Originally posted by: garritynet
ANYTHING that gives you decent performance (crysis on low at 40fps) for $1 or .01 is NOT w/o value.. stop your weak attempts at warping the truth, you'd make a terrible lawyer.
Who is talking about running Crysis at 40fps for $1. As I recall we were talking about running a game a 5 fps.
Everyone else seemed to be able to read it... you're saying you can't??? once again, stop your weak attempts at warping the truth.
It is very poorly composed. I think that you would be able to communicate your ideas more effectively with some editing and abridging of your original post.
Ok, another warping of truth! thanks garrity, love the support, man! LIKE I SAID: compare it to the average forum post and I think the standard of writing is significantly higher, what standards are you comparing it to, professional publications???
Once again, it plays a new game (crysis warhead) at more than 40fps on low. Stop your weak attempts at warping the truth.
What are you talking about here? I was referring to this comment by you:
SO, if card B gets an ethereal 45fps, which is 9x the performance of card A we might assume that it's running at the highest spec, in which case we can see that Low is gen 1 running at 1x speed, we multiply that by 2x2x2 and you get a factor of 8, therefore, you can take the 5 and multiply it by 8 and... voila! 40fps at lowest settings!
In this comment you did some bizarre math where you figured that this 'card A', the one dollar card with 5fps will now run Crysis at 40fps. I mean you are literally saying that because 2^3 multiplied by 5 equals 40 then the imaginary card that I4AT created will also run the most graphic intensive game available at 40fps on low. On what reality is this grounded?
You have to be able to READ first of all, but since you say you can't read my writing, I'm not surprised you didn't get it. My point was that IF we assume that card B can run crysis on very high at 45fps for $100 (which is certainly possible at lower rez's, especially a custom config, considering that a $100 can buy you a 9800gt or 4830; I can even run crysis on what looks very similar to High on my 8400m gt laptop card with 16 shader processors), you can therefore run the $1 card on low at least at 40fps assuming that each setting between low, medium, etc. corresponds to an increase in one generation between settings. So, yes, I would say that this is a very accurate assessment, that this entirely hypothetical card can run crysis/warhead on low for $1 at 40fps relative to card B.
You really seem to be missing the point about moore's law-- you constantly are getting large increases in performance for free, if you go ahead of that curve, you are always losing (UNLESS, like I said, you have certain specific NEEDS which must be met, these would outweigh the price/performance benefits and is at the discretion of the individual buyer to decide the best value for THEM
No one buys video cards just so they can stay in the optimal position of a curve some guy on the internet made up. We all have 'specific needs' which 'outweigh' any imagined 'benefits' that you might be referring to.
So Moore's law is imaginary now?! THIS IS THE LAST TIME I AM GOING TO RESPOND TO THIS POINT THAT YOU KEEP MAKING OVER AND OVER AND OVER AGAIN--- IT'S A TOOL TO HELP YOU, A THE VERY LEAST, SHOW YOU WHERE YOU STAND ON THE CURVE SO YOU DON'T GET SUCKED INTO POOR VALUE BASED ON SHODDY RECOMMENDATIONS, PRETTY GRAPHS; IT IS OBJECTIVE, RATHER THAN SUBJECTIVE AND AT LEAST SUGGESTS THE CONSUMER HOW TO AVOID CRAPPY 1:1 PERF. RATIOS; YOU THEN HAVE AN INFORMED CHOICE WHICH YOU LACKED PRIOR TO THIS METHOD. A LITTLE APPRECIATION WOULD BE APPRECIATED. WHATEVER, JUST LOOK AT MY RESPONSE TO LOUISSSS A FEW POSTS ABOVE.
(UNLESS, like I said, you have certain specific NEEDS which must be met, these would outweigh the price/performance benefits and is at the discretion of the individual buyer to decide the best value for THEM
Originally posted by: garritynet
That would be card A. It now runs COD3 at 100fps. Its Call of Duty 3 so multiply 3 X 5 to get 15. Now take away 5 for the AF and AA. So now we have 10. 10^2 is 100. See 100fps. I suggest getting Card A which should be on sale for $1 giving you 10fpd.
