The Baron said:
I use a program called FlashSwitch (http://www.flashswitch.com/). It's a little icon that sits in your systray, you just click the icon to turn flash on and off and reload the page, assuming you use IE of course.
ET article confirms ATi/Nvidia cheats
- Thread starter John Reynolds
- Start date
I use a program called FlashSwitch (http://www.flashswitch.com/). It's a little icon that sits in your systray, you just click the icon to turn flash on and off and reload the page, assuming you use IE of course.
Ok - thanks Dave.
Now - is it just me or does anyone else think the results are whacked? I mean, the scores are not even close. Hell - taking a quick swoop and we are looking at a 49% difference between the 1280x1024 scores and a 52% difference at 1600x1200...
This has probably already been asked - but exactly how different are the two settings between the manufactures? In regards to card settings (AA/FSAA, etc)...
BTW Dave, I am not questioning your tests or the results - or whoever provided the numbers. With a difference in speed across the board one would expect to see the same results on other benchmarks.
Maybe I have missed something - but I do not recall any result showing 40% or greater...
Something does not add up here....
Now - is it just me or does anyone else think the results are whacked? I mean, the scores are not even close. Hell - taking a quick swoop and we are looking at a 49% difference between the 1280x1024 scores and a 52% difference at 1600x1200...
This has probably already been asked - but exactly how different are the two settings between the manufactures? In regards to card settings (AA/FSAA, etc)...
BTW Dave, I am not questioning your tests or the results - or whoever provided the numbers. With a difference in speed across the board one would expect to see the same results on other benchmarks.
Maybe I have missed something - but I do not recall any result showing 40% or greater...
Something does not add up here....
Saem said:
Err...Javascript is client-side, nor does it have any graphics capability.
My suggestion? Well, of course you keep all the results in a simple database; MySQL would be an obvious choice. You have a CGI script in Perl (or something) that handles the user's request, selects the proper data from the database, and formats it all for a pass off to...
...jgraph, a utility which outputs a beautiful graph in...
...Postscript. Um.
So then, as per the handy suggestion on the jgraph site, we pipe the result to a command-line invocation of ghostscript, and pipe that result to pnmtopng (or ppmtogif), tie it all up with a ball of twine, and voila! A beautiful image file to serve up to our discerning viewer, assuming he has not gotten bored waiting for all of this to occur and gone to the kitchen to make himself a sandwich.
Alternatively, Perl has some native image producing support via Perl/Tk, although I wouldn't want to code up the means to produce graphs that way.
Alternaternatively, someone has probably already done it and posted it to CPAN.
Or, seeing as how the total amount of data surely won't be very large: a dozen or so video cards * a few PC configurations * a dozen or two tests * five output resolutions * four or five AA settings * four or five AF settings * two or three AF algorithm choices * four or five bytes to encode the score (ooo!--less if you use packed decimal!) = maybe a MB or so of data. Ok, that's actually larger than I thought. But anyways, the idea was to just download all the data, and have a client-side java applet generate the graphs or something.
Well...?
What are you waiting for, Wavey? Go to it! And make some custom timedemos for every review while you're at it! And speaking of, where is that NV35 review anyways?!
I certainly didn't mean Flash (I've complained already on AT's forums about that). I meant some script that would allow users to download only graphs for the settings they consider comparable. Generating them on the fly on the user's PC is very intriguing, tho, and would probably save both bandwidth and CPU time on the server--very interesting idea, especially given how ridiculously fast even the slowest new PC's are nowadays.The Baron said:
I remember Marco saying he was working on something like that, but that was a long time ago. I'm not sure if it's still on the way, or if it proved to be more trouble than it's worth.
BTW, I just noticed CNet's reviews only use 3DM03 and UT2K3 as benchmarks. I'm guessing it's a fairly popular site, so we can see that 3DM03 cheating can be a big issue. Nice to see CNet has taken down their overall 5900U score to reevaluate after they retest 3DM03. Funny how a seemingly more technical and reader-oriented site like AT can't even do that: post a blurb at relevant points througout the article notifying readers they are investigating suspect numbers.
Doomtrooper
Veteran
A much needed update..well done.
Deflection
Newcomer
Doomtrooper said:
rofl
I'm sorry, but this update seems pointless...
ET said:
Why are these different? And for a .5% gain?
An update that where the difference is within the magin for error that brings up more questions than it answers.
[edit- to explain my post a little better]
Myrmecophagavir
Newcomer
So what exactly does 3DMark03 do when you tell it to turn on AF that would cause this extra overhead?
