High Memory VPS for workloads that live in RAM.
30-day performance guarantee on your first order. High Memory plans bill at the normal monthly price; the €1 first month applies to Standard and D-4 to D-16.
Every vCPU on the H-series comes with 8 GB of DDR4 ECC memory, on AMD EPYC Milan, with NVMe RAID10 underneath and unlimited traffic. Pick 64, 128 or 192 GB of RAM and give it to Redis, a large Postgres or MySQL buffer pool, Elasticsearch, a JVM application or analytics. Full root and your own kernel. Amsterdam or New York, same price in both.
Transparent pricing
Plans and prices
| Plan | vCPU | Memory | NVMe | Traffic | Monthly | 1 year | |
|---|---|---|---|---|---|---|---|
| Standard | |||||||
| S-2Standard | 1vCPU | 2 GBMemory | 30 GBNVMe RAID10 | UnlimitedTraffic | $7.99/moFirst month $1 | $7.19/mo on 1 year | Start for $1 |
| S-4Standard | 2vCPU | 4 GBMemory | 60 GBNVMe RAID10 | UnlimitedTraffic | $13.99/moFirst month $1 | $12.59/mo on 1 year | Start for $1 |
| S-8Standard | 3vCPU | 8 GBMemory | 120 GBNVMe RAID10 | UnlimitedTraffic | $24.99/moFirst month $1 | $22.49/mo on 1 year | Start for $1 |
| S-16Standard | 4vCPU | 16 GBMemory | 240 GBNVMe RAID10 | UnlimitedTraffic | $44.99/moFirst month $1 | $40.49/mo on 1 year | Start for $1 |
| S-32Standard | 8vCPU | 32 GBMemory | 480 GBNVMe RAID10 | UnlimitedTraffic | $84.99/moFirst month $1 | $76.49/mo on 1 year | Start for $1 |
| S-64Standard | 16vCPU | 64 GBMemory | 720 GBNVMe RAID10 | UnlimitedTraffic | $159.99/moFirst month $1 | $143.99/mo on 1 year | Start for $1 |
| Dedicated CPU | |||||||
| D-4Dedicated | 1vCPU | 4 GBMemory | 45 GBNVMe RAID10 | UnlimitedTraffic | $37.00/moFirst month $1 | $31.45/mo on 1 year | Start for $1 |
| D-8Dedicated | 4vCPU | 8 GBMemory | 90 GBNVMe RAID10 | UnlimitedTraffic | $69.00/moFirst month $1 | $58.65/mo on 1 year | Start for $1 |
| D-16Dedicated | 4vCPU | 16 GBMemory | 160 GBNVMe RAID10 | UnlimitedTraffic | $118.00/moFirst month $1 | $100.30/mo on 1 year | Start for $1 |
| D-32Dedicated | 6vCPU | 32 GBMemory | 220 GBNVMe RAID10 | UnlimitedTraffic | $215.00/mo | $182.75/mo on 1 year | Deploy D-32 |
| D-64Dedicated | 10vCPU | 64 GBMemory | 400 GBNVMe RAID10 | UnlimitedTraffic | $377.00/mo | $320.45/mo on 1 year | Deploy D-64 |
| High Memory | |||||||
| H-64High Memory | 8vCPU | 64 GBMemory | 480 GBNVMe RAID10 | UnlimitedTraffic | $249.00/mo | $211.65/mo on 1 year | Deploy H-64 |
| H-128High Memory | 16vCPU | 128 GBMemory | 960 GBNVMe RAID10 | UnlimitedTraffic | $459.00/mo | $390.15/mo on 1 year | Deploy H-128 |
| H-192High Memory | 24vCPU | 192 GBMemory | 1.4 TBNVMe RAID10 | UnlimitedTraffic | $669.00/mo | $568.65/mo on 1 year | Deploy H-192 |
What actually needs 8 GB per core.
Most servers run out of CPU or disk first. Not these. Here RAM is the resource that matters, and extra cores would just sit there idle.
Redis and Memcached
An in-memory store is only as big as the RAM it sits in, and Redis needs headroom above the dataset for forks during snapshots. On H-64 a 40 GB dataset with a safe maxmemory fits comfortably.
Postgres and MySQL with big buffer pools
A database is fast when its working set is in shared_buffers or the InnoDB buffer pool and the rest in page cache. 128 GB lets a 100 GB database run almost entirely from memory, with NVMe RAID10 for the writes.
Elasticsearch, OpenSearch and JVM apps
The JVM asks for a fixed heap. The search engine then asks for at least that much again, as page cache for its segments. On H-128 that works out to a 31 GB heap plus 90 GB of file cache.
Analytics and build caches
ClickHouse or DuckDB queries that kept spilling to disk simply stop spilling. Then keep the Bazel, Gradle and ccache directories in memory, and a CI build takes the same time every run. Boring, in the good way.
H-64 to H-192: sizing, prices, and when Standard is the better buy.
Prices are per month in EUR, USD in brackets. Every plan comes with unlimited traffic, one IPv4 and a /64 IPv6 block. No setup fee, and Amsterdam and New York cost the same.
| Plan | vCPU / RAM / NVMe | Monthly | Fits |
|---|---|---|---|
| H-64 | 8 vCPU with 64 GB and 480 GB | €229.00 ($249.00) | A Redis primary, a mid-size Postgres, a single JVM app |
| H-128 | 16 vCPU with 128 GB and 960 GB | €422.00 ($459.00) | An Elasticsearch node, a large MySQL, several caches at once |
| H-192 | 24 vCPU with 192 GB and 1.4 TB | €616.00 ($669.00) | Analytics, in-memory datasets, several databases pulled onto one box |
Do the arithmetic before ordering H-64, and do it honestly. The Standard S-64 (16 vCPU, 64 GB, 720 GB NVMe) is €139.99 a month ($159.99), the S-32 (8 vCPU, 32 GB, 480 GB) is €74.99 ($84.99), and both sit on the same EPYC Milan and DDR4 ECC as the H-series. So if 64 GB is enough and the cores won't go to waste, S-64 hands you the same memory with more vCPU and more disk, for less money. That is just the numbers. Where the H-series earns its keep is at 128 GB and 192 GB, past the point where the Standard line stops, or when the 8 GB per vCPU shape suits a licensing reason of your own.
