Hi all,HardOCP today released their review of the new GF-FX 5950. What is most interesting to me here is not the review of the card itself, but the use of FS2004 as one of the benchmarks... In particular, their methodology in testing different cards and the Image Quality comparison using FS9.I think Brent did a credible job trying to keep the settings comparable between runs, as well as picking a fairly representative area to bench (comments?). The image comparison was informative as well in my opinion (lets remember, they didn't optimize the .ini file to extend filtering levels here - as most users outside of community boards will not).I think it would greatly behoof us as a community if we wrote to sites like and congratulate them (mailto:[email protected]) on their good use of FS as a bench (as many of us know, it can be very difficult). Unfortunately, [link:www.anandtech.com]Anand has stopped using FS in their latest reviews (mailto:[email protected]) - probably due to the difficulty of getting comparable runs. Providing encouragement and thanks to hardware sites that "do it right" and possibly helping out those who might be struggling is the only way to continue to get valid hardware results specific to our favorite hobby. I'd like to see sites use FS as a CPU bench as well, since we all know its extremely intensive in that aspect.So, what do you think of HardOCP's methodology and results?FS2004 specific resultshttp://www.hardocp.com/article.html?art=NTM3LDc=The start of the articlehttp://www.hardocp.com/article.html?art=NTM3LDE=Take care,Elrond

Thanks, Elrond, interesting post.As a complete dumdum when it comes to these matters, all I can say is I'm pretty horrified that the results - all close - show that these top-of-the-range, whoops-sorry-Mr-Bank-Manager cards only achieve 25-30 FPS with Autogen and other sliders at MID settings, when using default planes in a default sim from the default airport. Makes a bit of a mockery of all the airy-fairy claims of superb frame rates with everything maxed that you read in these forums. And also makes the MS claims for minimum spec very laughable: "Pass me another frame, Bill, I'm bored with this one"Mark "Dark Moment" Beaumonthttp://www.swiremariners.com/newlogo.jpg

Toms Hardware also started using FS9 for his benchmarking and this includes his latest Athlon 64 and Intel P4 Limited Edition benchmarks. Now when did I send them email/request that FS should be part of their benchmarking suite ? Probably over 2 years ago.Michael J.

Thats great of you to write them with your encouragement: lets hope many others here do so as well. I too have seen Tom's and other sites review with FS9 and it is indeed a step in the right direction. Encouragement wherever we see that and even more important where we don't can only do us all good.Take care,Elrond

Mark, this test was run at 4xAA and 8xAF settings... and still produced average framerates in the 30s.. If you really want to go for performance, lower AA to 2x and AF to 4x and I bet you that you see higher framerates.These cards make FS9 hum!

Hello Dark, "Makes a bit of a mockery of all the airy-fairy claims of superb frame rates with everything maxed that you read in these forums."Humm, see their screenshots and the setting they use, pretty good FPS with 8X anasotropy that take 8 FPS drop for nothing compare to 4x, they got 35 FPS. They use all slider to full except of scenery and autogen to mid. I don`t use Anasotropy 8x, but only 4x, make differences . I increase the autogen and scenery, disable the scenery shadows, mesh 80%, I get similar result to their FPS result.p4 3.2 GHZ, gforceFx 5600 256 MB Ultra, 52.16 nvidia driverI can reduce the AA to 2x, Another 5 fps gain :) Pretty good FPS compare to Fs2002 at his release :)BestEric

Hi Mark, "Makes a bit of a mockery of all the airy-fairy claims of superb frame rates with everything maxed that you read in these forums."Are you sure? This is the fact, see their setting and average FPS, it just confirm what most user that's have great fps. I am wonder why they have used 8XAF with the test, this get near 7 fps drop.Average framerates in the 30s with 4x antialasing and 8XAF, pretty good at this time with these actual market system available.ThanksChris WillisFSW GROUP[link:fsw.simflight.com/FSWMenuFsSim.html|Clouds And Addons For MsFs

Hi Chris,"I am wonder why they have used 8XAF with the test, this get near 7 fps drop."I'm sure its because no one would really be buying a $500+ (US) card and not expect to use the best filtering available. Beyond that, FS is such a slow moving game that Image Quality, as they state, is of paramount interest to most. 8x Aniso and 4x AA (or at the very least, 8x Aniso and 2xAA) is almost a requirement for the type of game FS is - specially at that type of price range.Take care,Elrond

