Graphics charts idea.

[imported_balzi]

Hi all,

I had an idea that is maybe old, but maybe it's not. I think it's worth a try,

I always find myself searching the web for a comparison between 2 particular cards...
say 9200 and 9600SE, X800pro and 6800 GT and 6800, 9600XT and 9500 pro, 5700ultra and 9800SE.. whatever.. I do this so often, and I figured that out there is the information I want but it's not in one spot.

I think this would work well...
Every time Anandtech 'bench's a new GPU they enter the corresponding scores into this database

GPUscores :
GPU type , manufacturer, test name, test type, score.

some example records.
6800 128MB , ASUS, Farcry, avg fps, 70 (the numbers are not real)
6800 256MB , ASUS, Farcry, min fps, 56
X800Pro 256MB, ABIT, DOOM3, avg fps, 45
X800XT-PE 256MB, Sapphire, homeworld2, max fps, 102

then someone just needs a bunch'o'scripts TM that give you a choice of 2-3-4 or more GPUs and it'll just print bar-graphs for each group of records.

like you'd ask for 6800 GT 128Mb vs. X700XT 128MB -- all records that compete. and you'd get a customised comparison that you can quickly browse over to get a idea of what's best to buy..

anyone??

Hell, if Anandtech gave me their results in a format so I don't have to extract them from the articles, I'd bash it together myself.

thoughts people??

 

[piesang]

I would like that a lot. But I think it would be too much work and every once in a while the entries will need to be updated with new drivers and they are not always benched on the same systems.
If possible it is a good idea.

 

[klah]

Providing CSV files of the results would facilitate this, but the platform and drivers change too often to make it worthwhile.

Easier to just check digit-life since they bench 40-50 cards at a time.

 

[Marsumane]

Dont keep changing the data, just add. It would also be good for comparison to see how much companies drivers improve in perticular games. I think it would be worthwhile to have these databases. Some site like anandtech should have a link to just a huge section of these in terms of gpu and cpu comparisons.

 

[imported_balzi]

piesang - if a new score was created with new drivers you'd include that seperately, the DB table would have a driver revision column. you could also have cpu speed, ram size columns.

klah - doing 40-50 cards at once is EXACTLY the reason I think this needs to be done. how long does it take to compare 2 cards out of 40 or 50.. 90% of you're time would be used finding the blasted things in all that info.
Are you saying you have access to CSV files of all results??

Marsumane - agreed, only add data..

the structure of the site??
My thoughts -
front page, "how many vid-cards are you comparing?" with drop-down lists of cards.

user chooses 3 cards. page updates - now gives drop-downs of possible OS choices, RAM sizes, CPU speeds.

you could, at your leisure choose two slightly diff CPUs if identical ones are not available - mostly it would be.
and after each choice the list of benchmarks available for all vid-cards is adjusted. So the user has feedback on what benchs will be displayed.

let's say you choose 5700, 9600 Pro, 9600SE.
WinXP for all
512MB RAM
but the 5700 was benched on a AXP 3000 whilst the ATI cards were both done on AXP 2400, A64 2800, P4 2.8 and P4 3.4EE. you decide to go for A642800 as a supposed equivalent to the AXP3000.

then it brings up the benchs that all have.. but hte list is only 4 long. So you choose the AXP2400, now the list is 12 long.. that's better.. CLick "SHow me these results".

whaddya reckon.. I reckon it needs javascripting, I could do the backend mysql-php-apache stuff. but the updating through choices needs javascript (don't know it well).. sometimes the page would need to refer back to hte database though.. so maybe a full php serverside is better.

comments??

 

[imported_balzi]

some more thoughts here
from Tech Report

also, it seems that I'll be looking for people willing to help out.. idea's are useless if there no action. it's even biblical.. that should motivate my fellow tech-heads.. or not!

I think eventually this'll need to be hosted by anandtech, techreport or some tech site cause I don't have the bandwidth to handle the use I believe this would get.

 

[imported_balzi]

BUMP -- anyone who actually would be willing put some thought into getting this going?????

 

[piesang]

OK, that is a better idea to just add entries. The database will however get HUGE.
Try and contact some of the bench sites and see if they will give it a go. Maybe they could pass on the scores to you and you can start creating a test site and scripts.

Definitely worth it.

 

[SonicIce]

What if they threw in all kinds of differant systems too, there would be hours and hours of testing involved.

 

[sfmedic]

You mean like they do with automobiles? Such as the Cars Direct site? You can compare up to a certain number of cars on spread sheet format. It would seem easy, but to be fair almost impossible to the limitless number of hardware/software configurations that are possible. If this were to happen and to be accurate, every card tested would need be done on a machine set to a standard configuration.

Maybe ATI or Nvidia could be approached and set up some beta testers. Now theres an Idea?

 

[Lonyo]

Like Futuremarks database, but with entries other than 3D Mark scores, and from review sites only?

