sysbench cpu --threads=1 run — single-thread throughput in events per second; compare against a known-good baseline.sysbench cpu --threads=N run — how well it scales; if 4 threads give under 2x the single-thread result, you are on hyperthreads or shared cores.vmstat 1 under load — the st column is steal time. Above 5% sustained, a neighbour is eating your CPU.Record those numbers before you install anything. They are your baseline, and they are the fastest way to prove a provider swapped you onto a busier host.
When to pay for cores instead of clock
Choose more cores when you run parallel builds, multiple containers, or a scheduled batch job that overlaps with live traffic. Choose clock speed when your source of pain is a single slow request: a heavy WooCommerce cart, an uncached admin page, or a database query that times out.
The cheapest mistake in this tier is buying 4 vCPU of old silicon for a WordPress blog. It scores well in a headline and performs like a 2-core from three generations later. Spend the difference on a modern CPU class and NVMe, and check the renewal price before you sign up for a year.
Bottom line
Clock speed sets your ceiling on responsiveness; cores set your concurrency. Most budget buyers need the first, a development environment needs the second, and everyone needs to verify both with a five-minute benchmark before prepaying.
lscpu— shows the model, base clock and thread count per core.sysbench cpu --threads=1 run— single-thread throughput in events per second; compare against a known-good baseline.sysbench cpu --threads=N run— how well it scales; if 4 threads give under 2x the single-thread result, you are on hyperthreads or shared cores.vmstat 1under load — thestcolumn is steal time. Above 5% sustained, a neighbour is eating your CPU.
Record those numbers before you install anything. They are your baseline, and they are the fastest way to prove a provider swapped you onto a busier host.
When to pay for cores instead of clock
Choose more cores when you run parallel builds, multiple containers, or a scheduled batch job that overlaps with live traffic. Choose clock speed when your source of pain is a single slow request: a heavy WooCommerce cart, an uncached admin page, or a database query that times out.
The cheapest mistake in this tier is buying 4 vCPU of old silicon for a WordPress blog. It scores well in a headline and performs like a 2-core from three generations later. Spend the difference on a modern CPU class and NVMe, and check the renewal price before you sign up for a year.
Bottom line
Clock speed sets your ceiling on responsiveness; cores set your concurrency. Most budget buyers need the first, a development environment needs the second, and everyone needs to verify both with a five-minute benchmark before prepaying.
lscpu— shows the model, base clock and thread count per core.sysbench cpu --threads=1 run— single-thread throughput in events per second; compare against a known-good baseline.sysbench cpu --threads=N run— how well it scales; if 4 threads give under 2x the single-thread result, you are on hyperthreads or shared cores.vmstat 1under load — thestcolumn is steal time. Above 5% sustained, a neighbour is eating your CPU.
Record those numbers before you install anything. They are your baseline, and they are the fastest way to prove a provider swapped you onto a busier host.
When to pay for cores instead of clock
Choose more cores when you run parallel builds, multiple containers, or a scheduled batch job that overlaps with live traffic. Choose clock speed when your source of pain is a single slow request: a heavy WooCommerce cart, an uncached admin page, or a database query that times out.
The cheapest mistake in this tier is buying 4 vCPU of old silicon for a WordPress blog. It scores well in a headline and performs like a 2-core from three generations later. Spend the difference on a modern CPU class and NVMe, and check the renewal price before you sign up for a year.
Bottom line
Clock speed sets your ceiling on responsiveness; cores set your concurrency. Most budget buyers need the first, a development environment needs the second, and everyone needs to verify both with a five-minute benchmark before prepaying.
Budget VPS spec sheets push two numbers at you: cores and gigahertz. A $5 plan with “4 vCPU @ 2.4 GHz” looks stronger than a $5 plan with “2 vCPU @ 3.5 GHz” until you realize the first provider is splitting two physical cores across four hyperthreads on older silicon. For most hosted workloads, single-thread speed decides how fast a page loads; core count only decides how many requests you can handle in parallel before things queue.
This is not an argument that cores do not matter. It is an argument that they matter differently for web hosting, databases and development work, and that you should pick the balance that matches your workload instead of the biggest number on the page.
