PDA

View Full Version : Cool Idea about Tweak guides - what do you think?


Pages : [1] 2

TB303
11-28-05, 07:08 AM
Hi people,
here is an Idea I have, I would like your feedback.

One of the most common question on this board and others goes like this:
"I have X video card and I play X game at Y resolutions, what settings would you reccomend?"

And of cousre there are a billion cards, systems, resolutions, settings and games out there. Now Tweak guides are a good start, but again you have to experiment for yourself... It's good to knw that using AF "may lower your fps" - but by how much?

What I propose is a site that has a database of settings and performance for a specific game (contributed by users), so people can log in, choose their game, insert thier system specs and see a break down of settings vs fps - so you can see in real time what effects this have on yoru FPS and maybe print a summary.

what do you think? I would love to hear from someone more clued up...

Thanks,

rudedog
11-28-05, 07:27 AM
I think it's a great idea, I'll start


BFG 7800GTX 512 in SLI Recommended Resolution(s): Any damn Resolution you want to play :D

Gaco
11-28-05, 08:35 AM
We'd have to get a good programmer or something for this one right? Can Mike C handle teh sh!t? :D

Good idea though, I'm in if everybody else is and if it turns out a possible and successful project...

TB303
11-28-05, 09:17 AM
We'd have to get a good programmer or something for this one right? Can Mike C handle teh sh!t? :D

Good idea though, I'm in if everybody else is and if it turns out a possible and successful project...

Cool,
thanks for the support.

I'm no programmer, but I don't think it should be very complicated.
We need a system similar to a web form, a bit like those "OC calculators" - that give various multi at different HTT and Mem frequencies - we could probably use a modified version of the same system.

My problem is how do we stop idiots from submiting bogus info (*like 3Dmark05 comparisons*)?

SDinc
11-28-05, 10:56 AM
You can just take a statistical average and throw out entries that vary too widely from the average. You need a lot of data before you can get a statistically accurate accessment though.

Even with the data, you have to factor player preference in as well. Some players like to have more eye candy and are willing to sacrifice frames per second for it. Others have performance minimimums. For example: Can't play any game under 60 frames per second.

TB303
11-28-05, 11:09 AM
You can just take a statistical average and throw out entries that vary too widely from the average. You need a lot of data before you can get a statistically accurate accessment though.

Even with the data, you have to factor player preference in as well. Some players like to have more eye candy and are willing to sacrifice frames per second for it. Others have performance minimimums. For example: Can't play any game under 60 frames per second.

That's exactly why this will be very useful,
instead of you doing trial and error, you can build on teh comulative experience of others.

Who else is up for it?

FraGTastiK
11-28-05, 12:09 PM
much simpler.look for reviews like this one.

http://www.nvnews.net/vbulletin/attachment.php?attachmentid=14630&stc=1&d=1133197485

http://www.nvnews.net/vbulletin/attachment.php?attachmentid=14631&stc=1&d=1133197526

people may get better or worse results depending on how different their system is compared to review's test system and possible OC or handicaps.but it would give a good estimation about what kind of performance one might see with his/her system.

Gaco
11-28-05, 12:10 PM
Also apps and other **** running in the background could cause some minor and misleading deviation =/

TB303
11-28-05, 12:23 PM
Of course this won't be 100% accurate, but it will give you an idea and a starting point.

Guys, if you just look at the sheer amount of posts in this forum about "WHat settings for my card", it immediatly becomes apparent how needed such a system would be.

Also, about those Jpegs |White|Mute| posted - they are far too little, and far too late. What I'm talking about is a far more detailed breakdown of settings and performance - you could see how much FPS would "cost you" to increase AA by 2, for example.

Come on...

SDinc
11-28-05, 11:31 PM
It's a pretty simple database with input/reporting type of application. Obviously the best way to provide this would be by hosting a web site. The web site would display stats for each game and allow users to input their favorite settings as well. The only problem is you will be relying on people to input their information in order to get good stats. (Unless you can get data from somewhere else). Without a lot of input, the information wouldn't be very accurate. You really need thousands of users to input information for each game in order to get good results. There are a ton of variable factors that would need to be taken into account. The number of combinations would be very large so you will need tons of data. Some HW factors would be:

CPU Brand
CPU Speed
Memory Speed
Amount of memory
GPU Model
Speed of GPU
GPU Memory
Speed of GPU Memory
Display Driver Version
Sound Card
Windows version

j0j081
11-29-05, 06:45 AM
quit talking about it and do it! :)

nutcrackr
11-29-05, 07:57 AM
Cool idea, I think submitting data from users would be the best way to go about getting all the performances from many different configurations. Then you could have a rating system and or some averaging system to get the general settings for the users system. They could specifiy resolution, AA, if they need real time shadows etc and then you could give the users a report. Essentially it would require building up a set of configurations for each game as a form and letting users alter and submit how well it runs. Maybe even provide a benchmark for games. You could have links to tweaks which may be user submitted also. Microsoft has the system from futuremark in which it detects your computer specs and tells you the performance but this would be more detailed and provide tweaks for easy access. It would start off slow but it could become a very good resource. (xmasmile)

TB303
11-29-05, 10:12 AM
Cool,
So how do we go about starting this?

