Originally Posted by Rollo
1. Why not show the cards at PAX? It's a week earlier? I'm sure NVIDIA will be at CeBit as well.
Actually,Cebit is 3 weeks before Pax,and the main point is that it's a far larger technology show,having far more visitors and press coverage than PAX ever could....Nvidia would get more benefit from a Fermi presentation there than they ever could at PAX,period.
2. You apparently didn't understand me, and you apparently don't understand business. Are you an adult?
There are choices to be made when manufacturing these products. If NVIDIA released game benchmarks (or to a lesser extent clocks) 3 weeks before launch, ATi can make choices about their refresh parts.
A more expensive heatsink fan might be needed to run the cards at competitive (or better) levels, or a better PCB. I'm not an electrical engineer but better quality capacitors, etc might be required. Even if the board layout and HSF is "set in stone" a choice might be made to use more voltage to run at higher clocks, or not, and give OEMs the freedom to market OCd variants,or not. They might opt for higher speed and factor higher returns in exchange for a "halo" part, or not.
This is how business works. You don't tip you hand and give useful information to competitors because "some guys on the forum want it" or "the press wants it".
Likewise, withholding the information doesn't imply the parts are bad, or good. Much like the HD2900XT and HD4870, or the FX5800 and 8800GTX were surprises of the good and bad flavor, lack of information does not imply anything. (contrary to what chromosome deficient stillborns like Charlie Demerjian post as they soil their cheap boxers, or ATi fans post as they think "Ooh boy! I'm helping ATi now! Down with NVIDIA!")
You're not getting it,ATI already went on the safe side and said that Nvidia might hold on to the speed crown for a bit with fermi,so the surprise effect from with holding final clocks on fermi buys Nvidia a few weeks extra at most,hardly any advantage that matters in the longer term,as the major damage to fermi already happened with the delays it's suffered so far.
It's ancient history basically,as we all know that computer products in general,already have a short lifespan as it is as the best thing on the market,made a lot shorter when they're 6 months late....It blew it's window of oportunity as the fastest thing out there and hold on to that title at least for 6~7 months.
You mentioned the FX5800 and the HD2900XT right,what happened with both products after they were released?....They both were replaced very quickly with the FX5900,having a 256 bit memory bus rather than the 128 bit the FX5800 shipped with,and on ATI's side,they moved to 55nm pretty fast to make the original R600 GPU cheaper to make and use less power and be easier to cool,now relabled the RV670 chip.
In both failures,that happened within 2~3 months after the original release,which were both very late as well,just like Fermi is.....In short,there's very little reason to keep the clocks on this first generation Fermi a closely guarded secret anymore....If you miss your launch window by that much,you're screwed basically.