Johannes Kepler Universit at Linz Spezialforschungsbereich F013 Graphical and Arithmetical Benchmarking of Workstations Fonds Zur F Orderung Der Wissenschaftlichen Forschung Graphical and Arithmetical Benchmarking of Workstations

نویسندگان

  • Gerhard Kurka
  • Bruno Buchberger
  • Heinz Engl
  • Hans Irschik
  • Ulrich Langer
  • Reinhard Lerch
  • Ewald Lindner
  • Peter Paule
  • Otmar Scherzer
  • Joseph Schicho
  • Jens Volkert
  • Franz Winkler
چکیده

This paper summarizes the results of a series of performance tests carried out as part of an evaluation of diierent graphics workstations. The graphics performance was determined by Viewperf 5.1. In addition, we present some results of a relatively simple graphics benchmark, which measures the raw performance with regard to Gouraud-shading. In order to get a good estimation of the processing power under real world conditions, a multi-grid software package was executed on all machines. 1 Systems and their connguration In this report we investigate the graphics and processor performance of three contemporary commercial graphics workstations. In addition, we will also compare these machines with an ordinary mainstream PC. 1.1.1 Connguration processor MIPS R10000 clockrate 250Mhz L1 cache size 32KB for instructions, 32KB for data L2 cache size 1MB for instructions and data spec95 integer / oating point 13.6 / 24 execution units / max. simultaneaus issues 5 / 5 pipe-stages (integer, load/store, fp) 5 / 6 / 7 main memory 640MB SDRAM number of processors 2 width of the data-bus | width of the address-bus | system interconnection system 7x7 crossbar graphics subsystem MXE with 4MB texture cache operating system Irix 6.4 1.1.2 Remarks Silicon Graphics, inc. claims that their Octane/MXE ooers the most powerful desktop graphics available, providing unmatched realism and polygon performance. However, a direct head-to-head comparison with the two competitors Sun and HP has shown that SGI's Octane/MXE ooers a relatively low graphics performance in the case of large Gouraud-shaded surfaces, only textured surfaces are drawn at high speed. The MXE graphics subsystem consists of two geometry-accelerators providing a peak performance of 1.344 GGops each, and two rasterizers that deliver together a peak ll rate of approximately 240 million Gouraud-shaded pixels per second. In spite of its texturing capabilities it provides only a 4MB texture-cache on the graphics board. This leads to a notable loss in performance if the desired textures do not t into the small cache.

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Algorithm Invention and Verification by Lazy Thinking

Acknowledgment: Sponsored by FWF (Österreichischer Fonds zur Förderung der Wissenschaftlichen Forschung; Austrian Science Foundation, Project SFB 1302, in the frame of the SFB "Scientific Computing" at the Johannes Kepler University, Linz, Austria. My PhD student Adrian Craciun implemented the case study in the frame of the Theorema system starting from my earlier versions of the induction prov...

متن کامل

Local gravitational supersymmetry of Chern-Simons theory in the vielbein formalism

We discuss the Chern-Simons theory in three-dimensional curved space-time in the vielbein formalism. Due to the additional presence of the local Lorentz symmetry, beside the diffeomorphisms, we will include a local gravitational supersymmetry (superdiffeomorphisms and super-Lorentz transformations), which allows us to show the perturbative finiteness at all orders. On leave from the Département...

متن کامل

Graphical and Arithmetical Benchmarking of Workstations

This paper summarizes the results of a series of performance tests carried out as part of an evaluation of di erent graphics workstations. The graphics performance was determined by Viewperf 5.1. In addition, we present some results of a relatively simple graphics benchmark, which measures the raw performance with regard to Gouraud-shading. In order to get a good estimation of the processing po...

متن کامل

The Erwin Schrr Odinger International Institute for Mathematical Physics the Construction of Non-commutative Manifolds Using Coherent States the Construction of Non-commutative Manifolds Using Coherent States

We describe the construction of non-commutative manifolds, which are the non-commutative analogs of homogeneous spaces using coherent states. In the commutative limit we obtain standard manifolds. Applications to the Fuzzy sphere and to the Fuzzy hyperboloid are discussed in more detail. *) Part of Project Nr. P8916-PHY of the \Fonds zur FF orderung der wissenschaftlichen Forschung in Osterreich".

متن کامل

Local Ill–Posedness and Source Conditions of Operator Equations in Hilbert Spaces

The characterization of the local ill–posedness and the local degree of nonlinearity are of particular importance for the stable solution of nonlinear ill–posed problems. We present assertions concerning the interdependence between the illposedness of the nonlinear problem and its linearization. Moreover, we show that the concept of the degree of nonlinearity combined with source conditions can...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 1998