CPU Quality indicator being useless

[Poch]

Note for mod : maybe i should have posted in suggestions or the cpu tool thread, but couldn't decide, feel free to move my topic if needed and sorry for inconvenience

Hi

I have what i think is a good idea to improve the CPU indicator. But maybe it's not and you will prove me why

First of all, let me say i was very happy about this feature, but after several months using it, i think most people will agree that is is useless as-is. What is the CPU indicator ? It's a decision-making helper tool : "should i kick this guy that seem to have an old CPU from my game ?". So its not "how good/bad is a computer", the only data we, players, care about is SIMSPEED. I don't care if someone has 199 cpu and me a 201. What i care about is whether or not all payers in a game will perform without too much difference in gamespeed.

Now, host a seton or any other cpu-heavy game, let everyone join, look at cpu numbers, and try to make a prediction for each of the 7 other players : "When my computer will be running simspeed at +0 , i think that this player X will be running at simspeed Y". Try to have an educated guess for each player in a decent time. You just can't because the tool final interface (numbers 100-300) doesnt help humans at all. Furthermore, the numbers are wrong so even if we could interpret them the prediction would just be wrong.

Therefore my conclusion is that the tool doesn't fill its purpose actually.

How to solve the main problem : ie. have a tool that player actually understand and can use to make decisions
Get rid of numbers between 100 - 300 who aren't meaning anything to anyone and replace them with something that actually means something in term of simspeed. Like letters. If i have a "C-grade" CPU, then i know i can play at the same simspeed as anyone having a "C" cpu. If in the game someone has a "D" cpu, then we know at some point the guy will go -1/-2 when we all are +0. Maybe i dont know exactly at which point it will differ, and maybe 2 guys with a "D" CPU won't go -1 at the exact same time but it doesn't matter here. You can use more letters or +/- to get more granularity. Like A+/A/A-/B+ etc...

That way i can host a seton game reserved to B or better CPU and be confident that simspeed will (probably) never drop before +0. And people could filter themselve from joining some game, having a better understanding of how bad/good their CPU actually is.

How to solve problem related to wrong benchmark numbers not being representative of simspeed
And don't use a custom LUA benchmark in the lobby for every game we play. Just launch a "reference" replay once and measure actual simspeed. Simspeed dont lie. I see that ZePilot made a small ingame benchmark to test the proxy feature. We need exactly the same thing but with a reference short-but-cpu-heavy replay. That replay should bring the best affordable CPU (like an oc'ed 4770K) from +10 to -2 gamespeed at least.
Then you can say being able to stay at +0 all replay long is A.
Then you can say being able to stay at -1 all replay long A-.
Then you can say being able to stay at -2 all replay long B+.
Or you could measure total replay running time, etc... Whatever is the best data to translate into an explicit scale.
Etc.. You get the idea.

That way no more guys around 300 score have more powerfull CPU that people around 200. I have friends with score x2 or x3 compared to mine but we run the games at the exact simspeed. Windowed or fullscreen doesn't help it, their score just won't improve.

It's a very simple idea and i really think something based on an arbitrary scale would definitely be better that random numbers. What do you guys think ?

[johnie102]

This has been suggested before. The problem will be that people with a B quality score will kick every C or D player even if the map is a 5x5 or a 10x10. If you really want to do this the right way, the displayed 'score' should depend on what map is selected and the amount of players present in the lobby.

[Poch]

Thanks for quick answer. I knew it had already been said somewhere but was too lazy to dig the conversation ^^

You could argue that people would kick 200+ or higher scores actually too if they own a 150 CPU. It just doesnt happen actually so how would that be different with letters ?

It doesn't need to be map dependant at all, you just need a benchmark replay heavy enough to be representative of big games.

People want that tool only for heavy games like setons, phantoms, 40*40, etc. No one actually looks or think of CPU when playing a 4v4 on a regular (non-turtle-fest) 10*10km map.

So i don't really agree that people would be arbitrary kicked of most games even by having a low grade CPU. If everyone is aware that the "E" CPU can still run for example a game like 4v4 canis in +0 to the end, then you won't get kicked of those games by having an "E" CPU. Now you may have a hard time sneaking into a B+ only seton, but it's the point of the tool

I have no interest kicking a E-CPU player on a canis even if i own a A-CPU (which i do). What's the point if i know the E-CPU will run the map just fine ? Game won't run better with everyone being "A", i just need no one to be worst than "E" if thats the lowest acceptable CPU for 10*10. It is hard enough to get 7 players to start a game, so you don't get too picky for a classic game if it is not justified imo.

I think actually on small map 4v4 team games there is more discrimination based around ping/location/rating than cpu. For big big games, ping is less a matter than being able to predict end-game simspeed.

