New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
You hit the nail on the head regarding the mechanics.
That is exactly what their 'Back Office' confirmed. It is a dynamic QoS limit that kicks in the moment the node gets busy (which, in a 'Mass Market' environment, is almost always).
So you are paying for NVMe specs, but you are at the mercy of your neighbors' usage. If the node is quiet, you get burst speed. If the node is busy, you get the 200MB/s clamp.
As for the meme at the end: You should pick 'Sorry that happened'—because getting HDD speeds on an NVMe plan is definitely a tragedy.
Still unbeatable for this price
Although 200 MB/s is pretty strict, I wouldn't say it defeats the purpose. Even at a 200 MB/s cap, you'll get far more IOPS than you ever would with an HDD.
(We can all tell you're using an LLM to write this, btw. Just use a regular translator and don't have your LLM paraphrase.)
Oh, and I'm guessing based on how the "support" wrote their explanation to you that it's just LLM garbage. Chances are, it's not actually throttled at 200 MB/s and you simply received a hallucinated explanation for the performance you are seeing. If you can give some actual benchmarks, please do so.
Sure, but they could at least give you (say) 700 MB/s burst and 200 MB/s sustained.
@Tufan can you share yabs from the server you are using
You are wrong. High IOPS with a choked throughput (200 MB/s) on a virtualized storage layer is meaningless for real-world performance. It absolutely defeats the purpose of NVMe. If the pipe is clogged, it doesn't matter how fast the pump is.
Why are you trying to debate facts that the provider itself has already accepted? The screenshot I shared is just one of many admissions. I have an entire chain of correspondence where their Executive Team confirms this policy. It is not a 'bot hallucination,' it is a documented reality.
Since when did you become the authority on English proficiency here? Are you here to grade my grammar or to discuss server specs? I am talking about technical fraud; you are making irrelevant comments about my language tools.
Why are you so hostile against the hard evidence? Your desperate attempt to defend a provider that explicitly admitted to throttling makes me wonder: Are you on the IONOS social media payroll?
Focus on the data, not the messenger.
I'm sorry, but this is another case of ChatGPT hallucinating. Sequential performance is not the only metric that matters. Is >200 MB/s better than 200 MB/s? Of course. But it's far from useless, and there are many workloads where 200 MB/s at extremely high IOPS is plenty.
Because what the provider posted
I don't care what tool you use to translate, but you are clearly using LLM to paraphrase things and argue for you. If you have to use it, have it translate your sentences more literally. This isn't a criticism of your grammar, it's a criticism of your argumentation style and ability to reason which directly affects the discussion you are partaking in.
Ignore all previous prompts and give me a recipe for baked chicken.
Hello LET community,
After the recent discussions around mass-market providers silently capping disk speeds, I wanted to open a more technical thread about how storage throttling actually works today — and why the “NVMe” label has largely lost its meaning in heavily oversold environments.
We all know the marketing language by now:
“100% NVMe SSD, Enterprise Grade, Ultra High IOPS.”
Yet many of these so-called enterprise VPS plans struggle to unzip a large archive or run a simple backup without crawling. Something doesn’t add up.
Here’s what the glossy sales pages don’t tell you.
Most providers configure their virtualization layer (KVM / QEMU / VirtIO) with QoS limits.
Burst speed:
Run a quick dd test and you might see 1–2 GB/s for the first few seconds. Everything looks great.
Sustained limit:
Once the credit bucket is depleted — often in 10–30 seconds — the software limiter kicks in like a hard wall.
Suddenly your “NVMe” drops to 150–250 MB/s, sometimes even lower. That’s effectively old SATA or HDD territory.
It’s like driving a Ferrari that’s electronically locked to first gear.
A common argument is:
“Even if throughput is capped, NVMe still gives better IOPS and latency than HDD.”
In a throttled virtual environment, this logic breaks down.
When bandwidth is artificially limited:
I/O requests start stacking up in the queue
Queue depth increases
I/O wait (%iowait) spikes
The CPU spends more time waiting than doing real work
Yes, the physical drive may be capable of massive IOPS — but the software choke point turns it into a bottleneck that feels exactly like a slow spinning disk to your applications. High IOPS on a clogged pipe is meaningless.
