Why Location Matters More Than the Average Score
Most benchmark articles give you one number per provider. Vultr's CPU score is 4,100. Done.
That number is technically correct. It is also misleading if you use it to make a deployment decision, because Vultr runs 9 different US datacenters on hardware that is not identical. The Silicon Valley location scores 4,240 on CPU. Honolulu scores 3,780. That is an 11% gap — larger than the difference between Vultr and DigitalOcean (2.5%), larger than the difference between Vultr and Linode (5.1%).
The same pattern holds for disk I/O. The best Vultr location (New Jersey, 54,200 read IOPS) outperforms the worst (Honolulu, 44,500 read IOPS) by 18%. At that point you are no longer comparing providers. You are comparing hardware generations within the same provider.
None of this means Vultr is bad. Every single location scored above average in our 13-provider test group. But if you care enough about performance to read a benchmark article, you should also care about which specific datacenter you deploy in. So that is what this article gives you: six months of data from every US location, not just the New Jersey number that everyone else tests.
Benchmark Methodology
I am going to be specific here because benchmark methodology is where most comparisons fall apart. Hand-wavy "we tested the servers" descriptions are not useful. Here is exactly what we did.
Hardware Under Test
Vultr Cloud Compute (Regular Performance) — the $5/month tier. 1 vCPU, 1 GB RAM, 25 GB SSD. We chose this plan because it is the entry point most people start with and the most directly comparable across providers. We deployed one instance in each of Vultr's 9 US locations:
- New Jersey (EWR)
- Chicago (ORD)
- Dallas (DFW)
- Seattle (SEA)
- Los Angeles (LAX)
- Atlanta (ATL)
- Silicon Valley (SJC)
- Miami (MIA)
- Honolulu (HNL)
Software Stack
Every instance runs Ubuntu 24.04 LTS, deployed from Vultr's stock image. No kernel tuning. No sysctl tweaks. No compiler optimizations. The goal is to measure what a real user gets when they spin up a server and start using it.
Test Tools & Parameters
| Category | Tool | Command / Parameters | What It Measures |
|---|---|---|---|
| CPU | sysbench 1.0.20 | sysbench cpu run --threads=1 --time=60 |
Single-thread compute throughput (events/sec) |
| Disk Read | fio 3.36 | fio --name=randread --ioengine=libaio --iodepth=32 --rw=randread --bs=4k --direct=1 --size=1G --numjobs=1 --runtime=60 |
Random 4K read IOPS (raw storage throughput) |
| Disk Write | fio 3.36 | fio --name=randwrite --ioengine=libaio --iodepth=32 --rw=randwrite --bs=4k --direct=1 --size=1G --numjobs=1 --runtime=60 |
Random 4K write IOPS |
| Network | iperf3 3.16 | iperf3 -c [test-server] -t 30 -P 4 |
TCP throughput to standardized US endpoints |
| Latency | ping / mtr | mtr -rwc 100 [gateway] |
Intra-datacenter round-trip latency |
Testing Schedule
Three sessions per week per datacenter, spread across different days and times. Each session runs every test 5 times. We take the median of each session, then report the median of all session medians over the 6-month window (September 2025 through February 2026). That is roughly 78 sessions per location, 390 individual test runs per metric per datacenter.
This approach eliminates noisy-neighbor outliers. When I say Silicon Valley scores 4,240, that number represents hundreds of data points, not one lucky afternoon. For the full cross-provider methodology, see our benchmark methodology page.
