[{"data":1,"prerenderedAt":662},["ShallowReactive",2],{"guides-index":3},[4,103,217,328,487,563],{"id":5,"title":6,"body":7,"date":93,"description":94,"draft":95,"extension":96,"meta":97,"navigation":98,"path":99,"seo":100,"stem":101,"__hash__":102},"guides\u002Fguides\u002Fbare-metal-vs-vps.md","Bare Metal vs. VPS: What's Actually Different",{"type":8,"value":9,"toc":83},"minimark",[10,14,19,22,25,29,32,36,39,43,46,59,66,69,73,76,80],[11,12,13],"p",{},"\"Bare metal\" sounds like the premium, expensive option next to a VPS, and sometimes it is. But the price gap between the two isn't as predictable as that assumption suggests, and the two aren't really on the same spectrum of \"better\" and \"worse\" so much as different tools built for different problems. Here's what actually separates them, and when it's worth paying for one over the other.",[15,16,18],"h2",{"id":17},"the-core-difference","The core difference",[11,20,21],{},"A VPS is a virtual machine: a hypervisor divides one physical server's CPU, RAM, and storage among several tenants, and you get an isolated slice of it. Bare metal is the opposite: you get the entire physical machine, no hypervisor, no other tenants, nothing else running on that hardware but your own workload.",[11,23,24],{},"That single difference cascades into everything else. On a VPS, your neighbors on the same physical box can, in theory, affect your performance if the provider oversells shared resources. On bare metal, there's nothing to share, because there's nobody else on the machine.",[15,26,28],{"id":27},"why-someone-chooses-bare-metal","Why someone chooses bare metal",[11,30,31],{},"Full hardware access is the headline reason. You can run your own hypervisor on top of it if you want to virtualize further yourself, install a custom kernel, tune hardware-level settings a VPS provider would never expose to a shared tenant, or meet a compliance requirement that specifically forbids multi-tenant hardware (some data-handling regulations and enterprise contracts do exactly this). Performance predictability is the other big one: with no hypervisor overhead and no other tenants, a bare metal box gives you the most consistent, ceiling-free performance you can buy, which matters for latency-sensitive workloads like real-time databases, game servers, or anything with a strict SLA.",[15,33,35],{"id":34},"why-most-people-still-pick-a-vps","Why most people still pick a VPS",[11,37,38],{},"Bare metal isn't free of trade-offs. Provisioning takes longer, sometimes measured in hours rather than the minutes a VPS spins up in, since a physical machine has to be racked, imaged, and configured rather than just carved out of an already-running hypervisor. Scaling is chunkier too: resizing a VPS is often a few clicks and a reboot, while scaling bare metal usually means provisioning an entirely new physical box. And for most workloads, especially anything that isn't consistently maxing out CPU or fighting noisy neighbors, a VPS is simply the cheaper and more flexible way to get equivalent day-to-day performance.",[15,40,42],{"id":41},"the-price-gap-isnt-what-youd-expect","The price gap isn't what you'd expect",[11,44,45],{},"Here's the part that surprises people: bare metal pricing varies enormously by provider, to the point that \"bare metal costs more\" isn't a safe assumption on its own.",[11,47,48,53,54,58],{},[49,50,52],"a",{"href":51},"\u002Fproviders\u002Fovhcloud","OVHcloud's"," cheapest bare metal server, 25skb012, is €9.99 a month (about $11.42) for 4 dedicated cores, 32GB of RAM, and 120GB of SSD storage. That's not a typo, and it undercuts a lot of VPS plans with fewer resources. ",[49,55,57],{"href":56},"\u002Fproviders\u002Fvultr","Vultr's"," cheapest bare metal plan, vbm-4c-32gb, has nearly identical specs, 4 cores and 32GB RAM, and costs $120 a month, about 10.5 times OVHcloud's price for a comparable box once you convert both to the same currency. Same category of product, wildly different price, purely down to how each provider positions and prices its bare metal line.",[11,60,61,65],{},[49,62,64],{"href":63},"\u002Fproviders\u002Fcherry-servers","Cherry Servers"," sits in between, with bare metal starting at €49 a month (about $56) for a 4-core, 16GB box (its e3-1240v3 plan), and it's worth noting that Cherry Servers treats dedicated hardware as its core product rather than a premium tier bolted onto a VPS lineup, which shows in how competitively its entry-level bare metal is priced.",[11,67,68],{},"The takeaway: don't assume bare metal is automatically the expensive choice. It's worth comparing actual listings, because the gap between the cheapest and priciest bare metal offer for near-identical specs can be an order of magnitude.",[15,70,72],{"id":71},"how-to-tell-which-one-youre-looking-at","How to tell which one you're looking at",[11,74,75],{},"On vpsudo, every offer carries a type badge: VPS, bare metal, or GPU. Bare metal listings are, by definition, dedicated hardware, since there's no hypervisor to share cores through in the first place. That's a useful shortcut if you're scanning the offer table for a guaranteed-isolated box without reading every provider's product description individually.",[15,77,79],{"id":78},"picking-between-them","Picking between them",[11,81,82],{},"If you're building something that can tolerate the normal give-and-take of shared virtualized infrastructure, a VPS is almost always the right starting point: cheaper, faster to provision, easier to resize as you grow. Reach for bare metal when you have a specific reason a VPS won't satisfy: a compliance rule about tenancy, a workload that needs every cycle of a physical machine, or a performance requirement where \"usually consistent\" isn't good enough and you need \"always consistent.