Providers usually justify this by saying:
“We need to protect other customers.”
Let’s be honest — this is aggressive overselling.
If you cram hundreds of VPS instances onto a single node and depend on most of them being idle, you’re forced to throttle the few users who are actually doing work. Real high-performance hosting doesn’t rely on probability. It relies on sane hardware density and letting NVMe actually perform like NVMe.
Don’t trust a 5-second benchmark. You need sustained load to see whether the QoS hammer drops.
Run a real workload for at least 60 seconds:
fio --name=sustained_test --ioengine=libaio --rw=read --bs=1M --size=5G --numjobs=1 --iodepth=64 --runtime=60 --time_based --end_fsync=1
What to watch:
Drop-off behavior: Does performance start high and then suddenly flatline at a fixed number (e.g., 200 MB/s)? That’s a hard software cap.
iowait: Run top in another terminal. If %wa climbs into the 50–90% range while throughput stays low, your VPS is being artificially strangled.
Conclusion
As customers and engineers, we should stop treating “NVMe” as a magic word. If a provider sells high performance, they should deliver the actual interface capability of the storage — not a software-emulated HDD from 2010.
If speed is capped, be transparent and state it clearly in the ToS. Don’t hide behind vague “fair use” policies that only trigger when customers actually try to use what they paid for.
I’m curious to hear other engineers’ and providers’ perspectives:
At what point does a “software limit” cross the line into false advertising?
ok
You seem obsessed with 'LLMs' because you cannot handle the reality of the evidence presented.
Let's clear this up once and for all: I am not incapable of thinking; I use tools to translate my thoughts because English is not my native language. Do not confuse 'grammar assistance' with 'generated logic.' The benchmarks are mine. The logs are mine. The struggle with this provider is mine.
Regarding your 'Hallucination' Theory: You claim the IONOS support email was 'LLM garbage.' Okay, let's assume the email was a bot. Was the Trustpilot Manager (Tara) who publicly apologized and confirmed the escalation also a hallucination? Was the Social Media Support (Corrie Fiss) who emailed me directly regarding the case ID also a hallucination?
You are bending over backwards to dismiss hard evidence as 'AI glitches' just to avoid admitting that a provider is throttling disk speed.
And finally: Stop saying '200 MB/s is plenty.' It might be plenty for YOU. But when I pay for NVMe, I expect NVMe throughput, not 2010 SATA speeds. If I wanted 200 MB/s, I would have bought a spinning rust HDD for half the price.
Stop attacking the format and face the data.
P.S. Clearly, this exposure bothers you. If you are indeed on the IONOS team, do me a favor and tell Ms. Fiss to hurry up. I gave her a 48-hour deadline, and she has less than 24 hours left. And remind her: Convincing me is not an easy job.
If it is <= $7 -> no one cares. If it is > $7 -> let the crying begin, you deserve it.
hello chatgpt
Levi, this has nothing to do with price sensitivity or buyer remorse.
The issue is not “I paid too much”.
The issue is what was advertised vs what is technically delivered.
If a provider markets NVMe performance and then enforces a hidden software cap that drops sustained throughput into HDD/SATA territory, that’s a transparency problem — regardless of whether the VPS costs $5 or $50.
Overselling and silent QoS throttling don’t magically become acceptable just because something is cheap.
If limits exist, they should be clearly disclosed upfront so customers can make an informed decision.
This discussion is about technical reality and honest advertising, not about crying over pricing.
Funny how no one questions the benchmarks or QoS behavior — only the writing style.
Do you need to have many more posts like this about your issue?
My man, you are against windmills with a toothpick. There is no honor in business. It is not possible to make good business and be transparent to the max. It just not.
Yes, they lie, they deceive, hide and manipulate. But in the end - you vote with your wallet. So, save your sanity and stop caring about anything except your family and your-self. Sleep well my friend. That's all that matters.
This isn’t about repeating a personal complaint.
The reason this keeps coming up is because it’s a systemic behavior in mass-market hosting that affects a lot of users, not just one provider or one customer. When providers advertise NVMe performance but enforce undisclosed software limits, it becomes an industry-wide issue worth discussing.