Originally posted by: LOUISSSSS
1. so what card according to your math would give my system the best performance/$$ to play cod3 @ 16x10 @ 8xaa 8xaf
see the system in my signature (minus the video card of course) for reference.
2. same question as above, but what if my system had a Pentium D 3.2ghz?
3. What card according to your math would give best performance/$$ to play cod3 @ 16x10 @ 0xaa 0xaf
Originally posted by: djayjp
ONCE AGAIN, the purpose of this guide is not to maintain a current price/performance guide, those are just EXAMPLES there to help you get familiar with the method for determine optimal price/performance... if I were to keep it current, it would have to be nearly continuously updated (due to the flux in prices and new cards upsetting ratios). What makes it so 'Ultimate' compared to any other guide is that it never becomes obsolete (the examples practically do though overnight!), that is until moore's law does!
Originally posted by: djayjp
The beauty of my method also is that instead of doing this uber detailed analysis looking at game after game, you can just look at the overall optimal price/performance ratio if you want-- so, due to the insane ratio of the 4670 for only $60 after rebate, I can't recommend going any lower than that (the 9500gt costs $5 less and offers almost 3x worse performance!). For your reference, in the chart you can see that the 4670 performs 2x as fast as a 7900gtx! Gotta feel sorry for those suckers who spent 10x that for one of those then and gets half the performance! Again, the beauty of optimal price/performance! And it's slightly faster than an 8800gts 640mb! (which costs more!)
Originally posted by: djayjp
Originally posted by: LOUISSSSS
1. so what card according to your math would give my system the best performance/$$ to play cod3 @ 16x10 @ 8xaa 8xaf
see the system in my signature (minus the video card of course) for reference.
2. same question as above, but what if my system had a Pentium D 3.2ghz?
3. What card according to your math would give best performance/$$ to play cod3 @ 16x10 @ 0xaa 0xaf
Thanks for the question!
1) Unfortunately, I can't give you DIRECT numbers of these newer cards being run on COD3... COD4 is another matter entirely, however. The best approach would be to estimate based on statistics the average decrease in performance of vid cards going from COD3 to COD4. Then adapt that over to the new benchmarks using the new cards and COD4 at that rez.
However, COD4 is a very non-challenging game to run even at #1's settings, fully maxed out (and of course COD4 is significantly more challenging to run than COD3). At the link below, you can see that the 4670 achieves buttery smooth 57fps at 16x10 w/ 4xaa and 8xaf.
The beauty of my method also is that instead of doing this uber detailed analysis looking at game after game, you can just look at the overall optimal price/performance ratio if you want-- so, due to the insane ratio of the 4670 for only $60 after rebate, I can't recommend going any lower than that (the 9500gt costs $5 less and offers almost 3x worse performance!). For your reference, in the chart you can see that the 4670 performs 2x as fast as a 7900gtx! Gotta feel sorry for those suckers who spent 10x that for one of those then and gets half the performance! Again, the beauty of optimal price/performance! And it's slightly faster than an 8800gts 640mb! (which costs more!)
http://www.tomshardware.com/ch...f-Duty-4-v1-6,745.html
As far as the CPU, I can't say for sure cuz the D ain't so hot . But if I HAD to guess, I'd say that it should be alright.