Two more notes. High Memory plans use shared vCPU; if the workload needs pinned cores for a latency target, that is the Dedicated tier, which tops out at 64 GB on D-64. And the €1 first month does not apply to the H-series; the 30-day performance guarantee on a first order does. Every tier is on the plans page, and the disk numbers behind them on the benchmarks page.
Setting up the kernel for an in-memory workload.
Step 1: Order, choose an image, update
Same images and same API as every other plan: Ubuntu 24.04, Debian 13, AlmaLinux 10 and the rest of the Linux list. The server is online 30-60 seconds after payment.
Step 2: Swap off, overcommit on
Swap on a server whose job is RAM turns a memory problem into a latency problem. Redis asks for vm.overcommit_memory=1 so background saves can fork without the kernel refusing.
Two sysctl values and no swap partition. Redis wants overcommit on, and a 64 GB box that starts swapping is worse than one that refuses an allocation.
Step 3: Transparent huge pages: off for Redis, explicit for Postgres
Redis and JVM stores want transparent huge pages disabled, because THP defragmentation stalls at the wrong moments. Postgres wants explicit hugepages sized for shared_buffers. Pick the setting for the main workload.
Make the THP setting survive reboots with the transparent_hugepage=never kernel parameter; you have your own kernel and control the boot line.
Step 4: Give the application the memory, but not all of it
Redis with a hard ceiling and an eviction policy, and a Postgres that actually uses the RAM it paid for. Numbers below are for an H-64.
Leave room for the kernel, the page cache and the fork during a Redis snapshot. maxmemory at three quarters of RAM and shared_buffers at a quarter are starting points; measure and adjust.
Step 5: Verify, then baseline
Write the numbers down somewhere you'll actually find them again. Then order. And should a resize be needed later on, RAM and vCPU change in place with one short reboot, while data and IP stay exactly where they are.
Pitfalls we see, and how to avoid them.
Redis with no maxmemory
Redis grows until the OOM killer ends it, often during a snapshot fork when memory use spikes. Set maxmemory and a policy on day one.
A JVM heap above 32 GB
Go past roughly 31 GB and the JVM drops compressed pointers, at which point a bigger heap can actually hold less. Odd, but true. Keep the heap under that line and hand the rest to page cache.
Buying RAM for a slow query
A missing index is not fixed by 128 GB. Run EXPLAIN ANALYZE and check the buffer hit ratio first; if it is already high, more memory changes nothing.
Forgetting the disk arithmetic
A 100 GB dataset in memory still wants a copy on disk, plus snapshots, plus backups. H-64 gives you 480 GB. Before you assume that fits, add up the RDB files, the WAL and a nightly dump.
Pick the region closest to your users.
A cache is only as fast as the round trip to it. Nothing else counts. From Amsterdam we measure 9.0 ms to London and 8.5 ms to Frankfurt. Out of New York, Boston sits at 5-8 ms and Washington DC at 7-12 ms.
Questions about High Memory VPS.
How is High Memory different from Standard?
Same hardware, to start with. Both run shared vCPU on EPYC Milan with DDR4 ECC memory and NVMe RAID10. The difference is shape. The H-series gives 8 GB of RAM per vCPU and climbs to 192 GB, where the Standard line carries less memory per core and stops at 64 GB on S-64.
Should I buy H-64 or S-64?
Can you put sixteen vCPU to work, and is 64 GB enough? Then S-64 gives the same memory with more cores and more disk at a lower monthly price, and that is what we'd tell you. H-64 is for when you specifically want the 8 GB per vCPU shape. H-128 or H-192 is for when 64 GB is not enough.
Does the €1 first month apply to High Memory plans?
No. The €1 first month applies to S-2 through S-64 and to D-4, D-8 and D-16, and stops there. High Memory plans bill at the normal monthly price from day one. What you do keep is the 30-day performance guarantee on your first order: full refund within 30 days of activation, once per customer.
Are the vCPUs on the H-series pinned?
No. High Memory sits on shared vCPU, exactly like the Standard tier. Got a latency target that needs pinned cores? Then look at the Dedicated tier instead. One catch, though. Dedicated tops out at 64 GB on D-64, so above that you are picking one or the other, not both.
Is the memory ECC?
Yes, every one of them. All BuyVPS nodes run ECC memory, DDR4 on Standard and High Memory, DDR5 4800 MT/s on Dedicated. For an in-memory database that detail is not a footnote. Flip a bit in a web request and the request gets retried, nobody notices. Flip a bit in a table that lives entirely in RAM and it lands in your data. And stays there.
Can I resize an H-plan later?
Yes. RAM and vCPU resize in place, usually at the cost of one short reboot, and your data and IP don't move. A one-off? Click it through in the panel. Prefer to script it? The REST API does exactly the same thing.
Deploy a measured VPS
From $6.79/mo on a 2-year term. KVM on AMD EPYC, NVMe RAID10, ECC memory. No setup fee, cancel monthly.
30-day performance guarantee. Not satisfied? Full refund, no admin fee. How it works