home *** CD-ROM | disk | FTP | other *** search
- In article <Norman-120495145637@red_knight.msfc.nasa.gov>, Norman@eisner.decus.org (Richard) writes:
- |> Back to discussing benchmarks...
- |>
- |> According to Digital News and Review, DEC has officially announced a new
- |> model Alpha: AlphaStation 250 4/266. This is the one I heard about
- |> that has the 128 bit bus and 2Meg cache. Here are the Spec92 benchmarks:
- |>
- |> AlphaStation 400 4/233 250 4/266
- |> Clock rate 233MHz 266MHz about 15% increase
- |> Specint92 157.7 198.6 about 25% increase
- |> Specfp92 183.9 262.5 about 43% increase
- |>
- |> Depending on the graphics card & monitor you are looking at about $12-14K.
-
- Did the article say which model of Alpha CPU was tested here?
- |>
- |> Questions:
- |> 1)What are the Spec92 numbers for a 100MHz pentium running WinNT 3.5?
-
- SPEC figures have been fairly closely associated with UNIX systems, at least
- up to this point. You might want to look up the real SPEC approved numbers for
- systems that closely match those you're interested in. It's worth noting here, too,
- that the common perception that SPEC measures CPU only performance only is
- a little misguided. When you read the SPEC newsletter you will get a listing of
- the exact hardware configuration, OS and compilation flags used to generate the
- FORTRAN and C tests that make up the SPEC suite. The figures I remember for
- a typical Pentium system are around 100 SPECint92 and somewhat less in the SPEC
- fp92 area. I could be way off here. The real competition will be the PowerPC 604
- based system that look to offer the best common return on investment. But we'll
- have to wait a few more months before we see any of them.
- |>
- |> 2)What effect does the 128bit bus and cache have on "application"
- |> benchmarks? The www.dec.com web server has a good WNT3.5 performance
- |> report which includes some interesting "application" benchmarks for
- |> Word, Excel, and Autocad (CADbench93), but they don't include the
- |> new 128bit bus machines yet. Note: a 90MHz pentium does pretty darn
- |> good in these tests, especially for the price difference.
-
- It can have a very positive effect, especially given that something like 3D
- rendering will show very good "locality of reference" and your cache loads
- will be 128 bits per cycle both to/from memory and to/from the CPU.
- |>
- |> 3)How is the work coming on developing the Lightbench95 benchmark? ;-)
-
- Actually, such a benchmark would be a lot more useful to you than SPEC or
- WINmarks. You would want a whole series of test renderings including some
- "cache-busters" so you could test more than just how the system runs a render
- out of cache. This technique figures prominantly in SPEC. Also, I suspect for
- a LightWave benchmark, floating point performance probably gets a greater
- weight than integer, hence the good potential for PowerPC 604 and Alpha chipsets.
- |>
- |>
- |> <<<<=======================================================================
- |> Richard Norman norman@eisner.decus.org
- |> AMIGA --- Amazing Multitasking Interactive Graphics & Animation
- |>
- |> Amiga Networking FAQ /pub/aminet/docs/help/anetfaq.lzh
- |> Inputs appreciated!
- |> =======================================================================>>>>
-
- --
- Joseph Crowe
- jcrowe@isd.tandem.com
-
-
-
-