see here and use the 9600gt as a stand-in for the 4670 (god, i love tom's hardware ):
http://www.tomshardware.com/re...pu-upgrade,1928-7.html
you can see that an e2160 at 1.8ghz (which is a core 2 duo with half the cache) can run it at 65.5 at 16x10, 4xaa, 8xaf. I would imagine that a D is about half as fast as an E series (CLOCK FOR CLOCK), so it should be able to run it fine at 3.2GHz
if we look at unreal tournament 2004:
http://www.tomshardware.com/ch...urnament-2004,398.html
Here you can see that a D 840 (3.2ghz) scores 46.4fps in this CPU based bench, compared to an e2160 (1.8ghz) which scores 53fps. SO, the d 840 is 86% the speed of the d2160. If we then go back to the previous benchmark of whether or not a cpu or gpu upgrade is worth it, we can see that in cod4, the e2160 scores 65.5fps, therefore, we can roughly estimate performance of your CPU in the cod4 bench to be 56fps! They're both dual core processors, so the game's enhanced multithreaded nature over UT2K4 shouldn't be too much of an issue. This should be fairly accurate. So, in conclusion, we can see that your CPU should be well matched up with the 4670 in cod4 (somewhat simplified/estimated: 56fps for CPU vs 57fps for GPU)! I just wouldn't try to run supreme commander on that CPU, in that case the CPU will be the bottleneck... UPDATE: I just looked it up, in supreme commander, the D 840 scores 14.5fps... ouch! My previous recommendation still stands though for COD4
For your reference, in the chart you can see that the 4670 performs 2x as fast as a 7900gtx! Gotta feel sorry for those suckers who spent 10x that for one of those then and gets half the performance!
Originally posted by: razor2025
Originally posted by: djayjp
ONCE AGAIN, the purpose of this guide is not to maintain a current price/performance guide, those are just EXAMPLES there to help you get familiar with the method for determine optimal price/performance... if I were to keep it current, it would have to be nearly continuously updated (due to the flux in prices and new cards upsetting ratios). What makes it so 'Ultimate' compared to any other guide is that it never becomes obsolete (the examples practically do though overnight!), that is until moore's law does!
But your methods are CLEARLY faulty to begin with. First, where do you find your pricing information? Because it's obviously either biased or lack effort in shopping around, as I pointed out in 4670 vs 9600GSO debate. I'd still love to find a link to your $61 4670 that doesn't involve a rebate. With faulting pricing information, your calculations are completely thrown off. Also your multiplier for performance/$ makes absolutely no sense. It's arbitrary. You can assign a high multiplier to make one card look incredibly "better" value, but the truth is to the contrary. This is even worse when you're shopping around for one card and not the other. You'll also need data from all common resolutions. Some card's performance shifts dramatically. You'll need to separate calculations based on data from different resolution, at the very least.
Originally posted by: razor2025
Originally posted by: djayjp
The beauty of my method also is that instead of doing this uber detailed analysis looking at game after game, you can just look at the overall optimal price/performance ratio if you want-- so, due to the insane ratio of the 4670 for only $60 after rebate, I can't recommend going any lower than that (the 9500gt costs $5 less and offers almost 3x worse performance!). For your reference, in the chart you can see that the 4670 performs 2x as fast as a 7900gtx! Gotta feel sorry for those suckers who spent 10x that for one of those then and gets half the performance! Again, the beauty of optimal price/performance! And it's slightly faster than an 8800gts 640mb! (which costs more!)
I don't know whether your ignorance is simply genuine, or you're misguided troll.
You can't compare a high-end video card from 2006 against a mid-range in 2008. How do you not see the logical fallacy in that? Do you know what a mid-range in 2006 (when 7900GTX came out) was? 7600GT that cost $190! 4670 wasn't available back in 2006, so are you suggesting those"sorry sucker" to travel to future to buy their video card? You know method is simply a lazy excuse to slap something together and try to "convince" others that you're right. Overall performance does not cover the detail in performance discrepancy. You HAVE to look at broad selection of games if you want your "ultimate guide" to work.
Originally posted by: LOUISSSSS
Originally posted by: nitromullet
Much like the theoretical $1/5fps card, your guide may be good in theory, but not terribly useful in practice.
1) If one was going for the best value, one would not buy a gaming video card to begin with.
Congratulations! You have made the highest quality, least unfair/impartial post so far in this thread! Thanks! No really, I mean it, I just want to say it again: Thank You!