\" And when you do reach for it, check more than one provider's bare metal pricing before assuming it's the pricier option, because as the numbers above show, that's often not true.",{"title":84,"searchDepth":85,"depth":85,"links":86},"",2,[87,88,89,90,91,92],{"id":17,"depth":85,"text":18},{"id":27,"depth":85,"text":28},{"id":34,"depth":85,"text":35},{"id":41,"depth":85,"text":42},{"id":71,"depth":85,"text":72},{"id":78,"depth":85,"text":79},"2026-07-10","The real difference between a bare metal server and a VPS, when each one makes sense, and why bare metal isn't always the pricier option, with real numbers from live provider data.",false,"md",{},true,"\u002Fguides\u002Fbare-metal-vs-vps",{"title":6,"description":94},"guides\u002Fbare-metal-vs-vps","4fbLy1IY1eos79tF4vqywy0QC2vY0J8G2v_DMmHXx-A",{"id":104,"title":105,"body":106,"date":93,"description":211,"draft":95,"extension":96,"meta":212,"navigation":98,"path":213,"seo":214,"stem":215,"__hash__":216},"guides\u002Fguides\u002Fhow-to-choose-a-server-location.md","How to Choose a Server Location",{"type":8,"value":107,"toc":203},[108,111,115,118,121,124,127,131,134,138,141,144,148,151,154,157,161,164,168,190],[11,109,110],{},"Most people pick a server region one of two ways: whatever's closest to them personally, or whatever shows up first in the dropdown. Neither is a great method, and the gap between a good regional choice and a bad one shows up in three separate places: how fast your app feels, whether you're on the right side of a compliance rule, and sometimes, though less often than you'd think, the invoice itself. Here's how to actually think about it.",[15,112,114],{"id":113},"the-physics-what-distance-costs-you","The physics: what distance costs you",[11,116,117],{},"Light in fiber-optic cable doesn't travel at the speed of light in a vacuum. Glass slows it down to roughly two-thirds of that, around 200,000 km per second. The rule of thumb that falls out of that number: about 1ms of one-way latency for every 200km of fiber path, or roughly 0.5ms per 100km. Round trip time, the number that actually matters for a request-response cycle, is about double that. And that's still the theoretical floor. Real routes aren't a straight line of fiber, they hop through internet exchange points and carrier backbones, so measured latency usually runs 1.5 to 2 times higher than the straight-line physics suggests.",[11,119,120],{},"Concrete example: a server in Frankfurt talking to a user in Berlin, a few hundred kilometers away, adds a few milliseconds, effectively unnoticeable. That same Frankfurt server talking to a user in Sydney, about 16,500 km away, has a theoretical fiber-only round trip of roughly 165ms, and real-world measurements on that route typically land closer to 280 to 320ms once actual routing is accounted for.",[11,122,123],{},"Whether that gap matters depends on what you're running. For anything interactive (a page load, an API call a user is waiting on, a form submit, a multiplayer game tick) keeping round trips under roughly 50ms is the threshold where a response starts to feel instant rather than laggy. That budget disappears fast, too: a page load isn't one round trip, it's a TLS handshake, a DNS lookup, and however many API calls the page makes, each one paying the same distance tax. Stack a few of those in sequence and 300ms of one-way latency becomes a second or more of perceived wait.",[11,125,126],{},"For batch work, none of this applies. A nightly backup, a cron job syncing a database, a video transcode, a log export, these are throughput-bound, not round-trip-bound. An extra 200ms on a job that runs for twenty minutes is nothing. If your workload runs unattended rather than making a person stare at a spinner, region choice barely affects the workload itself, though it can still matter for the other two reasons below.",[15,128,130],{"id":129},"optimize-for-your-traffic-not-your-address","Optimize for your traffic, not your address",[11,132,133],{},"This is the mistake that's easy to make without noticing: picking a region because it's wherever you happen to be developing from, or because it's the cheapest one a provider shows first, rather than checking where your actual users are. If you're building in Los Angeles but most of your traffic comes from the East Coast and Europe, provisioning in California because it's convenient during development is optimizing for the wrong audience. Pull up your analytics, look at where requests actually originate by traffic share, and pick the region that serves the bulk of that traffic.",[15,135,137],{"id":136},"data-residency-and-gdpr","Data residency and GDPR",[11,139,140],{},"This isn't legal advice, just the practical shape of the issue: if you're handling personal data belonging to people in the EU, a lot of organizations choose, or are contractually required, to keep that data on servers physically located in the EU, as part of complying with GDPR. Where the disk physically sits