I'm not quite sure that you recall AMD's financial condition in the late 2002/early 2003 timeframe.
After posting a loss of ~$600M in Q4 2002 (including restructuring charges), AMD went on to post a $150M loss in Q1 2003 and a loss of $140M in Q2 2003. It was not until Q3 2003 that AMD started to see some light and posted a loss of only $30M. Given this financial condition, how do you suppose Jerry would have opened the doors on a new $2B fab any earlier???
Well, the question is, did it get easier to secure the funds by waiting a year ? Did AMD's financials improve during that one year period ?
I think Hector, the business man, played it safe - wait and see what was necessary - an attack or a strategic retreat ?
Couldn't it be argued that the ATI acquisition essentially will get AMD the raw benchmark scores that this so called "brainiac" processor would have gotten them if they are able to successfully implement GPU (or several) on die as a coprocessor.
In late 2008 maybe, and only if the dev-tools are available early and some relevant benchmarks are re-written/compiled, will there be a few improved scores. But until then, the lost opportunities and beatings at Intel's hands, will only prove ATI even more expensive.
Finally, for many tasks, an improved CPU is the answer.
Also, keep in mind, putting a DSP/GPU hybrid on-die is a value proposition - not highend.
Very few, if any, applications will be able to leverage the tight integration. The rest will want tightly coupled, but discrete more powerful CPUs and GPUs. This also allows asymmetric memory bandwidth according to device needs.
they get the incremental revenue to boot... In this manner, the new design teams pay for themselves from the beginning rather than AMD having to finance the design teams until they actually have a shippable product.
What do you mean ?
AMD holds out on critical investments for years, to save up billions of dollars, to acquire a revenue stream, to pay for future CPU "brainiac" development ?
Apart from not making any sense, I was under the impression that ATI revenue is capable of paying for GPU development, and not much else.
Isn't buying this IP and design expertise virtually the same as having it in-house over the past few years once the interconnects are there? Is it possible that Hector may have taken a different road but gotten to the same or similar place?
I see, you've moved away from discussing CPU development.
AMD bought ATI to become a GFX company ?
Won't fly. AMD should have spent the money on CPU tech, not GPU tech.
First, as already stated - the acquired GPU tech has ended up vastly more expensive than any super CPU tech that AMD could have done themselves.
Second, it's not an either-or choice. Aggressive CPU development and build-out of capacity wouldn't exclude a shopping spree. ATI would just be bought with shares instead of cash.
Third, any immediate benefits of the acquisition could also have been gotten with a closer partnership with a few dollars thrown in. The benefits of integration and offloading the CPU are too longterm that they should take priority over CPU development.
Remember, the shift in priority took place a long time ago, when AMD began saving its dollars instead of developing CPUs.
But, I don't see the competitive landscape as being dire, and, I'm hoping K8L is everything that it should be...
Same here, but I doubt it. K8L sounds too much like another evolution of the original K7, albeit the biggest evolutionary step so far.
I put my limited faith in one thing in particular - the Core2 L1 is underpowered with half the per clock bandwidth of K8L. For many FP intensive applications the 128bit Core2 FPU is starved for data.
All other improvements in K8L are equal or inferior to Core2.
Finally, the newborn K8L will only face the mature Core2 at the beginning. Intel has improved cores and a completely new core in the pipeline.
Imo, AMD blew it when they failed to move in for the "kill". I doubt they'll get another chance. |
