Wait for DX10 or Upgrade?

[james_castle]

I'm pretty much on the fence as to whether I should go with a cheaper 7 series Nvidia or wait it out with my 6600gt and buy a reasonably priced DX10 card. Anyone know when the DX10 cards will start coming out? I've been considering the 7900gt and voltmodding it for overclocking and now that neweggs got it at $200 after rebate it seems like a good time to get serious. Thanks in advance.

 

[jaguarskx]

Reasonably priced DX10 GPUs will not likely appear until about spring 2007. Both ATI and nVidia will only be producing their high end DX10 GPUs for the time being.

 

[edmetalhead]

I don't think there is ever a "right" time to update. There will always be something better and/or cheaper coming out soon. Ultimately you have to decide. If you are happy with your current card, then why update? If, however, you want something more powerful and you're getting it at a good price then by all means go ahead.

 

[james_castle]

Thanks guys, I guess a better question would of asked about when DX10 cards will be in the bang for your buck market. Anyways I'll be upgrading to Allendale in a few weeks and I wont want to live that long with this 6600gt bogging me down. So I think I've got my answer. Thanks for the responses.

 

[LordPope]

I don't think there is ever a "right" time to update. There will always be something better and/or cheaper coming out soon. Ultimately you have to decide. If you are happy with your current card, then why update? If, however, you want something more powerful and you're getting it at a good price then by all means go ahead.

in this case i can't agree... DX 9 cards will not run dx10 functionality

and dx 10 cards will slay dx 9 cards in he same code due to less overhead in the dx10 structure

so i would wait it out..... or if the 6600 is really killing u....get a mid range to hold u over like a 7600gs for 120 ish ...then ebay it later

 

[purdueguy]

Does it have to be Nvidia? X1950Pro for $199 at either Newegg or Zipzoomfly. No MIR and no shipping charges at Zipzoomfly. Best bang for the buck. Check out my analysis, here.

 

[james_castle]

Hmmm. Thats a good point. The x1950Pro seems like a good deal but I dont want to be caught in the same boat with DX10 coming up. Though Ill definately consider it. I have to go to work here soon but right now Im leaning towards the 7600GTS at newegg for $115, $100 after rebate. It benches much like a 7600GT and saves you 20 bucks. Thanks a bunch.

 

[kaotao]

I would go for a 7600GT. It'll be a huge upgrade for you now. I'd wait until both Nvidia and ATI DX10 cards are out before making that upgrade instead of jumping on the first DX10 card from Nvidia.

 

[purdueguy]

What the X1950Pro is to the 7600GTS is what the 7600GTS is to the 6600GT.

So let's see. 7600GTS has roughly twice the performance of the 6600GT while the X1950Pro has more than twice the performance of the 7600GTS. This is turn means that the X1950Pro is more than 4 TIMES the performance of the 6600GT.

Upgrade to 7600GTS for $100 for twice the performance or spend twice that for more than 4 times the performance. This would mean that the X1950Pro should sell for $300 if the price to performance gain was linear.

For the money, the X1950Pro is the way to go. DX10 won't be somewhat worth it for at least 6 months. Though, I give it 9-12 months to be value priced.

 

[james_castle]

I was unsure as to whether or not I should upgrade with a $200 card. However now I've run across an option that costs half that and still give me roughly twice the performance. From the benches I've seen the x7600gts will perform well enough until DX10 is an option. I'll use the $100 else ware. Thanks.

 

[james_castle]

Upgrade to 7600GTS for $100 for twice the performance or spend twice that for more than 4 times the performance. This would mean that the X1950Pro should sell for $300 if the price to performance gain was linear.

It depends on how you look at it. I could go with the 7600gts for $100 or I could get twice the performance for spending twice as much ($200). So it seems like the benefit to price ratio is the same and it's just a matter of choice. As of now I'm looking more toward a "hold over" type of solution so the 7600gts seems fine.

 

[purdueguy]

To keep things in true perspective you have to compare it to what you already have and that's the 6600GT.

Spending 100 dollars gives you twice the performance.

Spending 200 dollars gives you over 4 times the performance.

Double the dollars for Quadruple the performance.

