Live data from Hacker News

Cloud VM benchmarks 2026

devblog.ecuadors.net

51–60 of 165 posts

Re: Cloud VM benchmarks 2026

#51
post #39
post #20

I just ran some massive tests on our own CI. I use AMD Turin for this on gcp, which was noted as one of the fastest ones in the article. The most insane part here is that the AMD EPYC 4565p can beat the turin's used on the cloud providers, by as much as 2x in the single core. Our tests took 2 minutes on GCP, 1 minute flat on the 4565p with its boost to 5.1ghz holding steady vs only 4.1ghz on the gcp ones. GCP charges…

I used to run a site that compares prices[0]. Not only is the ecosystem pull to the cloud strong, but many developers today look at bare metal as downright daunting. Not sure where that fear comes from. Cloud challenges can be as or more complex than bare metal ones. [0]: https://baremetalsavings.com/

The fragmentation and friction! Comparing prices usually requires 10 open browser tabs and a spreadsheet, which is what keeps people locked into their default cloud. I built a tool to solve this called BlueDot (ie, Earth, where all the clouds are)[0]. It’s a TUI that aggregates 58,000+ server configurations across 6 clouds (including Hetzner). It lets you view side-by-side price comparisons and deploy instantly from the terminal. It makes grabbing a cheap Hetzner box just as easy as spinning up something on AWS/GCP.

[0]: https://tui.bluedot.ink

Re: Cloud VM benchmarks 2026

#52
post #39
post #20

I just ran some massive tests on our own CI. I use AMD Turin for this on gcp, which was noted as one of the fastest ones in the article. The most insane part here is that the AMD EPYC 4565p can beat the turin's used on the cloud providers, by as much as 2x in the single core. Our tests took 2 minutes on GCP, 1 minute flat on the 4565p with its boost to 5.1ghz holding steady vs only 4.1ghz on the gcp ones. GCP charges…

I used to run a site that compares prices[0]. Not only is the ecosystem pull to the cloud strong, but many developers today look at bare metal as downright daunting. Not sure where that fear comes from. Cloud challenges can be as or more complex than bare metal ones. [0]: https://baremetalsavings.com/

> Cloud challenges can be as or more complex than bare metal ones.

Big +1 to this. For what I thought was a modest sized project it feels like an np-hard problem coordinating with gcloud account reps to figure out what regions have both enough hyperdisk capacity and compute capacity. A far cry from being able to just "download more ram" with ease.

The cloud ain't magic folks, it's just someone else's servers.

(All that said... still way easier than if I needed to procure our own hardware and colocate it. The project is complete. Just delayed more than I expected.)

Re: Cloud VM benchmarks 2026

#53
Disclosure: I work on VMs at Google Compute Engine :)

This was a really, really good write-up. I appreciated the breadth of VMs tested and the spread of benchmarks. A few random observations:

1. Turin is a beast.

2. The data on price-performance makes Hetzner look really fantastic, especially for small scale projects where region placement doesn’t matter much and big bursty scaling isn’t required.

3. I think the first ever cloud VM I ever provisioned was on DigitalOcean. I was surprised at how old their fleet was, but I guess they have some limited Emerald Rapids offerings now: https://www.digitalocean.com/blog/introducing-5th-gen-xeon-p...

Re: Cloud VM benchmarks 2026

#54
post #20

I just ran some massive tests on our own CI. I use AMD Turin for this on gcp, which was noted as one of the fastest ones in the article. The most insane part here is that the AMD EPYC 4565p can beat the turin's used on the cloud providers, by as much as 2x in the single core. Our tests took 2 minutes on GCP, 1 minute flat on the 4565p with its boost to 5.1ghz holding steady vs only 4.1ghz on the gcp ones. GCP charges…

Self-racking lets you rack a bunch of gear you'd never find in VM/dedicated rentals, like consumer parts or older, still very good parts. Overclocking options are available as well if you DIY. If you need single-threaded performance, colo is really the only way to go anyway. We have two full racks and we're super happy with them.

Or under clocking and under volting for even better performance to price/power/longevity ratios

Re: Cloud VM benchmarks 2026

#55
Needs: network performance and costs, desperately.

I wouldn't mind SAN/non-local storage performance and costs too.

VM cost isn't where AWS gets you. It's ALLLLLLLL the other nickel and diming that they kill you on, especially since outbound data transfer costs a log of money to get off of the platform.