What the two numbers actually measure
Clock speed is how many instructions one core completes per second. Cores are how many instruction streams run at once. Hyperthreading matters too: a “vCPU” on many budget plans is a hyperthread, which delivers roughly 25–40% of a full core under heavy mixed load, not 100%. A provider selling “4 vCPU” on a 2-core host is giving you two physical cores’ worth of throughput in a shape that benchmarks flatter than it performs.
Generation matters as much as either number. A 3.0 GHz Xeon Gold from 2019 usually outperforms a 3.4 GHz Xeon E5 from 2014 in per-clock work, because cache size and memory bandwidth improved by more than a third in that span. Neither figure appears on the order page.
Web hosting: single-thread speed wins
A cached WordPress page is a short serial chain: PHP builds the HTML, the template renders, the response is compressed and sent. Each request is handled by one worker on one core. If your PHP render takes 80 ms per request and you have two cores at 3.5 GHz, one core finishes 12 requests per second while the other picks up the next queue entry. Halving the clock speed to 1.75 GHz doubles that number to 160 ms and halves your capacity.
Adding cores helps once you saturate one. Going from 1 to 2 cores roughly doubles throughput for concurrent requests, but 2 to 4 gives diminishing returns because PHP workers, the database and the OS are all contending for the same memory bandwidth and disk. On a $5–$10 tier, the difference between 2 and 4 vCPU is often smaller than the difference between NVMe and SATA storage.
Practical rule: for a single site under 50,000 monthly visitors, prefer higher clock speed and 2 vCPU over 4 low-clock vCPUs. You can compare how providers describe their CPU class in our budget hosting comparison table.
Databases: mixed, but latency-sensitive
MySQL and MariaDB do use multiple cores, but per-query latency is dominated by single-thread execution. One slow query is one core’s problem. Where cores pay off is concurrency: dozens of simultaneous connections, each with its own short transaction.
PostgreSQL behaves similarly, with an added consideration for parallel query execution, which does split work across cores — useful for analytics queries, largely irrelevant for a blog’s read path. On a 2 GB VPS, memory for the buffer pool and shared buffers is usually the binding constraint before CPU.
| Workload | What helps most | Budget-plan priority |
|---|---|---|
| Cached WordPress / static site | Clock speed, disk IOPS | 2 vCPU @ high clock, NVMe |
| Uncached WordPress / WooCommerce | Clock speed + RAM | 2–4 vCPU, 2–4 GB RAM |
| MySQL/PostgreSQL OLTP | Clock speed, then IOPS | 2–4 vCPU, RAM for buffer pool |
| Analytics / reporting queries | Cores, memory bandwidth | 4+ vCPU, 4+ GB RAM |
| Builds, compiles, image batch jobs | Cores | 4+ vCPU, fast disk |
Development work: where more cores pay off
Build pipelines, test runners, Docker image builds and TypeScript compilation are genuinely parallel. A npm run build or a Gradle compile scales close to linearly with real cores. If you use a VPS as a remote development environment, core count matters more than it does for a production blog.
But watch the memory ceiling. Four parallel Node build workers at 500 MB each will exhaust 2 GB and push the box into swap, which makes the build slower than two workers would have been. More cores than you have RAM to feed is a waste of money.
If your main purpose is compiling and running containers, our plan sizing help can point you to tiers that pair 4 vCPU with enough memory, rather than the cheapest 4-vCPU plan that starves them.
How to measure what you actually got
Spec sheets lie by omission, so measure. Four commands cover most of it, all within the first hour of a trial:
lscpu— shows the model, base clock and thread count per core.sysbench cpu --threads=1 run— single-thread throughput in events per second; compare against a known-good baseline.sysbench cpu --threads=N run— how well it scales; if 4 threads give under 2x the single-thread result, you are on hyperthreads or shared cores.vmstat 1under load — thestcolumn is steal time. Above 5% sustained, a neighbour is eating your CPU.
Record those numbers before you install anything. They are your baseline, and they are the fastest way to prove a provider swapped you onto a busier host.
When to pay for cores instead of clock
Choose more cores when you run parallel builds, multiple containers, or a scheduled batch job that overlaps with live traffic. Choose clock speed when your source of pain is a single slow request: a heavy WooCommerce cart, an uncached admin page, or a database query that times out.