Per-Datacenter Breakdown
This is the section that does not exist in most Vultr benchmarks. Every location, all metrics, from six months of testing.
| Datacenter | CPU Score | Read IOPS | Write IOPS | Network (Mbps) | Latency (ms) |
|---|---|---|---|---|---|
| Silicon Valley (SJC) | 4,240 | 52,100 | 41,800 | 960 | 0.8 |
| New Jersey (EWR) | 4,180 | 54,200 | 43,100 | 965 | 0.7 |
| Chicago (ORD) | 4,160 | 51,400 | 41,200 | 955 | 0.8 |
| Dallas (DFW) | 4,120 | 50,800 | 40,500 | 950 | 0.9 |
| Los Angeles (LAX) | 4,100 | 49,900 | 40,100 | 948 | 0.9 |
| Seattle (SEA) | 4,080 | 49,200 | 39,400 | 945 | 1.0 |
| Atlanta (ATL) | 4,050 | 48,600 | 39,000 | 940 | 1.0 |
| Miami (MIA) | 3,960 | 46,800 | 37,500 | 935 | 1.1 |
| Honolulu (HNL) | 3,780 | 44,500 | 35,200 | 890 | 1.4 |
Green = top performer in that metric. Red tint = notably below average. All figures are 6-month medians.
What the Per-Datacenter Data Tells Us
Three patterns stand out.
First, the mainland locations cluster tightly. If you exclude Honolulu, the CPU variance drops from 11% to under 4%. Silicon Valley through Atlanta form a band of performance that is, for practical purposes, interchangeable. Pick whichever is closest to your users.
Second, New Jersey leads on disk I/O. It is the only location that cracked 54,000 read IOPS consistently. My hypothesis: EWR is Vultr's flagship US datacenter and likely gets hardware refreshes first. The write IOPS data supports this — New Jersey's 43,100 write IOPS is 9% above the network-wide average.
Third, Honolulu is a different tier. Every metric is meaningfully lower. The CPU gap alone (3,780 vs. the 4,100 average) would drop Vultr from #4 to #7 in our provider rankings if Honolulu were the only location tested. If your users are in Hawaii, Vultr Honolulu is still your best option — no other provider we test has a Hawaiian datacenter. But go in with calibrated expectations.
CPU Benchmark Results
Vultr's aggregate CPU score is 4,100 (sysbench events/sec, single-thread). That places it #4 out of 13 providers in our test group — solid top-tier territory, behind Hostinger (4,400), Hetzner (4,300), and Kamatera (4,250).
The 7.3% gap between Vultr and Hostinger sounds significant in a table. In practice, it means a PHP request that takes 48ms on Vultr takes 45ms on Hostinger — a difference that disappears into network latency and database queries. CPU score matters most for sustained compute: compilation, transcoding, ML inference. For web workloads, the more useful metric is consistency, and Vultr delivers sub-3% variance across 6 months.
| Provider | CPU Score | Price | CPU per $ | vs. Vultr |
|---|---|---|---|---|
| Hostinger | 4,400 | $6.49/mo | 678 | +7.3% |
| Hetzner | 4,300 | $4.59/mo | 937 | +4.9% |
| Kamatera | 4,250 | $4/mo | 1,063 | +3.7% |
| Vultr | 4,100 | $5/mo | 820 | — |
| ScalaHosting | 4,100 | $29.95/mo | 137 | 0% |
| DigitalOcean | 4,000 | $6/mo | 667 | -2.4% |
| Linode | 3,900 | $5/mo | 780 | -4.9% |
| BuyVM | 3,800 | $2/mo | 1,900 | -7.3% |
| RackNerd | 3,600 | $1.49/mo | 2,416 | -12.2% |
| Contabo | 3,500 | $5.49/mo | 637 | -14.6% |
| InterServer | 3,400 | $6/mo | 567 | -17.1% |
Notice the CPU-per-dollar column. Vultr's 820 points per dollar is respectable but not remarkable. Hetzner and Kamatera extract more compute per dollar. The budget providers (BuyVM, RackNerd) win on raw efficiency, though with trade-offs in features and support. If CPU throughput per dollar is your sole metric, Hetzner's benchmarks deserve your attention.
Disk I/O Results
Vultr delivered 50,000 read IOPS and 40,000 write IOPS on our 4K random I/O tests (median across all US locations). The read-to-write ratio of 1.25:1 is characteristic of SATA SSD-backed storage with a write-ahead buffer.
Solid mid-table numbers that comfortably clear every common workload. A busy WooCommerce site generates 200-400 random IOPS per page load. Vultr gives you 50,000. You need 10,000+ database inserts per second before write IOPS becomes a constraint.
| Provider | Read IOPS | Write IOPS | R/W Ratio | Storage Type | Price |
|---|---|---|---|---|---|
| Hostinger | 65,000 | 52,000 | 1.25 | NVMe | $6.49/mo |
| ScalaHosting | 58,000 | 46,000 | 1.26 | NVMe | $29.95/mo |
| DigitalOcean | 55,000 | 42,000 | 1.31 | NVMe | $6/mo |
| Hetzner | 52,000 | 44,000 | 1.18 | NVMe | $4.59/mo |
| Vultr | 50,000 | 40,000 | 1.25 | SSD | $5/mo |
| Linode | 48,000 | 36,000 | 1.33 | NVMe | $5/mo |
| Kamatera | 45,000 | 38,000 | 1.18 | SSD | $4/mo |
| BuyVM | 42,000 | 34,000 | 1.24 | SSD | $2/mo |
| Contabo | 28,000 | 22,000 | 1.27 | SSD | $5.49/mo |
The storage type column is worth attention. Several providers above Vultr in the IOPS rankings use NVMe, which has a fundamental speed advantage over SATA SSD. Vultr's Regular Compute tier still uses SATA SSDs in most locations. Their High Frequency Compute tier uses NVMe and closes much of the gap.
The raw IOPS numbers miss tail latency. Vultr's p99 I/O latency was 2.1ms read and 3.4ms write — clean numbers. Some providers with higher median IOPS show worse tail latency from noisy-neighbor effects. For databases, consistent low-tail-latency matters as much as peak throughput.
See full disk I/O comparison across all 13 providers →
Network Speed & Latency
Vultr achieved 950 Mbps on our iperf3 throughput tests, effectively saturating the advertised 1 Gbps port after protocol overhead. Intra-datacenter latency measured 0.9 ms median (0.7 ms at the best location, New Jersey).
Vultr maintains direct peering with over 20 tier-1 networks. For latency-sensitive applications — real-time APIs, WebSocket connections, game servers — the network layer is where Vultr genuinely competes with anyone.
| Provider | Throughput (Mbps) | Latency (ms) | US Datacenters | Price |
|---|---|---|---|---|
| DigitalOcean | 980 | 0.8 | 3 | $6/mo |
| Hetzner | 960 | 0.9 | 2 | $4.59/mo |
| Vultr | 950 | 0.9 | 9 | $5/mo |
| Linode | 940 | 1.0 | 4 | $5/mo |
| BuyVM | 940 | 1.1 | 2 | $2/mo |
| Kamatera | 920 | 1.2 | 4 | $4/mo |
| ScalaHosting | 920 | 1.1 | 2 | $29.95/mo |
| Hostinger | 900 | 1.3 | 2 | $6.49/mo |
| Contabo | 850 | 1.5 | 2 | $5.49/mo |
DigitalOcean has faster throughput (980 vs. 950 Mbps) but operates 3 US locations to Vultr's 9. If your users are in the Southeast, Vultr's Atlanta or Miami gets you 15-30ms closer than DigitalOcean's nearest option. That geographic advantage outweighs a 30 Mbps difference that only matters during bulk transfers.
See full network speed comparison →
Overall Performance Score
We calculate a weighted composite: 40% CPU + 30% Disk + 30% Network. Scores are normalized against the best performer in each category so the scale is 0-100.
| Component | Vultr Score | Category Best | Category Leader | Normalized | Weighted |
|---|---|---|---|---|---|
| CPU (40%) | 4,100 | 4,400 | Hostinger | 93.2% | 37.3 |
| Disk Read (30%) | 50,000 | 65,000 | Hostinger | 76.9% | 23.1 |
| Network (30%) | 950 | 980 | DigitalOcean | 96.9% | 29.1 |
| Overall Score | 89.5 / 100 | ||||
The 89.5 places Vultr #4 overall. Vultr never drops below 76.9% in any category — a floor higher than what many providers achieve as their ceiling. Per-datacenter, Silicon Valley would score 92.1 and Honolulu 81.3. Both above average, but a meaningful gap between them.
Detailed Provider Comparison
How does Vultr stack up against the providers you are most likely considering? Here is the full picture with every metric side by side.
| Provider | CPU | Disk Read | Disk Write | Network | Overall | Price |
|---|---|---|---|---|---|---|
| Hostinger | 4,400 | 65,000 | 52,000 | 900 | 94.8 | $6.49 |
| Hetzner | 4,300 | 52,000 | 44,000 | 960 | 93.6 | $4.59 |
| Kamatera | 4,250 | 45,000 | 38,000 | 920 | 90.2 | $4.00 |
| Vultr | 4,100 | 50,000 | 40,000 | 950 | 89.5 | $5.00 |
| DigitalOcean | 4,000 | 55,000 | 42,000 | 980 | 89.2 | $6.00 |
| Linode | 3,900 | 48,000 | 36,000 | 940 | 84.1 | $5.00 |
| ScalaHosting | 4,100 | 58,000 | 46,000 | 920 | 91.0 | $29.95 |
| BuyVM | 3,800 | 42,000 | 34,000 | 940 | 81.5 | $2.00 |
| RackNerd | 3,600 | 35,000 | 28,000 | 900 | 74.8 | $1.49 |
| Contabo | 3,500 | 28,000 | 22,000 | 850 | 68.2 | $5.49 |
Vultr and DigitalOcean are remarkably close (89.5 vs 89.2). The practical difference is datacenter count, API quality, and features — not performance. See our Vultr vs DigitalOcean comparison for the non-benchmark factors.
Value Score — Performance Per Dollar
At $5/month, Vultr's value score is 17.9 points per dollar (89.5 / 5). Here is where that falls in the rankings:
| Rank | Provider | Overall Score | Price | Value (pts/$) |
|---|---|---|---|---|
| 1 | BuyVM | 81.5 | $2.00 | 40.8 |
| 2 | RackNerd | 74.8 | $1.49 | 50.2 |
| 3 | Kamatera | 90.2 | $4.00 | 22.6 |
| 4 | Hetzner | 93.6 | $4.59 | 20.4 |
| 5 | Vultr | 89.5 | $5.00 | 17.9 |
| 6 | Linode | 84.1 | $5.00 | 16.8 |
| 7 | DigitalOcean | 89.2 | $6.00 | 14.9 |
| 8 | Hostinger | 94.8 | $6.49 | 14.6 |
| 9 | Contabo | 68.2 | $5.49 | 12.4 |
| 10 | ScalaHosting | 91.0 | $29.95 | 3.0 |
Budget providers dominate the value ranking because they deliver decent performance at rock-bottom prices. But "value" is misleading in isolation — BuyVM's 2 US datacenters and limited support are real trade-offs. Among mainstream providers, Vultr offers more performance per dollar than DigitalOcean or Hostinger, with more US coverage than Hetzner or Kamatera.
High Frequency Compute vs. Regular
Vultr sells two relevant tiers at the entry level: Regular Cloud Compute ($5/mo) and High Frequency Compute ($6/mo). The HFC tier uses AMD EPYC processors and NVMe storage. We benchmarked both.
| Metric | Regular ($5/mo) | High Frequency ($6/mo) | Difference |
|---|---|---|---|
| CPU Score | 4,100 | 4,680 | +14.1% |
| Disk Read IOPS | 50,000 | 64,000 | +28.0% |
| Disk Write IOPS | 40,000 | 51,500 | +28.8% |
| Network (Mbps) | 950 | 955 | +0.5% |
| Latency (ms) | 0.9 | 0.9 | 0% |
The HFC tier is a different product. A 14% CPU improvement and 28% disk improvement for $1/month extra is arguably the best dollar-for-dollar upgrade in the VPS market. The HFC CPU score of 4,680 would place Vultr #1 in our entire 13-provider ranking, ahead of Hostinger's 4,400. The disk IOPS would rank #2, just behind Hostinger.
The catch: HFC is not in all locations. As of March 2026, it is available in New Jersey, Chicago, Dallas, Los Angeles, Silicon Valley, and Atlanta. Seattle, Miami, and Honolulu only offer Regular Compute. If your target datacenter has HFC, take it.
Performance Consistency Over Time
A single benchmark is a snapshot. Six months of data shows you the movie. Here is how Vultr's scores moved over our testing window.
| Metric | Lowest Monthly Median | Highest Monthly Median | Variance |
|---|---|---|---|
| CPU Score | 4,040 (Nov 2025) | 4,170 (Jan 2026) | 3.1% |
| Disk Read IOPS | 48,800 (Oct 2025) | 51,200 (Feb 2026) | 4.7% |
| Disk Write IOPS | 39,100 (Oct 2025) | 41,000 (Feb 2026) | 4.6% |
| Network (Mbps) | 942 (Dec 2025) | 958 (Feb 2026) | 1.7% |
Sub-5% variance on every metric across 6 months. For comparison, some providers showed 15-20% disk I/O variance month-to-month from noisy-neighbor issues. Vultr's numbers suggest well-managed hypervisor density and storage provisioning.
The slight improvement from October to February likely reflects Vultr's ongoing hardware refresh cycle. Newer hardware entering rotation nudges the median upward — a good sign for long-term performance trends.
Best Use Cases Based on Benchmark Data
Not every workload cares about the same metrics. Here is where Vultr's specific performance profile fits best.
Multi-Region US Deployments
This is Vultr's strongest play. No other provider in our test group offers 9 US datacenter locations with consistent performance across all of them. If you need servers in both coasts plus central US, Vultr is the only option that does not require multi-provider complexity. The 4% mainland performance variance means your app behaves the same in New Jersey and Los Angeles.
Developer and DevOps Workflows
Vultr's API is the best in the industry. Combined with Terraform, Packer, and CLI support, you can spin up consistent-performance instances programmatically and trust the numbers. CI/CD pipelines and infrastructure-as-code workflows benefit from predictable benchmarks.
Web Applications and APIs
The balanced CPU/network profile handles request-response workloads well. A 4,100 CPU score processes Laravel, Django, Rails, and Express requests without bottlenecking. For standard web applications, Vultr's benchmarks are indistinguishable from the top 3 providers.
Game Servers
Low latency (0.9ms), geographic coverage (9 US locations), and free DDoS protection make Vultr one of the strongest game server platforms. The CPU score handles Minecraft, Valheim, and similar titles without issue. The network consistency — sub-2% variance — means players get reliable ping times.
Containers and Kubernetes
50,000 read IOPS handles image pulls efficiently. VKE comes with a free control plane. For production Kubernetes, consider the HFC tier for 28% better disk I/O.
Where Vultr Is Not the Best Fit
Storage-heavy workloads (Contabo offers 400 GB at $5.49 vs Vultr's 25 GB). Budget projects where every dollar counts (RackNerd starts at $1.49/mo). Raw disk I/O bottlenecks (Hostinger's 65,000 IOPS is 30% ahead).
Verdict
Vultr's aggregate scores tell a straightforward story: #4 out of 13, solid all-rounder, no weak spots. That is accurate but incomplete.
The more useful finding: your choice of Vultr location affects performance more than your choice between Vultr and its closest competitors. Silicon Valley Vultr outperforms average DigitalOcean. Honolulu Vultr underperforms average Linode.
My recommendation: pick the location closest to your users from the mainland locations (all within 4% of each other). If it offers HFC, take it — the $1/month upgrade is the best price-to-performance improvement in the VPS market. If you need Honolulu, expect 8-11% below mainland numbers.
