VM host: dual CPU vs single CPU - same CPU performance rating
-
@coliver said in VM host: dual CPU vs single CPU - same CPU performance rating:
@dashrender said in VM host: dual CPU vs single CPU - same CPU performance rating:
@pmoncho said in VM host: dual CPU vs single CPU - same CPU performance rating:
@dustinb3403 said in VM host: dual CPU vs single CPU - same CPU performance rating:
@pmoncho said in VM host: dual CPU vs single CPU - same CPU performance rating:
@jaredbusch said in VM host: dual CPU vs single CPU - same CPU performance rating:
@scottalanmiller said in VM host: dual CPU vs single CPU - same CPU performance rating:
@pete-s said in VM host: dual CPU vs single CPU - same CPU performance rating:
But for instance in my case 2 x E5-2630V2 (2x6 cores @ 2.5GHz) shows up as having about the same benchmark performance (single thread and multi thread) as 1 x E5-2670V2 (10 cores @ 2.5Ghz).
Comparing typical TDP: 2x6 cores is 2x80W=160W while 1x10 core is 115W.
This is partly why I push companies so hard to move to single proc servers. Often makes far more sense. There is way less overhead.
But in this case tool six scores versus Single Tencor you might as well get the dual processor advantages since you’re going to have to pay for the licensing anyway if he runs any Microsoft products at all
It seems to me, in these situations, software dictates hardware configuration negatively. That is frustrating if one wants to continue to use MS products. Choices, choices, choices!!
I would say the software isn't dictating the hardware, but the licensing terms that comes with the software is.
You are correct. I re-typed that a couple different times and forgot the word "licensing" in the final reiteration.
Sure that's true - but only on the minimum side. i.e. Windows licensing defaults to dual 8 core setups. IF you need more than that - you have to buy more.
Buying a single 10 core machine that will run Windows means if you ever need a second processor in there - you'll be required to buy 2 more two core license packs. Something to consider.
How often are people adding CPUs to servers? Is that something that generally happens? Of course @JaredBusch was faster then I was.
Never. RAM and disk, sure. Not CPU. Maybe if you buy a single proc server for a specific reason, the reassign it to be a busy SQL server or virtual host.
-
Yeah I have to agree with the others, you buy a box and generally just fill everything because you're buying for what you need, not what you might need.
Purchasing a board that supports dual-cpu's and not using the other is just wasting money on an empty socket.
-
@dustinb3403 said in VM host: dual CPU vs single CPU - same CPU performance rating:
Yeah I have to agree with the others, you buy a box and generally just fill everything because you're buying for what you need, not what you might need.
Purchasing a board that supports dual-cpu's and not using the other is just wasting money on an empty socket.
Except most vendors don't charge for the second socket. So that is very rarely the case.
-
@pete-s said in VM host: dual CPU vs single CPU - same CPU performance rating:
Thanks guys.
I decided to go with the single 10-core CPU in this case as it would leave open the possibility of adding another CPU and set of memory if needed.
No licensing to consider in this particular case.
Keep in mind that the SL code of the existing CPU should be recorded as when it comes time to add the second CPU the same SL code CPU would need to be acquired.
There's a reason why Tier 1 charges huge coin to add a second CPU at a later date. They need to keep them on the shelf.
-
@coliver said in VM host: dual CPU vs single CPU - same CPU performance rating:
How often are people adding CPUs to servers? Is that something that generally happens? Of course @JaredBusch was faster then I was.
Probably very seldom. But if you want to upgrade the memory you might have to install the second CPU just to get access to the rest of the memory slots. So if a dual CPU system have 16 memory slots you can only populate 8 of them if you only have one CPU.
This is a test system using older refurbished hardware so it might be more likely for me to add things as I go.
-
@pete-s said in VM host: dual CPU vs single CPU - same CPU performance rating:
@coliver said in VM host: dual CPU vs single CPU - same CPU performance rating:
How often are people adding CPUs to servers? Is that something that generally happens? Of course @JaredBusch was faster then I was.
Probably very seldom. But if you want to upgrade the memory you might have to install the second CPU just to get access to the rest of the memory slots. So if a dual CPU system have 16 memory slots you can only populate 8 of them if you only have one CPU.
This is a test system using older refurbished hardware so it might be more likely for me to add things as I go.
There is that. Some of the PCIe slots on the server board are not active unless there is a CPU in the second socket.
-
These are fringe scenarios, not normal.
Of course snowflake systems exist, but that is not the common scenario.
-
@pete-s
Also what is the defacto standard in bench-marking single CPU rating ?
How do you test ?
-
@emad-r said in VM host: dual CPU vs single CPU - same CPU performance rating:
@pete-s
Also what is the defacto standard in bench-marking single CPU rating ?
How do you test ?
Generally you test with your specific applications. Benchmarks are interesting, but tell you very little about how things will impact you specifically.
-
@emad-r said in VM host: dual CPU vs single CPU - same CPU performance rating:
@pete-s
Also what is the defacto standard in bench-marking single CPU rating ?
How do you test ?
I use the Passmark CPU test because they have a benchmark you can run on your machine and then automatically submit the test results. Then you can see the average results on the website. https://www.cpubenchmark.net/
A CPU like Xeon E5-2630 V4 has the result submitted from 59 machines and then averaged.
The CPU test consists of the following
- Integer Maths Test
- Compression Test
- Prime Number Test
- Encryption Test
- Floating Point Math Test
- Extended Instructions test
- String Sorting Test
- Physics Simulation Test
- Single Threaded Test
This is just a general benchmark for the CPU.
-
Alright, I have some updates now that I have two identical machines but with different CPU configurations.
Machine 1: 2 x E5-2630v2, 6 cores each = 12 cores @ 2.60 GHz, 64GB RAM
Geekbench: ~29000
Idle power: ~60 WattMachine 2: 1 x E5-2670v2, 10 cores @ 2.50 GHz, 64GB RAM
Geekbench: ~27000
Idle power: ~30 WattThe dual CPU machine is slightly faster but results are within 10% and there are some variation in the test when repeated.
The big difference however is in power consumption. Dual CPUs requires about twice the power when idling.
The machines both have the same dual socket motherboard, but the single CPU has fewer memory sockets but bigger (4x16GB) versus the dual CPU machine (8x8GB). But that is the reality as well - you need to populate more sockets with two CPUs.
