Sign in with
Sign up | Sign in
Your question

What is most important for gpu?

Last response: in Graphics & Displays
Share

what part is the most important in gpu design?

Total: 10 votes (7 blank votes)

  • raster pipeline(more rop)
  • 17 %
  • memory bus width
  • 17 %
  • ram speed
  • 0 %
  • core clockrate
  • 17 %
  • shader speed
  • 0 %
  • shader number count(more shader)
  • 0 %
  • wiring and architecture setup
  • 34 %
  • texture mapping unit (more tmu)
  • 17 %
  • floating point/computing power (for F@H)
  • 0 %
  • i/o interface/ramdac/ media decoding power
  • 0 %
December 8, 2010 7:45:28 PM

since 6970 is on the block....i was debate with some stupid hothead in techpowerup a while ago about what is most important in gpu design, debating the possibility of 512bit bus and 64rops...however they are insist shader power(more shader!!!) over rasterization(steroid of rop/tmu/bus) and they were saying future gaming will be more shader heavy because of limited console hardware that can't output higher texture quality and better pixel modeling(as they mentioned: console gaming taking place over pc gaming), while rops/tmu can be push by clock insanely high to save die space for more shader!! ram can also be push with higher frequency as well....which i brought the example of gtx 480 that beating 5870 in most of benchmark because of 48rop and 384bit bus....then i got kicked....

can you guy tell me are these idiot's theory are right? basically this is another war between clock speed vs architecture.......

please discuss

More about : important gpu

December 8, 2010 8:24:25 PM

cheesesubs said:
since 6970 is on the block....i was debate with some stupid hothead in techpowerup a while ago about what is most important in gpu design, debating the possibility of 512bit bus and 64rops...however they are insist shader power(more shader!!!) over rasterization(steroid of rop/tmu/bus) and they were saying future gaming will be more shader heavy because of limited console hardware that can't output higher texture quality and better pixel modeling(as they mentioned: console gaming taking place over pc gaming), while rops/tmu can be push by clock insanely high to save die space for more shader!! ram can also be push with higher frequency as well....which i brought the example of gtx 480 that beating 5870 in most of benchmark because of 48rop and 384bit bus....then i got kicked....

can you guy tell me are these idiot's theory are right? basically this is another war between clock speed vs architecture.......

please discuss

I would have to say its all about the architecture, for example a Pentium 4@3.0Ghz and a Core i7@3.0Ghz let say they both have the same Cache and FSB over all the i7 is faster because of its architecture/design or even say a HD4870 and HD5870 for example they both have the same memory size, clock and BUS...etc..etc the HD5870 is faster/better because of its new architecture.
December 8, 2010 8:28:28 PM

timmy121 said:
I would have to say its all about the architecture, for example a Pentium 4@3.0Ghz and a Core i7@3.0Ghz let say they both have the same Cache and FSB over all the i7 is faster because of its architecture/design or even say a HD4870 and HD5870 for example they both have the same memory size, clock and BUS...etc..etc the HD5870 is faster/better because of its new architecture.


cypress is better than r770xt is because it has double of rop/tmu/shader count. even under the same frequency/ram size/bus width 5870 will destroy 4870.....even 4870 clock very high 5870 still beat it even down clock alot.
December 8, 2010 8:32:29 PM

cheesesubs said:
cypress is better than r770xt is because it has double of rop/tmu/shader count. even under the same frequency/ram size/bus width 5870 will destroy 4870.....even 4870 clock very high 5870 still beat it even down clock alot.


Yes that's is what I'm saying, even if the 4870 and 5870 had the same numbers in spec the 5870 will beat it......all about the architecture.
!