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.
Declining benchmarks for providers
I’ve noticed with several providers that the CPU bench drops significantly after a month or two. Not sure if that’s normal across the board.
- Lowhosting - GB5 multi core scores dropped for me by over 2000 pts. From 6900 to 4400.
- The terrahost BF deal I have has dropped to 4500 from 5600. Single core however remains the same.
I guess the nodes get busier over time.
Thanked by 1toumi111
Comments
its shared environment, what do you expect?
I expect it to vary with load but the numbers have been consistent over the past few weeks.
Some hosts will keep adding more users as the node fills up. Then hits maximum capacity of users. Then it's really dependent on who your neighbors are.
So when you originally get the VM and run the benchmark you might be running it on a configuration that isn't full.
It's imho kinda useless to share benchmarks and look at them as a point-of-reference for resource allocation. I'd trust them more on different environments like AWS/Azure/GCP where those resources are guaranteed, but not on a VPS/VM from these hosts here.
Fair enough. I was actually only surprised at terrahost for the drop since they aren’t really your typical low budget provider.
Well benchmark score is just that, benchmark score. As long as the drop on benchmark doesn't affect my website and or other application performance so badly on that said vps I don't give a damn about it. Just my opinion..
Definitely normal and in the long description of the economics of it, mostly desirable. The secret recipe is the balance.
i have never seen any LE* provider in my usage who has purposely kept the allocation consistently "optimum" that there is always access to full cpu power, especially in multicore vpses.
have only seen only idle or busy nodes.
For a start, a single shot benchmark is just a very crude orientation point and proper benchmarking goes over days and strives for min. 50 result sets and preferrably >= 100 (which may not be achievable on a very low-end system).
Plus, of course many (most?) providers tend to maximize ROI (and many even simply profit) by filling nodes to the brim as @jar hinted correctly.
The bit I question is "after a month or two". My experience rather suggests "after couple of days or two weeks at most"
azure, aws, gcp are always low
for example in GCP (google cloud) epyc rome 2 core
Single Core | 825
Multi Core | 981
Full Test | https://browser.geekbench.com/v5/cpu/12545485
Can be increased, but i guess at an additional cost.
That looks like 2 threads rather than 2 cores
Yes
but
just thinking, those are reasonable balance....
and I don’t trust monster yabs … unless on dedicated core cpu
25%