 

[FallenKell]

To keep things in true perspective you have to compare it to what you already have and that's the 6600GT.

Spending 100 dollars gives you twice the performance.

Spending 200 dollars gives you over 4 times the performance.

Double the dollars for Quadruple the performance.

Actually "To keep things in true perspective", you have to compare the relative costs/benefits of the original card if you want to compare the benefits of the new cards in terms of performance gains over the current card. So in other words:

Spend 0 dollars, gives you 1x your performance
Cost/Performance ratio Infinity

Spend 100 dollars, gives you 2x your performance
Performance/Cost ratio: 1/50

Spend 200 dollars, gives you 4x your performance
Performance/Cost ration: 1/50


Wait a second, how is it that the ratio's are the same? Because You set the performance to a value, in this case 1. The ratio of Performance/Cost is (Performance / Cost) for keeping the card you have is (1x performance divided by cost of $0) = Infinity (more to the point undefined since even the limit of x*0 as x approaches infinity is still 0 and 0 is never going to equal 1).

Now for the Nvidia card, performance is 2x, and price is $100, so (2 divided by 100) = 1/50...

Now for the ATI card, performance is 4x, and price is $200, so (4 divided by 200) = 1/50...

Huh? Yes, that is right the performance/price ratio is THE SAME. So now the question becomes, do I want to spend more money or not. The ratio is linear between these two cards.

Don't do pseudo math with ratios. You can't say it is 4x performance for 2x price. It is a linear performance to price ratio between getting the Nvidia or the ATI at the prices quoted.

In other words, it is NOT "Double the dollars for Quadruple the performance", it is the same exact cost for performance. It IS double the dollars for DOUBLE the performance.

 

[ike2187]

Getting a 7900gt and volt moding it may seem like a good idea since you would be getting 7900gtx performance with a 7900gt price. But usually you need an aftermarket cooler which puts you up another 30 dollars and thats close to the price of the 7900gto. So I would say go with a x1950 pro or the 7900gto.

 

[purdueguy]

The ratios are the same because your logic is flawed.

You even proved yourself inconsistant by stating the performance is linear. By defintion, a linear system has equal slope along the ENTIRE system. If the system was linear, ALL THREE of your ratios would of been 1/50. 1/0 is infinite and not 1/50, thus proving that the system is non-linear.

Using your logic, a 50% improvement would be $50.

.5/50 = 1/100

This does not equal the 1/50 in your comment. Proving again that your logic is flawed.

RELATIVE PERFORMANCE is what's needed.

You must start off by having 100% performance that naturally goes with starting off with an initial condition. That being the 6600GT. The 7600GTS will give you 200% performance and finally the X1950Pro will give you 400% performance. See the progression here. 100%, 200%, 400%.

Subtracting the RELATIVE PERFORMANCE gives

7600GTS to 6600GT = 200%-100% = 100%
X1950Pro to 6600GT = 400%-100% = 300%

So you are really getting a 100% gain (double performance) in getting the 7600GTS while you are getting a 300% gain (quadruple performance) in getting the X1950Pro.

Now 100%/$100 = 1 and 300%/200 = 1.5

1.5 > 1, therefore the X1950Pro has a much better performance gain per dollar spent.

 

[FallenKell]

I don't know why I try to educate people, but I will...

<I>"The ratios are the same because your logic if flawed."</I>

hmm... ok if the logic is flawed how? Logic is by its very difinition, logical.

<I>"You even proved yourself inconsistant by stating the performance is linear. By defintion, a linear system has equal slope along the ENTIRE system."</I>

Did I say that that all three cards had the same cost benefit anywhere in my post? I said that "Huh? Yes, that is right the performance/price ratio is THE SAME." Immediately after I had spoken ONLY about the Nvidia and the ATI card. Those two cards have the same performance/price ratios. Of course the card you own will have the best performance/price ratio. You already own it. It has no cost. It is NEEDED however, to give the "base performance" number used to make all the calculations in the entire system.

Without a "base performance" number, I could say the ATI card has a performance of 80. And the Nvidia card has a performance of 160. But 80 "what" and 160 "what"? What is the UNIT used? What is the definition of the "unit" and what does it mean to have more of that unit.