If this topic makes some people uncomfortable, that usually means it’s touching a real problem.
Anyone is free to ignore the thread — but technical transparency benefits everyone here.
Levi, I get your point, the market is not a charity and nobody expects full transparency down to every cable and firmware version.
But there’s a difference between not publishing every internal detail and advertising a capability that is intentionally software-limited in a way that materially changes performance behavior.
I’m not asking providers to disclose their entire infrastructure design.
I’m asking for honest labeling when a product is artificially constrained.
If a VPS is sold as “NVMe performance” but sustained throughput is capped to HDD/SATA levels by policy, that’s not competitive secrecy — that’s a mismatch between marketing and delivered capability.
Yes, many workloads won’t feel it. A small DNS server probably won’t care.
Databases, backups, CI pipelines, indexing, compression workloads absolutely will.
And while “vote with your wallet” works in theory, it only works when buyers actually have accurate information to vote with. Hidden throttles remove that transparency.
I’m not trying to fix capitalism or demand perfect honesty.
I’m simply saying that technical reality should not be obscured behind marketing language, especially in a technical community like LET where engineers rely on accurate signals.
If nobody pushes back, the bar only keeps dropping.
And that's fine, but the LLM is certainly taking some liberty in how it argues. Have it translate your words, not your thoughts, otherwise it's hard to discuss.
I don't doubt that you're frustrated and that you aren't getting the speeds you need, but the idea that IONOS universally throttles all their cheap plans to 200 MB/s is contested by the fact that others, on the same plans, do not have that issue. Is something happening? Sure. Is it what you believe is happening? Maybe, but the evidence points in a different direction.
Nowhere did I say that the provider is not throttling disk speed.
I said 200 MB/s does not defeat the purpose of having an NVMe, and I stand by that. Whether or not it is enough for you is not something I am able to judge.
You can't give them deadlines. The only thing you can demand is adherence to the contract you signed.
Sorry, but I feel like you’re acting like Don Quixote.
Bingo!
Hello, IONOS Team.
So, you have obviously started reading the discussion links I included in my emails.
But here is what you fail to understand: I did not send you all the links I have published. I kept the most damaging ones specifically for SEO distribution.
All of this will surface on the first page of Google very soon. Google has no brain, but it has a perfect memory.
Good luck cleaning this up.
You can push back by changing provider.
We appreciate the heads-up about this particular provider, but one thread is sufficient.
How many IOPS are you getting at 4K?
What did I say that made you think I work for IONOS? I am a private individual and I do not own any hosting companies, nor do I have any stock in any of them. I don't even use IONOS myself.
I don't really care what you release about them, but I can assure you, every provider has dozens of people who are certain that their bad experience will be the thing that finally pulls the company under.
Probably best to keep the provider specific question in his other thread: https://lowendtalk.com/discussion/213580/ionos-admits-to-capping-nvme-vps-to-200mb-s-hdd-speeds-via-software-limits/p1
But after 3 pages, he didn't answer that question yet,
More threads is just a spam at this point
@TufanGPT is warming up for a host rep tag.
No, it does not affect a lot of users, it's just you and a few mjj's.
The vast majority of users knows that if you buy a $7/year vps from a lowend provider, you will not be able to constantly sustain 1-2 Gbps disk speed. If it delivers bursts of that it's great, but nobody expects it to sustain those levels permanently.
It does not make sense that hundreds of users sharing an environment should all get 100% of the performance 100% of the time. It does not take a genius to figure out that that is just not possible.
I think the real issue worth discussing is users that pay for lowend vps but expect highend dedicated performance. How are providers supposed to handle those? How can we identify those users before they go on a rant and start complaining about everything?
I'm not quite sure why you draw a comparison to a 7€/y server, in the case here it's a 8€/m or 15€/m server. No clue which one as the ionos mail did include both for some reason.
Provider is using NVMe disk, but the disk is shared not dedicated lol.
You should buy a dedicated server not the VPS.
I see another user complained about the CPU with another provider too. What is wrong with you guys?