Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!utgpu!water!watmath!clyde!rutgers!sunybcs!bingvaxu!leah!itsgw!steinmetz!davidsen From: davidsen@steinmetz.UUCP Newsgroups: comp.sys.ibm.pc Subject: Re: Which 80386 system to buy? Message-ID: <7393@steinmetz.steinmetz.UUCP> Date: Thu, 17-Sep-87 14:57:37 EDT Article-I.D.: steinmet.7393 Posted: Thu Sep 17 14:57:37 1987 Date-Received: Sat, 19-Sep-87 10:07:59 EDT References: <133@ncrcpx.UUCP> <3320054@hpsrlc.HP.COM> Reply-To: davidsen@crdos1.UUCP (bill davidsen) Organization: General Electric CRD, Schenectady, NY Lines: 43 In article <3320054@hpsrlc.HP.COM> waynec@hpsrlc.HP.COM (Wayne Cannon) writes: |Has anyone any experience with the Tandy 4000? Currently the Tandy and |Wyse machines both appear attractive. Wyse says that they will support |both the 287 and 387 coprocessors, while Tandy supports only the 287. It's good to have the choice. While the 387 delivers more performance, it may be desirable to have a boost at lower cost. The 387 is about twice as costly as the 287, and seems to deliver about four times the performance (from published figures, don't flame me). |Both support true zero wait-state RAM access via a two-plane memory |interleave over the entire 32-bit memory space while some competitors |achieve their memory speed only via memory caching in a smaller |high-speed memory (cost savings?!) and run the bulk of their 32-bit RAM |at a slower pace. This is a more complex topic than you imply. The cost savings is in not using cache. The 45ns static memory used in most caches makes the manufacturer's cost for a 1MB machine about the same. The addition of caching logic probably makes it cost more (the new Intel cache controller may change that). However, unless you use real static memory for main memory, the cache is a win. If you use 16 bit memory to save money, the performance hit is 40% without cache, 15% with (that *is* my measurement). The cache gives about 18-22% improvement for 100ns 1MB chips (my measurement). Note that Compaq has added cache to their new motherboard with the 387 on it. I suspect that cache with interleaved memory will become the most cost effective method as speeds increase. When I bought my PC Designs machine I got some benchmarks from friends at COMDEX, who found it was 20% faster than the (original) Compaq386. At some point I will want to get a revised motherboard with a 20MHz 386 and 387, but I am pretty happy for now. Hope this clarifies the tradeoffs involved. I'm looking at getting a VESTA 386, and will let you know what results I have if I buy it. I want to run a BBS on it, using VAT, but doubt that their serial drivers are strong enough. -- bill davidsen (wedu@ge-crd.arpa) {uunet | philabs | seismo}!steinmetz!crdos1!davidsen "Stupidity, like virtue, is its own reward" -me