|
PureBytes Links
Trading Reference Links
|
Ernie,
I've used a dual AMD for about 18 months now with scores of problems. This
was a reply I sent to Long a moment ago:
Long,
I've been using a dual AMD 1700+MP system with 512MB and an EIDE disk. The
motherboard is a K7D Master from MSI, very fast FSB, and it will take up to
4GB RAM.
A caution, however. It is my understanding that, with the exception of
TS7, none of the versions are dual processor aware. You'll still see a
performance boost over a single processor but it won't be as
noticeable. Our tests indicate an increase of 28% in performance over
wintel processors (optimization runs consisting of mostly floating-point
calculations).
One thing to be aware of, however. "Wintel," as I call it, has optimized
their processors for the Window$ operating system. Some function calls are
faster but for raw number crunching power the AMD we have absolutely smokes
wintel. Comparable machines (dual AMD vs. dual Wintel) running the same
strategy were over 4X faster! In fact the Wintel processors clocked
slightly faster (2.0GhZ Wintel vs 1.7Ghz AMD).
In light of the thread regarding TS7, here is our experience with the same
code, same machine:
TS7 Verify 12.3 minutes.
TS6 Verify 15 seconds.
2000i Verify < 1 second.
One year of data, data1 (4 minute) and data2 (daily)
TS7 Apply strategy with default settings 22-30 minutes
TS6 Apply strategy with default settings 12-14 minutes
2000i Apply strategy with default settings < 5 seconds
TS7, TS6 large run optimization (88K+ tests) HAHAHAHAHAHAHAHHAHAAAHAHAHAHAH!
2000i large run optimization (88K+ tests) 4hrs 15 minutes.
My suggestion: Go with a really fast Wintel processor if you're not doing
large amounts of calculations else go for the AMD. I know you'll get good
mileage out of the dual processor, if you choose to go that route. Just
loading a strategy on 16 months of tick data took > 3 hrs on a 1.4Ghz AMD
machine but less than 2 hours on our dual. And stick with 2000i. The
programs run in separate processes, giving you more control. If, for
example, the Tradestation app is frozen, you can kill that single process
without worrying about the Power Editor; unlike TS7 and TS6.
The bottom line:
TS6 and TS7 are inferior products (compared with 2000i) that are unstable
and provide unreliable report results. I've run the same strategy on
three different machines and looked at three different reports
(non-optimized).(The daily, weekly and monthly tabs don't match the equity
curve, the analysis page doesn't match the summary page). Often I'll get
different reports simply by turning status off then on! My solution was to
write a series of SQL statements to capture results pushed into .txt files
and perform my own analysis.
TS7 is a BETA I don't care what the guys at TS have to say. It's one thing
for Micro$oft to release beta software as "production" (can anyone say
fixpack?) but to offer a platform to the trading public with MAJOR problems
is risky at best.
Maceo
At 08:40 PM 4/17/03 -0500, Ernie Bonugli wrote:
Omegalist & Stephen,
I experienced severe TS7 performance problems running on
a very fast dual AMD system!! TS should NOT be led to believe that
faster cpu's will overcome the problems.
I am also concerned when you make no mention of any attempt to
improve the performance of TS7 and instead attempt to entice
people with "other" TS7 features.
I hope I have misinterpreted your message as it sure does not
sound good!
Regards,
Ernie
ebonugli@xxxxxxxx
|