 | View Full Version : Cobalt RaQ 4i perfomance results vs. 800 MHz Pentium III dabystru 08-13-2001, 04:14 AM Found this at http://www.transmeta.com/technology/benchmarks/server_bench.html 450 MHz Cobalt RaQ 4i/128 MB RAM - 208 pages/sec 933 MHz Cobalt RaQ XTR/512 MB RAM - 916 pages/sec 800 MHz Compaq DL320/512 MB RAM - 1706 pages/sec Dissapointing for Cobalt :( webbcite 08-13-2001, 11:23 AM Not really apples to apples... There is a big difference between the processors...450Mhz vs. 933Mhz... Big difference in RAM 128M vs. 512M... You really can't put the 4i in the same category with these servers... dabystru 08-13-2001, 11:57 AM I have Cobalt RaQ 4i myself and of course I did not expect 450 MHz/128 MB RAM to be as fast as 800 MHz Pentium III/512 MB RAM. However I was dissapointed that it is 8 times slower (if Webbench is the right tool to measure the perfomance). Before that I thought that putting 8 RaQs together is definetely better than investing into one 800 MHz P3. Now I am not sure. insequi 08-13-2001, 12:43 PM Beside the Compaq, the RLX System 324 is looking good. Seems to be a better choice than upgrading to a Cobalt 4R XTR. Of course, a benchmark alone can't saywhich server is better. Chicken 08-13-2001, 08:08 PM Originally posted by webbcite Not really apples to apples... There is a big difference between the processors...450Mhz vs. 933Mhz... Big difference in RAM 128M vs. 512M... You really can't put the 4i in the same category with these servers... Ummm, what is the point of *that* test? If there was a test that compared a 450Mhz (AMD K6/2-450) with 512 MB of ram to a RaQ4 then I'd be interested in seeing the results, but... A newer Ferrari will beat yo' momma's station wagon, news at 11. SI-Chris 08-13-2001, 08:40 PM If you need to be serving more than 200 pages per second on a steady basis, the RaQ isn't what you should be using in the first place. firstmark 08-13-2001, 09:57 PM Hardly any site serving 200 pages a second would even consider being hosted on a single machine regardless of its power. 200 pages per second x 60 = 12,000 per minute 12,000 x60 = 720,000 per hour 720,000 x 24 = 17,280,000 per day x 30 = 518,400,000 per month What kind of business approaching 518 million page views per month is hosted on a single machine? How long can an raq maintain 100 pages a second before it dies is a better question. firstmark 08-13-2001, 10:03 PM The study is for "peak request load" that means it is the maximum for short periods of time. So it is kind of misleading to think a single machine can crack out those numbers consistantly. dabystru 08-14-2001, 03:50 AM Originally posted by Chicken Ummm, what is the point of *that* test?To me this shows that upgrade path from a single Cobalt RaQ should be not adding additional RaQs and spliting the load, but changing to a high spec machine.If there was a test that compared a 450Mhz (AMD K6/2-450) with 512 MB of ram to a RaQ4 then I'd be interested in seeing the resultsMe too. Why don't we download WebBench from http://www.etestinglabs.com/benchmarks/webbench/webbench.asp and test our machines ourselves and post the results? microsol 08-14-2001, 06:35 AM Originally posted by dabystru To me this shows that upgrade path from a single Cobalt RaQ should be not adding additional RaQs and spliting the load, but changing to a high spec machine.Me too. Why don't we download WebBench from http://www.etestinglabs.com/benchmarks/webbench/webbench.asp and test our machines ourselves and post the results? This software is only for winblowz machines. You don't want to compare winblowz with unix/linux, don't you? :rolleyes: dabystru 08-14-2001, 07:03 AM Originally posted by microsol This software is only for winblowz machinesNot so. WebBench runs on Digital UNIX servers with an Alpha processor, an IRIX server, a Linux server with an x86 processor (including Cobalt), a Solaris for SPARC server, an IntranetWare server, a Mac OS server, a Mac OS X (Rapsody) server, or a Windows NT server with an x86 processor. You do need two more machines besides the server you are testing which run the controller program and client program. Those two machines have to be Windows NT 4.0 or 2000 for controller and Windows NT 4.0, 2000, or 95/98 for client. See also: WebBench FAQ (http://www.etestinglabs.com/benchmarks/webbench/we40faq2.asp) dutchie 08-14-2001, 01:29 PM To me this shows that upgrade path from a single Cobalt RaQ should be not adding additional RaQs and spliting the load, but changing to a high spec machine Ofcourse the difference is huge, but imagine, that your High End server breaks down, or gets hacked or whatever, you'll have 600 websites down, and nowhere to go, if you had 4 raq's instead you, could divide the 200 sites over the remaining 3 raq's until you have things sorted out. |