From owner-freebsd-hackers Mon Mar 22 7:17:37 1999 Delivered-To: freebsd-hackers@freebsd.org Received: from duke.cs.duke.edu (duke.cs.duke.edu [152.3.140.1]) by hub.freebsd.org (Postfix) with ESMTP id 65D631517D for ; Mon, 22 Mar 1999 07:17:33 -0800 (PST) (envelope-from gallatin@cs.duke.edu) Received: from grasshopper.cs.duke.edu (grasshopper.cs.duke.edu [152.3.145.30]) by duke.cs.duke.edu (8.9.1/8.9.1) with ESMTP id KAA00331; Mon, 22 Mar 1999 10:17:00 -0500 (EST) Received: (from gallatin@localhost) by grasshopper.cs.duke.edu (8.9.2/8.9.1) id KAA23655; Mon, 22 Mar 1999 10:16:27 -0500 (EST) (envelope-from gallatin@cs.duke.edu) From: Andrew Gallatin MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Date: Mon, 22 Mar 1999 10:16:27 -0500 (EST) To: Mike Smith Cc: Amancio Hasty , freebsd-hackers@FreeBSD.ORG Subject: Re: Gigabit ethernet -- what am I doing wrong? In-Reply-To: <199903212129.NAA00925@dingo.cdrom.com> References: <199903212013.MAA50370@rah.star-gate.com> <199903212129.NAA00925@dingo.cdrom.com> X-Mailer: VM 6.43 under 20.4 "Emerald" XEmacs Lucid Message-ID: <14070.22663.501605.138166@grasshopper.cs.duke.edu> Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Mike Smith writes: > > > > Hmm.... > > Does anyone know when the faster PCI busses (64bits or 66MHz PCI clock rate) > > are going to be availabe ? > > 64-bit PCI has been available on Alpha and Sparc systems for some time. Be very, very wary of 64-bit PCI implementations on Alphas and UltraSparcs. In many cases a 32-bit intel 440BX or similar chipset will outperform 64-bit Alpha or Sun chipsets. On alphas, the only workstation 64-bit PCI implentation which is free from performance problems with is the new Tsunami chipset, used in the new 21264 workstations & small servers (XP1000, DS20, DP264) (older server chipsets might be OK, but I don't know -- we don't have any 4100's or 8x00s). Machines based on the Pyxis (21174) chipset have great DMA write bandwidth, but bad DMA read bandwidth and early revisions are buggy as well. On the miata (Personal Workstation) this can be 'fixed' by removing the L3 cache. On the AlphaPC164lx, this can be 'fixed' by turning on DMA read prefetching in the PXYS_CTRL register -- which works great for Myrinet Lanai5 cards, but will lock a machine with an older Myrinet Lanai4, or Alteon card. The older CIA (2117{1,2}) chipset (used in AlphaStations and EB164's) has great DMA read bandwidth, but terrible DMA write bandwidth. Suns have similar performance problems. They do have the advantage of being able to run 64-bit cards a 66Mhz, which makes up for a lot if your NIC supports it (Alteon does). Take a look at http://www.myri.com:80/scs/PCI64X/performance for 64-bit, 33Mhz DMA performance numbers. And http://www.myri.com:80/myrinet/performance/DMAperf.html for 32-bit, 33Mhz numbers. They call DMA reads 'E2L', and DMA writes 'L2E'. Drew ------------------------------------------------------------------------------ Andrew Gallatin, Sr Systems Programmer http://www.cs.duke.edu/~gallatin Duke University Email: gallatin@cs.duke.edu Department of Computer Science Phone: (919) 660-6590 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message