View Single Post
  #13 (permalink)  
Old February 9, 2012, 02:12 PM
Dragonstongue Dragonstongue is offline
MVP
 
Join Date: Sep 2011
Posts: 432

My System Specs

Default

Quad 7970 would beat the tar out of Mars II. Mars II are very fast, they most surely are, but they do run into the problem of not having enough Vram(you want at least 3gb, 1.5gb can become a terrible memory capacity bottleneck) when pushed to the best performance they can deliver, SLI/Crossfire takes some Vram to operate the more linked the more Vram obviously running things like AA/AF and very high resolutions can tap out a cards potential performance because of running out of Vram long before they run out of potential horsepower. I guess, think of it like a huge motor with a very high boost turbo charger, if it runs out of fuel because the feul pump is unable to deliver enough, well then the motor/turbos can only give X of performance.

Nice overall system, but, I would be going dual 6990 or quad 7970/7950, in the case of 7970/7950 you can overclock them like mad, and they would really show what they have to deliver.

Course, if it were me, I would also go a 100% mass liquid cooled route. not a 360 rad, but rather an outside of case, self contained large capacity system, this way here you never run out of coolant, you have huge heat capacity, and might require less power as the system is cooler therefore more efficent in terms of energy consumption.

Wildfire SSD or Sammy 830s, a 1100+ watt power supply and a smaller say 550w dedicated to the liquid cooling side. EVGA SR2-x SR2 SR3 etc X79 based board, along with minimum 16gb DDR3 2000+ low timings low voltage memory

There are many ways to do this, but a "dream" system can be under $1k or +$15k all depends of course on how far you make the dream go.

Also on the cpu side, a i7 980X may have 6 core 12 thread, even overclocked it shouldn`t be bottlenecking, but there is only so much PCI-E bandwidth, the cpu itself is wow fast, but nothing is 100% perfect, and the way the MARS II are built, they have an internal NF200 bridge. So in a perfect world, it would not bottleneck ever, but do realize, hyper threading does not always give 100% scaling all it takes is for the computations to have a blip and it suffers, it could be anything IMO, heck it could be the MARS II not running proper or with enough speed.
Reply With Quote