wbmw,
You need exceptionally heavy lifting to get people to embrace a new technology. Just because 64-bits has a cool marketing prospect, it doesn't mean that real software developer engineers will see it that way. I am again losing faith in AMD's ability to provide a solid adaptation plan for x86-64.
The way I see it is that (ignoring Linux for the moment) there is a huge hurdle to overcome, which is to get Windows x86-64 released, which will get things moving.
My expectation is that it will be a while before we get any of the big apps such as Office, Photoshop etc. converted, so I am not counting on them at all. But there will be a floodgate of the small, utility type apps, many of the ones that actually get benchmarked. 64 bit registers + extra 8 64 bit registers should do wonders for programs such as WinRar/WinZip/WinAce, music encryption utilities, video ripping, extracting, converting, encoding utilities. This will get some people excited. Then, one of the game companies will release x86-64 compiled engine, and will get some insane FPS, or someone will write an x86-64 version of Seti@Home. Anyway, my point is that the stimulus does not have to and probably will not come from a big company like SAP, but from a bunch of small developers, shareware etc. This is on the desktop.
On the server, when it comes to performance, it comes down to databases. I was really surprised about the DB2 announcement / leak, so AMD still needs SQL Server and Oracle. I think SQL Server will happen, I have some doubts about Oracle. Database performance is what makes companies spend big bucks on servers, and they will make or brake Opteron in the server market. I think Opteron will have excellent database performance, as good or better than Itanium 2, without the baggage of IA-64.
Anyway, Windows.Net Server + SQL server or Suse Linux + DB2 is all that is needed for Opteron shortly after introduction. I generally agree with you that huge amounts of money will not be committed to Hammer software, but not a lot is needed to make it a success.
There was one sentence that struck me in the article I linked:
Design teams at PC manufacturers will need to begin addressing 4GB memory barrier fairly soon, analysts say. They will have to choose whether to move to 64-bit chips or use chips tweaked in other ways to address more than 4GB of memory.
news.com.com
So they are already talking about 4 GB barrier, well before I expected it. But the talk will become intense when the memory guys move to 512Mb parts mainstream and start initial 1 Gb shipments. A 2 sided, 2 high DIMM with 512Mbit chips results in 2 GB per DIMM, 4 of them result in 8 GB. When 1 Gbit chips become available, you will not need 32 of them just a 2 sided 16 chip DIMM, which is a mainstream product. So getting to >4 GB on a mainstram product is not a science fiction, it is within grasp, I would say within a year or 2, 2GB DIMMs will become mainstream, and as we approach it, the 4 GB limit will be more of an issue. It is not an immediate issue, but Intel will need to have an answer to 64 bit desktop within 2 years.
Joe |
