Show CPU model ID in lobbies

Jip

I'd like to add to this topic that the benchmark being used and throttling are two different problems that would require to be solved separately.

And, next to that, I'd love to add in that:

You can fake the benchmark score and therefore probably the CPU model too.
Considering we are 'power users' according to you, overclocking OR underclocking a CPU wouldn't be taken into account if you show the CPU model. Especially underclocking is a powerful technique for laptops to prevent the throttling you mention. Therefore your new technique is also unreliable, as I believe Pearl is trying to show to you

I thought of a version of 'benchmarking' that would be a lot better: every minute throughout the sim you compare the sim rates of all the players in the game. You get a number between -1 and 1, depending if you are below, on or above average.

At the end of the game this number can be merged with previous games - allowing you to show some form of pattern that is more reliable than the current test. If your CPU does perform on average well then this number would be positive. If your CPU performs averagely poor then this number would be negative.

This is all done locally and stored similarly to how the benchmark is stored. As of writing I am not sure how that is stored - there may be a flaw there.

I think it would help best in combination with the current benchmark, which needs to be retrofitted slightly to check for memory speed because that is in fact the major bottleneck for CPU's if no throttling is happening:

https://www.semanticscholar.org/paper/The-Gap-between-Processor-and-Memory-Speeds-Carvalho/6ebec8701893a6770eb0e19a0d4a732852c86256?p2df

From a practical point of view, @speed2 Do you know if there is a way to retrieve the information shown in the console command ren_ShowNetworkStats? It contains all the information this would require.

Pearl12

@Jip that sounds amazing.

How would it work across maps? I have an older CPU (2012) that doesn't lag except in huge games like Setons. Since we know the average is susceptible to outliers, I'm wondering if playing the occasional Setons would make it look as if I have a laggy CPU when on 95% of maps it's not a problem.

I'm guessing people would just learn that a certain number is likely to lag on certain maps? Like, if my average is -0.3 (to pull an example out of thin air), that's probably fine for up to 10x10s, but might be worrisome on a longer 20x20.

Jip

You can make it map (size) or player number dependent. I just need access to the data in that console command. It is in the client, just maybe out of reach.

Jip

I've looked into it, so far haven't found a way to retrieve all the current sim rates of all players. I tried with:

GetSessionClients()
Retrieves ping, authorized sources, ejected by, and how long the player has been 'silent' with regard to packages - but no sim rate
GetSimRate()
Retrieves the current sim rate, regardless of your own (maximum) sim rate
GetSimTicksPerSecond()
Always returns 10

Jip

@Pearl12 I spoke with both Speed2 and Balthazar about this, it doesn't appear to be promising. The data in the console command ren_ShowNetworkStats does not appear to be accessible, and without that no comparison can be done.

Balthazar mentioned that the following snippet may show that you are the one that is slowing down the game:

ForkThread(function()
	local time1, time2 = GetGameTimeSeconds(), GetSystemTimeSeconds()
	WaitSeconds(1)
	time1, time2 = GetGameTimeSeconds() - time1, GetSystemTimeSeconds() - time2
	LOG(time1 / time2)
end)

But it may not be reliable.

Another approach is to keep track of the current sim rate of a game on a per map basis. In the long run, but that would be a long run, players with good CPU's will have a better average than players with a poor CPU. Since players with a poor CPU (or throttling) will always slow down the game at some point, where as when you play with only good CPU's this wouldn't happen.

left_boy

There is a command from the game engine that tells the others that somebody is the bottle neck but there is nothing implemented to do something with that data. Sometimes I see messages in logs that say unknow message "Bottleneck" or "Bottleneck cleared" (or something similar not remember the name)

Jip

I can't find any message being made with the words 'bottleneck' or 'cleared', can you be more specific?

left_boy

I am not sure how they are called just know that I have seen such messages sent a couple of time and that they used to be logged by the client under Warning saying the command was not implemented

left_boy

I am no game dev and have never really looked into the lua files beyond the init.lua. So I can not help you on it. But I am 100% sure messages exist that are sent when a bootleneck appears (not sure how it works) but I have seen such messages in log files of the client. It is probably an GPG thing that was built before FAF era if you would let me guess.

left_boy

Maybe it is even part of the engine who knows...

