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.
vCores across providers
Hi,
I am interested if there is some basis of similarity between vCores, vCPUs, v(fill in the blank) between providers: particularly the bigger providers like linode, DO and aws.
This is what I found:
Aws ec2 can get up o 128 cores
Linode max 20 cores
DO at 32 cores,
Vulr 24 cores
Ramnode 4 cores
I know there is more to a VPS than cores and cpu speed, but can we roughly compare a linode core to an AWS core to a DP core, etc.
Comments
There are three specific questions you need to answer:
1) What is the underlying CPU
2) How much "access" do you have to that CPU (How much of Ghz can you use)
3) Are the CPU cores oversold or do you get a dedicated core/thread?
I would guess the performance will vary greatly across all providers especially when you consider clouds like Azure use AMD's and Intel CPU's.
Hence most large game producers/hosters/etc. use physical hardware, CPU performance in the cloud is all over the place to say the least.
Don't know exactly what you mean by "comparing" without the speed/amount but I'm sure AWS cores are slower than any of the others you listed.
dediserve > Max 40 vCPU
It all comes down to what the CPU is, and how many vCPU's they're running on top of it.
Eg, a provider might have 8 Logical CPU Cores, and sell 32vCPU's per server. That's an effective 400% oversell on processor usage.
Or, another way to work it out. An E3-1246 v3 has a passmark of 10,000 and 4 logical cores. A host that over provisions at 400% will split that 10,000 by 16 which is 625. Equivalent to a $40 Fanless Atom CPU.
Untill you know how the providers are splitting their resources, you can't gauge an equal comparison.
That said, services like serverbear (RIP) allowed you to compare VPS packages by benchmark score from most providers on the planet.
passmark is next to useless.. no point going off that.
@pbgen that's too much of a simplification, as the Client will likely be able to burst the CPU when others don't fully utilise their allocations so you can't say it will be limited to a fabless atom. More like fabless atom guaranteed with option to burst ;-).
Similar to the way cars are mesured by HP. Weight, Shape and Size also have an effect on top speed. But there are many vectors you could mesure.
Passmark generates a score based on a set of tests, it may not reflect the performance of the task you wish to see but all that matters is that all the other CPU tests are identical. It provides a guideline for overall performance.
This is true, but for "Production" cases, you need to be able to forcast performance to accurately accommodate your growth. Thus, the need for solid stats on what you're getting, a burst is not a guarantee and can not be relied upon.
This is a hole that will only get deeper, what the OP is looking for is totally subjective and not an easy thing to awnswer. Even if there was an awnswer, it would be specific to OPs requirements and not applicable to all other scenarios, like mentioned above with passmark.
not only are they unreliable but they do not share the hardware and such (such as board etc)
and they are not open about the methods. geekbench provides a more reliable test (you can search the motherboards/cpu etc)