Excel Benchmark 2011: An Excel Speed Test (with trading functions)

Live Analysis of Benchmark Results

Of 2694 Benchmark results, currently the Excel version with the highest *average Total Score of 62.75 is Excel 2013. Excel 2010 has an average Total Score of 49.3. Excel 2002 has an average Total Score of 47.12. Excel 2003 has an average Total Score of 41.05 and the worst average Total Score is Excel 2007 at 22.55.

Of the top 30 scores, 96.67% of the processors brand name is Intel. Followed by AMD which is the processor in 3.33% of the top 30 results. The exact processor with the current top score of 133.2100 is Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz_Sharaf. It's likely that this processor is overclocked beyond what is shown on the benchmark results page and is not necessarily the top performing processor at default settings.

Of the six tests used to determine Total score, Test 3 shows the most variance between Excel Versions. For example the best score for test 3 in Excel 2003 is 56785 while the best test 3 score for 2010 is 24137896 so for test three, Excel 2003 performs 0.24% better when only the best scores are considered. Test 3 measures how many ticks can be displayed on a stock chart within a 30 second time period. The other tests composing Total Score involve more intensive calculation tests than this charting display test. Although this difference is large, Test 3 should only be considered over Total Score for the most charting intensive Excel programs where VBA and/or formula based calculations are relatively minimal. For most Trading applications Total Score which is a composite of the 6 tests is the most useful score to use for evaluating Excel versions/Processors/OS etc.

Of the 2694 current results 455 are below 0. I would not consider any system scoring below 0 adequate for running an excel based trading system. A modern system scoring below 0 should be checked for unnecessary startup programs, spyware, viruses etc.

*Scores below 0 are set to zero for purposes of calculating average total score.

If your score is low and you’re considering a desktop upgrade, I currently recommend this high benchmark scoring Acer Predator Intel Core i7 3.4GHz Desktop with 12GB SDRAM and Microsoft Office Home & Business 2010.

BenchMark_ET.xls measures Excel performance with 6 trading related Excel/VBA tests. Once the benchmark is complete, your results can be submitted to the Database (just press “Submit”). All submitted results are available here (all results are sortable). The default view is most recently submitted result topmost (so your result should be on top until you sort). System Information is listed along with each test result (processor name, OS, Clock Speed, Cores, Excel Version, File type). A composite Scoring system is in place, but each individual test result is also listed in the results database. All results can be sorted in ascending/descending order by clicking each result’s title. For best results disable autosave and background error checking in Excel options.

Instructions:

  1. Download Benchmark_ET.xls.
  2. Open Benchmark_ET.xls and press the button “Start Benchmark”
  3. (optional) Submit and view your results.

Benchmark for Excel by Exceltrader

Below is an explantion of each test:

Test 1: Data Build Time

Measures the time it takes VBA to create 5 columns * 65535 rows of random tick data.

Test 2: MACD VBA Calculation

Measures the time it takes VBA to calculate the MACD indicator for the 5 columns * 65535 rows of data created in test one. (The VBA code for the MACD calculation is here.)

Test 3: Dispayed Ticks (30 Sec)

Measures the number of ticks (price changes) than can be displayed in 30 seconds. In my original benchmark this was the only type of test. The results varied so greatly that newer versions of excel took hours to complete the benchmark while older versions took just minutes making the benchmark annoying to run. By limiting this test to 30 seconds the total time to run the benchmark will be similar among all systems tested.

Test 4: 63000 ticks converted to OHLC

After 63000 ticks are created, this test converts the tick data to OHLC data. The result is measured in time.

Test 5: Live formula Test

This test begins with with 5000 bars of OHLC and a series of formula based indicators. The last price changes in E5000 as rapidly as possible but waits for all formulas to calculate between each price change.  The benchmark result is the number of price changes that can be completed with formula recalculations within 30 seconds.  Below is the list of indicators calculated with live formulas in this test.

  1. SMA (100,200,300)
  2. RSS
  3. RSI
  4. Fisher Transform of RSI

Test 6: Live formula Test

This test is identical to test 5 except that all formulas also refer to and are dependent on the changing cell, E5000. Test 6  compared to Test 5  is meant to test the “smart recalculation engine” promoted by Microsoft.  More info can be found in this Microsoft article under the section titles “Understanding Calculation Methods in Excel”.

Scoring:

A composite scoring system is in place and the result is listed in column two on the results data base page and also at the top of the benchmark results form upon completion. With the exception of test 3 all tests are counted equally. Test 3, which tests displayed ticks on a stock chart is counted 3 times because for trading it’s an important result to consider. To establish the composite score 20 tests were run on various systems and excel versions. The best and worst scores for each test were used to establish the benchmark score. The best scores from the initial tests resulted in a score of 100 and the worst scores resulted in a score of 0 therefore scores greater than 100 and below 0 are possible.  On the results page idle processor speed is shown and overclocked systems may not show the true processor speed. If you see two systems that look the same with significantly different scores, the higher score is likely from an overclocked system. Thanks for trying my Excel benchmark and submitting your system’s result!

 

Comments

  1. Your website is an oasis in the desert. I have been struggling to do something similar to your backtester for ... well, I am embarrassed to say how long. How did you ever figure out how to query the ichart website like that? And the little circles on the chart - a thing of beauty. You are amazing. Thank you.
    Check out hquotes.com. It's free also, and data goes back to 1973. Automatically dumps data into Excel, which is nice, but your method is slicker. Not sure what adjusted close is.
    I made an "Average Directional Index" macro on the model of your stochastics and MACD macros. I imagine you've done this already but if not, would you like me to post it here? Is this the right place to discuss what has been working and what hasn't for other home-brew Excel traders out there?
    Happy trading.
    Renaud

  2. Gerry Strohl says:

    Great website! FYI, I ran the Benchmark three times, each time I attempted to submit the results (results were in the -97 to -132 range for AMD Turion X2 Dual-core Mobile RM-70, Processor speed: 2000, 64--bit, memory 3.75GB with Windows Vista Home Premium & Excel Year: 2007), I receive a "Run-time error '-2147417848 (800 10 108}': Automatic error The object invoked has disconnected from its clients"

  3. I have just started working and I used ur stockbacktest.It was a great help for me but I am stuck at a point.How can I use money management in it. Please help asap.

  4. Nice benchmark.
    I tried to submit my results but it gives error because it can detect the number of processors, and I have a very common i7 930 processor.

  5. Andrew Williams says:

    I noticed a huge Benchmark score improvement when I disabled PowerPivot Add in. With PowerPivot Add-in ID2063 (-173 score) to without PowerPivot Add in ID2065 (+63 score).

  6. Amir S. says:

    Interesting! thanks for all the good work!
    I also second what Andrew Williams said:- disabling the PowerPivot Add In has made a huge different in the results
    -250 before. and +32 after disabling it.

  7. davemcwish says:

    Had some weird results with this. Initially I got a close to last -858.4345. Disabling my non- com-addins took it to -77.0595 and disabling the com addins (including PowerPivots) it got to the heady heights of +68.6105 which is on a par with similar spec'ed machines.. Putting all the addins back I now get ~ -350; it looks like ID #2425 was an abberation.

    I'm curious as to why the results vary so much for tests 1 &2 when I disable/enable addins.

Speak Your Mind