Thread: Need a New Computer

Tweet

Originally Posted by kitz

I am considering a MAC. But they are fairly expensive .................

Kitz

Expensive! That is the story in a nutshell. A few of our GE engineering folks had MAC's and had to move to the PC world to survive. But with the ability now of going both ways could be an option but too much more money for too little gain and for me, photo work. So......

Originally Posted by Don Shillady

Matt, you and Denny have aluded to a question I put to the tech guy at Velocity Micro who gave an unsatisfactory reply. That is, does the Duo or Quad hardware actually do parallel processing? My understanding of parallel processing usually requires special software to break up tasks into separate parts for each CPU and then needs a high speed cable to connect the CPUs. However a Duo or Quad chip has the CPU on the same chip so communication between them should be as fast as is possible with that close connection. If there is actual parallel computing then a Quad-CPU is obviously what I want. Thanks for the suggestion to look at other vendors. For all the hype surrounding Velocity Micro, the guy I talked to did not seem on top of the total picture. I used to get PC Week and reveled in the info when I was a consultant for a PC startup company, but they cancelled my subscription for some reason and I am now at least 7 years out of date in the PC world; such a time lag corresponds to hundreds of years in human history so I need to read more on the multi-CPU capabilities. Maybe you can answer a question? My Toshiba Satellite says it is a "ACPI Multiprocessor" under the System Hardware, but when I do Cntrl-Alt-Del I only get one task report as if there is only one CPU. I believe the CPU is a Pentium but do not know further specs on the CPU, maybe I need to "lift the hood" and see what is on there? Maybe there is some actual hardware parallel processing in the Quad-CPU? If so, I want one!

By the way, what is the best low priced engine software?

Don Shillady
Retired Scientist/teen rodder

Because of the design and intended use of the dual and quad core prosessors, being used for multi tasking, alone they will only do 1 prosess per core, but you can get software that will put them togther. older quad prosessor servers did use all prosessing power for all tasks, but some still do.

Thanks Matt and Denny. Matt points out the problem from the recent past in that Duo and Quad systems do great on a mix of various tasks but when the clock speed in the individual CPU is slower than that of a fast single CPU then a single processor will run faster due to clock speed if only a single number-crunching job is running. Thus a Duo or Quad setup with a slower clock speed will indeed run slower than a single processor when running a single job. The articles Denny sent (thanks for the research) shows that increasing amounts of true parallelism is being built into the Itanium Intel chips but the dates look like these chips won't be available for 18-24 months. Matt I would be interested to know what software you mention that might parallelize a Quad mother board. While we are chatting I can note that when I retired I went through a mental Cold Turkey process going from using UNIX-type systems for many years to WINDOWS and I must admit I had a lot of learning to do before I could use WINDOWS creatively; thank goodness for the WATCOM f77 compiler for PCs! About 15 months ago I had the "Restart Virus" but learned how the local "Geek Squad" fixes WINDOWS viruses. You can buy a "LINSPIRE" disk for $50 and install an alternate PC-UNIX system that can be booted up instead of WINDOWS. Since LINSPIRE is only on a few PCs nobody writes viruses for LINSPIRE. To my amazement all my files were there under LINSPIRE along with a lot of familiar UNIX commands so I just copied all the important files to a backup device using LINSPIRE and then scratched my C-drive. It still took me about ten days to reinstall all the WINDOWS software on the cleaned C-drive but I did get it back working again. As far as I know the LINSPIRE system would be an alternate system to WINDOWS that might be better than going to a more expensive MAC system and one can mentally migrate from a UNIX system in a research environment to a home PC system. I note that one of the key points of our/my recent book is that one can do a lot of research type chemistry calculations on a WINDOWS-PC system and that is why we spent a lot of time porting our UNIX software to a PC.

Matt, give me a hint on whatever parallel software you can recall for a Quad system.

Hey, once again what is the recommended best engine software to estimate dyno results for an engine?

