View Full Version : Mike's 3DMark Commentary
When I upgraded my CPU from 800MHz to 1.8GHz I had no performance benefit. I was using a Geforce 3, FSAA really dived on that card with newer games.
When I upgraded from my NV20 to and R300, my performance jumped more than 6 fold. :)
hordaktheman
02-13-03, 04:51 AM
Yes, I know what you mean but the point is that at least you have the possibility to turn off the fsaa.
On a high end vid card/midrange cpu, you can turn on high detail settings without a performance drop, but your performance is crappy anyway, whether it's at low detail or high detail.
On a midrange vid card/ high end cpu, you have at least the possibility to lower your resolution/detail, or turn off fsaa.
My point is that if you had upgraded your nv20 to a r300 on your 800mhz cpu, your performance would have sucked just as bad, while in your case, you had the OPTION to improve your framerates by turning off fsaa. If you had turned off the fsaa on your 1.8ghz cpu your performance WOULD HAVE improved.
That's why I'm saying that 3dmark2001/2003 are deceiving.
My point is that if you had upgraded your nv20 to a r300 on your 800mhz cpu, your performance would have sucked just as bad, while in your case, you had the OPTION to improve your framerates by turning off fsaa. If you had turned off the fsaa on your 1.8ghz cpu your performance WOULD HAVE improved.
Sorry, I wasn't talking about 3dmark in that post. :)
Anyway, I only play with FSAA. I don't believe turning off AA is an option despite the option being there in the driver control panel. :)
My brothers P3 800MHz takes just as large of a dip with FSAA as it had on my P4 1.8GHz. :)
He also doesn't believe turning off FSAA is an option. :)
Though at least his games are still playable.
I only care about FSAA+AF performance. Nothing less, I ignore all benchmarks that don't show FSAA+AF scores, as they are useless to me as toilet paper. :D
hordaktheman
02-13-03, 04:59 AM
No, I know you weren't, but the point is that you were already straining the vid card by running it with fsaa turned on. Had you turned fsaa off, your cpu upgrade would have been far more noticable.
At the same time your vid card upgrade absolutely helped your fsaa framerates, they wouldn't have if you hadn't already upgraded your cpu.
Originally posted by SnakeEyes
Smokey:
It goes back to what jbirney posted earlier (which I agree 100% with). Basically his point was that the final weighted score is worthless, but the tests themselves as well as their results are extremely useful.
Ignore the final '3DMark' score, and instead look at the details to see what the average framerates for the game scenes is. Those will tell you more about your card's performance than the score ever could. (For instance, getting 40+ fps for the 9700Pro vs. the FX's ~15 at the moment would tell me that for what that particular scene is testing, the FX is especially weak. The weighting of the scores to determine the final 3DMark combined score can hide things like that.)
Yes, but when im getting 2.6fps on GT2+3 Im screwed whichever way you look at it, and getting my new cpu soon isnt really going to help :p
jbirney
02-13-03, 07:30 AM
sebazve,
I think Brent and Kyle (dont know so just my worthless 2 cents input) have been wanting to get away from the 3dMark thing for some time now. Brent has posted in the B3D forums that he knows that 3dmark2001 was not very usefull for measuring card A vrs card B. Now why they did not do this awhile ago (stop using 3dmark 2001 with the overall results in reviews) I dont know. But I dont think they are playing favorites. Sometimes when a new rev of a tool comes out its the best time to pitch it.....
sebazve
02-13-03, 09:27 AM
Originally posted by jbirney
sebazve,
I think Brent and Kyle (dont know so just my worthless 2 cents input) have been wanting to get away from the 3dMark thing for some time now. Brent has posted in the B3D forums that he knows that 3dmark2001 was not very usefull for measuring card A vrs card B. Now why they did not do this awhile ago (stop using 3dmark 2001 with the overall results in reviews) I dont know. But I dont think they are playing favorites. Sometimes when a new rev of a tool comes out its the best time to pitch it.....
ok then so end of story. i wasnt really attacking hardop or anything since i like their reviews i just wanted to know why th e sudden desicion just after nvidias...
I think Brent and Kyle (dont know so just my worthless 2 cents input) have been wanting to get away from the 3dMark thing for some time now.
If I should be so bold as to give my take on HardOCP's position...I can assure you that you are correct. We ( especially KYle ) have been unsatisfied with 3dMark's illustration of a real gaming environments for quite some time. Long before 3dMark2003 was in any appreciable form, Kyle was already contemplating an article to outline the points he has recently made. If you read the article (http://hardocp.com/article.html?art=NDMw) you will find that his points cover all benchmarks and do not solely attack 3dMark2003. Rather, the arrival of 3dMark2003 was merely a convenient means of emphasizing the points being made.
