Forged Alliance Forever

Re: Lobby CPU Quality Indicator

2014-08-13T21:28:47+02:00

Oops, forgot about that. So actually both numbers are smurf-able.

Statistics: Posted by E8400-CV — 13 Aug 2014, 21:28

Re: Lobby CPU Quality Indicator

2014-08-13T21:06:03+02:00

LittleInferno wrote:
I think its runing preety well .

No it's not. The score can be manipulated by adjusting process affinity and priority. Example: viewtopic.php?f=2&t=7989&p=77311&hilit=overclock#p77215

I started a thread a long time ago where I let people benchmark a replay. One guy, Au_Manta-Aoid showed us that those factors don't make any difference in actual replay performance: viewtopic.php?f=2&t=3046&start=100#p42123

I'll start a revised version of that topic soon, with a realistic benchmark (random replay, probably Setons 4v4).

Statistics: Posted by Dr_Doolittle — 13 Aug 2014, 21:06

Re: Lobby CPU Quality Indicator

2014-08-13T20:28:30+02:00

Did you read that post before submitting?

Statistics: Posted by E8400-CV — 13 Aug 2014, 20:28

Re: Lobby CPU Quality Indicator

2014-08-13T19:55:35+02:00

I think there is only 1 thing miss, that is a fast stress benchmark iff you now what i mean.
1 thing its run the test without cpu use other its give you the image off what the cpu will be capable when got some stress.
That way prevent games lag because on lobby that (X) cpu give a decent rating but in the game with some work cause lot lag that actualy kills many games.
Isnt it possible we have a fast benchmark tool on lobby for that, is a big step iff we can find one.

Statistics: Posted by LittleInferno — 13 Aug 2014, 19:55

Re: Lobby CPU Quality Indicator

2014-08-13T19:21:49+02:00

At least there is no smurfing with the CPU number. Can't say that for the other number eh, IKatherine...

10 minutes to get a bench number is a bit long.

What would be nice of course is a way to detect those players that play on crappy laptops, do the CPU test with full Turbo boost and after 10 minutes in run the game at base clock due to overheating.

Statistics: Posted by E8400-CV — 13 Aug 2014, 19:21

Re: Lobby CPU Quality Indicator

2014-08-13T18:00:27+02:00

I think its runing preety well .

When i join give me a score 141

When i run the benchmark update the score to 130 or 135

Iff the bar isn´t full that might be a bug, then its not bad tool and i think its well accurate.

Statistics: Posted by LittleInferno — 13 Aug 2014, 18:00

Re: Lobby CPU Quality Indicator

2014-08-13T15:53:33+02:00

Is there anyone who'd like to redesign the cpu quality indicator? I'd like it to be more accurate.

I suggest it to bench a replay at +10 for 10 minutes and only show scores in lobby if you've benched the replay and had no hardware/overclock changes.

Statistics: Posted by Dr_Doolittle — 13 Aug 2014, 15:53

Re: Lobby CPU Quality Indicator

2014-02-26T19:03:03+02:00

Thank you for replying to my inquiry. ^_^

(Really, that's the most human way I could think to say that...)

Statistics: Posted by errorblankfield — 26 Feb 2014, 19:03

Re: Lobby CPU Quality Indicator

2014-02-26T05:53:37+02:00

@errorblankfield & Brainstrom,

Such a benchmark could be built into the FAF lobby and passed into FA just like the rating or country information (I believe this has been suggested earlier in this thread). Unfortunately, I don't have the time or motivation to implement that on right now.

The code is open source, so if someone else wants to design and build it, more power to them. The most accurate benchmark would probably be to run a representative game replay at max speed and score based on time required to complete. In the case of the game replay, running the benchmark infrequently would make sense. That would of course have downsides as well.

Regarding the current benchmark, I don't think it's an issue of duration. The main problem is knowing how to weight each aspect of the test. Right now, all aspects of the LUA benchmark are weighted equally. If I knew which LUA operations were used the most in FA, I could weight them more heavily and, as such, improve the accuracy of the existing benchmark.

Statistics: Posted by Duck_42 — 26 Feb 2014, 05:53

Re: Lobby CPU Quality Indicator

2014-02-25T08:15:27+02:00

^I really would like CPU test that doesn't have to be ran every match (every launch of FaF works for me).

Aside from background applications, my rating doesn't change that much match-to-match. (Even with background apps, +/- 20 pts depending)

Is it even feasible to stress test on launch? Or does the game need to be open or something? In which, that can be exported back to the client so until you close said client, the CPU score is saved. Though that might also be a lot of work...

Not a biggie, but still something that would be nice.

Statistics: Posted by errorblankfield — 25 Feb 2014, 08:15

Re: Lobby CPU Quality Indicator

2014-02-25T03:37:25+02:00

Hi Duck,

I somewhere read that the benchmark tool would be much better if it would have more time to stresstest your hardware. The (legit) argument against such an implamentation was that it would take to long, and your pc would be unusable for the time period.

As i'm pro good ratings, I suggest to build such a benchmark tool witch will run whenever you start the lobby client. BUT, and here's the idea, only when hardware changes are detected. This way most of us only have to bench once a year.

Greetings,
Brainstrom

Statistics: Posted by Dr_Doolittle — 25 Feb 2014, 03:37

Re: Lobby CPU Quality Indicator

2014-02-03T14:07:42+02:00

Using the list at http://www.cpubenchmark.net/singleThread.html, I'd say ~288 is within the margin of error for the E8400.

Statistics: Posted by Duck_42 — 03 Feb 2014, 14:07

Re: Lobby CPU Quality Indicator

2014-02-03T02:51:36+02:00

Hmm, it seems my E8400 (@3.6) got a nice decrease (lower = better) in CPU score; it's now ~288, it used to be ~320... I think. Not sure that's justified though?

cataclysm315 wrote:
I know its not about the number of cores, but its the fact that the cpu is in general quite high end. I would have thought a higher score would have been given? It doesnt matter i guess, i can still play at the end of the day Thanks for your reply.

No, the 2630qm is actually the slowest i7 in it's cohort. Turbo boost is only 6/6/8/9 and perhaps the cooling / power stuff in your laptop prevents it from achieving that. In the theoretical case that it does go to the full Turbo on 1 core, it runs at 2,9GHz, which is actually the same frequency the slowest standard-power mobile i5 of the same generation (i5 2410M) runs at on 1-core Turbo (2.3 GHz base frequency). Of course your i7 does benefit from more L3 cache, but I honestly don't know if FAF depends on that (my guess is it doesn't).

The good news is that if your's is in a socket, you might be able to upgrade to something like a 2960XM, or more realistically; 2860QM

Statistics: Posted by E8400-CV — 03 Feb 2014, 02:51

Re: Lobby CPU Quality Indicator

2014-02-02T19:31:55+02:00

The new CPU benchmark algorithm has been posted as part of the lobby 2.0 update. If you encounter any problems with it, please let me know.

Statistics: Posted by Duck_42 — 02 Feb 2014, 19:31

Re: Lobby CPU Quality Indicator

2014-01-22T21:54:31+02:00

@stormbeforedawn, as i know the laptop CPU's are of course not as good as a desktop PC (in most cases). However, I can cope with 4v4 matches with intense units (around 2000) and have tested this on rohans gap and on setons using cheats and online game play. I also run the game at full settings.
@Duck_42, i was thinking the same thing as you. Maybe the revises version will use more accurate tests to calculate CPU performance. As it stands im on par with most people right now so people dont scream and run when i join like i thought they would lol.

Statistics: Posted by cataclysm315 — 22 Jan 2014, 21:54