French
After publishing our article about Microsoft’s opt-in native NVMe feature on Windows Server 2025, we received numerous requests for a direct storage performance comparison between Windows Server 2025 (with native NVMe support) and a Linux-based server OS. One particularly enthusiastic Redditor even offered us beer to conduct the test! With such overwhelming demand, we had no choice but to run the same benchmark suite on Linux to deliver the head-to-head comparison our readers wanted.
Windows Server 2025 NVMe vs Linux performance ssds
A Long Time Ago, in an OS Version Far, Far Away
Linux has supported NVMe since kernel version 3.3, released in March 2012. Similarly, Windows Server has offered NVMe support (non-natively, via SCSI translation) since the 2012 R2 release around October 2013. More than a decade later, the debate over whether Windows or Linux delivers better storage performance still rages on—and we’re adding more data to the conversation with our benchmark results comparing the two operating systems.
Since we already had test results for Windows Server 2025 using both non-native and native storage stacks, we decided to evaluate two popular storage stacks on Linux for a fair comparison. For our FIO benchmarks, we used libaio and io_uring—two of the most widely used APIs for storage transactions. While io_uring is significantly newer and offers numerous improvements for asynchronous I/O, libaio remains a staple due to its flexibility and ease of use (Didona, Pfefferle, Ioannou, Metzler, & Trivedi, 2022). A full architectural breakdown of both stacks is beyond the scope of this article, but we’re providing comprehensive results to enable direct side-by-side comparison.
Testing NVMe on Ubuntu Server 24.04.4 LTS
We used the same hardware platform for this comparison as in our Windows Server 2025 native NVMe article, ensuring consistency and accurate performance comparisons. The server is equipped with two 128-core AMD EPYC 9754 CPUs, 768GB of DDR5 memory running at 4800 MT/s, and fifteen 30.72 TB Solidigm P5316 NVMe SSDs (PCIe 4.0) configured in JBOD mode.
As noted in our previous article, the Solidigm P5316 has a 64-kilobyte indirection unit size, which often results in lower-than-expected write performance for smaller block sizes (such as 4K tests). To provide a comprehensive view, we ran tests with block sizes of 4K, 64K, and 128K, measuring both read and write operations across all configurations.
We selected Ubuntu Server 24.04.4 LTS as our Linux representative due to its widespread popularity and long-term support (LTS) status. It ships with Linux kernel 6.8 by default—while not the newest available, this kernel version likely represents a large portion of real-world Linux server installations globally.
Highlights
-
Windows Server 2025 (native NVMe) outperforms Ubuntu in 3 out of 4 read performance benchmarks
-
Windows Server consistently exhibited lower CPU usage during most tests
-
Ubuntu Server 24.04.4 LTS wins in 3 out of 4 write performance benchmarks
Benchmark Results
Random Read Performance
|
Metric
|
Windows Non-Native
|
Windows Native
|
Linux libaio
|
Linux io_uring
|
Windows Non-Native
|
Windows Native
|
Linux libaio
|
Linux io_uring
|
|---|---|---|---|---|---|---|---|---|
|
Block Size
|
Random 4K
|
Random 64K
|
||||||
|
Bandwidth (GiB/s)
|
6.1
|
10.058
|
9.198
|
9.504
|
74.291
|
91.165
|
77.517
|
77.7
|
|
IOPS
|
1,598,959
|
2,636,516
|
2,411,000
|
2,491,000
|
1,217,176
|
1,493,637
|
1,270,000
|
1,273,000
|
|
Average latency (ms)
|
0.169
|
0.104
|
0.198
|
0.192
|
0.239
|
0.207
|
0.377
|
0.376
|
|
Total CPU Usage (%)
|
72.67
|
74.22
|
99.77
|
99.76
|
68.44
|
65.11
|
83.16
|
84.72
|
Sequential Read Performance
|
Metric
|
Windows Non-Native
|
Windows Native
|
Linux libaio
|
Linux io_uring
|
Windows Non-Native
|
Windows Native
|
Linux libaio
|
Linux io_uring
|
|---|---|---|---|---|---|---|---|---|
|
Block Size
|
Sequential 64K
|
Sequential 128K
|
||||||
|
Bandwidth (GiB/s)
|
35.596
|
35.623
|
31.867
|
31.433
|
86.791
|
92.562
|
97.05
|
97
|
|
IOPS
|
583,192
|
583,638
|
522,000
|
515,000
|
710,978
|
758,252
|
795,000
|
795,000
|
|
Average latency (ms)
|
0.809
|
0.812
|
0.919
|
0.932
|
0.613
|
0.608
|
0.603
|
0.604
|
|
Total CPU Usage (%)
|
44.89
|
37.11
|
53.94
|
41.74
|
61.56
|
49.56
|
75.14
|
76.90
|
Random Write Performance
|
Metric
|
Windows Non-Native
|
Windows Native
|
Linux libaio
|
Linux io_uring
|
Windows Non-Native
|
Windows Native
|
Linux libaio
|
Linux io_uring
|
|---|---|---|---|---|---|---|---|---|
|
Block Size
|
Random 4K
|
Random 64K
|
||||||
|
Bandwidth (GiB/s)
|
1.803
|
1.756
|
1.876
|
1.815
|
7.654
|
7.655
|
7.652
|
7.651
|
|
IOPS
|
472,725
|
460,383
|
492,000
|
476,000
|
125,391
|
125,406
|
125,000
|
125,000
|
|
Average latency (ms)
|
0.992
|
1.028
|
0.974
|
1.007
|
3.814
|
3.816
|
3.827
|
3.828
|
|
Total CPU Usage (%)
|
26.00
|
20.67
|
45.76
|
22.80
|
12.22
|
9.33
|
20.07
|
10.90
|
Sequential Write Performance
|
Metric
|
Windows Non-Native
|
Windows Native
|
Linux libaio
|
Linux io_uring
|
Windows Non-Native
|
Windows Native
|
Linux libaio
|
Linux io_uring
|
|---|---|---|---|---|---|---|---|---|
|
Block Size
|
Sequential 64K
|
Sequential 128K
|
||||||
|
Bandwidth (GiB/s)
|
44.67
|
50.087
|
52.283
|
52.25
|
50.477
|
50.079
|
52
|
52.083
|
|
IOPS
|
731,859
|
820,603
|
856,000
|
856,000
|
413,495
|
410,232
|
426,000
|
427,000
|
|
Average latency (ms)
|
0.399
|
0.558
|
0.560
|
0.560
|
1.022
|
1.149
|
1.126
|
1.125
|
|
Total CPU Usage (%)
|
70.44
|
57.78
|
61.88
|
62.75
|
58.44
|
47.33
|
61.49
|
44.27
|
Note: Our Linux IOPS results are rounded to the nearest thousand due to differences in FIO reporting between Windows Server 2025 and Ubuntu Server 24.04.4 LTS. Bandwidth, latency, and CPU usage results are rounded consistently across both platforms.