The cheapest mistake in this tier is buying 4 vCPU of old silicon for a WordPress blog. It scores well in a headline and performs like a 2-core from three generations later. Spend the difference on a modern CPU class and NVMe, and check the renewal price before you sign up for a year.
Bottom line
Clock speed sets your ceiling on responsiveness; cores set your concurrency. Most budget buyers need the first, a development environment needs the second, and everyone needs to verify both with a five-minute benchmark before prepaying.
lscpu— shows the model, base clock and thread count per core.sysbench cpu --threads=1 run— single-thread throughput in events per second; compare against a known-good baseline.sysbench cpu --threads=N run— how well it scales; if 4 threads give under 2x the single-thread result, you are on hyperthreads or shared cores.vmstat 1under load — thestcolumn is steal time. Above 5% sustained, a neighbour is eating your CPU.
Record those numbers before you install anything. They are your baseline, and they are the fastest way to prove a provider swapped you onto a busier host.
When to pay for cores instead of clock
Choose more cores when you run parallel builds, multiple containers, or a scheduled batch job that overlaps with live traffic. Choose clock speed when your source of pain is a single slow request: a heavy WooCommerce cart, an uncached admin page, or a database query that times out.
The cheapest mistake in this tier is buying 4 vCPU of old silicon for a WordPress blog. It scores well in a headline and performs like a 2-core from three generations later. Spend the difference on a modern CPU class and NVMe, and check the renewal price before you sign up for a year.
Bottom line
Clock speed sets your ceiling on responsiveness; cores set your concurrency. Most budget buyers need the first, a development environment needs the second, and everyone needs to verify both with a five-minute benchmark before prepaying.
Budget VPS spec sheets push two numbers at you: cores and gigahertz. A $5 plan with “4 vCPU @ 2.4 GHz” looks stronger than a $5 plan with “2 vCPU @ 3.5 GHz” until you realize the first provider is splitting two physical cores across four hyperthreads on older silicon. For most hosted workloads, single-thread speed decides how fast a page loads; core count only decides how many requests you can handle in parallel before things queue.
This is not an argument that cores do not matter. It is an argument that they matter differently for web hosting, databases and development work, and that you should pick the balance that matches your workload instead of the biggest number on the page.
What the two numbers actually measure
Clock speed is how many instructions one core completes per second. Cores are how many instruction streams run at once. Hyperthreading matters too: a “vCPU” on many budget plans is a hyperthread, which delivers roughly 25–40% of a full core under heavy mixed load, not 100%. A provider selling “4 vCPU” on a 2-core host is giving you two physical cores’ worth of throughput in a shape that benchmarks flatter than it performs.
Generation matters as much as either number. A 3.0 GHz Xeon Gold from 2019 usually outperforms a 3.4 GHz Xeon E5 from 2014 in per-clock work, because cache size and memory bandwidth improved by more than a third in that span. Neither figure appears on the order page.
Web hosting: single-thread speed wins
A cached WordPress page is a short serial chain: PHP builds the HTML, the template renders, the response is compressed and sent. Each request is handled by one worker on one core. If your PHP render takes 80 ms per request and you have two cores at 3.5 GHz, one core finishes 12 requests per second while the other picks up the next queue entry. Halving the clock speed to 1.75 GHz doubles that number to 160 ms and halves your capacity.
Adding cores helps once you saturate one. Going from 1 to 2 cores roughly doubles throughput for concurrent requests, but 2 to 4 gives diminishing returns because PHP workers, the database and the OS are all contending for the same memory bandwidth and disk. On a $5–$10 tier, the difference between 2 and 4 vCPU is often smaller than the difference between NVMe and SATA storage.
Practical rule: for a single site under 50,000 monthly visitors, prefer higher clock speed and 2 vCPU over 4 low-clock vCPUs. You can compare how providers describe their CPU class in our budget hosting comparison table.
Databases: mixed, but latency-sensitive
MySQL and MariaDB do use multiple cores, but per-query latency is dominated by single-thread execution. One slow query is one core’s problem. Where cores pay off is concurrency: dozens of simultaneous connections, each with its own short transaction.