is part of what regulators and enterprise contracts care about, separate from anything about encryption or access control.",[11,142,143],{},"The good news is there's no shortage of EU options across the providers listed on vpsudo. Frankfurt, Paris, Amsterdam, Warsaw, Stockholm, and Milan all show up as live regions across multiple vendors, and OVHcloud alone offers French datacenters in Gravelines, Strasbourg, and Roubaix on top of smaller \"Local Zone\" deployments in Prague, Madrid, Brussels, Marseille, and Vienna. EU residency is rarely the limiting factor in picking a provider. Check with someone who actually knows your compliance obligations before treating \"EU region\" as a full answer, but as a baseline, EU-soil hosting for EU personal data closes off an entire category of question before it gets asked.",[15,145,147],{"id":146},"regional-pricing-differences","Regional pricing differences",[11,149,150],{},"Here's something that surprised us when we checked vpsudo's own live offer data rather than assuming it: for most providers, the price of a given plan doesn't change by region at all. Vultr's vhf-1c-1gb dedicated plan is $6 a month whether you deploy it in New Jersey, Tokyo, or Mumbai, across all 30-plus regions it's offered in. The same holds for Linode, DigitalOcean, Scaleway, Cherry Servers, IONOS, and RackNerd in our data: which region you pick changes latency and compliance, not the number on the invoice.",[11,152,153],{},"OVHcloud is the exception, and it's a real one. Its vps-2025-model1 plan (4 vCPU, 8GB RAM, 75GB NVMe) costs €6.49 a month in OVHcloud's standard regions (Beauharnois, Frankfurt, Gravelines, Strasbourg, Singapore, Sydney, London, Warsaw, Mumbai, Milan, Roubaix). The identical spec under its \"Local Zone\" line, vps-2025-model1.LZ, costs €7.49 a month instead, deployed in a different, smaller set of European cities: Prague, Madrid, Amsterdam, Brussels, Marseille, Vienna, and Zurich. Same 4 vCPU, same 8GB RAM, same 75GB of NVMe storage, but a €1 (roughly 15%) premium purely for landing in one of OVHcloud's smaller regional deployments rather than its larger flagship datacenters.",[11,155,156],{},"Region availability itself also varies a lot by provider, separately from price. Vultr and Linode each cover dozens of regions; Netcup runs entirely out of Nuremberg; RackNerd sticks to about a dozen North American and European cities. Check that a provider actually operates in the region you need before you compare its price against anyone else's, because the cheapest plan on paper is worthless if it's not orderable where your users are.",[15,158,160],{"id":159},"bandwidth-and-peering-quality","Bandwidth and peering quality",[11,162,163],{},"Two servers with identical CPU, RAM, and bandwidth allowance in different regions of the same provider can still perform differently for traffic leaving the datacenter, because of what network they sit on rather than what hardware they run. Datacenters in major connectivity hubs (Frankfurt, Amsterdam, London) tend to have dense peering with other networks and internet exchanges, so traffic often takes a short, direct route to wherever it's going. A smaller or newer regional deployment can lean more heavily on transit providers to reach the same destinations, adding hops that don't show up anywhere on a spec sheet. This isn't something a plan listing tells you upfront, so if peering quality matters for your workload, a quick traceroute or ping test from a box in the region you're considering is worth doing before you commit to a year of it.",[15,165,167],{"id":166},"the-checklist","The checklist",[169,170,171,175,178,181,184,187],"ul",{},[172,173,174],"li",{},"Where does most of your traffic actually come from, by real numbers, not guesswork",[172,176,177],{},"Does the workload need sub-50ms round trips, or is it fine running unattended",[172,179,180],{},"Are you handling EU residents' personal data, and does that mean EU-soil hosting",[172,182,183],{},"Does the provider even offer a region where you need one, before you compare price",[172,185,186],{},"Is there a regional pricing tier (like OVHcloud's Local Zone) that changes the number",[172,188,189],{},"Is the datacenter in a well-peered hub, or a smaller build-out worth testing first",[11,191,192,193,197,198,202],{},"Once you've answered those, ",[49,194,196],{"href":195},"\u002F","vpsudo's offer table"," lets you filter by region directly, so you're comparing plans that are actually available where you need them rather than discovering a mismatch after checkout. And if the deciding factor between two regions on the same provider turns out to be shared versus dedicated CPU rather than location, the ",[49,199,201],{"href":200},"\u002Fguides\u002Fshared-vs-dedicated-vcpu","shared vs. dedicated vCPU guide"," covers that trade-off in full.",{"title":84,"searchDepth":85,"depth":85,"links":204},[205,206,207,208,209,210],{"id":113,"depth":85,"text":114},{"id":129,"depth":85,"text":130},{"id":136,"depth":85,"text":137},{"id":146,"depth":85,"text":147},{"id":159,"depth":85,"text":160},{"id":166,"depth":85,"text":167},"The physics behind server latency, why data residency and GDPR matter, and where regional pricing actually differs, with real examples from live provider