In my above post, the UNIT of performance is the performance of the already owned graphics card (hence the "1x" performance measurement used when showing the performance/cost ratio of the already owned card, the performance of itself will be 1... I could have set it to 3248623, but that would make creating a ratio very difficult, since in the mathematical use, a ratio is normally expressed in its "lowest terms", meaning the greatest common factor between the numerator and the denominator is 1). Now since we are dealing with a denominator of 0 in the first term, there is no common factor in one sense, and there is every common factor in another sense. When dealing with this situation, because of the infinite set, it is simply best to use the number 1 when we are setting the values ourselves as we are in this case.

The performance of the owned graphics card is the standard that everyone has already been using, including you in YOUR posts, even though your math is horribly flawed...

<I>"Using your logic, a 50% improvement would be $50"</I>

What? Boy you really do not understand number ratios do you?

A 50% improvement would mean a TOTAL PERFORMANCE of 1.5. So, a card with a 50% improvement would need to cost $75 if it had the same performance/cost ratio of 1/50 that we have been talking about.

To see how the value $75 is calculated, see below:

let X = unknown cost of the card that has 50% more performance then the owned card. Now we need to find the value of this card assuming a 1/50 performance/cost ratio. We set the value of the owned card's performance to one (1). 50% MORE then 1, is 1.5. Now we solve the equation:
1.5/X = 1/50
1.5*50 = X
X = 1.5*50
X = 75


But as your poor grasp of ratios shows, you also would have had to say that we are keeping the same ratio. You see, to get a ratio, you need 2 numbers, not one. Asking about a card with 50% more improvement would still need to have a cost associated with it in order to get the ratio, or you need to state that you are using the ratio that we have to determine the cost...

<I>"This does not equal the 1/50 in your comment Proving again that your logic is flawed."</I>

No it just proves that you do not know how to calculate a number that is 50% larger then some arbitrary value "X" (or "Y" or "Z", choose your variable name). You see, .5 is actually 50% SMALLER then 1, not 50% LARGER...

<I>"RELATIVE PERFORMANCE is what's needed."</I>

Sure, RELATIVE PERFORMANCE is great. But RELATIVE to WHAT? Do you even know what RELATIVE Performance means? It means the performance of an item RELATIVE to ANOTHER item. A "base" or "known" performance.


<I>"You must start off by having 100% performance that naturally goes with starting off with an initial condition. That being the 6600GT. The 7600GTS will give you 200% performance and finally the X1950Pro will give you 400% performance. See the progression here. 100%, 200%, 400%. "</I>

I agree entirely. In fact, my post had the same exact thing. In your above sequence, you set the performance of the 6600GT as 100%. In my post I didn't use percentages, I used their numerical equivlent. So, instead of 100% I used, one (1), and instead of 200%, I used two (2), and instead of 400%, I used four (4). You can see that in the 3 paragraphs that are immediately following where I first show the Performance/Cost ratio's (i.e. the paragraph that starts with "Wait a second").


<I>"Subtracting the RELATIVE PERFORMANCE gives

7600GTS to 6600GT = 200%-100% = 100%
X1950Pro to 6600GT = 400%-100% = 300%

So you are really getting a 100% gain (double performance) in getting the 7600GTS while you are getting a 300% gain (quadruple performance) in getting the X1950Pro."</I>

YES! We have a WINNER!!! DING DING DING! You just said it, the X1950Pro is 4x more powerful then the 6600GT! It is NOT 4x more powerful then the 7600GTS. It is NOT "double the dollars for quadruple the performance". Because the Quadruple the performance is Quadruple the 6600GT. The dollars of the 6600GT is ZERO (0), zip, nada, none, ziltch.



<I>"Now 100%/$100 = 1 and 300%/200 = 1.5

1.5 > 1, therefore the X1950Pro has a much better performance gain per dollar spent."</I>

huh? What speudo logic math did you just do. THINK about what you just did. What are the 100% and 300% numbers you have right there? They are quite literally the "performance of 7600GTS - performance of 6600GT" and "performance of X1950Pro - performance of 6600GT".

So now you are creating the performance cost ratio of "performance of 7600GTS - performance of 6600GT" over the price of the 7600GTS. And "performance of X1950Pro - performance of 6600GT" over the price of the X1950Pro.

What meaning will this new statistic even give someone? If you are meaning this value to represent the performance/cost ratio of the 7600GTS you have the wrong numbers in there, since you have just calculated the performance of the 7600GTS minus the performance of a 6600GT over the cost of a 7600GTS. That is <B>not</B> the performance cost ratio of a 7600GTS. The same is true with the X1950Pro.

The very definition of the performance/cost ratio of an item means that the performance of the item AND ONLY THAT ITEM is in the performance value, and the cost of that item AND ONLY THAT ITEM is in the cost value.

Your problem with this from the beginning is how to set a value to the performance of the items in question and how that relates to the cost of said item. The costs are a known value. The costs are $100 for the 7600GTS, $200 for the X1950Pro and $0 for the 6600GS. The performance value is a relative performance value relative to a known value (the known value being the performance of the 6600GS). The RELATIVE PERFORMANCE/COST ratios are then, the "relative performance number relative to the 6600GS" over "the cost of the item".

So, the relative cost performance ratio of the 6600GS is: 100%/$0

The relative cost performance ratio of the 7600GTS is: 200%/$100

The relative cost performance ratio of the X1950Pro is: 400%/$200

Now to compare frations, you set the denominators the same. Since you can not set the denominator of the 6600GS to anything other then zero it can not be compaired to the other two items (well, you can, but to do that you need to multiply their numerators with zero, and we all know that anything multiplied by zero equals zero, their values would be 0/0 for BOTH the 7600GTS and X1950Pro...)

So we can't actually compair the performance/cost ratio of the 6600GS against the 7600GTS OR the X1950Pro. But we CAN compair the performance/cost ratio of the 7600GTS and the X1950Pro.

To get their denominators the same, we can either multiply the 7600GTS by 2, or we can divide the X1950Pro by 2. We can also start by reducing the values we already have and then looking to see if their denominators are equal after reduction.

The 7600GTS value is 200%/$100 = 100%/$50 (or 1/$50) = 10%/$5 = 2%/$1

The X1950Pro value is 400%/$200 = 200%/$100 = 100%/$50 (or 1/$50) = 10%/$5 = 2%/$1

hmmm... Certainly looks like the relative performance/cost ratio of the 7600GTS and the X1950Pro are the same.

(it helps to do valid math)

 

[jabberwhacky]

Wow serious nerd fight goin on.

I too have a 6600GT and recently have been reading over these forums trying to find out what I want to upgrade to. This thread has helped a lot, and I wanted to get some specifics. You guys are saying that the performance would be doubled in a 7600GTS and quadrupled in an X1950Pro, but I'd like some more info on these two cards. Besides the fact that the X1950 is double the "performance," are there any other features it touts over the 7600GTS? I've had my card for almost exactly 2 years now and I'd be OK with paying about 100 bucks for a 2x upgrade that I ditch in 9-12 months for a DX10 card, but I wouldn't really want to ditch the $200 card so soon. That's why I ask if there's any other features in the X1950 that might lengthen it's lifespan.

edit: holy crap my account is old lol

 

[james_castle]

I just finished ordering my 7600GTS. Some people have been complaining about the HSF being a bit noisy but my VF700 should take care of that. I'll post some before and after 3dmark scores once it's in.

 

[purdueguy]

Go back to school. You're logic is still wrong. You can't claim linearity and then assume something esle in how to calculate the cost of what a card would give.

YOU MADE THE ASSUMPTION of a 1/50th RATIO. YOU MADE THE ASSUMPTION OF LINEARITY.

Don't do pseudo math with ratios. You can't say it is 4x performance for 2x price. It is a linear performance to price ratio between getting the Nvidia or the ATI at the prices quoted.

YOU CAN'T GO BACK AND NOW RECALCULATE IT TO 1/75th. BY DOING THAT YOU'VE JUST PROVEN IT'S NOT LINEAR.

IS IT QUADRUPLE PERFORMANCE FOR DOUBLE THE PRICE. When talking about double, triple, quadruple, etc., it is different than 100%, 200%, 300%, etc.

By definition:

Double = 100% performance gain
Triple = 200% performance gain
Quadruple = 300% performance gain

YOU CANNOT TAKE 200%/$100 & 400%/$200. It's not a 200% gain or 400% gain.

YES! We have a WINNER!!! DING DING DING! You just said it, the X1950Pro is 4x more powerful then the 6600GT! It is NOT 4x more powerful then the 7600GTS. It is NOT "double the dollars for quadruple the performance". Because the Quadruple the performance is Quadruple the 6600GT. The dollars of the 6600GT is ZERO (0), zip, nada, none, ziltch.

Exactly, 4x more powerful or 300% percent gain in performance.

It's 100% GAIN and 300% GAIN. Not 200% and 400% as you are putting it.

THEREFORE, I WILL RESTATE:

Formula = Performance Difference/Cost Difference

huh? What speudo logic math did you just do. THINK about what you just did. What are the 100% and 300% numbers you have right there? They are quite literally the "performance of 7600GTS - performance of 6600GT" and "performance of X1950Pro - performance of 6600GT".

THIS IS EXACTLY RIGHT AND THIS IS WHAT IS BEING COMPARED. EVERYTHING NEEDS TO BE NORMALIZED TO THE PERFORMANCE OF THE 6600GT.

So now you are creating the performance cost ratio of "performance of 7600GTS - performance of 6600GT" over the price of the 7600GTS. And "performance of X1950Pro - performance of 6600GT" over the price of the X1950Pro.

One critical thing missing from your equations. Once again your missing the point that EVERYTHING is a difference. It's a comparison between the "performance of 7600GTS - performance of 6600GT over the price of the of the 7600GTS - price of the 6600GTS" to the "performance of X1950Pro - performance of 6600GT over the price of the of the X1950Pro - price of the 6600GTS"

It just so happens that the cost of the 6600GT is a sunk cost and is zero masking its effect upon the above equation.

7600GTS vs 6600GT

(200%-100%)/($100-$0) = 100% PERFORMANCE GAIN/$100 = 1% PERFORMANCE GAIN PER DOLLAR SPENT

X1950Pro vs 6600GT

(400%-100%)/($200-$0) = 300% PERFORMACNE GAIN/$200 = 1.5% PERFORMACNE GAIN PER DOLLAR SPENT

YOU HAVE TO TAKE THE DIFFERENCE BETWEEN THE RELATIVE CARDS THAT YOU ARE ANALYZING.

You educate people? Please, go back to grade school.

Here's two equations. y = 2x and y = x^2

Plug in the number 2. y = 4 & y = 4

Now, are these two equations the same just because one point was picked and they gave the same answer. Of course they are not. You are trying to claim the cost/performance is like the equation y = 2x and it is not. The cost/performacne is like the equation y = x^2.

Here's a plot of the performance / cost of the 3 cards.



If you took calculus you will know that if you integrate an equation and plug in the boundary conditions, you will get the area underneath the curve.

Heck, just looking at the area under the curve, you can automatically see that the area between the cost of 100 & 200 is over twice the size as the area between 0 & 100. This by itself shows that you get more performance per dollar spent for the X1950Pro than the 7600GTS.

The curve is given approximately by

y = 0.005x^2 + 0.5x + 100

Integrating you get

y = 0.005x^3/3 + 0.25x^2 + 100x + constant

Plugging in the intial conditions of cost of 100 and 0 (7600 GTS vs 6600GT)

y ~ 14167 + some constant

Plugging in the intial conditions of cost of 200 and 0 (X1950Pro vs 6600GT)

y ~ 43333 + some constant

As you can see, 43333 is roughly 3 times bigger than 14167. The difference is 29,166. The constants cancel each other out.

This means that the difference in the area between 100 & 200 and 0 & 100 is twice the size as the area between 0 & 100. Thus PROVING that you get more performance value in the X1950Pro than the 7600GTS.

If the relative performance/cost ratio were the same then the difference in the area between 100 & 200 and 0 & 100 WOULD BE THE SAME as the area between 0 & 100. As shown, it's not the same.

Once again, RELATIVE PERFORMANCE and RELATIVE PRICE is what's necessary. Everything needs to be compared back to the 6600GT.

That is the 7600GTS vs 6600GT and the X1950Pro vs. 6600GT. Then compare the final 2 answers.

You cannot take the comparison of the 7600GTS vs 6600GT and the X1950Pro vs 7600GTS then compare these results. It's comparing apples to oranges.

If my little calculus demonstration doesn't prove my point to you, go take an ECON 101 class with basic calculus.

 

[JAYDEEJOHN]

I know a math problem when I see one.Hows this one?Say card A has 20 fps,card B costs 100 dollars and has a fps of 40,and finally we have card C which has a fps of 80 at a cost of 200 dollars Hmmmmmmm Got a problem?Purdue ya got my vote as anyone can see that 20 taken from 80 leaves 60......

 

[purdueguy]

Clear & Concise. Very well done.

I've been told to "Keep it simple, keep it simple". I must remember to do this. Then again, where would the fun be in that?

I was thinking about using FPS to prove the point but I couldn't resist when arguements against me were illogical and meaningless. And adding a little Calculus to the argument at the end was something I was only too trilled to provide. Ah, the days when doing multi-variable Calculus was the only thing I had to worry about.

FallenKell, is this simple enough for ya?

 

[JAYDEEJOHN]

As enthusiasts we focus on performance and cost,sometimes it does get blurry,but if youre like me and have some money.......FallenKell no disrespect dude,lets hope these low dollars = high performance when it comes to dx10 so we all can afford one,tho as crazy as I am Ill probly just buy one like alot of us cause Ill Just HAVE to have it no matter the cost

 

[FallenKell]

Go back to school. You're logic is still wrong. You can't claim linearity and then assume something esle in how to calculate the cost of what a card would give.

YOU MADE THE ASSUMPTION of a 1/50th RATIO. YOU MADE THE ASSUMPTION OF LINEARITY.

Where? I never made the assumption of a 1/50 ratio anywhere except for explaining the cost of a card that has 50% more perfromance.

The 1/50 ratio were calculated in the reduction of terms from the performance/cost calculations in both my posts. Specifically:

The relative cost performance ratio of the 7600GTS is: 200%/$100

The relative cost performance ratio of the X1950Pro is: 400%/$200

followed by their reductions:

The 7600GTS value is 200%/$100 = 100%/$50 (or 1/$50) = 10%/$5 = 2%/$1

The X1950Pro value is 400%/$200 = 200%/$100 = 100%/$50 (or 1/$50) = 10%/$5 = 2%/$1

The 1/50 was used when the value of performance was in terms of Y times the performance of 6600GS. I changed to percentage of performance of 6600GS when you failed to grasp the fact that 1 = 100%, 2 = 200%, and 4 = 400%.

If you look above, where in the ratio is there a flaw? What part of "performance of card X in terms of card Y divided by cost of card X" is incorrect in the standard known definition of performance/cost ratio? If we want to go to actual performance values of the cards themselves we could, but then that would entail us collecting a group of synthetic and practical tests and finalizing on what we concur as to the actual performance value of a card (something that review sites have been trying to do for years and years). We simply went on the decision that the 7600GTS is 2x faster then a 6600GS and that a 1950Pro is 4x faster then a 6600GS. You can use those numbers to create a statistic for performance values, and that is what we used. Once we have a performance value and a cost associated, you CALCULATE the performance/cost ratio of the product. It isn't my fault that 200%/$100 = 1/50, and that 400%/$200 ALSO = 1/50, that is just the facts of life and the system.

YOU CAN'T GO BACK AND NOW RECALCULATE IT TO 1/75th. BY DOING THAT YOU'VE JUST PROVEN IT'S NOT LINEAR.

I didn't recalculate ANYTHING other then the price of a card that has 50% more power then a 6600GS which assumes the same performance/cost ratio as the other two cards.

I DID NOT recalculate the PERFORMANCE/COST RATIO of a card that is 50% more power. I calculated IT'S PRICE! I USED the cost perfromance ratio of 1/50 to CALCULATE the price.

IS IT QUADRUPLE PERFORMANCE FOR DOUBLE THE PRICE. When talking about double, triple, quadruple, etc., it is different than 100%, 200%, 300%, etc.

In what world of mathematics do you live in? 100% is 1x performance. 200% is 2x performance or double. 300% is 3x performance or TRIPLE. <B>IT IS EXACTLY THE SAME</B> as using 100% or 200% or 300% or 400%, etc. You learn this in grade school math when you learn that you can convert a percentage to a numerical value by shifing the decimal place two places to the left (in other words, divide by 100), and you can convert a value into a percentage by shifting the decimal place two places to the right (multiply by 100).

By definition:

Double = 100% performance gain
Triple = 200% performance gain
Quadruple = 300% performance gain

umm... go see basic math 101...

For basic math with fractions and percentages read:
http://mathforum.org/dr.math/faq/faq.fractions.html

For a practice assignment you can go to the following:
http://mathforum.org/escot/fishfarm2.html

For more help on fractions as a ratio please read the following:
http://mathforum.org/library/drmath/view/58014.htm

As for percentages larger the 100%:
http://mathforum.org/library/drmath/view/58166.html
http://mathforum.org/library/drmath/view/62789.html

YOU CANNOT TAKE 200%/$100 & 400%/$200. It's not a 200% gain or 400% gain.

Yes, it is. It is exactly a 200% the performance or 400% the performance. We want the performance of the cards themselves. That is what 2 times the performance of something is and it is what 4 times the performance of something is. They are exactly 200% and 400% respectively.

Exactly, 4x more powerful or 300% percent gain in performance.

It's 100% GAIN and 300% GAIN. Not 200% and 400% as you are putting it.

THEREFORE, I WILL RESTATE:

Formula = Performance Difference/Cost Difference

AHHA! We have the "AH HA" moment. Your statistic that you are creating the the PERFORMANCE GAIN over cost. Not the PERFORMANCE over cost. This is entirely different and completely different equation. You are not longer looking at the performance/cost ratios you are looking at the gain ratios.

huh? What speudo logic math did you just do. THINK about what you just did. What are the 100% and 300% numbers you have right there? They are quite literally the "performance of 7600GTS - performance of 6600GT" and "performance of X1950Pro - performance of 6600GT".

THIS IS EXACTLY RIGHT AND THIS IS WHAT IS BEING COMPARED. EVERYTHING NEEDS TO BE NORMALIZED TO THE PERFORMANCE OF THE 6600GT.

Actually this is only when you are looking for the performance gain per cost, not the performance per cost.

So now you are creating the performance cost ratio of "performance of 7600GTS - performance of 6600GT" over the price of the 7600GTS. And "performance of X1950Pro - performance of 6600GT" over the price of the X1950Pro.

One critical thing missing from your equations. Once again your missing the point that EVERYTHING is a difference. It's a comparison between the "performance of 7600GTS - performance of 6600GT over the price of the of the 7600GTS - price of the 6600GTS" to the "performance of X1950Pro - performance of 6600GT over the price of the of the X1950Pro - price of the 6600GTS"

It is only a difference if you are calculating for a performance gain per cost. Not the performance/cost of a product.

It just so happens that the cost of the 6600GT is a sunk cost and is zero masking its effect upon the above equation.

You are correct. It is a sunk cost.

YOU HAVE TO TAKE THE DIFFERENCE BETWEEN THE RELATIVE CARDS THAT YOU ARE ANALYZING.

You educate people? Please, go back to grade school.

Here's two equations. y = 2x and y = x^2

Plug in the number 2. y = 4 & y = 4

Now, are these two equations the same just because one point was picked and they gave the same answer. Of course they are not. You are trying to claim the cost/performance is like the equation y = 2x and it is not. The cost/performacne is like the equation y = x^2.

Correct.

Here's a plot of the performance / cost of the 3 cards.

INCORRECT. That plot is the plot of the PERFORMANCE GAIN/Cost of the 3 cards.

If the relative performance/cost ratio were the same then the difference in the area between 100 & 200 and 0 & 100 WOULD BE THE SAME as the area between 0 & 100. As shown, it's not the same.

Yes, I agree, but you are using the wrong statistic for the performance/cost ratio. You are using a statistic for the performance gain/cost, which an entirely different animal.

Once again, RELATIVE PERFORMANCE and RELATIVE PRICE is what's necessary. Everything needs to be compared back to the 6600GT.

We already are comparing everything back to the 6600GS since we defined the performance value in TERMS OF the performance of a 6600GS.

You cannot take the comparison of the 7600GTS vs 6600GT and the X1950Pro vs 7600GTS then compare these results. It's comparing apples to oranges.

If my little calculus demonstration doesn't prove my point to you, go take an ECON 101 class with basic calculus.

Actually you need to go take a few classes in probability and statistics which teach you how to form and name a proper statistic.

And in fact I would never recommend someone take an ECON 101 class to gain the basics of calculus. I would tell them to take a basic calculus class to learn calculus (this from an engineer).

So the basics of this entire argument can be summed up real simple. Poor communications. I made the assumption that someone who is talking about statistics and ratios actually knew how to properly name and define said ratio. It took you until this final post to actually define the statistic that you were creating. The second you did that, the "ah ha" moment occurred.

 

[zenmaster]

I just have to laugh at all the complex math calculations going on.

Just go buy the card you need to play the games you want.

Even if you are a grown adult, you can walk the neighborhood one Saturday and offer to rake leaves, wash cars, etc... etc.. etc... and bring in $100 easy and likely more. (P.S., dont forget to declare it :>>>)

If you are too lazy to do these things, I would suggest saving your money because even if you have it now, it will soon be gone :>>>

 

[purdueguy]

Your links are completely pointless to this discussion.

And you're

AHHA! We have the "AH HA" moment. Your statistic that you are creating the the PERFORMANCE GAIN over cost. Not the PERFORMANCE over cost. This is entirely different and completely different equation. You are not longer looking at the performance/cost ratios you are looking at the gain ratios.

moment. Go back and read my initial statement.

Upgrade to 7600GTS for $100 for twice the performance or spend twice that for more than 4 times the performance. This would mean that the X1950Pro should sell for $300 if the price to performance gain was linear.

Take a quick look. I've been talking about GAIN this entire time which is what's relevant. You are the one talking about PERFORMANCE OVER COST which is completely irrelevant and incorrect to this discussion.

Who cares? You want performance GAIN. Best bang for the buck. Simple as that.

Quote:

Here's a plot of the performance / cost of the 3 cards.



INCORRECT. That plot is the plot of the PERFORMANCE GAIN/Cost of the 3 cards.

And your wrong on my graph. It is performance / cost. 0 cost for 6600GT give 100% performance, $100 gives you 200% and $200 will give you 400%.

If it was a performance GAIN, spending 0 dollars would give you 0 performance gain, $100 gives you 100% and $200 gives you 300%. Is that what's on the graph? I don't think so.

I will repeat, by definition:

Double = 100% performance GAIN
Triple = 200% performance GAIN
Quadruple = 300% performance GAIN

Tell me, what is a 1x in performance GAIN. Zero, zilch, nadda.

If you go from 100% to 200%. What's the difference? A 100% performance GAIN.

Everything is relative. You have to compare GAIN in one card to another and that's exactly what you are NOT doing.

Yes, it is. It is exactly a 200% gain or a 400% gain. We want the performance of the cards themselves. That is what 2 times the performance of something is and it is what 4 times the performance of something is. They are exactly 200% and 400% respectively.

If you believe this, then I want to know what engineering field you are in so I can stay away from what you are producing. Ever hear of safety margin. If something has a safety margin of 0.25, that means it's actually 125% percent of what is really needed. You'd probably multiply by 25% and the damn thing would break.

Do you understand what a 200% GAIN is? It means it's 3x faster, stronger or whatever you are measuring.

Just like jaydeejohn put it so simply:

Assume card A (6600GT) has 20 FPS, card B (7600GTS) has 40 FPS, and card C (X1950Pro) has 80FPs.

This means that Card C has double the performance of card B which in turn has double the performance of card A. Thus, Card C has quadruple the performance.

Now, do the simple math. Obviously, you didn't understand Calculus in the first place which proved my point.

You get a 20 FPS advantage of using the 7600GTS over the 6600GT for a cost of $100.

This is costing you $5 per extra frame.

You get a 60 FPS advantage of using the X1950Pro over the 6600GT for a cost of $200.

This is costing you $3.33 per extra frame.

Now tell me, which card offers the best bang for the buck?