PostgreSQL behaves similarly, with an added consideration for parallel query execution, which does split work across cores — useful for analytics queries, largely irrelevant for a blog’s read path. On a 2 GB VPS, memory for the buffer pool and shared buffers is usually the binding constraint before CPU.
| Workload | What helps most | Budget-plan priority |
|---|---|---|
| Cached WordPress / static site | Clock speed, disk IOPS | 2 vCPU @ high clock, NVMe |
| Uncached WordPress / WooCommerce | Clock speed + RAM | 2–4 vCPU, 2–4 GB RAM |
| MySQL/PostgreSQL OLTP | Clock speed, then IOPS | 2–4 vCPU, RAM for buffer pool |
| Analytics / reporting queries | Cores, memory bandwidth | 4+ vCPU, 4+ GB RAM |
| Builds, compiles, image batch jobs | Cores | 4+ vCPU, fast disk |
Development work: where more cores pay off
Build pipelines, test runners, Docker image builds and TypeScript compilation are genuinely parallel. A npm run build or a Gradle compile scales close to linearly with real cores. If you use a VPS as a remote development environment, core count matters more than it does for a production blog.
But watch the memory ceiling. Four parallel Node build workers at 500 MB each will exhaust 2 GB and push the box into swap, which makes the build slower than two workers would have been. More cores than you have RAM to feed is a waste of money.
If your main purpose is compiling and running containers, our plan sizing help can point you to tiers that pair 4 vCPU with enough memory, rather than the cheapest 4-vCPU plan that starves them.
How to measure what you actually got
Spec sheets lie by omission, so measure. Four commands cover most of it, all within the first hour of a trial:
lscpu— shows the model, base clock and thread count per core.sysbench cpu --threads=1 run— single-thread throughput in events per second; compare against a known-good baseline.sysbench cpu --threads=N run— how well it scales; if 4 threads give under 2x the single-thread result, you are on hyperthreads or shared cores.vmstat 1under load — thestcolumn is steal time. Above 5% sustained, a neighbour is eating your CPU.
Record those numbers before you install anything. They are your baseline, and they are the fastest way to prove a provider swapped you onto a busier host.
When to pay for cores instead of clock
Choose more cores when you run parallel builds, multiple containers, or a scheduled batch job that overlaps with live traffic. Choose clock speed when your source of pain is a single slow request: a heavy WooCommerce cart, an uncached admin page, or a database query that times out.
The cheapest mistake in this tier is buying 4 vCPU of old silicon for a WordPress blog. It scores well in a headline and performs like a 2-core from three generations later. Spend the difference on a modern CPU class and NVMe, and check the renewal price before you sign up for a year.
Bottom line
Clock speed sets your ceiling on responsiveness; cores set your concurrency. Most budget buyers need the first, a development environment needs the second, and everyone needs to verify both with a five-minute benchmark before prepaying.
lscpu— shows the model, base clock and thread count per core.sysbench cpu --threads=1 run— single-thread throughput in events per second; compare against a known-good baseline.sysbench cpu --threads=N run— how well it scales; if 4 threads give under 2x the single-thread result, you are on hyperthreads or shared cores.vmstat 1under load — thestcolumn is steal time. Above 5% sustained, a neighbour is eating your CPU.
Record those numbers before you install anything. They are your baseline, and they are the fastest way to prove a provider swapped you onto a busier host.
When to pay for cores instead of clock
Choose more cores when you run parallel builds, multiple containers, or a scheduled batch job that overlaps with live traffic. Choose clock speed when your source of pain is a single slow request: a heavy WooCommerce cart, an uncached admin page, or a database query that times out.
The cheapest mistake in this tier is buying 4 vCPU of old silicon for a WordPress blog. It scores well in a headline and performs like a 2-core from three generations later. Spend the difference on a modern CPU class and NVMe, and check the renewal price before you sign up for a year.
Bottom line
Clock speed sets your ceiling on responsiveness; cores set your concurrency. Most budget buyers need the first, a development environment needs the second, and everyone needs to verify both with a five-minute benchmark before prepaying.
Budget VPS spec sheets push two numbers at you: cores and gigahertz. A $5 plan with “4 vCPU @ 2.4 GHz” looks stronger than a $5 plan with “2 vCPU @ 3.5 GHz” until you realize the first provider is splitting two physical cores across four hyperthreads on older silicon. For most hosted workloads, single-thread speed decides how fast a page loads; core count only decides how many requests you can handle in parallel before things queue.