Let's take one game and focus on it, MFS:MW for example,
but we need some sort of a benchmark that we can measure against.

TB303
12-01-05, 06:03 AM
Anyone has any ideas?

supra
12-01-05, 09:47 PM
.NET + ADO.NET :D

Im a IT/Web Programmer for a risk management company that deals with alot of database systems.

First thing I would do is design the database model.

Infinity666
12-01-05, 10:10 PM
Awesome idea I hope it actually happens(xmasgrin)

Clay
12-01-05, 10:26 PM
.NET + ADO.NET :D

Im a IT/Web Programmer for a risk management company that deals with alot of database systems.

First thing I would do is design the database model.
I guess you don't follow the MSF, you skipped right over the entire envisioning phase. :p

supra
12-01-05, 10:32 PM
I guess you don't follow the MSF, you skipped right over the entire envisioning phase. :p

yea but this seems like a fairly simple system. Just large but not complicated.

;)

Clay
12-01-05, 10:34 PM
Yeah I was just messing with you. ;) It would be pretty straightforward.

SDinc
12-02-05, 12:22 AM
yea but this seems like a fairly simple system. Just large but not complicated.

;)

Famous last words! But don't look to MSF to solve all your problems. MSF is just a glorified waterfall approach. The best way to tackle these types of problems is using an iterative development process like the UP. All crappy software products neglect design and requirements analysis. Good product development requires involvement of the end user in every stage of the development process.

supra
12-02-05, 01:39 AM
ok seems like i have left the impression to jump right into development :D . Yes obviously you need to do analysis & design with this system. What i was suggesting was assumed that analysis has already been done.

I think designing the db model is most important in this system because the whole system seems to be just queries from the db.

keep in mind im talking about designing the db model as in ER diagrams. Not actually creating the db yet.

TB303
12-02-05, 10:20 AM
People,
thanks for the feednback, I'm going to try and do a small design document with a friend over the weekend.

Maybe you can come up with a solution to a problem we are stuck at:

If we take a game like NFS:MW that doesn't have a proper becnchmark, how can we ask people to bench it accuratly? We were thinking of asking them to do the "first race" and run fraps in the background and submit the first minute data, but that may be inaccurate - i.e someone may be staring at the sky for a minute and the FPS will skyrocket...

Do you have any idea on how to get the "Avarage joe" to bench a game easily and accuratly? thanks!

SDinc
12-02-05, 10:29 AM
I am not sure benchmarking really matters. Wouldn't the questions really be more like:

a) What settings do you like to use?
b) What is more important to you Image Quality or Performance?

TB303
12-02-05, 11:13 AM
I am not sure benchmarking really matters. Wouldn't the questions really be more like:

a) What settings do you like to use?
b) What is more important to you Image Quality or Performance?

Well,
What I had in mind was more of a "Scientific" approach:
You specifiy yoru hardware (CPU, GPU, OS, RAM) and of cousre the game and the res you want to play at, and you see a break down of FPS.

You can start high, say at 60 FPS - with the majority of graphics options set off, and if you enable say, AA you see a drop of 5 FPS, if you increase shadows, you see a further drop of 3 FPS.

We can also start backwards, with max quality and than increase the FPS as we reduce the quaility.

the idea is to have this knowledge by taking a distributed approach, each person submits his FPS under 3 different settings (i.e 2AA, 4AA, 6AA) and with about 50 people we will have a very inclusive breakdown of systems and settings.

But how do I "Bench" a game that doesn't have a Benchmark?

nutcrackr
12-02-05, 06:16 PM
I think it would be best to start simple, you can add more funtionality later on :)

You've already said that games without benchmarks are going to be problematic at the least. So start off with just the settings they play the game at, now when you get 100's of people submitting specs from their computers you will hopefully start to see a trend. If more people are running with AA you could imagine performance is acceptable.

I could see lots of lovely graphs coming into play, everybody loves graphs :D However there will be more relation between how well joe runs game A to how well joe runs game B for comparisons, so if bill can run game B at settings similar to joe he will have an idea of settings needed, then you can link tweaks within pages or within profiles etc. Later you can add benchmarks for specific games on top of the standard basic functionality that can be applied to all games.