data.",{},"\u002Fguides\u002Fhow-to-choose-a-server-location",{"title":105,"description":211},"guides\u002Fhow-to-choose-a-server-location","c9yHK6A5ieLDUZo27HmParwHnLT_ydp7SbmW1c5nWkg",{"id":218,"title":219,"body":220,"date":93,"description":322,"draft":95,"extension":96,"meta":323,"navigation":98,"path":324,"seo":325,"stem":326,"__hash__":327},"guides\u002Fguides\u002Fnvme-vs-ssd-vs-hdd.md","NVMe vs. SSD vs. HDD: Which Storage Do You Actually Need",{"type":8,"value":221,"toc":315},[222,225,229,237,243,249,253,256,260,266,272,276,282,285,289,292,312],[11,223,224],{},"Storage type is the spec people skip past fastest when comparing VPS plans, and it's one of the ones that changes how a server actually feels the most. \"50GB of storage\" reads the same on paper whether it's a spinning hard drive or an NVMe SSD, but the two perform nothing alike. Here's what each one actually is, and what the trade-off costs in real listings.",[15,226,228],{"id":227},"the-three-types-physically","The three types, physically",[11,230,231,232,236],{},"A ",[233,234,235],"strong",{},"hard disk drive (HDD)"," stores data on spinning magnetic platters, with a physical read\u002Fwrite head that moves across the disk to find data. It's mechanical, which is exactly why it's slower for anything that jumps around the disk rather than reading it in one long sequential pass.",[11,238,231,239,242],{},[233,240,241],{},"SATA SSD"," replaces the spinning platters with flash memory, no moving parts, but still connects over the same SATA interface originally designed for hard drives decades ago. That interface caps how much throughput can move at once, so a SATA SSD is dramatically faster than an HDD but still bottlenecked by the connection it's plugged into.",[11,244,245,248],{},[233,246,247],{},"NVMe"," is also flash memory, but it talks to the system over PCIe instead of SATA, the same high-bandwidth bus used for graphics cards. Without the SATA bottleneck in the way, NVMe drives are consistently the fastest option for the kind of small, random reads and writes that databases, application servers, and anything under real concurrent load generate constantly.",[15,250,252],{"id":251},"why-the-difference-actually-matters","Why the difference actually matters",[11,254,255],{},"The gap isn't just a benchmark number, it's what your server feels like under real use. A database doing lots of small random reads and writes will feel meaningfully snappier on NVMe than on a SATA SSD, and both will run circles around an HDD for that same workload. Sequential throughput (writing one big file start to finish, like a backup or a video export) narrows the gap somewhat, since that's the one scenario where spinning disks aren't at their worst. But most real server workloads, web apps, databases, anything with concurrent users, are dominated by random I\u002FO, which is exactly where HDDs fall furthest behind.",[15,257,259],{"id":258},"where-hdd-still-wins-raw-capacity-per-dollar","Where HDD still wins: raw capacity per dollar",[11,261,262,263,265],{},"None of this makes HDD storage obsolete. It's still the cheapest way to buy raw capacity, and for the right workload that's the only spec that matters. ",[49,264,64],{"href":63}," sells a plan, S1-1-2gb-700hdd-shared, with 700GB of HDD storage for $19 a month. Its SSD-backed plans in the same product line top out around 80GB for $9 a month. If what you need is bulk space (backups, media archives, log retention, cold storage you rarely touch) the HDD plan gets you almost nine times the capacity for about twice the price. That's a good trade if you're not hammering the disk with random reads.",[11,267,268,271],{},[49,269,270],{"href":51},"OVHcloud"," shows the same pattern on its bare metal line. Its cheapest dedicated server, 25skb012, is $9.99 a month for 4 CPU cores, 32GB RAM, and 120GB of SSD storage. Step up to 24sk402 at $16.99 a month, same 4 dedicated cores, and you get 4,000GB of HDD storage instead, over 33 times the space, for roughly $7 more a month (with less RAM as the trade-off on that particular plan). If your bottleneck is disk space rather than disk speed, that's a genuinely good deal.",[15,273,275],{"id":274},"where-nvme-earns-the-premium","Where NVMe earns the premium",[11,277,278,281],{},[49,279,280],{"href":56},"Vultr"," makes the NVMe upgrade almost trivial to justify at the entry level. Its shared SSD plan, vc2-1c-1gb, is $5 a month for 1 vCPU, 1GB RAM, and 25GB of SSD storage. Its dedicated NVMe plan, vhf-1c-1gb, is $6 a month for the same 1 vCPU and 1GB RAM, but with 32GB of NVMe storage and a dedicated core on top. That extra dollar buys faster storage, more of it, and a reserved core all at once, which is about as clean a case as you'll find for skipping the cheapest tier.",[11,283,284],{},"For anything running a database, a CMS with real traffic, or an application server handling concurrent requests, NVMe is worth paying for even when it costs more than a comparable SSD plan. The random I\u002FO performance difference shows up directly in how fast pages load and queries return.",[15,286,288],{"id":287},"how-to-pick","How to pick",[11,290,291],{},"Match the storage type to what the disk is actually doing, not to what's cheapest per gigabyte in isolation:",[169,293,294,300,306],{},[172,295,296,299],{},[233,297,298],{},"Bulk storage, backups, archives, media libraries",": HDD is the right call, and the capacity-per-dollar gap over SSD or NVMe is too large to ignore for this use