This is not an argument that cores do not matter. It is an argument that they matter differently for web hosting, databases and development work, and that you should pick the balance that matches your workload instead of the biggest number on the page.
What the two numbers actually measure
Clock speed is how many instructions one core completes per second. Cores are how many instruction streams run at once. Hyperthreading matters too: a “vCPU” on many budget plans is a hyperthread, which delivers roughly 25–40% of a full core under heavy mixed load, not 100%. A provider selling “4 vCPU” on a 2-core host is giving you two physical cores’ worth of throughput in a shape that benchmarks flatter than it performs.
Generation matters as much as either number. A 3.0 GHz Xeon Gold from 2019 usually outperforms a 3.4 GHz Xeon E5 from 2014 in per-clock work, because cache size and memory bandwidth improved by more than a third in that span. Neither figure appears on the order page.
Web hosting: single-thread speed wins
A cached WordPress page is a short serial chain: PHP builds the HTML, the template renders, the response is compressed and sent. Each request is handled by one worker on one core. If your PHP render takes 80 ms per request and you have two cores at 3.5 GHz, one core finishes 12 requests per second while the other picks up the next queue entry. Halving the clock speed to 1.75 GHz doubles that number to 160 ms and halves your capacity.
Adding cores helps once you saturate one. Going from 1 to 2 cores roughly doubles throughput for concurrent requests, but 2 to 4 gives diminishing returns because PHP workers, the database and the OS are all contending for the same memory bandwidth and disk. On a $5–$10 tier, the difference between 2 and 4 vCPU is often smaller than the difference between NVMe and SATA storage.
Practical rule: for a single site under 50,000 monthly visitors, prefer higher clock speed and 2 vCPU over 4 low-clock vCPUs. You can compare how providers describe their CPU class in our budget hosting comparison table.
Databases: mixed, but latency-sensitive
MySQL and MariaDB do use multiple cores, but per-query latency is dominated by single-thread execution. One slow query is one core’s problem. Where cores pay off is concurrency: dozens of simultaneous connections, each with its own short transaction.
PostgreSQL behaves similarly, with an added consideration for parallel query execution, which does split work across cores — useful for analytics queries, largely irrelevant for a blog’s read path. On a 2 GB VPS, memory for the buffer pool and shared buffers is usually the binding constraint before CPU.
| Workload | What helps most | Budget-plan priority |
|---|---|---|
| Cached WordPress / static site | Clock speed, disk IOPS | 2 vCPU @ high clock, NVMe |
| Uncached WordPress / WooCommerce | Clock speed + RAM | 2–4 vCPU, 2–4 GB RAM |
| MySQL/PostgreSQL OLTP | Clock speed, then IOPS | 2–4 vCPU, RAM for buffer pool |
| Analytics / reporting queries | Cores, memory bandwidth | 4+ vCPU, 4+ GB RAM |
| Builds, compiles, image batch jobs | Cores | 4+ vCPU, fast disk |
Development work: where more cores pay off
Build pipelines, test runners, Docker image builds and TypeScript compilation are genuinely parallel. A npm run build or a Gradle compile scales close to linearly with real cores. If you use a VPS as a remote development environment, core count matters more than it does for a production blog.
But watch the memory ceiling. Four parallel Node build workers at 500 MB each will exhaust 2 GB and push the box into swap, which makes the build slower than two workers would have been. More cores than you have RAM to feed is a waste of money.
If your main purpose is compiling and running containers, our plan sizing help can point you to tiers that pair 4 vCPU with enough memory, rather than the cheapest 4-vCPU plan that starves them.
How to measure what you actually got
Spec sheets lie by omission, so measure. Four commands cover most of it, all within the first hour of a trial:
lscpu— shows the model, base clock and thread count per core.sysbench cpu --threads=1 run— single-thread throughput in events per second; compare against a known-good baseline.sysbench cpu --threads=N run— how well it scales; if 4 threads give under 2x the single-thread result, you are on hyperthreads or shared cores.vmstat 1under load — thestcolumn is steal time. Above 5% sustained, a neighbour is eating your CPU.
Record those numbers before you install anything. They are your baseline, and they are the fastest way to prove a provider swapped you onto a busier host.