Vultr is not the cheapest or the fastest in any single metric. It is the most geographically flexible US provider with consistently good performance and the best developer tooling. For most use cases, that matters more than winning a benchmark by 5%.
Try Vultr for Yourself
Get $100 in free credit and benchmark your own workload across any of the 9 US locations. Deploy in under 60 seconds.
Claim $100 Free Credit →Read full Vultr review | Vultr vs DigitalOcean | Vultr vs Linode | All benchmark results
Frequently Asked Questions
Which Vultr US datacenter has the best benchmark performance?
Silicon Valley consistently scored highest in our 6-month testing: CPU 4,240, disk read 52,800 IOPS, and 960 Mbps network. New Jersey and Chicago are close behind. But picking the closest datacenter to your users almost always matters more than picking the fastest one.
How does Vultr's CPU performance compare to DigitalOcean and Linode?
Vultr's average CPU score of 4,100 beats both DigitalOcean (4,000) and Linode (3,900). But the gap is small enough that workload differences matter more. All three handle web servers, APIs, and containers without issue. The real differentiator is Vultr's 9 US datacenter locations versus DigitalOcean's 3 and Linode's 4. For a deeper breakdown, see our Vultr vs DigitalOcean comparison.
Is Vultr's disk I/O fast enough for production databases?
Yes. Vultr's 50,000 read IOPS and 40,000 write IOPS handle PostgreSQL, MySQL, and MongoDB production workloads comfortably. We ran a pgbench test sustaining 800 TPS on the $5 plan without IOPS saturation. For extremely write-heavy workloads like high-volume time-series ingestion, consider Vultr's High Frequency Compute tier which delivers 20-30% higher disk I/O.
Do Vultr's High Frequency Compute plans benchmark significantly better?
Yes, meaningfully so. In our testing, HFC plans scored 14% higher on CPU (AMD EPYC vs older Intel Xeon) and 28% higher on disk I/O (NVMe vs SATA SSD). HFC starts at $6/mo and is available in most but not all US locations. If your workload is CPU or I/O bound, the extra dollar per month is worth it. The HFC CPU score of 4,680 would make Vultr #1 in our entire provider ranking.
Why does Vultr rank #4 overall instead of higher?
Vultr scores consistently well in every category but leads in none. Hostinger beats it on CPU (4,400) and disk (65,000 IOPS), Hetzner on value per dollar, DigitalOcean on raw network speed (980 Mbps). Vultr's advantage is having no weak category combined with the most US datacenter locations of any provider we tested. If you value geographic flexibility and consistency over peak performance in any single metric, Vultr is arguably the best overall choice.
How much do Vultr's benchmark scores vary between datacenters?
Up to 11% CPU variance (Silicon Valley 4,240 vs Honolulu 3,780), 18% disk I/O variance, and 8% network variance. Excluding Honolulu, mainland variance drops to 4% CPU and 10% disk. Choosing the right datacenter can matter as much as choosing between providers.
How stable are Vultr's benchmark scores over time?
Very stable within a given datacenter. Week-to-week variance was under 3% for CPU and disk, under 2% for network. Month-to-month variance stayed below 5% across the entire 6-month testing window. This predictability is one of Vultr's genuine strengths — you can capacity-plan against our numbers with confidence. Some competing providers showed 15-20% disk variance due to noisy-neighbor effects.
What benchmark tools and methodology do you use?
CPU: sysbench 1.0.20 single-thread. Disk: fio 3.36 random 4K at queue depth 32. Network: iperf3 3.16 with 4 parallel streams. Each test runs 5 times per session, 3 sessions per week, over 6 months. We report median of session medians on stock Ubuntu 24.04 LTS. See our full methodology.
Should I pick Vultr based on datacenter location or benchmark scores?
Location, almost always. The benchmark variance between mainland locations (under 4%) is smaller than the latency penalty of choosing a far-away datacenter. The exception: pure compute workloads (batch processing, rendering) with no user-facing latency requirement, where you should pick the fastest datacenter regardless of location.