Tom's VGA Charts Part 3

[Bjorn]

FBucks are an interesting idea, but the system definitely needs some work IMHO if a 5200 can outscore a 9600 for high-quality gaming with AA and anisotropic filtering.

It's not that i don't agree with you but they didn't exactly recomend the 5200 for high quality gaming.

When image quality settings are raised, the Radeon 9600 XT is an easy recommendation

 

[Dave Baumann]

FBucks are an interesting idea

Well, remember, Valve did this at the Shader Day, so I would assume that this is where Lars got the idea.

 

[beyondhelp]

Of course 60fps getting 1 fbuck is fine with an LCD but for a CRT it would have to be 85fps (being the ideal Monitor Hz setting where, for the majority, the eye no longer pickup's on the scan redraw).

That's not really germaine. Your being silly right? 60fps is the accepted "Minimum" for playability of an FPS Game. Most people run their rigs with Vsync disabled so Monitor refresh rates have NO bearing on the game FPS. That argument is silly and superfluous. Like the Fbucks charts themselves.


Besides, I have a bigger problem with the Prices used to come up with their charts. Looks to me like some of the low ball cards are undervalued by about 40-50%. (I used Pricewatch). 40-50% increase in price would cut the fbucks score almost in half.

Using the absolute rock bottom price one can find for a cheap card doesn't reflect what the Consumer will be paying for the card. It is unlikley they will find the card for the same price.

Tom's Prices........... FPS - Price - Fbucks --------------Pricewatch $ - Revised Fbucks
GF4 Ti 4600............706.9 - $65 - 10.87...................$142 ---- 4.98
GF4 Ti 4200 128.....589.6 - $53 - 11.13................... $90 ---- 6.55
GF4 Ti 4200 64.......598.9 - $54 - 11.09.....................$74 ---- 8.09

Big Difference huh?

These 3 are really way off base, and inflate the fbucks scores for those cards tremendously. And it is totally bogus to include cards that are not readily available. They include the Ti4400 and MX460 which pricewatch couldn't supply a price on, cause no Vendors are selling them anymore.
Fbucks was the lamest Idea I've seen in a long time.

Tom's Best Of 2003 Awards should include this idea as the winner of the Tour de Farce Award for 2003.

 

[andypski]

It's not that i don't agree with you but they didn't exactly recomend the 5200 for high quality gaming.

I'm aware of that - I was only commenting on the nuances of the FBucks system rather than the qualities of the article as a whole - the 5200 managed to score more FBucks than a 9600 in the AA and aniso tests, which seems to indicate a problem with the system to me as I doubt there's much in the way of playable frame rates involved.

 

[digitalwanderer]

I think fbucks would have worked out a LOT better on THG's graph if they would have seperated it into 3 graphs; low-end, middle, and high-end cards.

It'd make it a lot more apples to apples in an apples to oranges comparison.

 

[nelg]

It's not that i don't agree with you but they didn't exactly recomend the 5200 for high quality gaming.

I'm aware of that - I was only commenting on the nuances of the FBucks system rather than the qualities of the article as a whole - the 5200 managed to score more FBucks than a 9600 in the AA and aniso tests, which seems to indicate a problem with the system to me as I doubt there's much in the way of playable frame rates involved.

Of course we could all play our games on a reference rasterizer, sans VPU. Since this would cost $0 it would have a tremendous score.

 

[Bouncing Zabaglione Bros.]

I'm aware of that - I was only commenting on the nuances of the FBucks system rather than the qualities of the article as a whole - the 5200 managed to score more FBucks than a 9600 in the AA and aniso tests, which seems to indicate a problem with the system to me as I doubt there's much in the way of playable frame rates involved.

That's the problem with systems such as these. Anything with really high numbers or really low numbers (either for cost or for frames) will wildly skew the results compared to products where price/performance is in a more reasonable ratio.

For instance, a card that can only manage 10 frames a second would come out way on top if it was sufficiently cheap enough, which indicates that the whole idea needs more work to make it sensible in all cases. Frames divided by cost (or whatever Tom's did) is too simplistic to handle all cases.

 

[beyondhelp]

I think fbucks would have worked out a LOT better on THG's graph if they would have seperated it into 3 graphs; low-end, middle, and high-end cards.