 

[imported_balzi]

OK, i can see that the database would get really big really quick. but a million records is manageable so long as you NEVER do a LEFT JOIN or something.

if there's say 100 diff. video cards. and 100 diff systems. that's only 10,000.
PLUS the GF2 benchs wouldn't be run on later systems, and you'd never expect 6800s and X800s to be benched on Athlon 700s and P3-1.26Gs.. so there's heaps of holes. The main motivation is to get the data that has been collected by HW sites across the globe and get it to be accessible in a new way, one that I personally take great interest in.

Users would, I admit, have to be aware that even scores obtained on identical systems (CPU, RAM, Mobo, OS, etc) may not be perfectly accurate.. but hey, it's still way better than going to Toms and writing down 3dmark2k3 Doom3 and ut2004 and farcry and sim city3000 scores for the 9800SE cause AMDzone left it out when they did 9600SE and 5700LE, AND THEN returning to AMDzone hoping that a 2.8GHz P4 really is similar to Toms 2800+ A64 (which it isn't )

so I ask again.. is anyone willing to help out. Can the message get around?? I might bash up something this weekend just to see if hte concept can work.. but it'll need overhauling once the data gets to large, I'm no MySQL genius and I can see 1000s of records almost immediately.

 

[Rio Rebel]

Aren't there too many variables for this to be a reasonable effort? Even if you compare two cards running at the same resolution and same AF and AA, you still could have varying levels of other settings which affect performance (bi/trilinear filtering, textures, etc.)

And what about the CPU? In many game, the difference from one card to the next would be masked or exagerrated by using a significantly different CPU.

To me, this is a wonderful wish but not very feasable.

 

[imported_balzi]

well thank you for another guy pointing out how this isn't possible. :disgust:

Obviously if it's seen as a vain attempt then no-one is gonna help are they??
Does anyone think that perhaps when you point out an issue, you could possibly consider giving a solution.. where's the developer in people?? I suppose most of us are just gamers and are completely satisfied with getting served whatever the tech sites will give us... ah heck the optimism, ah heck the dream. Maybe Aussie's really are different.

Might find an Aussie site to pitch this too.. cya lazy-boys.

Balzi

Note: tongue planted firmly in cheek.

 

[PrayForDeath]

That's a great idea IMO, but you should post this in the software forum where you can actaully find programmers who can help, not gamers or hardware enthusiasts. I hope I could help you but am not a programmer myself.
And if you don't wanna get confused with the amount of systems, you can make a different chart for each review website, since each review website (Tom's hardware for example) uses the same system in all its reviews (but it's different from the other websites' systems), with slight improvements on the system between each review (CPU speed or RAM size for example), so you could add the option on the frontpage to first choose in which website database you wanna search (Anandtech, X-bit labs, Tom's hardware, etc, etc...) where each website gives you benchmarks on the same system with different video cards.
But this method would eliminate your CPU/RAM comparison idea, while giving you the option to compare benchmarks between the different review sites, and you still can compare diffrerent driver revisions.
Good luck

 

[PrayForDeath]

Bump?
Looks like the OP left the forum.

 

[imported_balzi]

Thanx for hte +ve input PFD.. umm.. what's an OP.. is that me?? O(riginal) P(oster) maybe??

anyway, I'm still viciously throwing thoughts around in my head.. trying to get a structure.. once I've got something plausible I want to build a system with some dummy data and i think most things can be more easily solved once a little bit of ground work (actual programming is done)

BTW, how can i move this to the software forum?? Another shameless link??:Q heaven forbid..

 

[Rio Rebel]

Well, since I got bitch slapped by the newbie , I'll change my comments to something more constructive:

IF you find a way to do this, please take into careful consideration how you're going to ensure that you are comparing apples to apples. Make sure that you aren't comparing the same graphics card with the same game at the same resolution BUT different cpu speeds, amount of ram, etc.

Bad data is worse than no data. That's why I made the comment I did.

Good luck.

 

[PrayForDeath]

Originally posted by: balzi
Thanx for hte +ve input PFD.. umm.. what's an OP.. is that me?? O(riginal) P(oster) maybe??

anyway, I'm still viciously throwing thoughts around in my head.. trying to get a structure.. once I've got something plausible I want to build a system with some dummy data and i think most things can be more easily solved once a little bit of ground work (actual programming is done)

BTW, how can i move this to the software forum?? Another shameless link??:Q heaven forbid..

That's right, OP = Original Poster
Unfotunately you can't move this thread but you can ask the mods to move it, otherwise you'll have to link it (shameless link :Q)

 

[imported_balzi]

Sorry Rio Rebel -- I've never b-slapped someone.. it was totally unintentional.. and thanks for the input. When the site starts to build would you be willing to try it out and point out potential problems (preferrably with a solution of sorts.. not code, unless you want to, but just some explanation of how we could work around issues.)

that would be great.. need some people with ideas about Grpahics cards and the impact of drivers vs. cpu vs. RAM vs. mobo vs. OS vs. hd vs. case etc etc etc... this really is endless but some are non-issues -- for instance, Floppy drive brand probably WILL NOT affect results.