Now then, I think if you carefully consider this line of thought you'll see that it's mistaken. I wholeheartedly disagree with that statement. It's pretty obvious, isn't it? Ok, so a gamer buys a new desktop with shyte integrated gfx (which is essentially last gen's technology at best) and they are unable to play modern games. The idea with buying a modern gfx card is that it should enable one to play modern games (with varying settings depending upon how demanding a title may be). So, this guy is obviously going to buy a new card for a mere $50-60 dollars and upgrade his performance SEVERAL FOLD, thus enabling him to play modern games (all on High or better in fact with the 4670)! Case closed!
2) You are not doing apples to apples comparisons, but giving an edge to one card over the other to prove your point. For example, in your Card A/B Crysis scenario, you fail either to notice or point out:
- A. By making custom tweaks to Crysis to make it run better but "virtually indistinguishable" from stock High settings, you are allowing the card to do less work. As such, you can not compare card A with custom settings and card B with stock settings.
B. Any tweaks you can make to Crysis with Card A can be applied to Card B as well.
You're right here-- I didn't apply any config tweaks to the fps amount I quoted for EITHER card. Why did you assume such?
C. You assume that a reduction in quality level in Crysis will yield in a linear increase based on some bastardization of Moore's Law that you've cooked up. You assert that every level of quality correlates to a generation of cards, which will double performance if reduced. Moore's Law states that transistor count will effectively double over time, and while this is certainly going to impact the performance of our hardware, transistor count is not in direct correlation to performance. Especially not with respect to video cards. It is also debatable if the quality levels are meant to correlate to video hardware generations.
In my personal experience with Crysis Warhead, a drop in from all "Enthusiast" to all "Gamer" settings does not yield double the fps.
I absolutely agree with you! And I made this very point in the OP. Yes, I could have researched out every single quality level setting for a particular card and noted the decrease in fps from one setting to another, but I didn't, so shoot me! I DO suspect, however, that the increase/decrease in performance from one setting to another is likely to be rather close to 2x/one half. And it IS TRUE that games are designed to run on multiple generations of hardware, otherwise the pc game market wouldn't exist!
3) You don't take into account time.
For your reference, in the chart you can see that the 4670 performs 2x as fast as a 7900gtx! Gotta feel sorry for those suckers who spent 10x that for one of those then and gets half the performance!
This may be true, but the 7900GTX was running at this performance level back in March of 2006, while then HD 4670 just came out in Sept. 2008. If one took your theory at face value one could never buy a video card, since there will always be a less expensive card that offers equal or better performance for less cost sometime in the future.
Yes, but you'll soon see that the 4670 has a 20:1 performance/price advantage vs the 7900gtx-- not bad at all I'd say! But yes, your point is the same as the one I made in the OP regarding moore's law (or simply the continual increase in performance from one gen to the next FOR FREE), the point is to see where this optimal place on the curve lies within each generation AND THEN to see how that relates to one's disposable income and NEEDS as we all do... this is ONE MORE factor to add to your purchasing decisions to give you MORE INFORMATION to help you decide rather than rely on vague recommendations or pretty graphs while dispelling the common misconception that 1:1 ratios are a good deal!
Either way, welcome to AT forums... Always good to see someone trying to offer insight, and accepting healthy debate.
Originally posted by: AmberClad
Ok, let's all calm down please, or this thread is headed for lockdown.
- AmberClad
Now then, I think if you carefully consider this line of thought you'll see that it's mistaken. I wholeheartedly disagree with that statement. It's pretty obvious, isn't it? Ok, so a gamer buys a new desktop with shyte integrated gfx (which is essentially last gen's technology at best) and they are unable to play modern games. The idea with buying a modern gfx card is that it should enable one to play modern games (with varying settings depending upon how demanding a title may be). So, this guy is obviously going to buy a new card for a mere $50-60 dollars and upgrade his performance SEVERAL FOLD, thus enabling him to play modern games (all on High or better in fact with the 4670)! Case closed!