When to pay for cores instead of clock
Choose more cores when you run parallel builds, multiple containers, or a scheduled batch job that overlaps with live traffic. Choose clock speed when your source of pain is a single slow request: a heavy WooCommerce cart, an uncached admin page, or a database query that times out.
The cheapest mistake in this tier is buying 4 vCPU of old silicon for a WordPress blog. It scores well in a headline and performs like a 2-core from three generations later. Spend the difference on a modern CPU class and NVMe, and check the renewal price before you sign up for a year.
Bottom line
Clock speed sets your ceiling on responsiveness; cores set your concurrency. Most budget buyers need the first, a development environment needs the second, and everyone needs to verify both with a five-minute benchmark before prepaying.
lscpu— shows the model, base clock and thread count per core.sysbench cpu --threads=1 run— single-thread throughput in events per second; compare against a known-good baseline.sysbench cpu --threads=N run— how well it scales; if 4 threads give under 2x the single-thread result, you are on hyperthreads or shared cores.vmstat 1under load — thestcolumn is steal time. Above 5% sustained, a neighbour is eating your CPU.
Record those numbers before you install anything. They are your baseline, and they are the fastest way to prove a provider swapped you onto a busier host.
When to pay for cores instead of clock
Choose more cores when you run parallel builds, multiple containers, or a scheduled batch job that overlaps with live traffic. Choose clock speed when your source of pain is a single slow request: a heavy WooCommerce cart, an uncached admin page, or a database query that times out.
The cheapest mistake in this tier is buying 4 vCPU of old silicon for a WordPress blog. It scores well in a headline and performs like a 2-core from three generations later. Spend the difference on a modern CPU class and NVMe, and check the renewal price before you sign up for a year.
Bottom line
Clock speed sets your ceiling on responsiveness; cores set your concurrency. Most budget buyers need the first, a development environment needs the second, and everyone needs to verify both with a five-minute benchmark before prepaying.
lscpu— shows the model, base clock and thread count per core.sysbench cpu --threads=1 run— single-thread throughput in events per second; compare against a known-good baseline.sysbench cpu --threads=N run— how well it scales; if 4 threads give under 2x the single-thread result, you are on hyperthreads or shared cores.vmstat 1under load — thestcolumn is steal time. Above 5% sustained, a neighbour is eating your CPU.
Record those numbers before you install anything. They are your baseline, and they are the fastest way to prove a provider swapped you onto a busier host.
When to pay for cores instead of clock
Choose more cores when you run parallel builds, multiple containers, or a scheduled batch job that overlaps with live traffic. Choose clock speed when your source of pain is a single slow request: a heavy WooCommerce cart, an uncached admin page, or a database query that times out.
The cheapest mistake in this tier is buying 4 vCPU of old silicon for a WordPress blog. It scores well in a headline and performs like a 2-core from three generations later. Spend the difference on a modern CPU class and NVMe, and check the renewal price before you sign up for a year.
Bottom line
Clock speed sets your ceiling on responsiveness; cores set your concurrency. Most budget buyers need the first, a development environment needs the second, and everyone needs to verify both with a five-minute benchmark before prepaying.
Budget VPS spec sheets push two numbers at you: cores and gigahertz. A $5 plan with “4 vCPU @ 2.4 GHz” looks stronger than a $5 plan with “2 vCPU @ 3.5 GHz” until you realize the first provider is splitting two physical cores across four hyperthreads on older silicon. For most hosted workloads, single-thread speed decides how fast a page loads; core count only decides how many requests you can handle in parallel before things queue.
This is not an argument that cores do not matter. It is an argument that they matter differently for web hosting, databases and development work, and that you should pick the balance that matches your workload instead of the biggest number on the page.
What the two numbers actually measure
Clock speed is how many instructions one core completes per second. Cores are how many instruction streams run at once. Hyperthreading matters too: a “vCPU” on many budget plans is a hyperthread, which delivers roughly 25–40% of a full core under heavy mixed load, not 100%. A provider selling “4 vCPU” on a 2-core host is giving you two physical cores’ worth of throughput in a shape that benchmarks flatter than it performs.
