Servers at the CGL

Navigation Bar

Servers and Workstations at the Computer Graphics Laboratory

A variety of UNIX-based servers and workstations provide the infrastructure for the compute- and data-intensive needs of our user community. The table below sumarizes these systems. (Not included in this table are our desktop computers.)

HP AlphaServer GS1280 HP AlphaServer ES45 (x4) HP AlphaServer DS20E SGI Onyx2 SGI Octane (x4)

Number of Processors 32 4 x 4 2 2 1

Processor Type 21364 (EV7) 21264C (EV68cu) 21264 R12000 R12000

Clock Speed (Mhz) 1150 1000 667 400 300

Main Memory 64GB 16GB x 4 4GB 1GB 640MB

Theoretical Performance
(SPECint2000/SPECfp2000) 877/1482
(each cpu)² 679/960
(each cpu) 455/577
(each cpu) 347/343
(each cpu) 264/283

Graphics Hardware n/a n/a n/a Infinite Reality 2 VPro V12

Rendering Performance
(millions of shaded triangle meshes/sec) n/a n/a n/a 8.0 15.3

3D Vector Performance
(millions of anti-aliased vectors/sec) n/a n/a n/a 7.4 unpublished

Operating System Tru64 Unix TCS 5.1B Tru64 Unix TCS 5.1B Tru64 Unix 5.1B IRIX 6.5 IRIX 6.5

Hostname(s) socrates¹ socrates¹ cgl spinoza ada, euler, larson, pascal

¹ Host socrates is a cluster of symmetric multiprocessor (SMP) servers running HP's TruCluster Server software. Documentation on TCS is available here. Nodes on the cluster are interconnected using a high-bandwidth, low-latency interconnect technology known as Memory Channel, supporting 90MB/s channel bandwidth between any two server nodes and 2.1 usec end-to-end latency.

	HP AlphaServer GS1280	HP AlphaServer ES45 (x4)	HP AlphaServer DS20E	SGI Onyx2	SGI Octane (x4)
Number of Processors	32	4 x 4	2	2	1
Processor Type	21364 (EV7)	21264C (EV68cu)	21264	R12000	R12000
Clock Speed (Mhz)	1150	1000	667	400	300
Main Memory	64GB	16GB x 4	4GB	1GB	640MB
Theoretical Performance (SPECint2000/SPECfp2000)	877/1482 (each cpu)²	679/960 (each cpu)	455/577 (each cpu)	347/343 (each cpu)	264/283
Graphics Hardware	n/a	n/a	n/a	Infinite Reality 2	VPro V12
Rendering Performance (millions of shaded triangle meshes/sec)	n/a	n/a	n/a	8.0	15.3
3D Vector Performance (millions of anti-aliased vectors/sec)	n/a	n/a	n/a	7.4	unpublished
Operating System	Tru64 Unix TCS 5.1B	Tru64 Unix TCS 5.1B	Tru64 Unix 5.1B	IRIX 6.5	IRIX 6.5
Hostname(s)	socrates¹	socrates¹	cgl	spinoza	ada, euler, larson, pascal

² SPECint_rate2000 and SPECfp_rate2000 are more meaningful benchmarks in an SMP architecture, since these measure system throughput instead of individual processor performance. Our 32-processor GS1280 system has a SPECin_rate2000 = 285 and a SPECfp_rate2000 = 405. Complete SPEC2000 benchmark results are available here.

All SGI graphics workstations are equiped with "CrystalEyes" stereoscopic viewing hardware for visualizing molecular models in three dimensions.

Comments about this document should be directed to tef@cgl.ucsf.edu. This document was last updated on 7 May 2004.