case.",[172,301,302,305],{},[233,303,304],{},"General-purpose web hosting, small-to-medium apps",": SATA SSD is a solid default, fast enough for most traffic without paying an NVMe premium you won't notice.",[172,307,308,311],{},[233,309,310],{},"Databases, high-concurrency apps, anything latency-sensitive",": NVMe, and as the Vultr example above shows, the price gap to get there is sometimes smaller than people assume.",[11,313,314],{},"vpsudo's offer table shows storage type on every listing where the provider discloses it, so you can filter directly for NVMe (or explicitly avoid it if you're buying bulk storage and don't want to pay for speed you won't use).",{"title":84,"searchDepth":85,"depth":85,"links":316},[317,318,319,320,321],{"id":227,"depth":85,"text":228},{"id":251,"depth":85,"text":252},{"id":258,"depth":85,"text":259},{"id":274,"depth":85,"text":275},{"id":287,"depth":85,"text":288},"The real difference between NVMe, SATA SSD, and spinning HDD storage on a VPS, with real price and capacity comparisons pulled from live provider data.",{},"\u002Fguides\u002Fnvme-vs-ssd-vs-hdd",{"title":219,"description":322},"guides\u002Fnvme-vs-ssd-vs-hdd","ZkNq_9zfSOhyOQoAFyoFf4PtTI9NJe0YHy3X9RuSQYk",{"id":329,"title":330,"body":331,"date":93,"description":482,"draft":95,"extension":96,"meta":483,"navigation":98,"path":200,"seo":484,"stem":485,"__hash__":486},"guides\u002Fguides\u002Fshared-vs-dedicated-vcpu.md","Shared vs. Dedicated vCPU, Explained",{"type":8,"value":332,"toc":473},[333,336,340,343,347,350,354,357,361,364,430,433,437,440,446,453,459,463,466,470],[11,334,335],{},"Almost every VPS plan lists a vCPU count, and almost none of them explain what that vCPU actually is. That gap matters, because \"1 vCPU\" on a shared plan and \"1 vCPU\" on a dedicated plan can perform very differently under load, even though the listing looks identical. Here's what the terms mean, how to spot which one a provider is selling you, and what the price difference actually buys.",[15,337,339],{"id":338},"what-vcpu-means-in-the-first-place","What \"vCPU\" means in the first place",[11,341,342],{},"A vCPU is a virtual CPU core, a slice of processing time that a hypervisor hands out to a virtual machine. It's not a physical chip you can point to. The hypervisor is the layer of software that carves up a physical server's real cores and RAM among however many virtual machines are running on it.",[15,344,346],{"id":345},"shared-your-core-on-a-timeshare","Shared: your core, on a timeshare",[11,348,349],{},"On a shared-vCPU plan, the physical core backing your vCPU is also backing other tenants' vCPUs on the same host. The hypervisor schedules time on that core between everyone sharing it. Most of the time this is invisible: the host isn't fully loaded, and your slice of CPU time shows up when you need it. But if another tenant on the same physical machine spikes hard (a \"noisy neighbor,\" in the term everyone in hosting uses for this), your own performance can dip even though nothing changed on your end. Providers oversell shared capacity to keep prices low, which is exactly why shared plans are consistently the cheapest entry point into VPS hosting.",[15,351,353],{"id":352},"dedicated-the-whole-core-reserved","Dedicated: the whole core, reserved",[11,355,356],{},"A dedicated vCPU is a physical core (or a fixed slice of one) reserved exclusively for your virtual machine. Nobody else's workload competes for it. You get consistent, predictable CPU performance, which is what makes dedicated cores the standard recommendation for production databases, anything CPU-bound (video transcoding, build servers, compute-heavy APIs), or workloads where latency spikes are actually costly.",[15,358,360],{"id":359},"how-to-tell-which-one-youre-buying","How to tell which one you're buying",[11,362,363],{},"Providers disclose this differently, and it's rarely spelled out on the pricing page:",[169,365,366,384,406,416,425],{},[172,367,368,370,371,375,376,379,380,383],{},[233,369,280],{},": plan codes starting ",[372,373,374],"code",{},"vc2"," are shared; ",[372,377,378],{},"vhf"," (high frequency) and ",[372,381,382],{},"vhp"," (high performance) are dedicated.",[172,385,386,389,390,393,394,397,398,401,402,405],{},[233,387,388],{},"DigitalOcean",": plan slugs starting with ",[372,391,392],{},"s-"," are shared \"Basic\" droplets; ",[372,395,396],{},"c-",", ",[372,399,400],{},"g-",", and ",[372,403,404],{},"m-"," prefixes are dedicated CPU-optimized, general-purpose, and memory-optimized droplets.",[172,407,408,411,412,415],{},[233,409,410],{},"Hetzner",": exposes a ",[372,413,414],{},"cpu_type"," field directly on each server type in its pricing API, so there's no guesswork.",[172,417,418,411,421,424],{},[233,419,420],{},"Scaleway",[372,422,423],{},"hardware.cpu.shared"," boolean directly in its product catalog, same idea.",[172,426,427,429],{},[233,428,64],{},": sells dedicated cores by default across its bare metal line, which is the core of its product, rather than treating dedicated as a premium add-on tier.",[11,431,432],{},"vpsudo's own offer table shows a CPU type badge on every listing where the provider discloses it, so you don't have to go digging through plan-code conventions yourself.",[15,434,436],{"id":435},"what-the-price-gap-actually-looks-like","What the price gap actually looks like",[11,438,439],{},"This is where it gets concrete, and the gap is smaller than you'd guess in some cases and much bigger in others.",[11,441,442,443,445],{},"On ",[49,444,280],{"href":56},", the shared plan vc2-1c-1gb (1 vCPU, 1GB RAM, SSD storage) runs $5 a month. The dedicated equivalent, vhf-1c-1gb (1 vCPU, 1GB RAM, but on NVMe storage), is $6. That's a $1 premium that also happens to come with faster storage as a bonus, since Vultr's dedicated tier runs on NVMe while its shared tier runs on plain SSD.",[11,447,442,448,452],{},[49,449,451],{"href":450},"\u002Fproviders\u002Flinode","Linode",", the gap is proportionally bigger. The shared plan g6-standard-2 (2 vCPU, 4GB RAM, 80GB SSD) is $24 a month. The dedicated plan g6-dedicated-2, with the exact same 2 vCPU, 4GB RAM, and 80GB SSD spec, is $36. That's a 50% markup for the identical resource numbers, purely for the guarantee that those two cores are yours alone.",[11,454,455,456,458],{},"And then there's ",[49,457,64],{"href":63},", which doesn't really sell a \"shared vs. dedicated\" upgrade path so much as two different product lines: its shared VPS plans start around $3 a month, while its bare metal line, dedicated by definition since there's no hypervisor at all, starts at $49 a month for a 4-core, 16GB box. That's not really the same axis of comparison as the Vultr or Linode examples, but it shows how wide the range gets once \"dedicated\" means an entire physical server rather than a reserved slice of one.",[15,460,462],{"id":461},"when-shared-is-genuinely-fine","When shared is genuinely fine",[11,464,465],{},"Don't read any of this as \"always buy dedicated.\" A lot of workloads never come close to saturating a shared core: static sites, low-traffic APIs, dev and staging environments, personal projects, anything where a brief CPU dip during someone else's spike is a non-event. Shared plans exist because most VPS traffic doesn't need guaranteed cores, and paying the dedicated premium for a workload that never touches its ceiling is money spent on a guarantee you'll never redeem.",[15,467,469],{"id":468},"when-dedicated-earns-its-premium","When dedicated earns its premium",[11,471,472],{},"The cases where it's worth paying up: production databases, anything with a latency SLA, CPU-bound batch jobs, and workloads where a performance dip during someone else's traffic spike would actually cost you (a checkout flow, a real-time API). If your workload falls in that bucket, the premium in the examples above (a dollar on Vultr, 50% on Linode) is usually cheap insurance against a problem that's genuinely hard to debug after the fact, because a noisy-neighbor slowdown looks identical to your own code being slow until you know to look for it.",{"title":84,"searchDepth":85,"depth":85,"links":474},[475,476,477,478,479,480,481],{"id":338,"depth":85,"text":339},{"id":345,"depth":85,"text":346},{"id":352,"depth":85,"text":353},{"id":359,"depth":85,"text":360},{"id":435,"depth":85,"text":436},{"id":461,"depth":85,"text":462},{"id":468,"depth":85,"text":469},"What 'shared' and 'dedicated' actually mean on a VPS plan, how to tell which one you're looking at, and real price gaps between the two from live provider data.",{},{"title":330,"description":482},"guides\u002Fshared-vs-dedicated-vcpu","em2sfWrIkCXcLZIk9Ka17N48rpNdkP4aR_dOXP10CUk",{"id":488,"title":489,"body":490,"date":93,"description":557,"draft":95,"extension":96,"meta":558,"navigation":98,"path":559,"seo":560,"stem":561,"__hash__":562},"guides\u002Fguides\u002Funderstanding-gpu-rental-pricing.md","Understanding GPU Rental Pricing",{"type":8,"value":491,"toc":550},[492,495,499,502,506,519,522,526,529,533,540,543,547],[11,493,494],{},"GPU rental is a different market from VPS and bare metal hosting, priced differently, sold differently, and worth understanding on its own terms before you compare a listing against anything else on vpsudo. Here's what actually drives the price you see, and why two listings for what looks like the same card can cost different amounts.",[15,496,498],{"id":497},"billed-by-the-hour-not-the-month","Billed by the hour, not the month",[11,500,501],{},"Almost every VPS and bare metal plan on vpsudo is priced monthly. GPU listings are priced hourly, which is the first thing to notice when you're scanning the offer table. That's not a quirk, it reflects how the workloads differ: GPU compute is usually rented for a training run, a batch inference job, or a rendering task with a defined start and end, not a server you expect to leave running indefinitely. The practical consequence is that you need to actually stop or terminate a GPU instance when you're done with it, since the meter runs continuously the way it wouldn't on a monthly VPS plan you're already paying for regardless of usage.",[15,503,505],{"id":504},"the-price-spread-is-enormous-and-it-maps-to-hardware","The price spread is enormous, and it maps to hardware",[11,507,508,509,513,514,518],{},"At the low end, consumer-grade cards rent for well under a dollar an hour. On ",[49,510,512],{"href":511},"\u002Fproviders\u002Fvast-ai","Vast.ai"," and ",[49,515,517],{"href":516},"\u002Fproviders\u002Frunpod","RunPod",", cards like the RTX 4070, RTX 5070, and RTX 3090 show up around $0.08 to $0.15 an hour. At the high end, data-center GPUs built specifically for AI workloads, the H100, H200, and B200, run roughly $2.60 to $8 an hour for a single card on the same two platforms. Multi-GPU clusters climb even higher: an 8x B300 cluster listed on Vast.ai priced at $53.51 an hour at the time of writing. That's close to a 700x spread between the cheapest single consumer card and the priciest multi-GPU cluster, on the same two marketplaces.",[11,520,521],{},"What explains the gap is mostly VRAM and memory bandwidth. Consumer cards top out with far less onboard memory than data-center cards, which matters enormously for AI workloads because the model's weights and activations have to actually fit in that memory to run at all. A large language model that needs 80GB or more of VRAM simply won't run on a consumer card no matter how cheap it is, so the price difference isn't arbitrary markup, it's paying for capacity that's a hard requirement for the workload rather than a nice-to-have.",[15,523,525],{"id":524},"match-the-card-to-the-job","Match the card to the job",[11,527,528],{},"Not every workload needs a data-center GPU. Inference on a smaller model, image generation, or general experimentation often runs fine on a consumer card at a fraction of the price. Training a large model from scratch, fine-tuning something with a big parameter count, or running workloads that need multiple GPUs talking to each other over a fast interconnect (NVLink, on the data-center side) is where you actually need the more expensive hardware. Renting an H100 to run a workload a $0.10-an-hour RTX card would handle just fine is the GPU-market equivalent of buying a bare metal server for a static website: technically fine, but not what your money should be going toward.",[15,530,532],{"id":531},"why-the-marketplace-model-looks-different-from-vps-hosting","Why the marketplace model looks different from VPS hosting",[11,534,535,513,537,539],{},[49,536,512],{"href":511},[49,538,517],{"href":516}," both operate closer to a decentralized compute marketplace than a traditional single-vendor hosting company. Individual hardware owners (and data centers) list their idle GPU capacity, and pricing reflects each individual host's own rate rather than one fixed provider-wide price list. That's part of why you'll see multiple listings for what looks like the exact same GPU model at different prices: they're genuinely different hosts, sometimes in different regions, with different reliability track records, competing on price for the same hardware spec.",[11,541,542],{},"It also explains an odd-looking detail if you dig into vpsudo's raw offer data: some RunPod listings show 0 for RAM and storage, because RunPod prices purely by GPU rather than bundling a fixed CPU, RAM, and disk allocation the way a VPS plan does. The GPU is the product; the surrounding resources are negotiated or included separately depending on the host.",[15,544,546],{"id":545},"putting-it-together","Putting it together",[11,548,549],{},"If you're shopping GPU listings on vpsudo, the two numbers to actually compare are VRAM (does the card have enough memory for what you're running) and hourly price, in that order. Everything else, like region or the specific host behind a listing, matters less than those two. And because billing is hourly rather than monthly, remember that the habits that work for a VPS (leave it running, it's a flat monthly cost either way) don't apply here: an idle GPU instance is still burning money every hour it's up, so shut it down the moment the job is done.",{"title":84,"searchDepth":85,"depth":85,"links":551},[552,553,554,555,556],{"id":497,"depth":85,"text":498},{"id":504,"depth":85,"text":505},{"id":524,"depth":85,"text":525},{"id":531,"depth":85,"text":532},{"id":545,"depth":85,"text":546},"Why GPU compute is billed by the hour instead of the month, what actually drives the price spread from consumer cards to data-center GPUs, and real numbers from live provider data.",{},"\u002Fguides\u002Funderstanding-gpu-rental-pricing",{"title":489,"description":557},"guides\u002Funderstanding-gpu-rental-pricing","mZtK5UmvBg8tx6WXnlOfe25m7EOFrtqm4h3Z4TzL3_Q",{"id":564,"title":565,"body":566,"date":93,"description":656,"draft":95,"extension":96,"meta":657,"navigation":98,"path":658,"seo":659,"stem":660,"__hash__":661},"guides\u002Fguides\u002Fwhat-to-look-for-when-buying-a-vps.md","What to Look for When Buying a VPS",{"type":8,"value":567,"toc":647},[568,574,578,584,590,594,604,608,611,615,618,622,625,629,640,644],[11,569,570,571,573],{},"The sticker price on a VPS listing tells you almost nothing on its own. Scaleway will rent you a box for $0.43 a month. ",[49,572,280],{"href":56}," has a shared-vCPU plan at $2.50. Neither number means much until you know what's actually inside: how many cores, what kind of cores, how much RAM, what the storage is built on, and what happens the day you outgrow it. This guide walks through the specs that actually separate a good deal from a bad one, so you're comparing plans on the numbers that matter instead of just the number at the top.",[15,575,577],{"id":576},"start-with-vcpu-type-not-vcpu-count","Start with vCPU type, not vCPU count",[11,579,580,581,583],{},"Two plans can both say \"1 vCPU\" and mean very different things. A shared vCPU is a slice of a physical core that's also being handed out to other tenants on the same machine. A dedicated vCPU is a core reserved for you and nobody else. ",[49,582,57],{"href":56}," cheapest shared plan, vc2-1c-1gb, runs $5 a month