Vultr High Performance vs High Frequency: $6 Plans Measured

On Vultr’s pricing page (checked 2026-10-28), Regular Performance is previous-generation Intel on regular SSD from $5 for 1 GB, while High Performance (AMD EPYC or Intel Xeon) and High Frequency (3GHz+ Intel Xeon) both use NVMe and start at $6. In our 2026-09-25 Tokyo test, High Performance AMD scored 4,303.83 sysbench events/s vs 1,170.04 for High Frequency and tied it on random reads (107k vs 110k IOPS).

This post contains affiliate links. If you purchase through our links, we may earn a commission at no extra cost to you.

Vultr Cloud Compute vs High Frequency: what actually changes

Vultr’s Cloud Compute line has three “classic” shared-vCPU tiers, plus a new fourth tier that most comparison posts haven’t caught up to yet. Regular Performance (what Vultr used to just call Cloud Compute) runs on previous-generation Intel CPUs with regular SSD storage. It’s the budget option: fine for a personal blog, a staging environment, or a Discord bot that doesn’t need to think fast.

High Frequency swaps in 3GHz+ Intel Xeon CPUs exclusively (no AMD option here) plus NVMe SSD. It also ships with more storage than High Performance at the same price point (32 GB vs 25 GB at $6), though you get less bandwidth in return (1 TB vs 2 TB).

High Performance sits in between on positioning but not necessarily on power: Vultr describes it as “new generations of AMD EPYC or Intel Xeon CPUs” at the same price for either, still on NVMe, with the most included bandwidth of the three. Looking at Vultr’s tier lineup as a whole (see our Vultr Review 2026), the naming here is genuinely confusing. “High Performance” doesn’t automatically outperform “High Frequency” in every metric, and the benchmarks below show exactly why that matters.

joshtronic’s 2025 benchmark vs its own conclusion

The most complete public benchmark I could find is joshtronic’s January 2025 four-way sysbench comparison, testing Regular, High Frequency, High Performance AMD, and High Performance Intel on identical 1 vCPU/1GB instances in the same region. The raw numbers:

The article’s own conclusion says High Frequency “shined in every category”, but its own table shows High Performance AMD scoring almost double High Frequency’s raw CPU throughput (2,620.69 vs. 1,339.47 events/sec). High Frequency does win on memory ops, MySQL reads, and fsyncs, so it’s not a bad instance. It just isn’t the flat CPU winner the prose claims, and I’d rather point that out than repeat a conclusion the author’s own data doesn’t fully support.

Our own test: Regular vs High Frequency vs High Performance AMD in Tokyo

To check those numbers on current instances, we ran the three entry tiers side by side ourselves. On 2026-09-25 we created three fresh 1 vCPU / 1GB instances in Vultr’s Tokyo (nrt) region, all on Ubuntu 24.04: Regular Performance (vc2-1c-1gb, $5/month), High Frequency (vhf-1c-1gb, $6/month), and High Performance AMD (vhp-1c-1gb-amd, $6/month). The same script ran on all three at the same time, and each benchmark ran exactly once for 30 seconds. That makes this one short sample from one region on one day. It’s useful for comparing the tiers against each other, but it says nothing about how any of them behave over weeks.

lscpu | grep -E 'Model name|^CPU\(s\)|Hypervisor vendor'
vmstat 1 30; mpstat -P ALL 1 10    # idle; vmstat repeated under sysbench load
sysbench cpu --threads=1 --time=30 run
fio --name=rr --filename=/root/fio.test --size=1G --rw=randread --bs=4k --iodepth=32 --ioengine=libaio --direct=1 --runtime=30 --time_based --group_reporting
fio --name=sw --filename=/root/fio.test --size=1G --rw=write --bs=1M --iodepth=8 --ioengine=libaio --direct=1 --runtime=30 --time_based --group_reporting

Our test: Vultr Tokyo (nrt), Ubuntu 24.04, fresh instances, one 30-second run per benchmark, 2026-09-25. Steal is the vmstat st column (30 one-second samples idle, 30 under sysbench load) plus mpstat %steal while idle.

The relevant lines from the raw output, trimmed:

## High Frequency (vhf-1c-1gb)
Model name:                              Intel Core Processor (Skylake, IBRS, no TSX)
    events per second:  1170.04
  read: IOPS=110k, BW=432MiB/s (452MB/s)(12.6GiB/30001msec)
  write: IOPS=1017, BW=1017MiB/s (1067MB/s)(29.8GiB/30009msec); 0 zone resets
# vmstat during sysbench load (us=100, st=0):
 r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa st gu
 1  0    268 101408  22916 648496    0    0     0     0 1008   53 100  0  0  0  0  0

## Regular Performance (vc2-1c-1gb)
Model name:                              Intel Core Processor (Broadwell, no TSX, IBRS)
    events per second:   817.61
  read: IOPS=59.7k, BW=233MiB/s (244MB/s)(6994MiB/30001msec)
  write: IOPS=307, BW=308MiB/s (323MB/s)(9237MiB/30029msec); 0 zone resets

## High Performance AMD (vhp-1c-1gb-amd)
Model name:                              AMD EPYC-Rome Processor
    events per second:  4303.83
  read: IOPS=107k, BW=418MiB/s (438MB/s)(12.2GiB/30001msec)
  write: IOPS=1816, BW=1816MiB/s (1904MB/s)(53.2GiB/30003msec); 0 zone resets

High Frequency beat Regular clearly on disk: 110k vs 59.7k random-read IOPS (1.84x) and 1,017 vs 308 MiB/s sequential write (3.3x). On CPU the gap was moderate: 1,170.04 vs 817.61 events/sec, about 43% more. That’s what the extra dollar actually bought in this run, and it fits the NVMe-versus-older-storage difference described above.

