Microsoft has recently introduced the new V6 SKUs into Azure, bringing a shift in performance standards. In today’s blog post, we’ll give an overview of how the V6 compares to its predecessor, the V5. We will cover both performance and cost. We’ll also discuss whether the upgrade is worthwhile for your end users.
Exploring the Performance Differences
The V6 series boasts enhanced hardware support. This is a noticeable improvement compared to the V5. It’s akin to the difference between a new and an old PC. Our testing focused on workloads common within the Azure Virtual Desktop environment, explicitly using AMD SKUs. A side-by-side comparison between V3, V5, and V6 virtual machines (“VMs”) reveals the evolution and improvements in each version. The V5 VMs use AMD’s EPYC 7763v processor with Zen 3 architecture. They feature DDR4 memory and PCI Express 4. These elements offer robust performance. In our tests, the V5 configuration had four CPUs. It also had 16GBs of RAM. The local and remote storage capacity supports high throughput. This setup is still widely used and preferred by many. Moving on to the V6, significant enhancements come into play. The V6 uses the AMD EPYC 9004 with Zen 4 architecture. It also features DDR5 memory. This provides a leap in memory speed. The PCI Express 5 ability further boosts performance. These advancements signal a much faster, and thus better, performance for VMs. Moreover, the transition from SSD to NVMe storage in the V6 further enhances VM efficiency, notably when using ephemeral disks.
For the testing we used the D4ADS_V5 and D4ADS_V6 SKU’s for side-by-side comparisons.
Benchmark Results and Analysis
In our tests using benchmarking tool Geekbench, the performance differences between V3, V5, and V6 VMs were clearly delineated. The single-core and multi-core scores improved dramatically from V3 to V6. The V6 showed a significant 32% increase in speed compared to the V5. Specifically, the single-core score jumped from 1048 (V3) to 4442 (V6), demonstrating the ability of the updated architecture. The comprehensive score changes are mainly attributed to the updated clock speeds and architectural designs. The V6 features a peak clock speed of 3.6 GHz compared to 3.2 GHz in the V5, alongside enhanced cache capabilities that promise improved performance for intensive workloads.
Cost Analysis: Is the Upgrade Worth It?
Enhanced performance naturally comes with a question of cost viability. The price per hour for a D4ADS_V5 VM is 39 cents. The V6 version is slightly higher at 41 cents. Based on an average of 160 operational hours per month, the V6 incurs an extra $3.20 monthly, a reasonable increase given the performance boost. For environments running a thousand VMs, the added cost is significant. It totals an extra $3,200 monthly. This cost is justifiable given the performance benefits. This is especially true in multi-session desktops where user experience is significantly improved.
| Single-Core Score | Multi-Core Score | |
| D4S_V3 | 1048 | 2190 |
| D4ADS_V5 | 1513 | 3372 |
| D4ADS_V6 | 1994 | 4442 |
The individual scores can be seen here:
D4S_V3 – https://browser.geekbench.com/v6/cpu/9154521
D4ADS_V5 – https://browser.geekbench.com/v6/cpu/9153603
D4ADS_V6 – https://browser.geekbench.com/v6/cpu/9153473
Conclusion
In conclusion, transitioning from V5 to V6 Azure VMs offers significant performance enhancements due to advancements in architecture and storage. While there are associated costs, the cost-to-performance ratio appears favorable, especially for environments with higher workloads. Future explorations into Intel SKUs can further refine the decision-making process. Direct AMD-Intel comparisons would also help organizations seeking optimal VM performance. If you are interested in visual demonstrations, you can stay tuned to our virtualmanc YouTube Channel. We will dive deeper into specific configurations and continue to explore the latest Azure developments.
If you have topics you’d like us to cover, feel free to reach out. See you in the next post! You can see a YouTube video in which we discuss this topic here: