Y
YKhan
David said:
As I said, AMD will get it close enough so that it doesn't matter; it
may still be behind, but it won't be anything that will be important to
people. In the server world, there is only a specific segment that
worries about absolute performance, and that would be the HPC/animation
crowd; everybody else is interested in infrastructure and balance of
performance. Benchmark superiority didn't win AMD any sales in its
first year. However, it got AMD's name in the news which eventually won
it sales in the next bunch of years, when it became clear that Intel
couldn't come up with an answer. If AMD doesn't have a good answer
within the first year, then it will have to worry. But AMD wasn't
caught off-guard with the exact wrong architecture at the exact wrong
time, so it's not going to suffer the 3 year catchup lag of Intel.
Either ramping up speed or increasing the cache size. Woodcrest has a
2:1 cache size advantage on Opteron. More than likely AMD will be using
the 65nm transition to ramp up cache size more than speed.
Actually, before K8L they have Rev G coming out. Rev G will be the
first one at 65nm, and there is some talk that there are some
incremental architectural improvements in the works for that, even
before K8L (Rev H) comes. One of the chipsites believes that there is
an additional integer execute unit inside Rev G.
The point is that these are coming down the line -- soon. The memory
manufacturers are holding competitions within themselves to see how
fast they can get their DDR2 modules going at, and who will get there
first, so they're chomping at the bit to introduce these things. The
DRAM makers want to sell some high-end modules to make some gross
margins (any gross margins!) -- something they don't seem to do too
much. We've seen a fairly lethargic pace of specs improvement on DDR2
so far, now that AMD is onboard -- everybody is onboard, so it's time
to open up the ride.
I don't think so, my feeling is that it's the desktop market where
Intel is going to hit the hardest. Intel won't have enough time to get
enough momentum going before AMD is firstly close enough, and then
secondly at par again. Not with the server market as slow to react as
it does. The desktop market is extremely dynamic, so that where we're
going to see the biggest ups and downs for both Intel and AMD.
Will Intel be able to upclock their FSB enough to take advantage of
these, even when they are official spec modules? And even if they can
increase the clocks on the FSB to take advantage of the bandwidth,
would they be able to take advantage of the latency? The faster these
modules get the lower their absolute latencies are. Will Core 2's magic
latency hiding technology be able to keep hiding the real latency, as
the real latencies keep going down? Or is there a tipping point where
it can no longer hide the real latencies?
Well the stories are a little convoluted here depending on which site
you read.
In this story, they don't mention any specific sub-market here (e.g.
consumer and sub-10-employee business market), so it's likely they're
talking about the overall desktop and laptop marketshares:
"The company's growth in the desktop and mobile markets was just as
strong in the second half of 2005. AMD's desktop processor share went
from 20.4 percent in the third quarter to 24.3 in the fourth, and its
mobile share went from 12.2 percent to 15.1 percent."
http://news.com.com/AMD+once+again+hits+the+roaring+20s/2100-1006_3-6030509.html
In this story they say that the 15.1% was the retail marketshare in
2005, but in 2006, AMD's notebook retail share had gone upto 44.7%:
"Intel has traditionally had a massive advantage in market share, with
83.13 percent compared to AMD's 15.14 percent of the U.S. retail market
for notebook PCs in April 2005, not counting sales by Dell or Wal-Mart
Stores, according to a survey of national retailers by Current
Analysis.
By April 2006, that lead had nearly vanished, with Intel at 54.71
percent compared to AMD's 44.66 percent."
http://www.infoworld.com/article/06/05/17/78397_HNamdchallenges_1.html
Anyways, I'm willing to concede that between the two stories one
mentions a sub-market and one doesn't, so it's possible that in the
absense of detail in one of the stories, that the other story with the
greater detail is right. So 15.1% may have been AMD's retail notebook
marketshare in 2005. But that same story now says that in 2006, AMD's
retail notebook marketshare is about 45%, which higher than the
estimate I gave which said 30% retail marketshare! So AMD does quite
definitely have a presense in the retail notebook market.
And Intel still has 4 million Sonomas to get rid of, from two years
ago.
So where exactly are we disagreeing here?
Oh, please do keep feeling sorry for yourself, "oh whoa is me, nobody
believes me!"
I've done a fair amount of Googling for you up above to show you
stories that I've read in the past which shows why I'm lead to believe
certain things. It's quite obvious that you have been blinded to the
huge upheaval in the consumer notebook market that's happened. The
consumer notebook market has taken off finally, but no thanks to the
efforts of Intel. You're not going to see how popular AMD notebooks
have become, if all you see are corporate notebooks. I know very few
people who own personal Centrino notebooks at home, but I know a lot of
people who have got a Centrino notebook from work (including myself).
At the consumer level, if it's an Intel notebook, I see more people
owing Celeron or even Pentium 4 notebooks than Centrino-class.
It's also been said that over 50% of HP notebooks are now AMD based,
although HP doesn't break it out itself. Again that's likely the retail
consumer marketshare, but it's impressive none-the-less.
Yousuf Khan
to 7/24.