+353-1-416-8900REST OF WORLD
+44-20-3973-8888REST OF WORLD
1-917-300-0470EAST COAST U.S
1-800-526-8630U.S. (TOLL FREE)

PRINTER FRIENDLY

Signals Ahead: Chips And Salsa XVI – Sweet 16 And Never Been Benchmarked

  • ID: 2483987
  • Report
  • February 2013
  • Region: Global
  • 49 Pages
  • Signals Research Group, LLC
A comprehensive performance benchmark study of LTE chipsets from Qualcomm, Samsung, GCT, Altair, Sequans, Intel, Renesas Mobile, and Invidia.

This report is included as part of a corporate subscription to Signals Ahead or it can be purchased separately.

KEY HIGHLIGHTS

- A staggering eight different chipsets from eight different chipset suppliers have been benchmarked, including both pre-commercial and commercial chipsets.

- All 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.

- Three chipsets vied for top honors but ultimately we had to declare one the winner.

ABOUT SIGNALS AHEAD SUBSCRIPTION

Signals Ahead is a research-focused product that is published on a periodic basis. Its clientele include all facets of the wireless ecosystem, including some of the largest mobile operators, the top handset suppliers, the major infrastructure vendors, subsystem suppliers, semiconductor companies and financial institutions, including Wall Street, Private Equity and Venture Capitalists, spread across five continents.
Note: Product cover images may vary from those shown
Table of Contents

Executive Summary

Introduction
key Conclusions and observations

The results – Sliced and diced Several different ways

Detailed results and Analysis
Detailed Results and Analysis – Static Channel Conditions
Detailed Results and Analysis – Extended Vehicular A (EVA5)
Detailed Results and Analysis – Extended Pedestrian A (EPA5)
Detailed Results and Analysis – Extended Typical Urban (ETU70)
Detailed Results and Analysis – Extended Typical Urban (ETU300)
Detailed Results and Analysis – Bi-directional stress tests

Test methodology

Final Thoughts

Appendix 1 – results from other Test Scenarios
Appendix 2 – results for all Test Scenarios

List of Figures
Index of Figures
Figure 1. Static Channel Conditions, Transmission Mode 2 (TS1 and TS6)
Figure 2. Static Channel Conditions, Transmission Mode 4 (TS21 and TS26)
Figure 3. EVA5 Channel Model, Transmission Mode 2, SNR = 10, MIMO Correlation = Mid; (TS3 and TS8)
Figure 4. EVA5 Channel Model, Transmission Mode 3, SNR = 10, MIMO Correlation = Mid; (TS13 and TS18)
Figure 5. Transmit Diversity (TS 3) versus Open Loop MIMO (TS 13) – deep dive analysis
Figure 6. EPA5 Channel Model, Transmission Mode 3, SNR = 20, MIMO Correlation = Mid; (TS12 and TS17)
Figure 7. EPA5 Channel Model, Transmission Mode 4, SNR = 20, MIMO Correlation = Mid; (TS22 and TS27)
Figure 8. NVIDIA versus Intel – detailed analysis of Test Scenario 12
Figure 9. ETU70 Channel Model, Transmission Mode 2, SNR = 0, MIMO Correlation = Low; (TS4 and TS9)
Figure 10. ETU70 Channel Model, Transmission Mode 3, SNR = 0, MIMO Correlation = Low; (TS14 and TS19)
Figure 11. Test Scenario 14 – Deep dive analysis
Figure 12. ETU300 Channel Model, Transmission Mode 2, SNR = 0, MIMO Correlation = Low; (TS5 and TS10)
Figure 13. ETU300 Channel Model, Transmission Mode 3, SNR = 0, MIMO Correlation = Low; (TS15 and TS20)
Figure 14. Extended Period FTP Stress Test, Static Channel Conditions, FTP, Transmission Mode 3, MIMO Correlation = Low (TS31)
Figure 15. Bi-directional Stress Test, Static Channel Conditions, UDP, Transmission Mode 3, MIMO Correlation = Low, (TS32)
Figure 16. Static Channel Conditions, Transmission Mode 3, (TS11 and TS16)
Figure 17. EPA5 Channel Model, Transmission Mode 2, SNR = 20, MIMO Correlation = Mid; (TS2 and TS7)

