Shyamal Roy of GEOMATE Company has years of experience working with advanced 64-bit IBM mainframe computers, and currently writes analysis software for MCAD. He provided WorldCAD Access with his opinion of the benefits of 64-bit computing to CAD:
When developing CAD software solutions, we need to deal with four fundamental objectives: usability, speed, depth, and domain of the solution. Let's look at the 64-bit computing possibilities in this
context.
Usability has noting to do with size of computing, whether 1 bit or 64 bits. Usability does not change significantly if computing speed changes. Faster computing may bring more haptic senses into play some day, if the vendors so desire.
Speed in CAD software consists of three distinct elements: graphical performance, data handling, and computation performance. Graphical performance is more affected by the display-list software design and GPU performance, rather than the performance of CPU (whether 32 bits or 64 bits). Data handling will improve with 64-bit performance. Computational performance will improve when software is modified to take advantage of 64 bits, a non-trivial task.
Depth of solution has been stagnant in most CAD applications for several years and is function of business goals of developers, rather than CPU path size (32 bit or 64 bits).
Domain of solution of CAD applications should increase, because 64-bit computing will introduce much better data handling capabilities.
I had a casual conversation with a fellow software developer in Silicon Valley, currently working on the 64-bit processor development. He commented that the throughput increase with 64-bit computing, without major architectural changes in applications, are expected to be less than 7% overall.
CAD users may be better off not holding their breath on 64-bit computing for at least 5-7 years.
No, 64-bit is actually closer to 0% faster.
The most significant performance bottlenecks in modern CAD programs come from memory allocation overhead, and operating system calls -- things which 64-bit processors do nothing to improve.
Floating point (math) operations have mostly been 64-bit for the last 20 years. So there's no gain there.
The only certain advantage that 64-bit processors have is the ability to handle large amounts of memory.
Since companies such as AMD are putting all their heavy development behind their 64-bit processors, you can generally expect that they will be faster than 32-bit processors. But not because of having more bits to work with.
I have real-world experience with 64-bit optimized software architecture. I see literally *no* performance gain with 64-bit computing versus 32-bit computing.
Where I do see a big gain is with multi-core processors -- but this is when using a multi-threaded object model. You won't find that kind of technology in any mid-market CAD products that I know of.
The only CAD vendors who are likely to have any kind of multi-threaded support in their products are those with a Unix heritage -- Pro/E, Catia, and NX, in particular. I believe UGS has demostrated some performance improvements with multi-core processors, but I'd have to look it up to be sure.
In contrast, CAE vendors tend to develop multi-threaded programs. The nature of the problem lends itself to this. So, if you're an FEA or simulation user, you can certainly justify a multi-core machine.
Posted by: Evan Yares | Dec 15, 2005 at 07:09 AM