home *** CD-ROM | disk | FTP | other *** search
Text File | 1993-04-18 | 4.3 KB | 105 lines | [TEXT/SNTE] |
-
- Tuesday, April 13, 1993 6:39:03 PM
- comp.sys.mac.hardware Item
- From: Derek A Fong,thewho@athena.mit.edu,Internet
- Subject: Excel Benchmark: 486/dx2/66V versus Quadra 800
- To: comp.sys.mac.misc
- comp.sys.mac.hardware
- comp.sys.mac.advocacy
-
- Well, I hope this is both informative and useful. No flames please. I
- don't mean to say this is the best way to test a Quadra 800 or a 486, but
- this is one test that might give someone a basic idea about the floating
- point abilities of the two machines.
-
- The test: Solve Laplace's eqn using a 2nd order central differencing
- scheme and having Microsoft Excel iterate 100 times. using a domain
- of 100 rows and 26 columns.
-
- Methodology: the boundary conditions or cells are set to 1 for cells A1:Z1,
- A2:a100,b100:z100,z2:z100. The rest of the cells are given the formula
- which takes the average of the adjacent cells . For instance for the cell
- B2, use the formula: (A2+B1+c2+b3)/4. Then fill this formula right and down
- for all of the "interior cells." (click ok when Excel says it cannot resolve
- circular references) The under options, go for the Calculations selection,
- click on the iterate "dialog box button" click ok and start your stopwatch
- simulatneously.
-
- The results: I did the test twice for two separate cases (being sure to
- complete reset the spreadsheet before each trial. It seems that the speed
- of iterations is STRONGLY dependent on how many cells are visible during the
- iterate process. I haven't figured out how to tell Excel not to perform
- screen updates during iteration (if anyone knows or could point me to the
- proper place in the manual...please do so). In any case, as you shall see
- below, screen updating takes Excel a long time and makes the iteration take
- over 4x longer on both machines.
-
- The machines:
- o a Quadra 800 giving Excel a 2000 kB partition in memory (actually, I don't
- think this matters too much as 100x26 isn't really a large spreadsheet)
- Microsoft Excel 4.0 was used.
- o a Gateway 486DX2/66V running Windows 3.1 and the latest version of Excel
- for Windows.
-
- Benchmark times:
-
- Gateway 486dx2/66V Quadra 800
- 100x26
- Laplace's eqn
- with 100 iter.
- showing
- cells L-Z,
- 1-35 48.86 seconds 50.08 seconds
-
- same problem
- not showing
- the cells
- involved in
- iteration 11.91 seconds 12.19 seconds
-
-
- As one can see, the times are very similar with the Gateway being slightly
- faster than the Quadra. Also, as mentioned above, Excel does much better if
- it doesn't have to do screen updates. I was quite surprised by the fourfold
- increase in speed when screen updates needed to be performed. The choice of
- the cells above was identical on both machines (ie, 15x35 cells were visible
- on both screens).
-
- Conclusions:
- At least for floating point operations, the Quadra 800 fairs very well
- against the the 486DX2/66. For those who don't know much about 486 machines,
- the Gateway is one of the fastest around (I think it was rated the fastest
- 66 mhz 486dx2 by PC magazine).
- For those of you who do iterations in Excel, it would be wise to
- move the "calculated cells" out of sight when doing the iteration. The time
- savings is tremendous.
-
- <BTW, the numbers above represent the averaging of each test run twice>
-
- anyways, I hope people enjoyed these results, and I would love to hear other
- benchmarks people might have regarding the speed of the 040 vs the 486. I have
- heard that Matlab on the Q950 is about 15% faster than a 486DX2/66, but I
- don't have any concrete numbers.
-
- Feel free to make any comments and thoughts on my results. Questions are
- welcome too, just send me some email.
-
- Derek
-
- thewho@plume.mit.edu
-
-
- ------ Internet Message Header Follows ------
- Xref: iat.holonet.net comp.sys.mac.misc:25029 comp.sys.mac.hardware:35171 comp.sys.mac.advocacy:5926
- Path: iat.holonet.net!news.cerf.net!usc!enterpoop.mit.edu!senator-bedfellow.mit.edu!athena.mit.edu!thewho
- From: thewho@athena.mit.edu (Derek A Fong)
- Newsgroups: comp.sys.mac.misc,comp.sys.mac.hardware,comp.sys.mac.advocacy
- Subject: Excel Benchmark: 486/dx2/66V versus Quadra 800 (Laplace's Eqn)
- Date: 13 Apr 1993 03:56:08 GMT
- Organization: Massachusetts Institute of Technology
- Lines: 81
- Distribution: world
- Message-ID: <1qddkoINNfe3@senator-bedfellow.MIT.EDU>
- NNTP-Posting-Host: alfredo.mit.edu
-
-