With Q in decline and disarray, Carly (Fiorina) might well be acquiring the island of Atlantis - James C. Blasius
Lets look at what was delayed and why. It was R600, the top dog of the bunch, not the lower end RV630 and RV610. If you look at the lust factor, there is R600 and other products, by the time you get to the RV610, you are in the territory of low-end stuff that no one really actually wants - think checkbox on the store shelf territory.
Conversely, the high-end stuff sells comparatively little, mid-range more, and low-end parts sell like mad. Dev costs follow the lust scale, the R600 cards are a pain to engineer and don't sell well, the low-end stuff is more 'pawn it off to the intern' territory. Profits are fattest in the mid range, followed by the low end, and then the top end makes a little bit, if any.
Basically, the R600 is a halo product, one that PR raves about then tries to sell you an RV630 because that is where they make the money. If you don't want an R600, you might not realise you can't afford it and buy an RV630 instead. So it goes.
By delaying the R600 from March to May, ATI suffered a big black eye, but the numbers on the bottom line won't really matter much. The defections should be minimal, Nvidia still does not have a functional G80 driver out, so the pressure is not all there. Neither side has the DX10 mid and low range parts out, and the R600 delay does not seem to have affected this schedule. There is nothing here that can not be fixed with a good deal of kissing up.
As I have been saying since we broke the AMD buys ATI story, I have one hope for the merger, that is AMD will instill the engineering discipline in ATI that they should have had over the last few years. I will forget for a moment that AMD's engineering discipline has seriously fallen off the rails lately, the potential is still there.
What do I mean? The high-end graphics parts are a bellwether for a graphics company, this is the part you look at to see what they are capable of designing unfettered by most constraints imposed by reality and marketing. It is a showcase that makes little bottom line impact but costs a lot.
The cold hard reality is that ATI has badly screwed up the last three generations of its showcase product. The R520 was so late it was pointless, the R580 less so but still not good, and the R600 was supposed to be a late 2006 product that is currently looking like a mid-07 launch. It was about where the 8800 is in performance, roughly equal in DX9 and way ahead in DX10 games that don't exist.
Unfortunately, the card completely missed an entire cycle and is now facing the half-step from Nvidia called the 8900. It won't beat that part. The R600s are also fraught with other problems if the 240W power consumption numbers are to be believed. Basically, it is a late mess.
Why did this happen? It was scheduled for March, now it is set for May, a last minute change. In digging for answers I found out that there was no technical problem, it was almost entirely marketing. ATI was set for a low volume if at all launch at Cebit, and real introduction later on. Saner heads seem to have prevailed and pushed the launch out closer to availability. I could not get a solid answer as to why the parts that were final in January are not shipping until May though.
This is where I think the good is starting to happen. If AMD is lessening the graphics games being played, all the better. If they are holding people's feet to the fire, great. Hopefully they will keep turning the screws until things snap into line, both at ATI and at AMD.
That brings us to the looking forward part. R600 is the last ATI part. It was done under ATI with almost no AMD input. By the time of DAAMIT, things were too far along to make a difference, and things continued as they were. You don't make radical changes at this late stage, be it in tools, personnel or management.
With R600 done, more or less, it is time for AMD to clean house. Having one chip go bad is a problem, a big problem. Three in a row is unacceptable, and changes have to be made. What those changes are, I don't have a clue, but someone at DAAMIT does.
R700 will be the first product that has AMD DNA in it, and will point the way forward. If it is out on time, on target, and all the other ons that one needs, it will be pretty clear that AMD did right. If they fumble again, the whole concept of graphics at AMD should be simply packed in and the money spent elsewhere.
So, R600's problems are of little direct impact, but can be a huge, huge pointer for things to come. Let's see if the changes are and have been made, it should be readily apparent in a few months. Until then, anything without an R7- prefix is looking backwards. µ
Sign up for INQbot – a weekly roundup of the best from the INQ