PDA

View Full Version : Back to basics


Ayahuasca
11-20-06, 03:37 PM
i have a conceptual idea as to why Quad SLI performance is not well (Maximum PC 2006.) it is already understood that Quad SLI would be extremely cpu limited, and apparently suffers from allusory driver inefficiencies. the bigger picture to me is that perhaps Quad SLI has approached the law of diminishing returns. for example to highlight the inefficiencies of the Pennsylvania Derpartment of Transportation the suggested scenario is considered to display diminishing returns in action. you have a small (relatively speaking, anyone know about hole sizes? lol.) hole in the highway, and PennDOT sends approximately thirty-some men to repair this. it is extremely inefficient to have a crew of thirty-some men to repair this as it increases the time to repair it. diminishing returns can also be shown in a basic assembly line. increasing the amount of people in the assembly line can improve productivity up to a point. after this point it is inefficient, and diminishing returns is experienced. i feel that the hole or work to be done represents the work load by the Quad SLI configurations, and that driver inefficiencies are only more pronounced by this. just my two cents.

einstein_314
11-20-06, 03:40 PM
...thanks for that...

Bearclaw
11-20-06, 03:42 PM
Uhm. I think you should follow this up with a thesis and type the whole thing out in a post.

Ayahuasca
11-20-06, 03:58 PM
from the sentiment i gather you guys find something wrong with trying to make an objective observation about general technology? i didn't mean to insult anyone's intelligence with my long, and drawn out attempt to explain diminishing returns. i would rather have been safe than sorry from misconceptions.

Ayahuasca
11-20-06, 04:07 PM
well its only a problem of my attempts to describe diminishing returns, but not a fallacy of my idea? i was only trying to describe the results of the Quad SLI being beaten by SLI 7900 GTXs. it would make sense if the scores were closer to being equal, but not less than the 7900 GTXs, so that's where diminishing returns comes into play.

Inf1n1tyComplex
11-20-06, 06:03 PM
well its only a problem of my attempts to describe diminishing returns, but not a fallacy of my idea? i was only trying to describe the results of the Quad SLI being beaten by SLI 7900 GTXs. it would make sense if the scores were closer to being equal, but not less than the 7900 GTXs, so that's where diminishing returns comes into play.
Allow me to translate for everyone... "Why send in 4 people to do the job of 2..." Thats bascially all he is saying... So if thiers someone stuck in a hole that only 2 people can fit in why send in 4 people when 2 are only needed leaving the other 2 in the way of the tight cramped space... So im guessing he believes that Quad sli is trying to do the job that only normal sli would be required too and in doing so is causing some kind of latency due to the inefficient spreading of the task...

But it is my belief that if you fix the quad sli drives and make it more efficent to spread the work load then quad sli would not recieve any diminishing returns... especially in the case where thier is an increased input.. as you said about the factory workers thier is a point in how many you need before you recieve diminishing results but if you increase the amount of labour required by adding more work by increasing speed or quantity of the assembly line then wouldnt more mean better...

Bascially what im trying to say is quad sli metaphoricaly has the workers but the assembly line is inefficient... if the assembly line was better planned and made more efficient (im talking about the driver here) then the output of the assembly line would increase dramatically...

Its as if your saying if the hole got bigger then those 32 workers would be more efficient but in the case of quad sli the games are getting bigger and are in need more processing power but the processing power of quad sli isnt efficient enough to deal with it. Its like trying to make a group of people who half the time dont know what the hell thier doing work together to try to produce something.. It just doesnt work unless everyone knows thier place thats when things get efficient... So technically what im saying is that current quad sli setup is either driver or structual bound and that it has nothing to do with diminishing returns... Diminishing returns is only a factor if the input is the same... in that case then adding more workers would increase the inefficiency but if you increase the input then adding more workers and reworking the assembly system would be the only way to go...

Ayahuasca
11-20-06, 06:08 PM
exactly, but before driver improvements would be noticable, any other bottleneck would need to be eliminated first (time. cpu catch up, etc.) theoretically Quad SLI could equal or best (marginally) SLI, but the economics are very bad (in terms of bang for buck.)

Inf1n1tyComplex
11-20-06, 06:30 PM
exactly, but before driver improvements would be noticable, any other bottleneck would need to be eliminated first (time. cpu catch up, etc.) theoretically Quad SLI could equal or best (marginally) SLI, but the economics are very bad (in terms of back for buck.)
I guess those bottlenecks would have to be tested... who knows it could all be down to crappy inefficient drivers that are creating all those other bottlenecks or it could be the architecture in which case it cant be fixed by drivers... or any other numerous reasons etc... So for now all we can do is hope the guys nvidia pays to fix this kind of crap find the problems out and fix the inefficienies etc so that future quad sli is worth the investment... until then we can theorise until the world ends or until we are all dead and burried as much as we like and we wouldnt end up getting anywhere unless ofcource we got ourselves a degree at somewhere inbetween those times and got a job working for nvidia so we can then test and fix what the hell is going on with quad sli..

As for Bearclaw i dont think he was trying to be rude or anything its just that seeing as we have no hard facts or specialisations in the field we are attempting to theorise then it would be just mere sophistry... Basically all we are doing is making up theories based on a source of knowledge that has little credibility in which is much akin to conspiracy theories...