Generation matters as much as either number. A 3.0 GHz Xeon Gold from 2019 usually outperforms a 3.4 GHz Xeon E5 from 2014 in per-clock work, because cache size and memory bandwidth improved by more than a third in that span. Neither figure appears on the order page.
Web hosting: single-thread speed wins
A cached WordPress page is a short serial chain: PHP builds the HTML, the template renders, the response is compressed and sent. Each request is handled by one worker on one core. If your PHP render takes 80 ms per request and you have two cores at 3.5 GHz, one core finishes 12 requests per second while the other picks up the next queue entry. Halving the clock speed to 1.75 GHz doubles that number to 160 ms and halves your capacity.
Adding cores helps once you saturate one. Going from 1 to 2 cores roughly doubles throughput for concurrent requests, but 2 to 4 gives diminishing returns because PHP workers, the database and the OS are all contending for the same memory bandwidth and disk. On a $5–$10 tier, the difference between 2 and 4 vCPU is often smaller than the difference between NVMe and SATA storage.
Practical rule: for a single site under 50,000 monthly visitors, prefer higher clock speed and 2 vCPU over 4 low-clock vCPUs. You can compare how providers describe their CPU class in our budget hosting comparison table.
Databases: mixed, but latency-sensitive
MySQL and MariaDB do use multiple cores, but per-query latency is dominated by single-thread execution. One slow query is one core’s problem. Where cores pay off is concurrency: dozens of simultaneous connections, each with its own short transaction.
PostgreSQL behaves similarly, with an added consideration for parallel query execution, which does split work across cores — useful for analytics queries, largely irrelevant for a blog’s read path. On a 2 GB VPS, memory for the buffer pool and shared buffers is usually the binding constraint before CPU.
| Workload | What helps most | Budget-plan priority |
|---|---|---|
| Cached WordPress / static site | Clock speed, disk IOPS | 2 vCPU @ high clock, NVMe |
| Uncached WordPress / WooCommerce | Clock speed + RAM | 2–4 vCPU, 2–4 GB RAM |
| MySQL/PostgreSQL OLTP | Clock speed, then IOPS | 2–4 vCPU, RAM for buffer pool |
| Analytics / reporting queries | Cores, memory bandwidth | 4+ vCPU, 4+ GB RAM |
| Builds, compiles, image batch jobs | Cores | 4+ vCPU, fast disk |
Development work: where more cores pay off
Build pipelines, test runners, Docker image builds and TypeScript compilation are genuinely parallel. A npm run build or a Gradle compile scales close to linearly with real cores. If you use a VPS as a remote development environment, core count matters more than it does for a production blog.
But watch the memory ceiling. Four parallel Node build workers at 500 MB each will exhaust 2 GB and push the box into swap, which makes the build slower than two workers would have been. More cores than you have RAM to feed is a waste of money.
If your main purpose is compiling and running containers, our plan sizing help can point you to tiers that pair 4 vCPU with enough memory, rather than the cheapest 4-vCPU plan that starves them.
How to measure what you actually got
Spec sheets lie by omission, so measure. Four commands cover most of it, all within the first hour of a trial:
lscpu— shows the model, base clock and thread count per core.sysbench cpu --threads=1 run— single-thread throughput in events per second; compare against a known-good baseline.sysbench cpu --threads=N run— how well it scales; if 4 threads give under 2x the single-thread result, you are on hyperthreads or shared cores.vmstat 1under load — thestcolumn is steal time. Above 5% sustained, a neighbour is eating your CPU.
Record those numbers before you install anything. They are your baseline, and they are the fastest way to prove a provider swapped you onto a busier host.
When to pay for cores instead of clock
Choose more cores when you run parallel builds, multiple containers, or a scheduled batch job that overlaps with live traffic. Choose clock speed when your source of pain is a single slow request: a heavy WooCommerce cart, an uncached admin page, or a database query that times out.
The cheapest mistake in this tier is buying 4 vCPU of old silicon for a WordPress blog. It scores well in a headline and performs like a 2-core from three generations later. Spend the difference on a modern CPU class and NVMe, and check the renewal price before you sign up for a year.
Bottom line
Clock speed sets your ceiling on responsiveness; cores set your concurrency. Most budget buyers need the first, a development environment needs the second, and everyone needs to verify both with a five-minute benchmark before prepaying.




