From owner-freebsd-hardware Tue Nov 14 14:33:59 2000 Delivered-To: freebsd-hardware@freebsd.org Received: from endplay.inode.org (cpu1532.adsl.bellglobal.com [206.47.27.13]) by hub.freebsd.org (Postfix) with ESMTP id E21AE37B4CF for ; Tue, 14 Nov 2000 14:33:56 -0800 (PST) Received: from coup.inode.org (coup.inode.org [172.16.2.203]) by endplay.inode.org (8.9.3/8.9.3) with SMTP id RAA19871; Tue, 14 Nov 2000 17:33:50 -0500 (EST) From: David Jones Date: Tue, 14 Nov 2000 17:33:48 -0500 X-Mailer: KMail [version 1.1.99] Content-Type: text/plain; charset="iso-8859-1" To: Jin Guojun (DSD staff) , freebsd-hardware@FreeBSD.ORG References: <200011141626.eAEGQoC27593@portnoy.lbl.gov> In-Reply-To: <200011141626.eAEGQoC27593@portnoy.lbl.gov> Subject: Re: Luna 340 MIME-Version: 1.0 Message-Id: <00111417334807.00296@coup.inode.org> Content-Transfer-Encoding: quoted-printable Sender: owner-freebsd-hardware@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org On November 14, 2000 11:26 am, Jin Guojun (DSD staff) wrote: > The issue is that what will be the performance dropped if the card is > plugged in a 32-bit 33 MHz PCI slot. According to the data sheet, the > MD5 performance is 100 Mbps. On a 32-bit 33 MHz PCI system, it all mean= s > that the performance will be dropped to 50Mbps, right? There are two factors involved here: the CPU power on the chip required t= o do=20 MD5, and the PCI bus bandwidth required to transfer the information. 32-bit 33 MHz PCI has a theoretical throughput of 132 MBps, or just over = 1=20 Gbps. This is more than enough to handle the MD5 requirement: 100 Mbps j= ust=20 to suck the data in. Note that you will never get near this in the real=20 world; PCI reads are slow. The real issue is the performance on multiple streams of 3DES+MD5, which = is=20 where the PCI system will be stressed. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hardware" in the body of the message