Don Shillady
Retired Scientist/teen rodder

I have never owned a dual or quad core so I don't know exactly what software there is because I have never needed it.. I have seen some searching for the dual core Athalon I'm getting but I just kinda glanced over it. however somthing like this might help you http://www-03.ibm.com/systems/p/soft...ssl/index.html

Matt, thanks for your diligence/research on this topic. This may be a situation where we are getting too far from the main function of this FORUM so let me know of comments on what engine dyno simulation software seems best at a reasonable price (under $100?). For those interested in the "other world" of computation using many PCs in parallel just look up the words "Beowulf LINUX RedHat" and you will find a wealth of information on linking many PCs in parallel using high speed switches, but to my knowledge there are no direct applications to Quad-core CPUs at present. I am aware of the ESSL and parallel-ESSL routines but that requires using their routines. Many years ago I decided that I did not have time to recode large programs every three years or so when a new computer system came in so I decided to avoid canned routines and just write/find my own routines as part of my own code rather than depend on linking external routines. Thus I achieved independence of a sort, but now would find it a big chore to use the external ESSL routines instead of my own code. For those interested, RedHat LINUX is another operating system for PCS that is favored by the Beowulf/parallel folks.

If it warms up today I may mess around with installing my steering column?

Don Shillady
Retired Scientist/teen rodder

I use desktop dyno 2000 which is $39-$50 depending on the version but it is not the most accurate.. Engine Analyzer is expensive but is by far the most accurate. it is $100 for the basic version..

Matt, Thanks for the lead on the Dnyo simulation. Maybe I can put the PC-computation discussion to rest for Denny and IC2 with the following address

http://www.cacr.caltech.edu/beowulf/.../building.html

That is a site which explains the basic setup for a 16-node Beowulf "Supercomputer" first pioneered at NASA. I am wondering if the extra nodes could be the other three CPU on a Quad chip for a 4-CPU Beowulf system? It is interesting that the Hot Rod phenomena has a large psychological component of a "need for speed" and that is very similar to the idea of building a supercomputer out of many PCs. Another similarity is that the Beowulf architecture wants to use many cheaper units from previous generations of technology; does that look like rebuilding an SBC from the '70s?

Don Shillady
Retired Scientist/teen rodder

Heres on from 1956 ,when everyone was first amazed with the display scope...
http://video.i-am-that-i-am.com/nsv/...e=filename.pls
Keep in mind ,the gamers have pushed computing faster than any other group,theres millions of em and they buy the latest greatest stuff for speed ,just like a hotrodder.
A computer like Denny is talking about is much more powerful than the multi story building full of computer.

Denny, for the individual user we both might be right. I chatted today with some hot shots on the West Coast who have a company based on Don Beckers's original Beowulf design and they gave me a rough estimate for a dual-Quad Xeon box (8 total cores) running the Beowulf RedHat LINUX software for about $2000. The two quad Xeons (total of 8 CPU cores) would be on the same mother board in the same case.

http://www.penguincomputing.com/inde...2&vid=scyldcom

The Xeon cores would only be clocked at 1.8 GHz but using one as the controller and the 7 others as computers they can use the same g77 fortran that is common in scientific computation. Maybe my request will stimulate further development of parallel computing on a single motherboard and we will see what their response is. I assume that price is without a monitor but I can use an existing monitor. The Red Hat LINUX will allow Internet interaction from the controller core so I will be very interested to see what they spec for $2000. Compared to my present 3.0 GHz laptop the performance theoretically could be (8x1.8)/3.0=4.8 times faster but of course you never really achieve that, although a factor of 4 might be possible. Comparing game images to MFLOPS (million floating point operations) in computation is like comparing apples to oranges. One famous calculation of a few years ago was a controversial evaluation of the dipole moment of the CO (carbon monoxide) molecule. The experimental value is about 0.13 Debyes but "common sense" expects the Oxygen end to be negative and the C end to be positive. After a humongous mult-day calculation it was found that the C end is really negative and the O end positive so that calculation represented many days of CPU time just to get a single number whereas in games the trick is to update images as you say. Interestingly one of the popular (free from US. Gov.) programs is called GAMESS (General Atomic and Molecular Electronic Structure System) so those guys do have a sense of humor at Ames Nat. Lab. in Iowa!