The bigger surprise is the plan at the same price. High Performance AMD costs the same $6/month as High Frequency and scored 4,303.83 events/sec (3.68x High Frequency) and 1,816 MiB/s sequential write (1.79x), while matching it on random reads (107k vs 110k IOPS, too close to separate with one run). One caveat on that CPU figure: sysbench’s prime-number test rewards newer CPU generations heavily, and EPYC-Rome is a newer design than the Skylake and Broadwell parts the other two plans reported. It’s one synthetic metric, not a prediction of how fast your PHP or Node app will run.

How this squares with joshtronic’s table above: joshtronic’s January 2025 run had the same order (844.61 / 1,339.47 / 2,620.69 events/sec for Regular / High Frequency / High Performance AMD), with our High Frequency result lower and our High Performance AMD result higher, a spread you’d expect across different hosts, regions and dates. Our 0% steal under a 30-second load also can’t show what happens over hours on a busy host.

For other plan sizes and providers, our free VPS cost calculator runs the monthly math for a whole fleet.

VX1: the tier most comparison posts are missing

While joshtronic’s benchmark and older posts on this exact topic were being written, Vultr quietly added a fourth Cloud Compute tier. VX1, launched October 21, 2025, runs on AMD EPYC processors with dedicated (not shared) vCPU cores, up to 50 Gbps networking, and either boot-from-NVMe or Block Storage. Vultr also claims provisioning takes under 15 seconds.

If even VX1 is not enough, we cover the bare-metal option in Vultr Bare Metal vs Cloud Compute.

The catch: Vultr’s VX1 docs (updated 2026-04-15) say automated backups are disabled and Windows isn’t available. VX1 started in New Jersey, and the docs still say New Jersey only, but Vultr’s plan API listed the cheapest VX1 plan in 9 locations on 2026-10-28, so check the console for your region. It starts at $43.80/month for 2 vCPU/8GB ($55.48 with 120 GB of local NVMe), a different price bracket from the $5-$6 tiers this article is mainly about.

Vultr Cloud Compute vs High Frequency pricing in 2026

High Frequency costs $1 more than Regular at 1 GB ($6 vs $5) and $4 more at 2 vCPU / 4GB ($24 vs $20), the same as High Performance, per Vultr’s pricing page (checked 2026-10-28). Regular also has a $2.50 IPv6-only 0.5 GB plan ($3.50 with IPv4). Our Vultr pricing breakdown covers the add-ons that raise the bill.

Prices from vultr.com/pricing and Vultr’s plan API, checked 2026-10-28

Which one should you actually pick?

• Running a single-thread-sensitive app: High Performance AMD at the same $6. In our Tokyo run it matched High Frequency on 4k random reads (107k vs 110k IOPS) and beat it on CPU and sequential writes.
• Running a database-heavy WordPress site or forum: test both. joshtronic’s January 2025 test had High Frequency ahead of High Performance AMD on MySQL reads (115,152 vs 69,627) and fsyncs (9,103 vs 3,309), which our run didn’t measure.
• Needing more disk on a 1GB plan: High Frequency has 32 GB of NVMe vs 25 GB for High Performance, at the cost of 1 TB of transfer instead of 2 TB.
• Running a personal project, staging site, or dev sandbox with no real traffic: Regular Performance is fine, and paying extra for High Frequency here is money you won’t notice the benefit of.
• Running sustained compute (video encoding, batch jobs, CI runners, anything that pins a core for more than a few seconds): High Performance AMD is the better bet: it scored 3.68x High Frequency’s sysbench result in our run, and joshtronic’s January 2025 test had the same order. Neither test ran for hours.

High Frequency beats Regular, High Performance AMD leads on CPU and writes

For $1 more than Regular, High Frequency pays off on disk: in our 2026-09-25 Tokyo run it read 110k vs 59.7k IOPS and wrote 1,017 vs 308 MiB/s. At the same $6, High Performance AMD led on single-thread CPU (4,303.83 vs 1,170.04 events/s) and sequential writes (1,816 vs 1,017 MiB/s) and tied on random reads, so pick it for CPU-bound or write-heavy work. For database-heavy sites, joshtronic’s January 2025 test had High Frequency ahead on MySQL reads and fsyncs, which our run didn’t measure. That was one 30-second run per test in one region, so rerun the commands above on both plans before you commit a real workload.