Index of Tables
Table 1. List of Participants
Table 2. Participants and Supporting Information
Table 3. Summary of Results - Overall
Table 4. Summary of Results - Transmission Mode 2 Test Scenarios
Table 5. Summary of Results - Transmission Mode 3 Test Scenarios
Table 6. Summary of Results - Transmission Mode 4 Test Scenarios
Table 7. Summary of Results - FTP Test Scenarios
Table 8. Summary of Results - UDP Test Scenarios
Table 9. Summary of Results - Static Channel Test Scenarios
Table 10. Summary of Results - EPA5 Channel Test Scenarios
Table 11. Summary of Results - EVA5 Channel Test Scenarios
Table 12. Summary of Results - ETU70 Channel Test Scenarios
Table 13. Summary of Results - ETU300 Channel Test Scenarios
Table 14. Test Scenarios
Table 15. ITU Channel Models
Table 16. CQI Mapping to Transport Block Size
Table 17. Detailed Test Results - Test Scenarios 1, 6, 11, and 16
Table 18. Detailed Test Results - Test Scenarios 21 and 26
Table 19. Detailed Test Results - Test Scenarios 2, 3, 4, and 5
Table 20. Detailed Test Results - Test Scenarios 7, 8, 9, and 10
Table 21. Detailed Test Results - Test Scenarios 12, 13, 14, and 15
Table 22. Detailed Test Results - Test Scenarios 17, 18, 19, and 20
Table 23. Detailed Test Results - Test Scenarios 22, 23, 24, and 25
Table 24. Detailed Test Results - Test Scenarios 27, 28, 29, and 30
Table 25. Detailed Test Results - Test Scenarios 31 and 32
Note: Product cover images may vary from those shown

Loading
LOADING...

In December 2011 we published the industry’s first performance bench-mark study of LTE baseband modem chipsets. In that study we tested fivecommercially-procured chipsets from four chipset suppliers. We testedtwo different Qualcomm chipsets. Fast forward fourteen months and weare finally out with the results from our most recent study in which threecompanies vie for top honors. Intel’s pre-commercial solution was thetop-performing solution that we tested.

This report is our sixteenth Chips and Salsa report since 2004, with the overwhelming majorityof these reports focused specifically on performance benchmarking. Over the years, we’ve benchmarked UMTS (call reliability) HSDPA, HSPA+, Mobile WiMAX, A-GNSS and LTE chipsets, with the results always providing the industry with a fully independent and objective assessment of how the chipsets compare with each other for the given set of evaluation criteria. For the eighth time, we have collaborated with Spirent Communications to get access to their 8100 test system and engineering support in order to obtain highly objective results.

The significant advantage of conducting lab-based tests is that we can easily replicate and repeat each test scenario in an automated fashion, thus ensuring a common and consistent set of test scenarios for each device/chipset that we tested. And with the Spirent 8100 test system that we used for the tests, we know that we went with a test platform that is widely recognized and being used in several early LTE deployments. SRG takes full responsibility for the analysis and conclusions associated with this benchmarking exercise.

In the most recent round of chipset testing, we tested a seemingly staggering number of solutions – we tested solutions from eight different chipset suppliers (reference Table 1). We attempted to test a solution from HiSilicon, but through no fault of their own we ran into some difficulties and faced
time constraints with MWC just around the corner. We reserve the right to publish their results in the near future and provide updated rankings. Many of these solutions were pre-commercial chipsets and/or the chipsets that came directly from the chipset suppliers. This approach ensured that the results that we are providing in this report are very forward looking and highly differentiated. It would be virtually impossible for any single organization to get
access to all of these chipsets and replicate this study.

Worth noting, we personally invited all companies with LTE chipset aspirations to participate in this study, and given our history in doing these tests, companies recognize the importance of supporting our efforts. Needless to say, if we didn’t include a company’s LTE chipset in this study then they probably don’t have a solution that is ready to be benchmarked against their peers. It is one thing to issue a press release, demonstrate a working PHY Layer without any upper protocol layers, or show a chipset operating under ideal conditions. It is another situation all together to put your proverbial money where your mouth is and allow a third party to benchmark your solution and publish the results for all to read. Sweet 16 and never been benchmarked!

As previously alluded to in this report, we used throughput as the primary criteria for evaluating the chipsets. We recognize that device manufacturers and operators use other objective and subjective criteria to select their chipset partners. The criteria includes support for multiple RF bands and legacy
technologies, power consumption, time to market, price, engineering support, and the inclusion of peripherals (e.g., application processor, connectivity
solutions, etc.). However, no one can dispute the importance of throughput and the ability of the chipset to make the most efficient use of available network resources.

We subjected the chipsets to 32 different test scenarios that combined a mix of fading profiles (Static Channel, EPA5, EVA5, ETU70 and ETU300) and transmission modes (Transmit Diversity, Open Loop MIMO and Closed Loop MIMO). All of the chipsets that we tested performed quite well with the less challenging test scenarios but we observed a fairly large separation of results with the more challenging test scenarios. In many cases the performance difference was in excess of 20% between the top- and bottom-performing solutions.

Based on our highly objective evaluation criteria, Intel had the top-performing solution by a very slight margin. This result may surprise some readers, but we point out that the Infineon 3G solution was always a strong contender in our previous benchmark studies. That scenario is in stark contrast
to its application processor which has continuously struggled to be competitive and to attract market share. Don’t throw the baby out with the bathwater. All this and more in this issue of Signals Ahead.
Note: Product cover images may vary from those shown
Adroll
adroll