OK, I'll throw in my results in the mix:
System:
Intel P4 1.8A GHz @ 2.25 GHz
Radeon 9700 with Catalyst 3.4 drivers
Control panel Direct3D changed settings:
- forced performance 4x ansisotropic filtering (I play my games with that )
- forced wait for vsync to off
3DMark settings were left at default, only resolution was changed to 1600x1200.
The results for game test 4:
3DMark 2003 build 313 = 16.6 fps
3DMark 2003 build 320 = 16.6 fps
3DMark 2003 build 330 = 14.9 fps
So:
- Build 330 is (100 - ((14.9/16.6) * 100)) = 10.24% slower than build 320.
- Build 320 is (((16.6/14.9) * 100) - 100) = 11.41% faster than build 330.
(Both points under 'so' are valid. You just can pick the one that suits you the best. Rule number one of letting benchmark difference look less or more impressive. )
As far as I am concerned, the ExtremeTech figures for gaming test 4 are completely valid.
System:
Intel P4 1.8A GHz @ 2.25 GHz
Radeon 9700 with Catalyst 3.4 drivers
Control panel Direct3D changed settings:
- forced performance 4x ansisotropic filtering (I play my games with that
)- forced wait for vsync to off
3DMark settings were left at default, only resolution was changed to 1600x1200.
The results for game test 4:
3DMark 2003 build 313 = 16.6 fps
3DMark 2003 build 320 = 16.6 fps
3DMark 2003 build 330 = 14.9 fps
So:
- Build 330 is (100 - ((14.9/16.6) * 100)) = 10.24% slower than build 320.
- Build 320 is (((16.6/14.9) * 100) - 100) = 11.41% faster than build 330.
(Both points under 'so' are valid. You just can pick the one that suits you the best. Rule number one of letting benchmark difference look less or more impressive.
)As far as I am concerned, the ExtremeTech figures for gaming test 4 are completely valid.
Joe DeFuria
Legend
Re: OT: Question to John Reynolds
I don't remember him, or anyone else, being against 32 bit color. I do recall that I, for example, was for 22 bit color: near 32 bit quality at near 16 bit performance. John was probably similar i his views, but he'll have to speak for himself.
sonix666 said:Hi John,
out of curiousity, are you the same John Reynolds that was quite against 32 bit color in the early TNT2 and G400 years?
I don't remember him, or anyone else, being against 32 bit color. I do recall that I, for example, was for 22 bit color: near 32 bit quality at near 16 bit performance. John was probably similar i his views, but he'll have to speak for himself.
Yeah yeah, the '22' bit of the Voodoo3. That post filter still makes me laughs. I was convinced in 32 bit and have gotten me a G400 then. And yeah, it was very fast in 32 bit color. Some people that went for the Voodoo3 in the end were quite jealous of the image quality of my G400.
Now, the big question, what do you think of floating point?
Now, the big question, what do you think of floating point?
Re: OT: Question to John Reynolds
I was opposed to 32-bit being discussed outside of performance, just as I argue against people who constantly bring up ATi's AF algorithm without mentioning how fast it is. Maybe it's just me, but I've always liked trade-offs that give both good IQ and good performance (22-bit post filter, 24FP, etc.).
But as for 32-bit output, I never started using it until November of '01 when I bought a GF3 Ti.
sonix666 said:Hi John,
out of curiousity, are you the same John Reynolds that was quite against 32 bit color in the early TNT2 and G400 years?
I was opposed to 32-bit being discussed outside of performance, just as I argue against people who constantly bring up ATi's AF algorithm without mentioning how fast it is. Maybe it's just me, but I've always liked trade-offs that give both good IQ and good performance (22-bit post filter, 24FP, etc.).
But as for 32-bit output, I never started using it until November of '01 when I bought a GF3 Ti.
Joe DeFuria
Legend
sonix666 said:
Um, why? Was a very good trade-off. 22 bit isn'at as good (qulity wise) as 32 bit, but it was significant;y better than 16 bit, the performance was better than the comparably spec'd / priced competition at 32 bit.
Glad you were happy!
And yeah, it was very fast in 32 bit color. Some people that went for the Voodoo3 in the end were quite jealous of the image quality of my G400.
Of course, the G400 wasn't nearly as "compatible" as the Voodoo3 back in the day, nor was it as cheap. It's all about the trade-offs.
Now, the big question, what do you think of floating point?
Um, what about it?
Only the post filter made it look somewhat better, however, internally it was still just as 16 bit as ever before, resulting in banding and ugly colors as soon as transparency was used.
I haven't ever noticed that it was less "compatible". The ending days of 3dfx and Glide were already happening bigtime.[/quote]