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.
Discrepancies in performance between the exact same Netcup Root Servers
I got x2 RS 2000 G11 servers in NUE. both are the same specs and same price so they are technically identical
HOWEVER
GB6 Scores for the first one
https://browser.geekbench.com/v6/cpu/9132260
the other one
https://browser.geekbench.com/v6/cpu/9387035
that's about 10% difference in single core and 15% difference in multicore
is this normal ? or should i contact support ?!
My Main concern is that they're Dedicated cores so i didn't expect to see variance between different nodes
Comments
If you feel there's a problem, yes, contact support, they're the only ones that can help.
It's perfectly normal. Single thread performance will get worse when other threads are in use. That's also a very small difference. Expect 15 % difference on the same server at different times. Maybe more.
If you can't take CPU resources impacted by guests your only option is a dedicated server.
yea i thought neighbors won't affect me at all because of "dedicated cores"
install alma 9
Would the silicon lottery affect vcpus on a kvm? Say the 2 vpses are on different physical servers
Not by that much
Please have mercy on the support team, they hear enough every day. Unless your application is controlling a nuclear reactor, there’s probably no need to open a ticket.
Are they on the same node?
Plus I feel the cpu frequency may vary based on the internal thermals and the likes?
Dedicated cores just would mean you can hammer them 24x7. But that would not guarantee consistent performance.
Oh yeah a few months ago some bright minds were accusing avoro and netcup of false advertising because there were trying to mine on vds.
nope i don't think so as i bought the other one around 2 weeks later
Buying at different times may not always mean different nodes, just saying.
i believe their interpretation of "dedicated" is actually "unlimited".
only wishosting advertises this accurately.
No this is not accurate. First of all this is not about advertisement but product specs.
Now, even if your v process gets a pinned core, it's out of any provider control within today's tech to thermally isolate your core or give you an exclusive memory controller or L3-cache etc
A dedicated core can and will have differences in performance depending on what your neighbors do, even if you get an exclusive instruction pipeline. These differences will affect synthetic benchmarks like geekbench, as these not only measure computational speed but they are also sensitive to shared caches and memory speed, i.e. compression tests.
In some cases (see when some idiot is mining), these differences can be huge, but providers try to minimize this by rebalancing nodes.
10% is far within reasonable expectations of a "dedicated core"
In addition, the EPYC 9634 has 2.25GHz base clock, 3.7Ghz turbo frequency. When all cores are utilized with variety of intructions it is roughly 3.1GHz, with only AVX2 its a little bit lower.
Even with pinned, 100% dedicated cores one node csn have up to 20% faster cores just because of turbo behavior.
And thats just a one variable that is mixed up with thermal headroom, cache performance, memory performance like you mentioned.
With miners in your node you can expect ~1100 singlecore instead of ~1900 on Genoa, because they will peg cache, even if node has just like 40% CPU usage. Pinned cores wont solve the source of the problem.