SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Technology Stocks : Advanced Micro Devices - Moderated (AMD)
AMD 246.76-0.5%Nov 14 9:30 AM EST

 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: Ali Chen who wrote (227061)3/1/2007 11:10:24 AM
From: pgerassiRead Replies (1) of 275872
 
Dear Ali:

Show me one application running on servers where only one thread is active. Even on Windows on a single user's PC there are many threads active. There are 12 threads running if one is just sitting at the desktop and many times more than one thread is activated when opening a simple thing like a browser window. I run Linux and I have 103 threads active with 10 or so receiving CPU cycles currently. And this is just one single core CPU PC. And these active threads talk to each other to get work done.

On the server for a small 100 employee job shop (machining metal), the server has 354 threads active and 56 are receiving CPU cycles at this time. 12 users are actively using the system doing things like payroll, accounts receivable, MRP, ERP, CAD/CAM, order processing and inventory. The RDBMS is the one using the most cycles as the various data sets and their respective engines are running all talking to each other to maintain consistency.

This is typical server operations throughout business. It is not a single thread running at base optimizations on a single chip. So what has gotten you to think that SPECxxx_base2006 has any bering on this type of use? SPECxxx_rate2006 more closely resembles this than any single thread test. In fact running all subtests siimultaneously would better approximate the typical workload, not one identical copy for every core. 95-99% use the default compiler of the OS on which the servers run with lots of switches and work arounds. More for reliability and not optimizations. Of the tests in SPECmine, SPECxxx_rate2006 comes closest to a valid apples to apples comparison between systems.

Since Intel will not supply the OS, Intel's compilers will not be used. Not so with Sun Solaris and Studio, IBM AIX and the IBM C Compiler, OS/X with Apple C Compiler and any Linux flavor and gcc. And if you use Windows, MS Visual C is the default choice, not Intel. And DOS is inactive for XP and beyond.

Again, irrelevant. When I buy a PC to run AUTOCAD or compile FPGAs, and I have a licence only for WinDos platform, your Linux "same percentage" does not buy me any performance. As I said to Petz, you are misguiding yourself by looking at a narrow field of applications. Sure, servers are a nobel application, but the bulk of money is in mass production of media PCs, general PCs, and workstations.

Not the bulk of the profits. That goes to the server makers and those of Notebooks. The former needs SPECxxx_rateyyyy and ther latter doesn't need multicores 95% of the time. And workstations generally do not use Windows due to memory limitations. They use some form of UNIX, typically Linux.

As for your need to use Windows, that is your choice but, don't think you are typical of users out there. Most users run many applications simultaneously on their PC, not get one PC per task. They can't afford that. Funny you should mention Autocad, but in a business that I do some work for, AUTOCAD is running under Linux. And CADCAM runs off a Linux based workstation and uses Oracle RDBMS running off of the server to store drawings, parts, inventories and CNC programs. The application comes from a VAR which delivered a turnkey system.

How many VARs have you developed for? I have done so at many different VARS in many fields over the years. In not one case has a server, turn key or otherwise, did only one simultaneous task. They had many users doing many different tasks at the same time. For instance, a warehousing system running off a turn key server will have a dozen entering stuff to be warehoused, two dozen picking things out needed for various jobs, a dozen shipping out finished product out the door and a few supervisors running the show. They will be in twenty different programs in 50 different places all completing tasks. And that is all going on in just one server with a hot failover backup system identical to the first. You don't want 50 employees twiddling their thumbs when a saerver goes down. That gets expensive real fast.

So SPECxxx_base2006 is just plain irrelevant for most users whether thay are on one PC or on a server. You are likely to be atypical.

Pete
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext