manipulate three data sets. price performance and third value

G

guanza

New member
Joined
Apr 28, 2021
Messages
7
hello
I'm not sure if data set is the correct term so correct me if I'm wrong.

Let's say I have three data sets. Price, performance and Watts.
If I divide price/performance I get a quotient. I'd like to factor in Watts too though. How would I go about it?

this wouldn't be done through a weight function right? https://en.wikipedia.org/wiki/Weight_function
I don't prioritize Watts or the performance.

I've multiplied and divided the quotient with the Watts but each of the list looks a bit different.
 
guanza said:
hello
I'm not sure if data set is the correct term so correct me if I'm wrong.

Let's say I have three data sets. Price, performance and Watts.
If I divide price/performance I get a quotient. I'd like to factor in Watts too though. How would I go about it?

this wouldn't be done through a weight function right? https://en.wikipedia.org/wiki/Weight_function
I don't prioritize Watts or the performance.

I've multiplied and divided the quotient with the Watts but each of the list looks a bit different.
Click to expand...
What is your objective? It's unclear.
 
guanza said:
hello
I'm not sure if data set is the correct term so correct me if I'm wrong.

Let's say I have three data sets. Price, performance and Watts.
If I divide price/performance I get a quotient. I'd like to factor in Watts too though. How would I go about it?

this wouldn't be done through a weight function right? https://en.wikipedia.org/wiki/Weight_function
I don't prioritize Watts or the performance.

I've multiplied and divided the quotient with the Watts but each of the list looks a bit different.
Click to expand...
I suppose you are looking at an electrical machine. Watts is a measure of performance. How are defining Performance in your data-set?

Please show us the data set you want to analyze.

Please show us what you have tried and exactly where you are stuck.

Please follow the rules of posting in this forum, as enunciated at:


Please share your work/thoughts about this problem
 
Thanks for answering.
The topic is GPUs.

I have $430/130 fps = $3.30/fps
and $550/150 fps = $3.66/fps

The card uses a certain amount of watts. let's say the avg is 180 and 230 respectively.
How would I factor this into the price/performance. Because as I've said multiplying 180 by $3.30 or dividing 180 by $3.30, in the whole, gives two very distinct outcomes. A card that was at top20 all of sudden is at top6.
 
guanza said:
Thanks for answering.
The topic is GPUs.

I have $430/130 fps = $3.30/fps
and $550/150 fps = $3.66/fps

The card uses a certain amount of watts. let's say the avg is 180 and 230 respectively.
How would I factor this into the price/performance. Because as I've said multiplying 180 by $3.30 or dividing 180 by $3.30, in the whole, gives two very distinct outcomes. A card that was at top20 all of sudden is at top6.
Click to expand...
What you're trying to do is the weighted average. You'd need all of your "data" to be in the same units.
 
alright. So it indeed is the link I pasted above. ? https://en.wikipedia.org/wiki/Weight_function

Would I need to divide $/fps then do watts/fps, and then add them? If correct, is it imperative to multiply one of them?
Would I need to prioritize watts aka *2 the watts/fps column?

also what's with the quotes and data ? is it because it's so limited or what
 
alright so I assume it's this? https://en.wikipedia.org/wiki/Weighted_arithmetic_mean

Either way, I color-coded the columns and extracted it visually. At least I think I am.
I still applied an additional column though with: (($/fps)+(watts/fps)) which I also color-coded.

I'd still appreciate formal feedback on the veracity of my adventure.
 
guanza said:
Thanks for answering.
The topic is GPUs.

I have $430/130 fps = $3.30/fps
and $550/150 fps = $3.66/fps

The card uses a certain amount of watts. let's say the avg is 180 and 230 respectively.
How would I factor this into the price/performance. Because as I've said multiplying 180 by $3.30 or dividing 180 by $3.30, in the whole, gives two very distinct outcomes. A card that was at top20 all of sudden is at top6.
Click to expand...
Figuring out a sensible goal function is one of the most typical challenges in engineering and applied sciences.
If I understand correctly your preference is for lower power units. If that is the case you can try factoring the power "drain" into the price of the card. E.g., if you expect to use the card for 2000 hours and your electricity cost is $0.25/kWh then the difference of 100 watts will correspond to 2000 x 0.25 x 0.1 = $50 of additional price.
 
guanza said:
Thanks for answering.
The topic is GPUs.

I have $430/130 fps = $3.30/fps
and $550/150 fps = $3.66/fps

The card uses a certain amount of watts. let's say the avg is 180 and 230 respectively.
How would I factor this into the price/performance. Because as I've said multiplying 180 by $3.30 or dividing 180 by $3.30, in the whole, gives two very distinct outcomes. A card that was at top20 all of sudden is at top6.
Click to expand...
This is not directly related to your question, but nevertheless: is the ratio of price per FPS always a sensible criteria? Say card C2 is twice as powerful as C1 and costs twice as much. But if you install 2 C1 cards will your algorithms run as fast as on one C2 card? Or will they spend a sizeable overhead on inter-card communications? This depends on algorithms, but in most cases 2 C1 cards will run slower.
 
>If I understand correctly your preference is for lower power units.
my goal is price/performance and just any additional metric really. I don't like to overpay plus I'll be rewarding the better product.
blamocur said:
But if you install 2 C1 cards will your algorithms run as fast as on one C2 card?
Click to expand...
Well yes but I'm calculating strictly single GPU ie as consumer not data center. SLI/nvlink for desktop use has been dead for a while now. But thanks for the tip.
E.g., if you expect to use the card for 2000 hours and your electricity cost is $0.25/kWh then the difference of 100 watts will correspond to 2000 x 0.25 x 0.1 = $50 of additional price.
Click to expand...
Oh yea that's a good way of looking at it. thanks
The thing is I thought I could somehow idk include this in a nice package. Because there's also the question of one or two fans. If a one-fanned version costs less, how much less it would have to cost for me to consider buying it given the noise it will produce etc.
 
You must log in or register to reply here.
Top