amuck-landowner

Dear Provider, I'd be more likely to buy your VPS...

DomainBop

Dormant VPSB Pathogen
...if you listed the Unixbench in the offer description because on those occasions when I need something with a little power, the "2 cores" you have listed in your description doesn't tell me sh!t. "2 cores" could mean I'm getting something with nearly the power of an older i3, or it could mean I'm getting something with less power than an Atom N2800. Vultr, VPSDime, and a few others list the unixbench scores (somewhere) on their sites but they're the exception.


 2 core VPS unixbench comparison
------------------------------------------------------------
 3946 Vultr, 4GB KVM (2GB plan w/double RAM coupon), Amsterdam
 3316 Hosthink, 1GB KVM, Istanbul
 1230 Edis, 2GB OpenVZ/250GB storage, Reykjavik
 1189 Linode, 2GB Xen, Tokyo
  844 UltaVPS (ProviderService), 2GB Xen, Dusseldorf
  702 XenPower (Prometeus), 2GB Xen/400GB storage, Milan

for reference:
--------
4158 Kimsufi i3-2130
 856 Kimsufi Atom N2800

note: all benchmarks are from today with the exception of the Vultr one which is from March 8
 

drmike

100% Tier-1 Gogent
So proposing unixbench be ran in a container for each plan, right?

Unsure what is potentially wrong / to be disputed if such is ran in container...  Somene more versed ideally will comment.

Face value here, Unixbench tests would be done on empty gear... so when you get such, going to be good bit less unless node is idle....
 

KuJoe

Well-Known Member
Verified Provider
As a customer, I would much rather see Unixbence tests from a client versus in an advertisement from a provider.

Keep in mind that different nodes will have different benchmarks so if I ran Unixbench from an empty node in one location and posted that in an advertisement (like @drmike said) and you order a VPS on a half-filled node in another location (or maybe even the same location) with different hardware specs and ran Unixbench there, that would make for a fun support ticket.
 
Last edited by a moderator:

tonyg

New Member
...and 2 cores or what ever number a provider will throw out there are meaningless if they throttle the cpu(s).

Two of the "in vogue" providers out there are throttling the cpus to 50%.
 

DomainBop

Dormant VPSB Pathogen
Face value here, Unixbench tests would be done on empty gear... so when you get such, going to be good bit less unless node is idle....
Ideally they'd be tested from active nodes but I think we all know there are some providers who cheat on the benchmarks they list in their ads (like the openvz providers who list dd scores in their offer that they tested on an empty node and then 2-3 months later after they've overloaded the node those 250MB/s dd scores have dropped to 25MB/s).

The Vultr benchmark I posted is from shortly after they launched in Amsterdam 4 months ago.  The most recent benchmarks I've seen have been about 20% lower in the 3300 range so there has been a performance hit as they've filled up their nodes.  The biggest hit would probably be on overloaded openvz nodes.  The other benchmarks I posted are from nodes that I've had for awhile and so they probably give a more accurate view of what can be expected.

...and 2 cores or what ever number a provider will throw out there are meaningless if they throttle the cpu(s).
Yeah, listing 2 cores is about as informative as some of the dedicated providers on WHT who list "Quad core Xeon" but don't tell you which quad core Xeon you're getting.
 

SkylarM

Well-Known Member
Verified Provider
Yeah, listing 2 cores is about as informative as some of the dedicated providers on WHT who list "Quad core Xeon" but don't tell you which quad core Xeon you're getting.
But it's a quad core, bro!


To me it seems to make more sense to mention what CPU is used, and how many cores you will get. We do our ads that way. That way there's no worry of us over-inflating tests on an empty node and then sticking you on a full one and you being surprised by different results. I'm all for benches, but they are very limited in scope and practicality compared to the relevant information (what type of CPU, what you consider as CPU abuse, so forth) being put up front.
 

willie

Active Member
Maybe it's just me but I've always felt hesitant to use substantial amounts of cpu on any budget vps (budget = the ones we talk about here for the most part).  So when I see someone looking for high cpu resources in a cheap vps, I think "ok, there's an overloaded node, better stay away".  VPS's are best for relatively low resource network services, occasional bursts of cpu and i/o, and for storage.  For heavy cpu it's better to use a single tenant server or a dedi.
 

KuJoe

Well-Known Member
Verified Provider
To me it seems to make more sense to mention what CPU is used, and how many cores you will get.
Yeah, listing 2 cores is about as informative as some of the dedicated providers on WHT who list "Quad core Xeon" but don't tell you which quad core Xeon you're getting.
But what if there's no guarantee what node you're put on and not all of the nodes have the same specs?
 

Aldryic C'boas

The Pony
Not to mention that certain 'providers' will misuse various methods of caching in order to spoof 'good looking results' that you could never actually achieve in production.
 

wlanboy

Content Contributer
Performance is not everything - for me durability is much more important.

I need a specific amount of performance but for more than a few months.

My personal key performance indicator is the page loading time over time.
 
Top
amuck-landowner