Don Shillady
Retired Scientist/teen rodder

Hi Denny, you are becoming a Quantum Chemist too! I have used GAMESS (free from US Gov, your tax dollars paid for it!) for many years in various forms and we reviewed a lot of their examples in our recent book. There is another (expensive) competitor called GAUSSIAN which is also available for PCs and I have that too. In order to charge high prices GAUSSIAN has to stay ahead and is regarded as the "Best" in the field but the guys at GAMESS try to keep up with the free version and there is the usual competition at work there. In fact the main example of my part of the book is my own code for similar calculations (and some for "twisted light" that nobody else has) called PCLOBE but it only runs under WINDOWS on a single processor. I am trying to figure out how to get it to run under an inexpensive form of parallel setup for more speed. There is software to make "molecular movies" and as you say the graphics capability is there to do "slow motion" movies of atoms that move in about 10**-14 seconds (0.00000000000001 seconds) to a slow movie at about the same number of frames in a computer game. This technique is called "Molecular Dynamics" but the current problem is that the calculations behind the positions of the many atoms are too complicated for most computers and that is where the speed is needed for accuracy. Before Brent throws me off this Forum for discussing non Hot Rod stuff let me say that a hot area of research at a nearby lab in the heart of Richmond is "tribology". I call it "Engine Heaven". They run Diesel V8s, Buick V6s and OHC-four cyls with a lot of instrumentation and strip them down to look for wear. Then they rebuild them and run them again. However the basic science questions have to do with lubrication and the effects of squeezing oil additives between metal parts at the molecular level. There is big money and big science involved in the study of lubrication. Another test they run is for rear gear additives and I know without those additives, the rear gears in big trucks hauling heavy loads would simply burn up in less than an hour of use! Sometimes when I talk about this stuff I wish I was 40 again and doing calculations but then again there was too much pressure in that business for me to go back to it except to talk over the good old days. Still if the Hot Rod Forum could visit that Ethyl lubrication lab in Richmond, the folks here would probably slobber like a baby with a candy cane!

Don Shillady
Retired Scientist/teen rodder

don where is this ethyl lab in richmond and can we visit?

jyardgirl, I have been in there three times on tours and for contacts but I cannot provide entry. I was not employed there but once had a small research contract with them. Maybe Google can show the buildings as located one block north of where US Route 1 crosses the James River in Richmond. On the map the bridge should be the "Lee Bridge". When I was there last I think the test beds were the Cummins Diesel V8s, Buick V6s and the Pinto 2300. They also have a large storeroom with many 50 gallon drums of various grades and type of base oils in which the additives are blended. I previously worked for a summer at a Socony-Mobil Lab working on what are now ancient formulations (1962) of some very smelly gear oil additives which are still used. That summer experience gave me some credibility with the science staff at Ethyl as well as the fact that my university office was only about six blocks away. The additives are thio-phosphono-esters and basically break down under pressure and react with an iron surface. Those compounds are responsible for the characteristic odor of rear gear oil and believe me they really extend the life of the ring and pinion gears thousands of times over what just plain heavy oil would. Of course I spilled some when I cleaned up my 8" Maverick rear for my roadster and forever the wooden floor in the garage will have that gear oil smell which I am sure is familiar to everyone here on this Forum! The Ethyl lab is working on more advanced lubricants and the old thio-phosphono-esters are now considered "classic gear additives". Maybe if someone at next year's NSRA National event at the RIR raceway could contact the Ethyl Corp. there is a long shot that a tour could be arranged. They do give tours to professional groups and the Virginia Section of the American Chemical Society has met there in the past so they are eager to put on a good PR presentation and what better folks than folks from NSRA. Unfortunately what I have just said is just an idea and you would need some highly placed representative from the NSRA to approach Ethyl Corp. regarding the NSRA event next August. Richmond is pretty lively with blown street cars that time of year and many neat rods!