One problem i see, is that you could'nt spot people lagging at a specific time (let say overheat, or cpu used by other app). But even with the actual "last minute real time" quality indicator tool, i still get from time to time a guy on a 5 year old overheating laptop that fucks a game only 5 minutes after starting it by going to -5 every 30s or so. So even the guys that SHOULD get spotted by the tool aren't So, again maybe it's time to try for a change even if it doesn't sounds "ideal" because actually the tool doesn't fulfill any purpose ?

[prodromos]

The problem with Setons(or any similar map), is that there isn't such a thing as.an A cpu; maybe some B- and a lot of Cs.
People playing Seton's have accepted apriori, that their game will lag terribly at some point, but because they like the whole concept so much, they really don't care.
Furthermore the mobile "revolution", has guaranteed that companies will think.twice before.making more powerful cpus, because they're already too powerful, lol!
Continue supporting this trend, and after some years you'll be happy if you can play some pathetic browser app.

[FireMessiah]

Its an indicator, it shouldn't be used as a set rule.
A general Idea, not a statement of fact.

If someone is showing a high score, I ask what CPU and GPU they are running.
For setons, I usually recommend an i5 or above/equivalent cpu, and gtx450 or above.

If someone is lower, but says they know it can handle it, that's acceptable.
If they don't reply, then boot them.

Courtesy is part of good hosting. If you have to kick someone, explain to them why

[Poch]

Maybe there are no "A" CPU, and of course, they may get better in a year, so keep room for improvement. You can update the scale every year if needed it doesn't matter.

I agree with what you said about being a decent host but that's not the point. People can't exchange on level of performance with simple words in lobby :s And what someone call "decent", i call it "crap".

The CPU tool has been made so that we don't have to ask to every player if he thinks he will be OK.
I can maintain a list of every player i encounter and how well their CPU perform but again why having such a tool then...

People playing Seton's have accepted apriori, that their game will lag terribly at some point, but because they like the whole concept so much, they really don't care.

You say that there are no CPU capable of handling it but i assure you there are. Most time i play seton after 25m we will get everyone at +0/+2 and two guys being at -2. Of course they say they can handle it. I just wish i could have spotted them back in lobby. But if 80% of cpu-heavy map player have high-end rigs, maybe they actually care about decent gamespeed... I agree there are regular seton player lagging like hell, but most of them have very good CPU imo.

You must add to that i'm not the host in most of those games i just dont want to play with laggers from time to time and i can never tell after joining lobby.

For example, typical seton from a few weeks ago :


This was a big game and the screen is taken while massive amount of units move at same time. That is why the best players are at -2. Excepting for massive ASF fight and some specific moments, a lot of people can handle such big games in +0 even after 30 minutes. But some people can't.

Now look at people simspeed. Don't tell me it can't be categorized in a very simple and intelligible way
To the simplest, let say everyone has a "B" CPU, PascalP got "D" and Zephire got "E".

Its not seton specific, usually in most cpu heavy games you just get 1-3 guys dropping below +0 and they never tell you in lobby. In my example, PascalP was at -2 while everyone else was still at +0. When i asked him a few days later (on a Seton he hosted) about his comp he said he could handle Seton well. When i know for a fact it will drop -1/-2 with the rest being at +0.

[sasin]

This has been suggested before. The problem will be that people with a B quality score will kick every C or D player even if the map is a 5x5 or a 10x10. If you really want to do this the right way, the displayed 'score' should depend on what map is selected and the amount of players present in the lobby.

Well then couldn't you use Poch's system but apply a different letter grade based on when the person lags in a replay? Basically just adjust for map size. So, if you make an hour long replay where more units steadily come onto the screen, you could do

sim speed never drops - A for all maps
sim speed drops after 50 minutes- B for 20x20, A for all other
sim speed drops after 40- C for 20x20, B for 10x10, A for 5x5
sims speed drops after 30- D for 20x20, C for 10x10, B for 5x5

etc.

I realize you could make it a bit more nuanced of course, but the scale need not be the same for all maps.

[E8400-CV]

A short benchmark will still give problems either way. All those people playing on Apple stuff, Ultrabooks, etc. all get cpu throttling after a sustained high load for the simple reason the thermal design of these products does not meet the requirements for full time full-load.

Also; a lot of people are doing other stuff at the same time. Just to name some things:
-playing music
-watching video
-parring / unzipping downloads

I play at a 2500K and a E8400; at the E8400 I was lagging at -2 once; I killed Windows Media player (192 kbps *wma) and I went immediately to +0. Now I OC'd it 20% and just don't run music anymore in Setons on the E8400.

I would like just to see the type number of the CPU. Lower end mobile CPU's are probably more likely to be in a chassis that can't cool them.

In addition to that there should be an option to kick someone if they are persistent running it slower than the others. In de last days I've had a game were someone apparently just went afk and was running at -5 (all others >0) for extended periods. Especially in Full Share games they should just be kicked if even their own team agrees.

To label the scores in groups 'A/B/C' like energy class of refrigerators might be nice, just like we have the Portuguese flag to indicate likely smurfs.