BF4 VPS performance

[underclocked]

I am considering ordering a 4 or 6 core VPS package and was curious if any of you have noticed improvements in the VPS performance when running BF4 servers. I remember reading early on that even smaller servers were having issues.

Is anyone out there running a 6 core VPS? How many FULL slots are you able to run if I were to limit each server to say, 48 players or less?

[soja]

In our experience on a dedi, each server uses almost a core at 64p(full).

for a VDS this will likely be a full core at 48p.

I don't think many people run BF4 on a vds, due to the problems when the game released.

[underclocked]

I do remember hearing about problems early on...wondering if any of this has been fixed?

[FAST]

VPS is not recommended for large BF4 servers.

[IcEWoLF]

I ran a 48 slot BF4 VPS server on 6 core, honestly it ran like crap. No pun intended NFo, but this game runs like crap on HT's.
If you would have had a full core, I think the server would run better.

[kraze]

I ran a 48 slot BF4 VPS server on 6 core, honestly it ran like crap. No pun intended NFo, but this game runs like crap on HT's.
If you would have had a full core, I think the server would run better.

BF4 did and still does have some performance issues across the board. However, the OS doesn't know what's an HTcore or not. Our VDS plans are also just broken up into HT cores as it makes it easier for selling, your BF4 server could've actually been on a "real core", persay.

There are many things which in the BF series which can drastically affect performance. Specifically the maps you're running, Punkbuster and the load of the VDS will heavily play a factor.

[underclocked]

KRAZE would you recommend 48 being the upper limit where performance won't be degraded? 32?

[kraze]

KRAZE would you recommend 48 being the upper limit where performance won't be degraded? 32?

48 is currently the slot cap as we have to give a baseline which works for all of our VDS machines. That number may be able to be pushed depending on the type of machine your VDS is on.

[soja]

I ran a 48 slot BF4 VPS server on 6 core, honestly it ran like crap. No pun intended NFo, but this game runs like crap on HT's.
If you would have had a full core, I think the server would run better.

BF4 did and still does have some performance issues across the board. However, the OS doesn't know what's an HTcore or not. Our VDS plans are also just broken up into HT cores as it makes it easier for selling, your BF4 server could've actually been on a "real core", persay.

There are many things which in the BF series which can drastically affect performance. Specifically the maps you're running, Punkbuster and the load of the VDS will heavily play a factor.

Right, however, my understanding is if you put a workload on one thread, and a game server on another, but those 2 threads use the same physical CPU core, the performance on the game server could decrease greatly, as it is sharing the CPU core with another process. Please correct me if I am wrong, I am genuinely curious if this is a correct interpretation of how it works.

[kraze]

I ran a 48 slot BF4 VPS server on 6 core, honestly it ran like crap. No pun intended NFo, but this game runs like crap on HT's.
If you would have had a full core, I think the server would run better.

BF4 did and still does have some performance issues across the board. However, the OS doesn't know what's an HTcore or not. Our VDS plans are also just broken up into HT cores as it makes it easier for selling, your BF4 server could've actually been on a "real core", persay.

There are many things which in the BF series which can drastically affect performance. Specifically the maps you're running, Punkbuster and the load of the VDS will heavily play a factor.

Right, however, my understanding is if you put a workload on one thread, and a game server on another, but those 2 threads use the same physical CPU core, the performance on the game server could decrease greatly, as it is sharing the CPU core with another process. Please correct me if I am wrong, I am genuinely curious if this is a correct interpretation of how it works.

There really isn't any way of knowing. The OS/our system can't tell the difference since an HT core is treated as a real core. A VDS could've been made up of all HT cores or a combination of both, and with Windows throwing that process around to different cores it's isn't feasible to say that bad performance is caused by the HT cores since you don't know what your server is running on.

We talk a bit more about what an HT core is and why it's good for performance here.
http://www.nfoservers.com/forums/viewto ... =47&t=5099
http://www.nfoservers.com/forums/viewto ... =47&t=4940

[Edge100x]

A "real" (physical) core contains two HT "cores". It is not true that one customer might be given a "real" core and another customer a HT core, because there are no "real" ones on the system with hyperthreading enabled.

When two customers have, or one customer has, two HT cores (threads) that share the same physical core, each one will only have part of that physical core, so its performance will be less than if it weren't sharing. Xen is good about balancing virtual CPUs in order to ensure that busy ones share with less busy ones in order to reduce this effect, and we so severely underload our hardware that it rarely comes up.