Don Shillady
Retired Scientist/teen rodder

Dave,

I too have always built my own machines with the exception of my very first one. The reasons for it were, as others like Denny have already mentioned, the fact that the cheaper computers that you can buy for the most part have mother boards that are not expandable. Also in some cases, they'll have built-in components (like video chipsets) that are hardwired to the mother board, (Dell used to do this, I don't know if they still do). One of the Dells that I worked on, the owner wanted to put different video card in, and went out and bought an AGP card, well, because the motherboard already had an AGP chipset hard-wired in, the manufacturer saw no reason to include a separate AGP port. The user had to get a PCI card instead with lower performance. Like all things, you get what you pay for, or more correctly, I guess, you don't get what you don't pay for. I've always preferred the ASUS motherboards becuase I have had good luck with them and the older ones had an Award Bios on the rom that allowed a few more tweaks and clocking options than AMI bios versions that I have seen. I was not aware of the problems Matt mentioned as mine is about 6 1/2 years old.

One of the other main reasons that I've always stayed away from these "complete systems" is that, in the early days, at least, the manufacturers didn't provide you with an actual copy of the OS that was loaded on your machine only a recovery disk that had an image of the data stored on a disk (Toshiba for example) or on a separate partition of the hard drive (Compaq for example). Along with the provided recovery method, came all of the what I refer to as Junk Software, so that anytime that you re-imaged the machine, you got the same adware junk software that was originally on the machine when you first bought it. In my mind, that software, eats up space and a lot of it seemed to always want to run in the background which just slowed the machine down.

Sure you can remove the junk software, but it gets old having to uninstall that stuff after the third or fourth time you have re-imaged the machine and it only adds to the time spent configuring it how I wanted it.

So short story long, I have, as I said earlier, always built my own. I also early on got my own imaging software so that I could configure the software installation the way I wanted it and then save the image for my recovery.

I did this because even if you defrag the drives every day, over time little file corruptions occur and slow the machine down. Before I got the software, I used to have to reformat the drive (I'd end up doing this about once every 8-10 months) because the machine would just stop performing like it used to. I would then spend the next 10-12 hours re-installing all of the software defragging it, downloading updates, configuring etc. At the end of the process though my machine would be blazing away again and I'd be happy for a few months. Now with the imaging process, I have it back up and running like new in about 45 minutes.

Anyway, best of luck to you in your search.

Cheers,
Dutch

http://www.calvin.edu/~adams/research/microwulf/

Denny here is somethong worth building for $1500! In case you don't appreciate 26 Gigaflops recall that the first recognized "Supercomputer" was the CRAY-1S which pumped out about 12 Megaflops back in the '80s. Since a Gigaflop is roughly 1000 Megaflops this little beauty is roughly 2000 times faster than the first supercomputer. Of course you have to run some sort of LINUX instead of WINDOWS but you can get away from VISTA and the writeup says the "0" node master core can access the Internet. Of great interest to me is the availability of GAMESS along with the free support from the Ames National Lab and the acquaintences I have there for the GAMESS program. This computer compared to my 3.0GHz laptop is like comparing a 239 flathead to a blown Chrysler Hemi!

Don Shillady
Retired Scientist/teen rodder

Originally Posted by Don Shillady

http://www.calvin.edu/~adams/research/microwulf/

Denny here is somethong worth building for $1500! In case you don't appreciate 26 Gigaflops recall that the first recognized "Supercomputer" was the CRAY-1S which pumped out about 12 Megaflops back in the '80s. Since a Gigaflop is roughly 1000 Megaflops this little beauty is roughly 2000 times faster than the first supercomputer. Of course you have to run some sort of LINUX instead of WINDOWS but you can get away from VISTA and the writeup says the "0" node master core can access the Internet. Of great interest to me is the availability of GAMESS along with the free support from the Ames National Lab and the acquaintences I have there for the GAMESS program. This computer compared to my 3.0GHz laptop is like comparing a 239 flathead to a blown Chrysler Hemi!

Don Shillady
Retired Scientist/teen rodder

Thanks for the link Don.

That's not too many steps above Torrodial Core Memory...