What the article doesnt say is from how many concurrent processes the pentium3 starts to win against the opteron. If it is 1000 processes I would go for the opteron as my workstation since it is not so often I have 1000 processes working at the same time.
From the article you can conlucde that the pentium4 is better with alot of threads but not if it makes a difference in normal usage. I.e with normal nr of threads running.
In my company we evaluated the opteron and it run twice as fast as our pentium4 servers with modelsim. But 50% slower with our backend tools. So the opteron doesnt win allways, but IMO it does win in enough areas with a big margin to get 20% market share in servers. As an AMD investor I would be quite satisfied with 20% (for now)
BTW since we our processor cycles are going mostly for modelsim we are only going to buy opteron servers for now.
Doing further research into this CSA, do you know what pops up on an analysis page by the firm? Yes, a link to Intel.com as the final choice. You can see it by going to: http://analyzer.csaresearch.com/ . Others have remarked on this from Anandtech Forums: http://forums.anandtech.com/messageview.cfm?catid=28&threadid=1348215&STARTPAGE=2&entert... (Checking out CSA, I seem to find that you are serving Dell and Intel as an advanced sales organization? "Helping" business to "upgrade" their hardware?)
This looks to be another dirty trick by Intel to make their machines look good. When I looked at this so called "OfficeBench", do you know what I found? Yes half of the workload was in "specifically, ActiveX Data Objects, workflow (MAPI), Windows Media playback, and Windows Media encoding". How many offices do media encoding on their workstations? Or Office PCs? Not many to make it half of the workload. But it does overweight Xeon's strengths. Of course the Clarity Benchmark also comes from CSA. No other benchmark sources were used.
I find it interesting that a workstation with a good graphics card isn't tested with such standard tests like SPECview and POV-ray. Or any renderers, 3D applications or any heavy processing like ScienceMark 2.0 or SPEC. In fact it looks like they used PC office apps gussied up to look like server tests instead of real server tests. To test what the machine is to be used for, you would think they would use tasks that typical workstations are used for. If you wanted to so what they tested for, any PC would have done as well. When you buy a dual CPU workstation, you process things like a 1,000,000 polygon wireframe or shaded render. Or you do a place and route on a million node VLSI ASIC. You don't encode a CD to a MP3 file (well maybe once in a blue moon).
One of the reviewers was clearly biased by financial conflict. This was not a workstation test. This was a HyperThreading PC test (and hatchet job). Perhaps then the reviewer would have been sent an Opteron 2P server e325 optimized for that work. And should have compared it to either a HP DL145 (or DL545 with 2 CPUs), a SUN Sunfire V20z or both. But as another on the thread above remarked, IBM IntelliStation A Pro comes in 6 configurations. All but one, uses SCSI320 drives. Funny that IBM sent them one without them.
News 
Market Data 
Discover 
