旼컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴� � Date: 03-07-96, 08:44 Number: 363 � � From: Steve S Jackson Refer: 362 � � To: Tino Lazaroff Recvd: No � � Subj: SCSI drive slow Conf: Ibm.General � 읕컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴� TL> interesting thing is that under DOS, I ran a TL> benchmarking s/w called 'CORE' and it tested the 42100 TL> as roughtly 13ms and the 1480 as 17ms. However OS/2 TL> benchmarking s/w seem to pin the 42100N at 21ms. Oh Hi Tino: The Core test is amusing, but is easily fooled. The SCSI disk has a cache, probably at least 512K that fools Coretest. I just ran the Byte Unix benchmarks last night, doing two runs. 1 with an IDE, rated at 11ms, and 1 with SCSI rated at 12ms. The Byte tests on Unix measure multi-tasking file I/O under varying loads. The SCSI consistently performed 60 % better, on the same system. When I used 2 SCSI drives, the difference over 2 IDE drives was over 150%, a MASSIVE difference on the same system. BTW, the ncr PCI scsi chipset (< $100.00) outperformed the $300.00 Adaptek 2940 by 15%. So, for a $100 controller, and an extra $50.00/drive, there is no reason to use IDE. Also, the above tests were done on an Intel Triton board, that has Busmastering IDE on the PCI bus, and fast DMA access. If I use the ISA bus (by mapping the controller to the ISA instead of the PCI, and disabling the Triton code in the drivers), IDE drops another 40%, making it a major bottleneck on any system faster than a 386-33. TL> well. I dont really care about the benchmarks too much TL> since the drive that is reporting the slower access TL> time seems to be MUCH faster than the other one. Yes, benchmarks can be useless, but some well rounded ones like the Byte Unix benchmarks, can give you a fair evaluation of certain system aspects running certain benchmarks. The DOS benchmarks are highly skewed to single tasking systems, and therefore completely useless on anything other than a complete ISA system running DOS only. You should see what the Core test says about my drives! Access times = 0.0 ms, seek times 0.1 ms, throughput at 8 Mbs. I only wish! Faster than SRAM! Take Care, and stick with SCSI sj --- Maximus/2 2.01wb * Origin: IBM Toronto - (905)-316-4255 - (40:649/313)