for 1 CPU and 1GB of RAM. Its dedicated equivalent, vhf-1c-1gb, is $6 for the same CPU and RAM count, just on a core that isn't shared. That's a $1 premium for a real, measurable difference in consistency: no other tenant's traffic spike eats into your CPU time.",[11,585,586,587,589],{},"Whether that dollar is worth it depends on the workload. A static site or a dev sandbox won't notice the difference. A database taking real traffic will. We've got a full breakdown of this in the ",[49,588,201],{"href":200},", including how to spot which is which on providers that don't spell it out in the plan name.",[15,591,593],{"id":592},"storage-how-much-and-what-kind","Storage: how much, and what kind",[11,595,596,597,599,600,603],{},"\"50GB of storage\" reads the same whether it's sitting on a spinning hard drive or an NVMe SSD, and the difference in how that storage performs is enormous. ",[49,598,64],{"href":63}," sells a plan with 700GB of HDD storage for $19 a month, next to a plan with just 80GB of SSD storage for $9. Neither is the \"better\" plan in the abstract: the HDD box is for bulk storage where raw capacity matters more than speed (backups, media libraries, archives), and the SSD box is for anything that reads and writes constantly, like a database or an application server. Buying the wrong one for the job is the single most common VPS mistake we see, and it's covered in more depth in the ",[49,601,602],{"href":324},"NVMe vs. SSD vs. HDD guide",".",[15,605,607],{"id":606},"bandwidth-and-data-transfer","Bandwidth and data transfer",[11,609,610],{},"Every plan comes with a monthly transfer allowance, and it's worth checking before you commit, especially if you're running anything that serves media, video, or a lot of API traffic. Overage policies vary by provider: some throttle your connection once you hit the cap, others bill per GB past it. Neither is disqualifying, but you want to know which one you signed up for before your bill shows up bigger than expected, not after.",[15,612,614],{"id":613},"where-the-server-actually-lives","Where the server actually lives",[11,616,617],{},"Region availability matters more than people expect. A server two continents away from your users adds real, noticeable latency, and some workloads (payment processing, data residency requirements for EU customers) have legal reasons to care where the disk physically sits. Most providers list their available regions right on the plan, and vpsudo's offer table shows region flags per listing so you can filter to what's close to your actual users before you compare anything else.",[15,619,621],{"id":620},"the-renewal-price-is-the-real-price","The renewal price is the real price",[11,623,624],{},"This one isn't specific to VPS hosting, but it applies with force here: a lot of providers lead with an introductory price that jumps significantly at renewal. It's a well-worn pattern across the hosting industry generally, not something particular to any one vendor. The fix is simple: before you commit to a plan because of an attractive first-month or first-year number, check what the same plan costs on its second billing cycle. If a provider doesn't make that easy to find, that's itself useful information.",[15,626,628],{"id":627},"what-offer-type-tells-you-at-a-glance","What \"offer type\" tells you at a glance",[11,630,631,632,635,636,639],{},"vpsudo tags every listing as VPS, bare metal, or GPU, and that label answers a lot of your questions before you read a single spec. VPS means you're sharing a physical machine with other virtualized tenants. Bare metal means the whole physical box is yours, no hypervisor involved. GPU listings are compute rented by the hour rather than the month, built for AI training, inference, or rendering rather than general hosting. If you're not sure which category fits what you're building, the ",[49,633,634],{"href":99},"bare metal vs. VPS guide"," and the ",[49,637,638],{"href":559},"GPU rental pricing guide"," go into when each one makes sense.",[15,641,643],{"id":642},"put-it-together","Put it together",[11,645,646],{},"None of these specs matter in isolation, they matter next to each other. A $2.50 plan with 0.5GB of RAM and a $5 plan with 1GB might both be \"cheap,\" but they're not comparable unless you're actually going to use that extra RAM. The fastest way to see this side by side is to pull up vpsudo's offer table, sort by the spec you actually care about (price, vCPU, RAM, or storage), and filter down to the provider type and region you need. That's a faster way to shop than reading marketing copy on ten different provider sites, because every number on the table comes straight from each provider's own pricing data.",{"title":84,"searchDepth":85,"depth":85,"links":648},[649,650,651,652,653,654,655],{"id":576,"depth":85,"text":577},{"id":592,"depth":85,"text":593},{"id":606,"depth":85,"text":607},{"id":613,"depth":85,"text":614},{"id":620,"depth":85,"text":621},{"id":627,"depth":85,"text":628},{"id":642,"depth":85,"text":643},"A checklist for comparing VPS plans past the sticker price: vCPU type, storage, bandwidth, and the traps that make one cheap plan a much worse deal than another.",{},"\u002Fguides\u002Fwhat-to-look-for-when-buying-a-vps",{"title":565,"description":656},"guides\u002Fwhat-to-look-for-when-buying-a-vps","eCrhkobWgwIqq74DR9gSWBWc_Qs6bTl6cCOSsXNP7ek",1785111492831]