Recent performance testing on LTE chipsets by Spirent Communications revealed some interesting results: the top performer wasn’t market leader Qualcomm, but Intel.
Rather than focus solely on conformance testing, the test scenarios used by Spirent on its 8100 test system involved both static assessments and performance in fading scenarios that simulated a device being carried at walking speed, and while a user was driving or riding on a train.
Spirent director Brock Butler said that the more realistic test scenarios showed clear differences in chip performance.
“Half the tests were done in static conditions for the standard, very benign. In those cases, there were some differences between chips, but not a lot. If you were to test that way … you could pick a winner, but you couldn’t see a lot of room between them.
“I think you really need to test in a more realistic way, to show the performance differences that are going to be out in the real world,” Butler added.
The test involved eight different chipsets from eight suppliers: Altair Semiconductor, GCT (which has partnered with LG on LTE chipsets), Intel, Nvidia, Qualcomm, Renesas Mobile, Samsung Electronics, and Sequans Communications.
Signals Research said that Intel’s pre-commercial chips were the top performer, but that differences were small among the top three chipsets.
The performance testing was Spirent’s eighth collaboration with Signals Research, where the full report is available.
According to Signals Research’s summary of the report, “All of the chipsets performed well under less challenging conditions, but with the more challenging conditions, there was a wide variance in the results with more than a 20% difference between the top- and bottom-performing chipsets.”
Nigel Wright, vice president of corporate marketing for Spirent, said that many other chipset vendors were invited, but declined to participate. Some LTE chipsets are still in early development, he noted.
“In the most recent benchmark study, three chipsets distinguished themselves from their peers with Intel’s pre-commercial solution capturing top honors,” said Michael Thelander, CEO of Signals Research Group. “At the same time, all of the chipsets performed well in certain test scenarios, suggesting that with additional optimization they could achieve meaningful performance gains. The big question mark remains for those solutions which we were not able to include in our study, since there is a major difference between a product demonstration under ideal conditions and a chipset that is ready to participate in an independent benchmark study.”
The tests measured downlink throughput and application layer performance indicators of the chipsets under a range of network conditions, with test scenarios based on 3GPP technical reports.
The performance differences in chips are of interest to mobile operators as well as OEMs, Wright noted. Since devices report conditions to the network and request resources based on that information, if a chipset has a high error rate and can’t actually make use of such network resources, it affects the network’s overall resource sharing and capacity.
“It impacts network capacity,” he said. “If you have millions of good-performing devices on your network, it actually requires less network capacity than millions of poorly performing devices.”
Previous tests in the collaboration between Spirent and Signals Research have included performance differences between HSPA+ chipsets and devices.