ZLO

can do like this guy http://prntscr.com/uit0pk

jcvjcvjcvjcv

@Pearl12 said in Show CPU model ID in lobbies:

@jcvjcvjcvjcv said in Show CPU model ID in lobbies:

@Pearl12 said in Show CPU model ID in lobbies:

Only if you have all the CPU models memorized. And even then, are you sure every laptop CPU is a distinct model from a tower? And even then, who is to say that every laptop CPU can't play every game?

Showing CPU make and model would be an improvement over the current situation. It is not unreliable. Currently there are too many fake low bench scores. And most of those are laptop players.
You do not need to memorize all the models, nor are there that many.

If you are asking for cpu model in addition to speed score, then sure. Go ahead.

I never asked for the removal of the bench. So yes; in addition to.

@jcvjcvjcvjcv said in Show CPU model ID in lobbies:

Your argument about that some average does not know something goes pretty flat on it's face as this whole game is something very few people master. The people that do not care about lag can continue their normal business of playing with players with 400 cpu score.

It doesn't have to fall flat on its face, because you're not standing up to begin with. You have no data to back up your argument. You have, once again, made a generalization. "Oh, people master this game, therefore they must know all the CPU models." Orly.

Where. is. your. data?

People not knowing the cpu model can just ignore it, just like 90% of the hosts ignore full cpu bench benc bar...

@Azraeel said in Show CPU model ID in lobbies:

@jcvjcvjcvjcv

You must be playing with some very bad cpus are smth, but I do understand that thermothrottling is a bitch.

Well, my foelist of people that lag disproportionally bad compared to their bench score is quite big... Some people do PM me when they have fixed their shit to have themselves scrapped from that list.

But you always have to tab out and see in the client window if there are foes in the lobby; it will not show in the game window itself, which makes no sense whatsoever.

jcvjcvjcvjcv

@Jip said in Show CPU model ID in lobbies:

I'd like to add to this topic that the benchmark being used and throttling are two different problems that would require to be solved separately.

And, next to that, I'd love to add in that:

You can fake the benchmark score and therefore probably the CPU model too.

Considering we are 'power users' according to you, overclocking OR underclocking a CPU wouldn't be taken into account if you show the CPU model. Especially underclocking is a powerful technique for laptops to prevent the throttling you mention. Therefore your new technique is also unreliable, as I believe Pearl is trying to show to you

I thought of a version of 'benchmarking' that would be a lot better: every minute throughout the sim you compare the sim rates of all the players in the game. You get a number between -1 and 1, depending if you are below, on or above average.

At the end of the game this number can be merged with previous games - allowing you to show some form of pattern that is more reliable than the current test. If your CPU does perform on average well then this number would be positive. If your CPU performs averagely poor then this number would be negative.

This is all done locally and stored similarly to how the benchmark is stored. As of writing I am not sure how that is stored - there may be a flaw there.

I think it would help best in combination with the current benchmark, which needs to be retrofitted slightly to check for memory speed because that is in fact the major bottleneck for CPU's if no throttling is happening:

https://www.semanticscholar.org/paper/The-Gap-between-Processor-and-Memory-Speeds-Carvalho/6ebec8701893a6770eb0e19a0d4a732852c86256?p2df

From a practical point of view, @speed2 Do you know if there is a way to retrieve the information shown in the console command ren_ShowNetworkStats? It contains all the information this would require.

That would be nice. Effectively you don't only get a 'skill' rating, but also a cpu rating.

However, the only real thing that counts is people being the slowest below 0.

Jip

There is a post right below that explains that as far as I am aware, that approach is not possible

Mvk_-

I believe the general point involved is to reduce or eliminate possible means to reasonable identify someone based on hardware or software config is where say a bottleneck might possibly be.

Is a big reason say modded games have such lack of mainstream interests if not mistaken.

One of the best alternative ive ever come across, in idea of course, is to just automate a foe like system to a point and just realize you dont have to manual just add people you have no interest in matching with/playing due to high constant sync/lag issues.

Otherwise to just be able to foe people within match would say be second best alternative to that one, another idea to say help not having to wait.

When a match lobbies and someone is foed they can be highlighted. Helpfully can remember why they are foed of course but again, improvement is the idea, yes?

Thanks