It'd make it a lot more apples to apples in an apples to oranges comparison.

Very good point dw. Maybe Under $100, $100-$250, and $250 and up. Or perhaps seperated into DX9 and Pre-DX9 catergories.

And throw out any contestant that can't maintain at least 30 fps in 3/4's of the benches. If it's unplayable, It shouldn't score.


I still have a problem with the fact that there's no weighting for Longevity of service.(FutureProofing?) That Ti4200 may be a Great buy right now with the predominately DX8 mix of Benchmarks Tom's used, But what about 6 mos. to a year down the road? Buy a $75-$100 card every 6 months just to maintain the minimum acceptable performance point? And always about 2 Gens behind the newest stuff? Yuck! That's no Fun.
Caveat Emptor!

 

[digitalwanderer]

And throw out any contestant that can't maintain at least 30 fps in 3/4's of the benches. If it's unplayable, It shouldn't score.

I like that rule, let's add it in. 30fps in 3/4's of the benchies seems more than fair for a minimum standard.

 

[cthellis42]

It just needs more attention and breakdowns. For current gaming, for future trends, for higher resolution at quality, for lower resolutions without, for playable frames...

Of course then that just gets "too complex" and people will get tired and go elsewhere. Typical users want simple charts they can look at and go "heyyyy!"

 

[beyondhelp]

Too bad the real world bears little resemblance to the Fantasy land of Review sites System setups. People looking to buy a $75 GF4 Ti4200 prob. don't own a P-IV 3.2C with a High end Dual Channel DDR400 Motherboard either, so the scores they are posting at Tom's is Best Case scenario. What's the poor end user with a P-IV 1.8 or XP1600+(or less) and PC2100 DDR going to end up with for fps in those benchmarks with those low ball cards?

It's all very misleading.

 

[Althornin]

It's not that i don't agree with you but they didn't exactly recomend the 5200 for high quality gaming.

I'm aware of that - I was only commenting on the nuances of the FBucks system rather than the qualities of the article as a whole - the 5200 managed to score more FBucks than a 9600 in the AA and aniso tests, which seems to indicate a problem with the system to me as I doubt there's much in the way of playable frame rates involved.

Of course we could all play our games on a reference rasterizer, sans VPU. Since this would cost $0 it would have a tremendous score.

ack! divide by zero error!

 

[nelg]

ack! divide by zero error!

But if you do the math in your head you will conclude that this way offers an infinite value.

 

[AlphaWolf]

For the people that have proposed revised Fbucks systems.

You would need to evaluate each game differently imo. Some games such as a sim is playable at a much lower framerate where as fps multiplayer games require a much higher rate. Also scores beyond a certain framerate are meaningless.

 

[OpenGL guy]

ack! divide by zero error!

But if you do the math in your head you will conclude that this way offers an infinite value.

Division by zero is undefined for a reason. However, I will agree that the limit as price goes to zero is infinity since price is always non-negative (unless someone paid you to take the CPU off their hands!)

 

[howdy b]

check out this article ,along same lines http://www.digit-life.com/articles2/over2003/index.html

 

[Bjorn]

I think fbucks would have worked out a LOT better on THG's graph if they would have seperated it into 3 graphs; low-end, middle, and high-end cards.

It'd make it a lot more apples to apples in an apples to oranges comparison.

And also leave out cards that are unavailable (or close to it, i checked 3 online stores here in Sweden and none hade a GF 4 4600 or a Radeon 9700 for sale) or add a separate graph for these cards.

Might also be a good idea to lower the resolution for the low end cards and adjust the F-Bucks accordingly. At least in the quality tests.

 

[MuFu]

The charts have been updated to reflect the correct prices now.

http://www.tomshardware.com/graphic/20031229/vga-charts-16.html

The FX5200 remains an FSAA + AF champ.

MuFu.

 

[ben6]

But strangely, the charts haven't been upgraded

 

[beyondhelp]

But strangely, the charts haven't been upgraded

Strange? We're talking about Tom's here? Nothing they do surprises me.

Replacing a couple of prices in a spreadsheet chart is easy. Making up new bar Graphs is something else...

Besides it doesn't fit in with the plan to Confuse, Obfuscate, and Mislead.