I honestly can't see why anyone would try to attack Kyle or HardOCP. At the end of the day, these articles and editorials are intended to benefit the consumer...For those grasping at straws and claiming that [H] is an NVIDIA-fanboy...I must remind you of the dozens of times [H] has been on NVIDIA's case for an issue. In addition, you might remember that Radeon 9700 Pro's are in each review testbed and that we were less than "blown away" ( ironic ) by the GeForce FX.
In my opinion, many of us need to take a few steps back and look at the big picture. We need to lose the delusions of conspiracies and shady partnerships and realize that the situation at hand is very black and white. You have a large website with a big name in your corner...trying to keep the big vendors honest and ensure you get the best quality for your dollar.
I trust that this will help clarify this situation and bit and put things in the proper perspective...
:D
batterbrain101
02-13-03, 11:55 AM
Originally posted by sebazve
what i dont understan from you guys is that for years you haven been using 3dmark as a benchmark but now since nvidia says it sucks you think that too.
Nvidia and ATI do spend time optimizing their drivers for 3dmark so what???
It's not that 3dmark sucks, its that drivers get optimized for increased performace for the benchmark but the performance of games doesn't improve (not with the last dozen or so det relaese anyway) I personally have no prob with 3dmarks etc., I paid for a video card so I can play games at full tilt with no issues, etc, who wants to spend serious cash on a card only to get home and find that while the benchmark ran just fine, that awesome game (s) aren,t because of the driver /game/hardware are having compatibility issues. That's always a nice feeling right? Now your thinking great X amount of money for this. Nice.
sebazve
02-13-03, 12:27 PM
Originally posted by batterbrain101
It's not that 3dmark sucks, its that drivers get optimized for increased performace for the benchmark but the performance of games doesn't improve (not with the last dozen or so det relaese anyway) I personally have no prob with 3dmarks etc., I paid for a video card so I can play games at full tilt with no issues, etc, who wants to spend serious cash on a card only to get home and find that while the benchmark ran just fine, that awesome game (s) aren,t because of the driver /game/hardware are having compatibility issues. That's always a nice feeling right? Now your thinking great X amount of money for this. Nice.
ok then so stop optimizing drivers fo 3dmark and spend more time with games but then dont trash 3dmark or any bechmark only because your driver and product sucks at it....:argh:
Originally posted by sebazve
ok then so stop optimizing drivers fo 3dmark and spend more time with games but then dont trash 3dmark or any bechmark only because your driver and product sucks at it....:argh:
But if you didn't optimize for 3DMark, what stops your competitor to do so and crush you in it?
And if you optimize a lot of games to run faster on your GPU via drivers, wouldn't it be reasonable to give such a boost to 3DMark too, to reflect real-world performance?
There's really no easy fix to all this...
Uttar
sebazve
02-13-03, 12:57 PM
Originally posted by Uttar
But if you didn't optimize for 3DMark, what stops your competitor to do so and crush you in it?
And if you optimize a lot of games to run faster on your GPU via drivers, wouldn't it be reasonable to give such a boost to 3DMark too, to reflect real-world performance?
There's really no easy fix to all this...
Uttar
or you can also try to make understand people that a 3dmark score doesnt mean ****! pretty much like amd does with mhz thingy...
The Baron
02-13-03, 03:48 PM
Ah, but AMD uses a "megahertz equivalent" performance rating or some stupid crap like that. So no. There's really been no education to speak of here.
It's still "higher number on hardware automagically equals way higher performance." This SOMEHOW has to change. But it won't, at least not for a while...
LORD-eX-Bu
02-13-03, 04:18 PM
Originally posted by pelly
I honestly can't see why anyone would try to attack Kyle or HardOCP. At the end of the day, these articles and editorials are intended to benefit the consumer...For those grasping at straws and claiming that [H] is an NVIDIA-fanboy...I must remind you of the dozens of times [H] has been on NVIDIA's case for an issue. In addition, you might remember that Radeon 9700 Pro's are in each review testbed and that we were less than "blown away" ( ironic ) by the GeForce FX. :D
Hey Pelly, personally, I love [H]. I think its one of the best, if not the best hardware sites around. Very informitive, only gripe is the forums, you think you could get any more of em' in there?:p But really, I think that site beats this one, I just don't know anyone there:D
Come over and introduce yourselves...you'll find an amazing knowledge base...
We can always use some good guys like you man...I'll keep an eye out for you!
:D
StealthHawk
02-13-03, 05:13 PM
Originally posted by The Baron
Ah, but AMD uses a "megahertz equivalent" performance rating or some stupid crap like that. So no. There's really been no education to speak of here.
It's still "higher number on hardware automagically equals way higher performance." This SOMEHOW has to change. But it won't, at least not for a while...
it never will. i still remember hearing people bragging about their "awesome" P4 systems even last year that had SDRAM and a TNT2 in them.
vBulletin® v3.7.1, Copyright ©2000-2012, Jelsoft Enterprises Ltd.