them luck on that one. They may however go for the low hanging fruit of re-physicalising the virtualised installations.
Ah, those expensive days, I remember them well...
So, the *catch-up* continues as it has since GLQuake was first released way back in 1997.
I recall it well, as I splurged out what was for me, then, a vast sum of money on a Voodoo1 video card.
That's when my personal literal armada of graphics cards started. Matrox G200 (fail), TNT1 (awesome - got several of those), TNT2 M64 (fail), several other TNT2 cards after that and then a vast succession of GeForce cards.
Some 10 years on and a few thousand quid later, I finally gave up playing catch-up and am currently stuck with a couple of 8600GT cards, which suit me just fine for now.
The amount of cash I spent on video cards far outstripped the amount of cash I spent on the games themselves.
It's ridiculous, but I was hooked.
Memory for ARM
Actually, all the ARMv7 chips I have (OMAP4, Tegra2) have 40-bit memory addressing, the same as Intel Xeons from a year or two ago. That means 2^40 as opposed to the 2^32 you stated simply because the registers are 32-bit. So that's a terabyte of RAM that could be addressed by it. The ARMv8 is going to have the 64-bit registers, better floating point (in fact possibly standardized, which is better and can hopefully put an end to soft-float being the standard), as well as the possibility of Mali/NVidia GPGPU stuff onboard. So while the future of ARM looks great, the present is better than people know.