CPU Power Comparisions?

Ask questions about dedicated servers here and we and other users will do our best to answer them. Please also refer to the self-help section for tutorials and answers to the most commonly asked questions.
Post Reply
mukunda
A regular
A regular
Posts: 42
https://www.youtube.com/channel/UC40BgXanDqOYoVCYFDSTfHA
Joined: Fri Nov 16, 2012 2:42 pm

CPU Power Comparisions?

Post by mukunda »

We are running a server right now that has a bit more than the normal amount of plugins, and we're looking for a solution to let it run smoothly. (I'd say the plugins take maybe 20% more than vanilla CPU power--pure speculation, to be safe lets say 30%)

Ideally we would want it to be 36 slot with 128 tick but we could maybe live with 102.4 or 96 tick, or 32 slot if we go normal gameserver route

As of now its running on our 6core VDS at 64 tick and it is overloading the core that it is running on.

We're having a little trouble deciding what's the best route, here's a few questions...
Can a 32slot ultraaccelerated normal game server handle the extra plugins we have and still run smoothly at 128 tick?
How much more power would a core from a dedicated server (199$ package) have?
Do you think a dedicated server core can handle 40 man 128 tick?
How much does tickrate increase cpu? would you say 128 tick runs at twice the cost of 64 tick? or maybe 50% more?
ImageImageImageImage
User avatar
Edge100x
Founder
Founder
Posts: 13156
Joined: Thu Apr 18, 2002 11:04 pm
Location: Seattle
Contact:

Re: CPU Power Comparisions?

Post by Edge100x »

You didn't say a game, but I'll assume CS:GO because of your tickrate references.
mukunda wrote:As of now its running on our 6core VDS at 64 tick and it is overloading the core that it is running on.
Does it hit the wall when you have plugins disabled, as well? What OS are you using, and have you tried a different one?
We're having a little trouble deciding what's the best route, here's a few questions...
Can a 32slot ultraaccelerated normal game server handle the extra plugins we have and still run smoothly at 128 tick?
This is difficult to predict because I don't know your plugins. But consistently running at 128 tick is difficult to achieve for a large server even without them.
How much more power would a core from a dedicated server (199$ package) have?
It would be similar to an ultraaccelerated server's core.
Do you think a dedicated server core can handle 40 man 128 tick?
I would not consider it likely.
How much does tickrate increase cpu? would you say 128 tick runs at twice the cost of 64 tick? or maybe 50% more?
It depends on your plugins and what they're doing on each tick. But in general, the CPU usage should be expected to scale linearly with the tickrate, meaning that a 128 tickrate server should use about double the CPU of a 64 tickrate server. That's because every tick is a full simulation recalculation, and they should take about the same amount of time regardless of how often they are occurring.
mukunda
A regular
A regular
Posts: 42
Joined: Fri Nov 16, 2012 2:42 pm

Re: CPU Power Comparisions?

Post by mukunda »

Thank you for the answers they're really helpful

er yeah it is CS:GO

I wrote most of the plugins myself and I'm pretty sure they don't cause too much load, only a few of them actually are doing constant operations but they don't scale with tickrate. At one point during a valve patch we had most of our plugins disabled and the VDS was still stating 90-95% CPU usage from the stats command. (although it was running 36 players)

It sounds like the best route would be to avoid a dedicated box (our community is really concerned with the budget) and just go with a normal ultraaccelerated gameserver. Ultraaccelerated basically has more CPU allotted for it right? (along side your special FPS locker), and we could reduce the tickrate to 96 perhaps to leave some extra space for expensive plugins.
ImageImageImageImage
mukunda
A regular
A regular
Posts: 42
Joined: Fri Nov 16, 2012 2:42 pm

Re: CPU Power Comparisions?

Post by mukunda »

also its a shame source servers have the tickrate locked, otherwise it would be great to reduce the tickrate at the beginning of the round and ramp up towards the end
ImageImageImageImage
User avatar
Edge100x
Founder
Founder
Posts: 13156
Joined: Thu Apr 18, 2002 11:04 pm
Location: Seattle
Contact:

Re: CPU Power Comparisions?

Post by Edge100x »

Ultraaccelerated means that it runs on Linux and we allow 128 FPS (though the game won't necessarily be able to achieve it), as well as that it has our FPS locker. If 100 FPS would be enough, it might also make sense to try a regular hyperaccelerated (Windows, 100 FPS max) server first.

You may be able to muck with fps_max on the fly through "sm_cvar", but it's not something I've experimented much with. Your idea is an interesting one and I'd say it's worth testing. (The game already limits its FPS/tickrate simply because it can't keep up, but you might see some benefit to having a more stable FPS.)
mukunda
A regular
A regular
Posts: 42
Joined: Fri Nov 16, 2012 2:42 pm

Re: CPU Power Comparisions?

Post by mukunda »

Are you saying that ultraaccelerated isn't necessarily more CPU power, but rather just a stable FPS? (and I assume the stability is required to have a smooth experience above 100 tick)
ImageImageImageImage
User avatar
Edge100x
Founder
Founder
Posts: 13156
Joined: Thu Apr 18, 2002 11:04 pm
Location: Seattle
Contact:

Re: CPU Power Comparisions?

Post by Edge100x »

Ultraaccelerated means three things for a CS:GO server:
  • A different OS -- Linux instead of Windows.
  • The "FPS Locker". This does not do as much for games now as it used to, because Valve started including its primary functionality in the game itself. We still have various tweaks that apply, and the system timer is still a little more stable than on Windows, but the Locker is no longer a major performance factor.
  • The availability of the 128 tickrate option.
Our ultraaccelerated machine hardware is similar to the hyperaccelerated machine hardware, so that's not a factor.

The benefits of ultraacceleration primarily help small servers that are able to maintain 128 tickrate, such as competition servers, by making jitter as small as possible. For large public servers, the regular Windows hyperaccelerated option is often a better one, because these servers can't frequently maintain 128 tick anyhow, and the hyperaccelerated option is less expensive.

In your case, I would recommend starting with a Windows hyperaccelerated configuration and 100 tick. If that works well and you don't see dips from 100, you might consider testing it with an ultraaccelerated configuration. It would be best to make changes near the end of your paid month, in case you decide to switch back to hyperaccelerated or to switch back to a VDS; that could save you money.
mukunda
A regular
A regular
Posts: 42
Joined: Fri Nov 16, 2012 2:42 pm

Re: CPU Power Comparisions?

Post by mukunda »

Ah, thank you, I will keep all of this in mind during our meeting
ImageImageImageImage
User avatar
Edge100x
Founder
Founder
Posts: 13156
Joined: Thu Apr 18, 2002 11:04 pm
Location: Seattle
Contact:

Re: CPU Power Comparisions?

Post by Edge100x »

I also forgot to mention one other thing that ultraacceleration (and being on Linux) gets the customer, and that is access to the server console (in addition to rcon). With CS:GO, that can make troubleshooting the Workshop a little easier, but in general, it's not needed for this game.
Post Reply