How does Xeon E 5645 2.4 GHz compare to i-7 2600K 3.4 GHz ?

Discussion in 'Hardware' started by trader12, Mar 30, 2012.

  1. trader12

    trader12

    I bought Dell workstation T-7500 with Xeon E 5645 2.4 GHz with other configuration like RAM, HDD, Win-7, etc same as Dell XPS 8300 desktop with i-7 2600K 3.4 GHz

    But I am finding that workstation seems slower than desktop.

    So is it the processor that is slow or something else ?

    Percentage wise How much slower it should be ?

    They want extra $ 500.00 to upgrade to Xeon X5650. 2.66 GHz. Would that be as good as i-7 2600K 3.4 GHz or better?

    OR

    Getting dual processor with two Xeon E5645 2.4 GHz better ? Both costs extra $ 500.00

    OR

    Other better options ?
     
  2. Yes, CPU is slower.. but as you usually don't use the CPU all that much in trading (most functions run from the RAM once loaded), a faster CPU should have zero impact on trading... might have an impact on some heavy number crunching if you're doing that sort of thing.

    Passmark CPU scores....

    Single E5645.. 6923 (that's fast)

    X5650... 7928

    i7-2600K... 9411

    Dual E5645... 14,040
     
  3. trader12

    trader12

    I am keeping 12 charts open, each with 26 Indicators.

    All charts are tick charts.

    How intensive computation is that ?

    I have 12 GB ECC RAM

    Is E5645 good enough ?
     
  4. Tick charts can use some/lots of CPU time. Same is true for "custom" indicators.

    1. Check your Task Manager to see how much of your CPU is in use.

    2. If the percentage is high, consider changing tick charts to 1-minute charts.. MUCH less CPU usage.

    If you require lots of tic charts, you'll probably end up going for "all the horsepower you can afford"...
     

  5. energy efficient 6 core server processor designed for multi-threaded applications that are not found in retail trading apps or any consumer apps. 2600k blows it away for trading. and for Scat, memory does nothing but store data, a mouse click or bar tick does'nt happen unless the cpu tells it to happen.
     
  6. Nobody can answer that. It depends of your setting for the tick charts (1000 ticks versus 10 ticks - huge difference). Also it is not the number of indicators that matters. It is what the indicators themselves do. I have developed some "neat" experimental indicators that is similar to Bollinger Bands to square-root the summation of many squares, except to push out the look back period to close to 80 instead of 20.... Kaboom... one indicator brought the CPU to a crawl.

    There is no "average", "roughly", "rule of thumbs" about indicators. Indicator is a generic term. Much like arithmetics to trigonometry to calculus. They are all maths. But the degrees of sophistication vary widely.

    I agree with Scat that you need to monitor yor CPU usage to determine if your processor is fast enough. During the normal trading hours (pick the most busy hour such as shorting after the market opens)... see what the overall CPU usage is. So if it is up to 80% (which is rare)... then you can figure, roughly, using a dual processor or i7-2600k would probably bring it down to 40%, etc..
     
  7. RL8093

    RL8093

    Yikes - very ... :eek: (depending on your settings)

    To add to other responses:
    - check for non-value-added programs running in background
    - some charting programs appear to handle tick charts more efficiently than others. With my last box, a tick chart on eSig slowed the unit to a crawl while same charts on Ninja didn't have any impact
    - hopefully you're using a charting program that utilizes multiple cores. If not, you've got to switch to one that does.
    - as other's said, monitor cpu usage. Windows has a bare-bones approach but there are many free programs that let you monitor temps, usage, etc (Piriform Speccy is one >> download.com is your friend)
    - check charting program refresh rate

    R
     
  8. dholliday

    dholliday

    I use a old i7 (1.8GHz) laptop for trading. It is auto-trading with no charts. Watching every tick (IQFeed) for 1000 symbols takes about 14% of CPU usage (7% for DTN data component and 7% for my trading app running lots of calculations on each tick). Assuming you are running a well written, multi-threaded app, charts are the bottle neck. Get faster graphic card(s).
    -David
     
  9. trader12

    trader12

    I am using ATI FirePro 2460 ( 512 MB ) that AMD says they have designed it especially for day traders.

    Is it good & true ?
     
  10. Don't know about the "specifically for day traders"... but for the financial markets in general... where multi-monitor is common. It's much like the Quadro NVS 450.
     
    #10     Apr 1, 2012