From owner-freebsd-hackers Thu May 18 02:38:43 1995 Return-Path: hackers-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id CAA03172 for hackers-outgoing; Thu, 18 May 1995 02:38:43 -0700 Received: from gndrsh.aac.dev.com (gndrsh.aac.dev.com [198.145.92.241]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id CAA03150 for ; Thu, 18 May 1995 02:38:38 -0700 Received: (from rgrimes@localhost) by gndrsh.aac.dev.com (8.6.11/8.6.9) id CAA14057; Thu, 18 May 1995 02:38:18 -0700 From: "Rodney W. Grimes" Message-Id: <199505180938.CAA14057@gndrsh.aac.dev.com> Subject: Re: Adaptec 2940? To: rgrimes@gndrsh.aac.dev.com (Rodney W. Grimes) Date: Thu, 18 May 1995 02:38:17 -0700 (PDT) Cc: FreeBSD-hackers@FreeBSD.org In-Reply-To: <199505180853.BAA13876@gndrsh.aac.dev.com> from "Rodney W. Grimes" at May 18, 95 01:53:29 am X-Mailer: ELM [version 2.4 PL24] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 2614 Sender: hackers-owner@FreeBSD.org Precedence: bulk > > > > > > > > > > > > On Wed, 17 May 1995, Justin T. Gibbs wrote: > > > > > > > The drivers are identical, so the interupt time should be the same for > > > > driving either card on the same machine. Your benchmark is not really > > > > valid since they were run on different motherboards. > > > > > > To some extent. It is instesting that a good EISA system can best a > > > poor PCI system. Woe to those buying cheap PCI motherboards. > > > > My old 486DX33 ECS EISA/VLB Sis chipset with write back cache performs > > better at memory speed benchmarks than most cheap PCI motherboards > > by a large margin (29MB/sec on the EISA board, I have seen as low > > as 20MB/sec on some PCI boards for the same test, same memory size, same > > CPU chip]) > > > > The fastest 486 PCI motherboard I have tested is the ASUS PCI/I-486SP3G, > > it uses 72 pin simms and memory interleaving (Ie, you *must* install > > simms in pairs). With a DX4-100 CPU chip in this board you can beat > > almost every P5-60 out there in ``time make CLOBBER=true world'' given > > identical memory and disk setup. > > > > My data on the ASUS PVI-486AP4 is not comparible as it was done using > > a DX33 chip, but the gut feeling of the box is that it has okay, but > > not great memory bandwidth. I was also running it with 1 8MB simm. > > > > Guess I should go configure the standard DX2/66 16MB on that board and > > run the test to see where it stacks up in the pile. > > Okay, I stuck 16MB in it, a DX2/66 chip, and ram Poul's little test: > thump:rgrimes {103} ./ram-speed > 49005fb0 0.758 uS/op 1.32e+06 op/S 5.033 Mb/S > 8938c0df 0.432 uS/op 2.32e+06 op/S 8.833 Mb/S > thump:rgrimes {104} > > Can you say dog ass slow!!! Defanitly not a compile engine :-) AHhh... ooopsss... this board has a backward turbo switch, I was running it in the bench set up with out a real switch on it or the LED. Turns out you have to put a jumper on the turbo pins to get it into turbo mode. That and a a few tweaks to the memory system (Memory speed set to 60nS/max, DRAM CAS Pipeline on, SRAM to FAST) got me this: thump:rgrimes {103} ./ram-speed 49005fb0 0.433 uS/op 2.31e+06 op/S 8.801 Mb/S 8938c0df 0.185 uS/op 5.39e+06 op/S 20.569 Mb/S thump:rgrimes {104} Running the BDE/LJO/RWG memory test program it does about 33MB/sec writting to main memory and 40 reading. This is now actually faster than my ECS SIS VLB/EISA motherboard slightly (26MB/29MB). -- Rod Grimes rgrimes@gndrsh.aac.dev.com Accurate Automation Company Custom computers for FreeBSD