Chris Rutter writes: > On a RiscPC with a 233MHz StrongARM and 78MB of core memory, the > `Building Dependency Tree' stage of Debian's `apt-get' progresses > /significantly/ slower than on a Netwinder with 64MB, for the same > sources.list set. Can anyone imagine why? > > (Significantly meaning taking about four or five times as long; both > machines weren't doing anything else.) Yes, there are several reasons: 1. The RiscPCs bus is slower than the NetWinders bus. The NetWinder bus runs at 50MHz. The RiscPC memory bus runs slower, at 16MHz. 2. The processor core runs slower on the RiscPC than on the NetWinder. The RiscPC is about 200 or 233MHz, the NetWinder is 275MHz. 3. Slower IDE - NetWinders can manage around 4-6MB/s, the RiscPC can only manage around 2MB/s. As an example of the slow-down, take the cache cleaning code: It takes 9 bus cycles to clean one cache line on a StrongARM-110. At 50MHz (NetWinder), that is 180ns. At 16MHz (RiscPC), that is 563ns. This means that flushing the cache takes over 3 times as long on the RiscPC as the NetWinder. Now, factor in the slower IDE and the slower core, and a figure of around 4 or 5 times slower sounds about right. _____ |_____| ------------------------------------------------- ---+---+- | | Russell King [EMAIL PROTECTED] --- --- | | | | http://www.arm.linux.org.uk/ / / | | +-+-+ --- -+- / | THE developer of ARM Linux |+| /|\ / | | | --- | +-+-+ ------------------------------------------------- /\\\ | _______________________________________________ http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm
Re: Netwinder / RiscPC speed disparity
Russell King - ARM Linux Admin Thu, 28 Sep 2000 05:11:38 -0700
- Netwinder / RiscPC speed disparity Chris Rutter
- Re: Netwinder / RiscPC speed disparity Chris Rutter
- Re: Netwinder / RiscPC speed disparity Philip Blundell
- Re: Netwinder / RiscPC speed disparity Russell King - ARM Linux Admin
- Re: Netwinder / RiscPC speed dispa... Rod Stewart