Why not have a separate computer designated for processor intensive functions? If you as a trader know that you need to do complex simulations or extensive number crunching; why not dedicate one computer to handle those tasks while you otherwise keep other trading operations, charting software, order entry etc. on the other computer. Or if you can perform the complex simulations and extensive number crunching at night after market hours. When your focus is not divided. Let's face it; we as traders miss far to many opportunities that the market presents just because our attention is elsewhere. Many times we complicate trading with so much other stuff that our focus is divided. The Keep It Simple Method of trading not only should apply to our entry and exit rules of trading, but it should also apply to how much work we pile on our machines.
Would't want to come off as cynical, but I can't imagine what in the hell anyone could gain from a "number crunching" operation that would be relevant to ST trading. After all, day (and even swing) trading is really just trading noise. Can a high powered calculation tell you which noise wiggle is significant and which isn't?. An old timer once told me, "trading should be reduced to the smallest number of variables. 3 maximum, 2 even better. (Let's see... that's the price chart + 1 other). And 20 years ago, I read a paper which advocated using "price only" for all investment decisions. That's about as simple as it can get, and is probably what's appropriate for trading.
Do that over 100,000 symbols (without the charts - if you had 100,00 charts open, you would need a cluster and 100 GB of RAM to keep up) and now a dual 3.0GHz computer would run at 95% load. nitro
now that HT is out, more software will be written to take advantage. If I were buying a new pc today, I would go for HT even if I didn't see any difference.
if you want max performance on a singe threaded cpu intensive app., XEON is the way to go rather than dual cpu.
I honestly don't believe that anyone here on ET has an active, open universe of 100,000 SYMBOLS that they are trading in with enough regular consistency to warrant having 100,000 charts open. Not only would I question their real abilities for analysis. But I would have serious concerns as to whether or not real quality reviews were happening on a jury rigged setup of software/hardware. Now I am sure that there are competent traders out there/here. Many of whom have the ability to manage and create mesmerizing algorithms for tracking highs, lows, and numerous channels of opportunity. But we should not kid anyone, this is not that easy. First, I really question the understood value of multi-threading and its benefits by the masses. Having worked in this industry for over 17 years now, I have questioned many of the ballyhooed improvements and the logic it was supporting. I have several multi processor machines here and I have played with them extensively. There are several schools of thought out there. The two main ones though are from the sometimes opposing camps. The first is from the vendors which tout the lovely benefits. All hailing the great new speeds and abilities to process multiple streams of data that is manicured for the mass public drooling. Without that, there will be no new purchases. The second is from the actual tech/user community which is fractured almost along interest distinctions also. But generally, from different perspectives. One view from the guys who exist on the "bleeding edge" of technology driving change. The other from the guys who operate on "real world applications and budgets" making things all work together in a global perspective. Striving for a defined, regularly achievable consistency is a task that usually escapes the purview of many of the discussions today. In the January edition of PC World, there was a great article (Two CPU's In One, by Ramon G. McLeod) on the topic. He actually defined many of the plusses and minuses pretty accurately. But none of the quests covered the arena of financial analysis software. In most of the publications that you read, you will find very little data in that regard. That is because to date, the results are NOT CONSISTENTLY IN FAVOR of better, measurable performance improvements of MOST software AS IT IS WRITTEN TODAY! IMHO, having tried many packages over a vast array of off-the-shelf-hardware with off-the-shelf software, the results don't warrant the action at this time. That's my story and I'm sticking with it!