Gpu 102

The best way to determine a cards performance is to look at benchmarks and game tests. The only other closest way is to look at the actual specs, such as vertex operations and such, but even that isnt always the case in determining the best card.
 
i thought it was most pixel pipelines, then core clock then memory clock then RAM in determing performance? Isnt this right?

If its just benchmarks, then why do they even give these common specs?
 
so is this the order in determining performance?

pixel pipelines, then core clock then memory clock then RAM


Bobo said:
Because it would be rather useless to just have a card and nobody know what is in it.
Click to expand...

well by looking just at benchmarks, it doesnt matter whats in the card, unless you are a "computer geek"


i mean it really cant be that hard to tell me how the pixel pipelines, core clock, memory clock, and RAM relate
 
joeswm8 said:
well by looking just at benchmarks, it doesnt matter whats in the card, unless you are a "computer geek"
Click to expand...

What if the benchmarking program is highly optimized for that specific card? Then when you take that card and play games you'd be surprised on how much worse it performs than you've expected.
 
You cant look just at a single aspect. The x1900/x1950 series for instance has much fewer pipelines then the 7 series does, but they generally perform better.
 
joeswm8 said:
well by looking just at benchmarks, it doesnt matter whats in the card, unless you are a "computer geek"


i mean it really cant be that hard to tell me how the pixel pipelines, core clock, memory clock, and RAM relate
Click to expand...

lol.. erm.. well, game benchmarks are the real life situations of a card. I mean... pixel pipelines x clock, and that substracted from the square root of the surcafe size of the 3rd pin from the left upper corner of the GPU.. I mean.. cmon, that helps noone. (and yes.. im beeing sarcastic here :| )
You know.. maybe you should read up about micro controllers first.. because there is a whole lot more to it than just pipelines, clocks, and memory size. and yes.. it is that hard..

Benchmarks are the best way of looking at a cards performance.. besides that,.. why would you be interested in anythign else? since thats what the card is used for after all.
 
The details such as memory on a card shows how much info a card can hold for stuff like the virtual world it has, some recommend 128MB, some 256MB, some even more so that lets us know before buying the game how much memory is needed, now when it comes to the speed of the memory or the GPU it gives us a base line for that card and gives us a better idea of how far to go when it comes to overclocking. In order to truly understand why some cards are better than others you might want to take up computer classes in college for, oh say, 8 or so years.
 
yeah, but i remember someone telling me about pixel pipelines, and the more pixel pipelines the more data gets transfered, and that amount of data is the number of pixel pipelines times the core clcok. Does anyone know this?

I think it was Jet that said this
 
joeswm8 said:
yeah, but i remember someone telling me about pixel pipelines, and the more pixel pipelines the more data gets transfered, and that amount of data is the number of pixel pipelines times the core clcok. Does anyone know this?

I think it was Jet that said this
Click to expand...

That's a very simplified way of seeing a very general difference in performance. gpureview.com does that for you, so you can look at their stats. However, Memory speeds and other things also affect performance. The best way is to look at performance comparisons, for instance, Tomshardware.com's Video card charts. There you can see real world performance.
 
if its all about comparing on charts and graphs, then why do the statistics of the cards even matter?

is this right:

as the pixel pipelines increase as the performance of the card increases. Then after the pixel pipelines comes the core clock which as it gets higher, the performance gets higher, and after that is the memory clock and as this gets higher performance gets higher, then after that is the RAM which after the pixel pipelines, core and memory clocks is the last determiner of performance, correct?
 
from the statistics...so what do they mean??

is this right???

as the pixel pipelines increase as the performance of the card increases. Then after the pixel pipelines comes the core clock which as it gets higher, the performance gets higher, and after that is the memory clock and as this gets higher performance gets higher, then after that is the RAM which after the pixel pipelines, core and memory clocks is the last determiner of performance
 
joeswm8 said:
from the statistics...so what do they mean??

is this right???

as the pixel pipelines increase as the performance of the card increases. Then after the pixel pipelines comes the core clock which as it gets higher, the performance gets higher, and after that is the memory clock and as this gets higher performance gets higher, then after that is the RAM which after the pixel pipelines, core and memory clocks is the last determiner of performance
Click to expand...

They all overlap, and each is dependant on the other. For instance, if you had a 8800GTX core, but didn't have enough memory bandwidth, it would perform poorly. Inversly, if you had an X300 core, but had the 8800GTX's memory bandwidth, it wouldn't perform much better (if at all) than a normal X300.
 
There is no such thing as best.

You keep trying to look for absolute rules of how to tell one GPU from another performancewise. There are no such rules.
 
there are always some rules. ok then, so can you please tell me if what i said was right, and this is in a very general sense if it makes you feel better, so yes or no in the most general of senses of what to look at when determining performance level, in order of determining performance in a very general sense:

pixel pipelines
core clock (dependant on memory)
memory clock (dependant on core)
RAM
 
Last edited:
No there aren't, because you're dealing with something like 6 variables here. You can get rough estimates, which is what the benchmarks are.
 
i give up, this thread can now be closed

what i have learned today is:

Benchmarks are all and the only things to look at when trying to determine GPU performance
 
You must log in or register to reply here.
Back
Top