Originally Posted by szeged
they need to, but they arent going to. They always overclock new cards and leave other cards at stock and claim victory for the new card.
Don't agree with you here. What I always see at nearly every website is testing everything at stock, and THEN devoting one page to show the overclock potential of the new GPU. For instance, Guru3d tested the 280X reference and compared it against reference cards from NV. Then they devoted one page to overclocking to show the potential of the card - this is what most websites do.
As seen here:
Now, post LAUNCH they will test pre overclocked models and compare it to the reference 280x. As an example, they may test a DC II 280X and compare it to reference. This is similar to how the 680 was reviewed - at launch, it was all stock clock GTX 680s. Then a couple weeks later, the factory overclocked models were tested to show how much faster they were versus other cards *and* the stock 680. This is the formula that most websites follow, and is a fair testing methodology for launch cards - launch cards for *new GPUs* are *usually* tested at stock clocks. Again - with one page dedicated to OC potential.
Like I said, though, they may test pre overclocked models such as the MSI lightning - but for new launch SKUs they test reference across the board for the most part with one page dedicated to OC potential of the new GPU architecture. Now, for what I like to call the "nutty insane" type of testing - where someone does a 2 minute suicide run in unigine or 3dmark firestrike using 16 gabillion volts - you won't find stuff like that on most websites. For suicide testing, you'll want to ask here, heh. I don't personally find suicide run benchmarks to be useful as far as real world performance.Edited by xoleras - 10/12/13 at 2:51pm