Quote:
Originally Posted by purpleannex
I really want Bulldozer to succeed. But i find myself constantly going back and forth between belief and disbelief.
AMD's reluctance to give anything away is disheartening, what could Intel do with the knowledge anyway? It's too late to change sandy Bridge / Ivy bridge, and Intel will know all in a couple of months for future designs (not that i think Intel will be copying AMD) anyway.
I do believe AMD is losing a lot of customers to Sandy Bridge, a loss that could be stemmed if customers had something to believe in.
|
This is pretty standard not only with our launches, but also our competitors' launches. There were no official sandybridge benchmarks until their launch.
While you believe we are losing customers to sandybridge, others may feel differently.
I have been in this business for about 20 years and I can tell you that the position that AMD is in relative to the launch dates of our product and our competitor has been played out hundreds of times, we all understand the challenges.
Quote:
Originally Posted by lonewolf371
Do you see the game changing at the server level? Let's say Intel is able to get 33+33 gain (78%) optimally and 33+20 pessimistically (60%), that leaves AMD a little behind in the performance race either way. Now, the question becomes one of yields and manufacturing costs; can you comment on whether it's going to be easier to make a Bulldozer vs. a Sandy Bridge?
It seems quite important even to us clients, as server profits would filter back into research and lead to better products for us.
|
OK, let's look at your assumption.
When intel went from 4 cores to 6 cores with westmere, they increased core count by 50% and increased performance by ~33%.
We have seen on the client side ~10-15% increase in performance.
Sandybridge will have 33% more cores.
If you take the old scaling (33% performance on 50% core increase) you can assume that adding 2 more cores gets them ~22% increase.
Today they are 5% behind AMD in spec int rate (x5680 vs. 6176 SE). So they have a deficit of 5%, but they get ~22% for more cores and ~15% for architecture improvments.
That comes out to ~ 32% greater than our EXISTING products, right? But bulldozer will be ~50% greater than our existing products, which nets ~18% performance advantage for AMD.
I have no idea where you are getting the 60-78% performance increases, those numbers just aren't real.
Mine probably aren't right, but at least they are built on two known factors (how thier cores have traditionally scaled in their architecture and the sandybridge scaling that they just delivered.)