Re: Cloud VM benchmarks 2026

#56

Earlier quoted context omitted.

Self-racking lets you rack a bunch of gear you'd never find in VM/dedicated rentals, like consumer parts or older, still very good parts. Overclocking options are available as well if you DIY. If you need single-threaded performance, colo is really the only way to go anyway. We have two full racks and we're super happy with them.

Or under clocking and under volting for even better performance to price/power/longevity ratios

No doubt. Especially for GPU inference at scale. We overclock/overvolt for training and tune way down for inference.

Re: Cloud VM benchmarks 2026

#57

Anyone have experience with Oracle Cloud and ease of moving away? This benchmark seems to recommend Oracle Cloud, but I’ve heard that Oracle has historically used aggressive licenses and legal terms to keep customers locked-in.

I've been using it professionally for years.

As long as you don't use their exclusive DBaaS, moving away is easier than from other places, as egress traffic is free.

The user experience though, stuff of nightmares...

Re: Cloud VM benchmarks 2026

#58

Disclosure: I work on VMs at Google Compute Engine :) This was a really, really good write-up. I appreciated the breadth of VMs tested and the spread of benchmarks. A few random observations: 1. Turin is a beast. 2. The data on price-performance makes Hetzner look really fantastic, especially for small scale projects where region placement doesn’t matter much and big bursty scaling isn’t required. 3. I think the firs…

Hetzner really shines with their dedicated stuff.

Re: Cloud VM benchmarks 2026

#59
Awesome write up. I use EC2 t4g instances extensively as I tend to scale workloads horizontal (using Kubernetes, Docker) and admittedly the workloads aren’t cpu constrained. Would be interesting to see how t4g’s compare to other low end comparable VMs on GCP, DigitalOcean, Hetzner. Though I get the difficulty testing as it requires maxing out CPU credits (waiting) for accurate results before starting testing for each VM.

Looking forward to t5g’s whenever (if ever) they release.

Re: Cloud VM benchmarks 2026

#60
There are definitely ways to get some really good hardware from lowend shops as well.

Hetzner is pretty decent too actually but I actually think that OVH might be even cheaper while still being competitive, but in the case of OVH, I think one minor issue people can have is the setup fee at times but time to time like during black fridays or special deals, there are ways to get 0 setup fees.

If someone wants something stable, I think that OVH is pretty great as well and is comparable to Hetzner in terms of pure price but I have heard that given their scale, their support can be 50/50 but I recommend joining their discord and maybe even using (twitter oof) to message them as thsi was something which worked for someone on hackernews the last time something like this was discussed.

If you are okay with some more steal factor, use netcup.

You should also probably look at gaming type servers too. I know a person who is a one man shop who was more passionate about these stuff and did have high-end hardware (200k$) in investment in his provider.

Going more into the specifics of finance from provider side, usually the idea is that they recoup the costs in 5 years if running at decent capacity/having sold quite a bit but the first few years so as in example of the 200k$, they currently make IIRC 40k-60k$ per year, you have to somehow find your way to customers but that is being messed up because of AI ram prices as their costs to fix any broken hardware has absolutely skyrocketed eating much of their profits.

It's an extremely competitive market at times so if you are looking for something specific. You can actually ask about it in forums like lowendtalk, lowendspirit and these people/providers can respond matching what you are looking for price/performance in for and they can also provide test servers if need be and there is a unified form of benchmarking within this space called yabs (yet-another-benchmark-script) which you can ask for a provider.

But in all regards, if someone doesn't want to go through this hassle, I will just say that the same provider that I mentioned earlier had also in public mentioned how hetzner prices are pretty competitive and in all honesty I agree with that too.

If someone doesn't want to go through some/any part of this hassle, then hetzner/OVH are my go-to safe options. They are big enough that their downtime shouldn't be blamed on you for picking them while being really good in themselves. (Something which I have actually heard quite often mentioned on hackernews)

And within this space, if you don't absolutely know what you want, there is no generic winner as there are niches that are occupied. So if you ever want to find any alternative to hetzner for even more cheaper, its best that you first decide what are the things exactly that you want and then ask rather than see for pre-existing if possible.

Also if possible, make deals during blackfriday/cyber-monday. Not sure about Hetzner but in OVH/other providers, you can get recurring deals and the sever setup costs etc. removed sometimes.

Post reply on HN