HP’s Intel® Xeon® Dual Quad-Core Workstation - The ability to reduce time itself!
HP’s latest Dual Quad-Core entry into the high performance workstation arena is delivering cost saving results that have created more than just buzz in the CAD/CAM industry. It’s more like a loud roar! For this article I contacted vendors who make the design tools used in architecture, mechanical design, manufacturing, digital content creation, and electronic design automation and to a fault each of them either has products specifically designed to take advantage of this powerful threaded 64-Bit computing technology or has products in the wings awaiting imminent release.
HP loaned me a brand-spanking new HP xw8600 workstation loaded with dual 64-bit Quad-Core Intel® Xeon® processors, 4GB of RAM, a high-end NVIDIA Quadro FX4600 video accelerator, a 320 GB of hard drive, and equipped with an MS Windows XP operating system. I booted up the machine and after being surprised by the lack of any noise --- this is the quietest machine I’ve ever used --- I began my search for a test case that would help me quantify the buzz.
While doing my research I located a perfect set of benchmark tests from CAM (Computer Aided Manufacturing) software manufacturer HSMWorks and put my test computer through its paces by emulating a real-life scenario that clearly shows the benefits of supporting multiple cores in a sophisticated application.
In the world of CAM, generating a toolpath for a CNC milling machine includes many tasks, such as feature loading, linking, ordering and more. Each of these tasks must be completed before staring the next, so to efficiently utilize multiple cores each, step in the process must also support multiple cores.
Using HSMWorks software I ran contour toolpath calculations on a very complex mold similar to the one shown in Figure 1. The detail and distance between each pin on the mold is an extremely small 1/16 of an inch, the customer wanted to machine the part with a very small 1/21 inch diameter tool and as a result the amount of material removed as the tool moves around each pin must be very small to achieve a good surface finish while avoiding breaking the tool. Much of this toolpath is generated at very tight tolerances which resulted in a very large NC file of more than 10 million lines, and before using the HP xw8600 workstation this led to very long calculation times. Also, each pin on the model is slightly different in shape and size, making it impossible to simply calculate one and then repeat the toolpath.
Figure 1. One half of a 10 x 10 inch plastic injection mold
For the benchmark I used a contour pass since it is the best strategy for finishing steep walls, but can be used for semi-finish and finish machining on the more vertical areas of a part.
Figure 2. Detail - One Mold Pin
Processing Time Reduced from 2 Days to Just 26 Minutes
The part itself is the core half of a 10 x 10 inch plastic injection mold supplied by one of HSMWork’s customers. After trying, unsuccessfully, to create a toolpath with several other CAM systems, the customer was forced to split the part into 4 smaller parts requiring less memory, and then calculate each segment separately, each taking approximately 12 hours to calculate. This resulted in a total of 2 days calculation time. The customer, obviously frustrated, turned to HSMWorks’ multi-core accessible software to see what they could do with the part. By taking full advantage of HP’s xw8600 workstation HSMWorks was able to calculate this complex toolpath in just 26 minutes!
Using my HP loaner I calculated the contour tool path on a 10 x 10 mold similar to the one shown above and was rewarded by these amazing, results:
This test’s most revealing result was the differential between running the job on 1 core verses 8 which resulted in a whopping 6.5X improvement in time saving performance! The 40 minute job for 1 core was reduced to just over 6 minutes using all 8 cores.
Note: So I didn’t have to wait a day or so to process the part on a single CPU I increased the tolerance and tool size for my test part. The ratios, however, are in exact relation and proportion to the size of the original part, the tolerances used, and the diameter of the tool. These results show the impact that multi-core processing has on toolpath generation.
32 vs. 64-bit Computing
As the use of multi-core PCs grows it is important to understand the difference between 32-bit and 64-bit applications. First, the difference is all about RAM memory and has nothing to do with speed from an application perspective. A 32-bit application is not able to use or access more than 2 to 4 GB of RAM, regardless of how much RAM is installed in a PC. 64-bit applications do not have this limitation. In the real world, 32-bit applications that need to execute more complex operations, and therefore require more than 2 to 4 GB of RAM, must manually divide the operation into smaller operations. This must be done by the user, often requiring lots of guesswork and trial-and-error. The result is wasted time when an operator guesses wrong and must rerun the scenario when it is discovered that one or more tasks could not be completed due to memory limitations.
In the machining example I used above, without the proper software the customer had to split their part into four sections because their 32-bit application couldn’t take advantage of the extra RAM installed on their machine. With HSMWorks they were able to process the entire part because the software fully supports 64-bit operations and effectively uses the maximum RAM available on the xw8600 allowing most jobs to complete in just one pass.
The shift to Multi-Core
Computer processor manufacturing evolved to multi-core a few years ago as it became more and more difficult to increase CPU speed, and today multi-core processors are the standard. Although multi processor support is now very common in server applications there are few, but a definitely growing number of applications, on the market today that can effectively use multiple cores. As a result, some users today are disappointed when they realize that they are unable to take full advantage of their newly purchased powerhouse PC, and questions like “Why is my CAM system only using 50% of the horse power when calculating complex operations,” are still common.
This disappointment is rapidly evaporating as more and more vendors introduce products that take full advantage of multi-core threaded computing. As these software manufacturers bring their multi-core capable software to market time saving benefits similar to those realized by HSMWorks are clearly gaining prevalence across an increasing spectrum of industries.
Let me know about your threaded multi-core software
As this trend expands I plan to follow up this article with
more reports from many diverse industry sectors, and if you are a software
manufacturer, and your software takes advantage of threaded multi-core
so I can feature your experience in an upcoming report.