I absolutely agree with you! And I made this very point in the OP. Yes, I could have researched out every single quality level setting for a particular card and noted the decrease in fps from one setting to another, but I didn't, so shoot me! I DO suspect, however, that the increase/decrease in performance from one setting to another is likely to be rather close to 2x/one half. And it IS TRUE that games are designed to run on multiple generations of hardware, otherwise the pc game market wouldn't exist!
Yes, but you'll soon see that the 4670 has a 20:1 performance/price advantage vs the 7900gtx-- not bad at all I'd say! But yes, your point is the same as the one I made in the OP regarding moore's law (or simply the continual increase in performance from one gen to the next FOR FREE), the point is to see where this optimal place on the curve lies within each generation AND THEN to see how that relates to one's disposable income and NEEDS as we all do... this is ONE MORE factor to add to your purchasing decisions to give you MORE INFORMATION to help you decide rather than rely on vague recommendations or pretty graphs while dispelling the common misconception that 1:1 ratios are a good deal!
Originally posted by: nitromullet
Yes, but there will be a card that comes out next year that will have a 25:1 performance/price advantage vs the 7900GTX, so you better not spend your money on that 4670 just yet... Oh wait... The year after there's going to be a card with 40:1 advantage to the 7900GTX. I wonder what will happen the year after that...? Better hold on to your cash, unless you want to be a sucker. The point being that you have to factor in time. Time has its own value.
Yeah, I really like this point that you make regarding time. Perhaps I could add a T variable and another one for fps' around the 30 mark... hmmm. I 'll have to think about that one. But yeah, I think that regarding time, one's needs to upgrade would tend to outweigh an ever-increasing value rate..., and a little something called common sense heheh . My personal suspicion is that it is optimal to upgrade every other generation, waiting for a performance/price ratio of 2.2x or higher and increase one's spending as long as that ratio is maintained (looking at generational statistics, I concluded that a real-world performance increase of 1.5x occurs on average from one gen to the next, so 1.5x1.5 for two gens= 2.25x; and upgrading every gen only yields 1.5x which is like going from 30fps to 45fps... not exactly stellar value for buying an entirely new card).
Originally posted by: LOUISSSSS
how about cheaper CF performance than single card performance? such as 2 x 4850 cards in CF work better than a single gtx280.
The best approach would be to estimate based on statistics the average decrease in performance of vid cards going from COD3 to COD4.
Originally posted by: garritynet
The best approach would be to estimate based on statistics the average decrease in performance of vid cards going from COD3 to COD4.
Call of Duty 3 and Call of Duty 4 are both completely different games developed by different companies. Just because they have sequential titles dose not mean that they offer a linear progression in graphics.
You completely misunderstood me on that point. What I was saying was that because these newer cards haven't been benchmarked on COD3 (duh), we can look at other, similar, cards which were benched on COD3 and on COD4 and then extrapolate based on results we already have for these new cards on bench's running COD4. Limitations include that you would have to compare two different benchmarks (which I don't like doing) which run COD3 and COD4 on the same cards, on the same settings. And then compare these results obtained with the older, though similar, cards and compare them with the newer ones and use statistics to (rather accurately) determine their relative performance differences and compensate (e.g., you could use a 3870, 3850, 9600gso or gt, 8800gs, etc. to stand-in for/compare with a 4670). But the point is moot b/c the 4670 can play the more intense game, COD4, at nearly 60fps on the highest settings that he had in mind.
However, COD4 is a very non-challenging game to run even at #1's settings, fully maxed out (and of course COD4 is significantly more challenging to run than COD3). At the link below, you can see that the 4670 achieves buttery smooth 57fps at 16x10 w/ 4xaa and 8xaf.
And then compare these results obtained with the older, though similar, cards and compare them with the newer ones and use statistics to (rather accurately) determine their relative performance differences and compensate
Then why are you wasting time by thread crapping??Originally posted by: Kelvrick
This guide is horrible. I don't even want to spend the time saying how horrible it is.