Oh I am aware of that, and it it a real head scratcher. for every one that I can find that concludes that side of things, I find a group that finds the opposite
Well, PC Perspective only tested Sleeping Dogs, so it may be an isolated case...although I would think it should easy enough to test for if you have the right hardware, a copy of FRAFS and some game benchmark software.
Although I am trying to keep an open mind. It does seem odd (and I cannot ignore) that the "okey-dokey" was 1)done by group consensus.
A little harsh considering a lot of the European based tech sites have included qualitative as well as quantative measurements for years. Even mainstream English speaking sites have explored the issues in the past (e.g.
this from four years ago)
2) provoked a lot of people who were otherwise not complaining about this problem before this came along
Yes, quite the cause célèbre. Likewise, how many people that did complain about the issue got shouted down by the overprotective fanboys? Even now, you can find deniers that claim there were never issues and that AMD's
mea culpa and consequent driver revision are nothing more than a placebo to placate the press.
3) That it was quite a scoop to 'innovate ' a new and exciting metric for benchmarking.
Are you surprised when a new paradigm enters the technology arena and everyone jumps on it?
Once upon a time, efficiency, and even power consumption itself weren't even considered part of graphics card reviews.
Then you might also remember that I did a lot of digging (and did not stop until I got to the right person when I noticed that supertiling did not seem to forgo effects like advanced DOF
No reason why it should. DoF is Gaussian blur filter applied to a completed frame (computed at the end of the pipeline after pixel shading). Geometry and Tessellation are applied during the frames construction. Even simple geometry requires a certain amount of overdraw to allow congruency between adjacent pixel blacks for transformed vertices, so each GPU is responsible for mapping vertices not only of its own pixel blocks but those of the adjacent ones as well- a workload that increases markedly once you introduce tessellation- a huge increase in vertex sets
As an aside, of all the issues concerning SuperTiling, the one that bewilders me is that it doesn't work with OpenGL, yet the
original rendering workloads used by Evans and Sutherland were exclusively OpenGL.
346.41?? What the heck xD.
HH are always good for a laugh. They are also the only review site that seems disappointed with the 13.8b driver
So what do we make of that lot? Well overall it's hard not to be a little underwhelmed
Seems they were expecting some large performance increases, as opposed to fixing the frame pacing issue that everyone else was expecting (and thankfully received).