The Numbers Don’t Lie
Our benchmarks make it clear that Ubuntu does not outperform Windows in every category. While libaio and io_uring delivered strong throughput in random-read tests, they could not match the performance of Microsoft’s native NVMe stack. The Windows NT kernel outperformed the Linux kernel by approximately 17% in our random-read 64K tests, with Windows native NVMe achieving 91.165 GiB/s versus io_uring’s best of 77.7 GiB/s.
However, Linux is not without its wins. Ubuntu Server narrowly edged out Windows Server in one read performance benchmark: the sequential 128K test. Here, Linux’s libaio performed best at 97.05 GiB/s, compared to Windows’ native NVMe at 92.562 GiB/s—a difference of about 5%. This suggests Linux may hold a slight advantage when managing block sizes larger than the drives’ indirection units.
Random write bandwidth was consistent across both operating systems, particularly in 64K benchmarks. The best and worst results in these tests differed by only 0.05%, indicating that all storage stacks were able to fully utilize the drives’ potential.
Interestingly, the Linux 6.8 kernel claimed victory in sequential write bandwidth tests for 64K and 128K block sizes. While the difference was not dramatic, the open-source storage stacks outperformed Windows Server’s native NVMe by approximately 2 GiB/s in both cases.
Latency results generally mirrored throughput performance, most notably in random read averages. Unfortunately for Linux, libaio and io_uring exhibited higher latency, with the largest gap seen in 64K random reads: Windows Server native NVMe had an average latency of 0.207 ms, compared to libaio’s 0.377 ms—a difference of 0.17 ms.
Perhaps the most surprising finding from our benchmarks is the significant difference in CPU usage between Windows Server 2025 and Ubuntu Server 24.04.4 LTS. In 3 out of 4 random and sequential-read benchmarks, Windows Server native NVMe had the lowest CPU usage. The most striking result was in the sequential read 128K test, where Windows used 27.34% less CPU than Linux.
Libaio and io_uring performed slightly better in random and sequential write tests, but not well enough to prevent Windows Server’s native NVMe from winning 3 out of 4 write CPU usage benchmarks. A notable exception was libaio’s CPU usage during the random write 4K test, which reached 45.76%—far higher than the ~20% CPU usage seen with other storage stacks.
Winner Winner, CPU Dinner
Our results show that Windows Server and Ubuntu Server perform closely in head-to-head random and sequential performance tests across different block sizes. In terms of bandwidth, Windows Server 2025 with native NVMe generally outperformed Linux in most read tests, while Linux delivered slightly better results in write tests. Latency figures followed a similar pattern, but the standout advantage was Windows Server 2025’s CPU efficiency when using native NVMe.
Microsoft has clearly invested heavily in refining its newest storage stack, and while it does not win every category against libaio and io_uring, it puts up a strong fight. These results are not definitive across all use cases and server configurations, but they provide valuable insights for server administrators deciding whether to deploy Windows or Linux when storage performance is a top priority—above OS compatibility.
Let us know what you think about these results by commenting on our social platforms or the SR Discord! Did you expect Windows Server to perform so well, or were you rooting for Linux? Would you like to see more Linux distributions or kernels tested? We’re always eager for your feedback, and reader-requested tests like this one are often our favorites.
Beijing Qianxing Jietong Technology Co., Ltd.
Sandy Yang/Global Strategy Director
WhatsApp / WeChat: +86 13426366826
Email: yangyd@qianxingdata.com
Website: www.qianxingdata.com/www.storagesserver.com
Business Focus:
ICT Product Distribution/System Integration & Services/Infrastructure Solutions
With 20+ years of IT distribution experience, we partner with leading global brands to deliver reliable products and professional services.
“Using Technology to Build an Intelligent World”Your Trusted ICT Product Service Provider!
Sandy Yang/Global Strategy Director
WhatsApp / WeChat: +86 13426366826
Email: yangyd@qianxingdata.com
Website: www.qianxingdata.com/www.storagesserver.com
Business Focus:
ICT Product Distribution/System Integration & Services/Infrastructure Solutions
With 20+ years of IT distribution experience, we partner with leading global brands to deliver reliable products and professional services.
“Using Technology to Build an Intelligent World”Your Trusted ICT Product Service Provider!