Another cnet article bashing apple

Fahrvergnuugen · #1 January 12th, 2004, 03:28 PM

http://news.com.com/2001-7339_3-0.ht...gutspro#behind

Quote:

Find out why Apple's supercomputing dreams will probably remain just that.

how is it just a dream when there is an apple based system at #3?

Is it me? Or do some of the editors at cnet really have a problem with apple. They're always bashing them in one way or another in my opinion.

malexgreen · #2 January 12th, 2004, 04:00 PM

Quote:

Originally Posted by Fahrvergnuugen

http://news.com.com/2001-7339_3-0.ht...gutspro#behind

how is it just a dream when there is an apple based system at #3?

Is it me? Or do some of the editors at cnet really have a problem with apple. They're always bashing them in one way or another in my opinion.

His bias is clear. Also the cost ratio of the top 2 compared to the Big Mac are 10:1, so even if they paid the students this would not justify the large price differential. And he kind of contradicts himself or makes another good point for the Big Mac:

Even though other companies are spending more to use RLX, Dell, HP, and Sun they STILL were NOT in the top 3!

**btoneill** · #3 January 12th, 2004, 04:13 PM

The article is right on the money. Most supercomputers are not custom designs such as the top 10, the other 490 on the list are standard models sold by IBM/Sun/HP/Cray/SGI. Apple will not be offering a $2.6M computer off the shelf. You can go to http://store.sun.com and order a $2.6M computer, which is not even maxed out. Apple will never be competeing on this level, nor do I believe they should.

It's a pretty feather in Apple's cap that their is a computer in the top 3 that was built from their hardware, but it's just that. The facilities alone to handle what VaTech did is far beyond what most supercomputer users are able to get.

Brian

Fahrvergnuugen · #4 January 12th, 2004, 04:54 PM

I can't say I totally agree with you.
Clustering is a valid way to build a supercomputer. No apple will never build a single machine that costs ridiculous amounts of money... the idea is ludicrous when you consider that you can take a bunch of cheap nodes and build one thats faster. In theory, if Virginia Tech had 3 times the budget, they could have bought 3 times as much hardware and thus been #1 on the list, and it STILL would have been a fraction of the price of its competitors. My point is, you get a lot more bang for your buck with nodes, regardless of who makes it.

I think the days of "custom designed" supercomputers, like Cray and Sun will come to an end because of advances in clustering technology. I'm not saying that Apple in particular is any threat at all to supercomputer manufacturers, I just think we're going to see a change in the way your standard supercomputer is constructed.

Just for kicks

http://darklotus.dyndns.org/temp/applestore_G5_node.pdf

**fryke** · #5 January 12th, 2004, 05:02 PM

Well, however you look at it, the journalist just seems 'pi*** o**' with the fact that Apple made it to the top 10, and at number 3. It's quite a short article, and it basically says nothing but: "I whine if I want to about Apple."

**btoneill** · #6 January 12th, 2004, 05:11 PM

Quote:

Originally Posted by Fahrvergnuugen

I
I think the days of "custom designed" supercomputers, like Cray and Sun will come to an end because of advances in clustering technology. I'm not saying that Apple in particular is any threat at all to supercomputer manufacturers, I just think we're going to see a change in the way your standard supercomputer is constructed.

Not for corporations. You can buy a 100 processor Sun box, that is one single server, running one single OS image, with one single shared memory segment. You can take any multithreaded application and it will run using as many processors as threads that the application will run, using as much memory as the application can use. In order to use a cluster, you have to write your software to work with a cluster API which enables it to run on different nodes, but it can be very difficult to change existing applications to use a cluster API, as the processing model is completely different. Not to mention issues with shared storage with clusters. It's easy to take a 100 processor Sun box, stick Oracle on it, and a big honking storage array, and you can take advantage of it. Try doing that with a cluster of 100 different boxes.

Clustered supercomputers have a very limited use, mainly for scientific calculations. When it comes to things like real time applications, their use goes down significantly without very specialized applications, and having the staff to write and administer them.

Brian

superfula · #7 January 12th, 2004, 05:13 PM

No it's a very valid arguement. I whole heartedly disagree that custom designed supercomputers are coming to an end. The only reason VT was able to do it so cheaply was because of numerous hours of volunteer help. Businesses would be forced to pay people a large some of money that schools can do for free. Node supercomputers are great for some areas, but custom designed supercomputers are still the best cost-effective way.

Dominyo · #8 January 12th, 2004, 05:17 PM

Quote:

Originally Posted by superfula

No it's a very valid arguement. I whole heartedly disagree that custom designed supercomputers are coming to an end. The only reason VT was able to do it so cheaply was because of numerous hours of volunteer help. Businesses would be forced to pay people a large some of money that schools can do for free. Node supercomputers are great for some areas, but custom designed supercomputers are still the best cost-effective way.

Even if VT had had to pay all those helping set up the suprcomputer, it would have still turned out to be significantly cheaper than a custom supercomputer of similar power...