_____________________________________________________________________In Flight Simulator 2004 we were quite surprised by the GFFX 5950Ultra surpassing the Radeon 9800XT for such an extended amount of time. For the best FS 2004 performance with AA enabled it appears the 5950Ultra with the 52.16 driver is very competitive performance-wise. However, the image quality differences between the Radeon 9800XT and 5950Ultra cannot be ignored so easily. In a slower moving game such as this, one notices IQ more and the 9800XT produced an easily identifiable Image Quality that was much better than our GFFX 5950's_____________________________________________________________________Since the 52.16 drivers are reported to limit AF to 2x regardless of what you want or where you set it and also limit the filtering to some sort of mixed bilinear, I'm not surprised at these specific results. His statement about FS2004 being to choppy at 1600x1200 res shows his lack of ability to properly set up and tweak the sim. I have to give praise though for trying to get accurate benchmarks in a game/sim where it's virtually impossible to do and pointing out the ATI IQ quality comparisons since ATI's AA is much better and ATI is actually using 8xAF that's not surprising.

Hi Elrond :-)Great post! As for their methodology, it will probably be at least as good as everything else. On the one hand they are trying to use the sim in as realistic manner as possible (ie no pre-recorded flights). On the other hand, I sometimes have to wonder if a synthetic test (like a pre-recorded flight) should still be done as well, because it might tend to isolate the specific strengths of the video card as opposed to the CPU. My theory being that playing back a pre-recorded flight is far less CPU intensive than ordinary flying, therefore the fps on playback are far more GPU dependant.My only concern is that even when trying to duplicate flights exactly, the cloud positions in an otherwise identical weather scenario are still going to be different from one flight to the next. Yes, the cloud coverage, type, number of layers, base, height, etc may be exactly the same from flight to flight, but you may have more or less clouds in your particular field of view at any given time from one flight to the next. This is going to impact the results. But I'm not suggesting clear weather either. We do need to know how GPUs cope with the 3D clouds.Notwithstanding all of that, my main interest in these tests is image quality. A quick look at the screenshots shows (to me) that FS9 anti-aliased on the Nvidia card does not look nearly as good as it does with the ATi cards. Look at the runway edge lines and the tail of the Learjet as proof. Those Nvidia screenshots look just like my own on my Geforce 4, which is quite noticeably inferior to the Voodoo 5 I used to own (except of course in relation to texture size). The ATi card, on the other hand, appears to have the same anti-aliased image quality as my old Voodoo 5 but has the speed, texture size and DX9 capabilities reqired for modern gaming.I'm thinking the next card I purchase (scheduled in 6 months time) will be an ATi based card unless Nvidia pulls something pretty darned amazing out of the hat to improve their relatively poor image quality.

>I have to give>praise though for trying to get accurate benchmarks in a>game/sim where it's virtually impossible to do and pointing>out the ATI IQ quality comparisons since ATI's AA is much>better and ATI is actually using 8xAF that's not surprising. >It sure is. The side by side comparisons rest the case in the Nvidia versus ATi IQ debate as far as I am concerned. Nvidia should hang their heads in shame.

>A quick look at the screenshots shows (to>me) that FS9 anti-aliased on the Nvidia card does not look>nearly as good as it does with the ATi cards. Look at the>runway edge lines and the tail of the Learjet as proof. Yes, this is the very first thing I noticed. Looks much better with ATI card.Michael J.

Hi Jon,I'd have to agree that using a pre-recorded flight would probably be the best answer... In FS, pre-recorded doesn't actually mean what it sounds as you know, since everything is rendered in exactly the same way as if manual control was being used. All FS does is control the flight path instead. I'm not sure it would make any difference on the CPU usage however, as the rendering would be identical. If anything, a pre-recorded would push the CPU just a tad higher since the sim has a bit more work to do. Beyond that, one can cover a more realistic flight pattern than straight out with a pre-recorded, since its playback would be identical unlike what one could do manually.I don't think Brent was aware that FS supports playback of a previously flown flight path... Thats probably why he didn't choose that route. Would you kindly write him and let him know that the option is available (while praising his current method as a great start)?As for the clouds, indeed that is a bit of a problem. Hopefully, the distribution is random enough that over the course of two flights with identical settings, the processing required to render them would be about the same. The same could be said about the frequency and distribution of AI craft, but again that isn't controllable (outside of something like Project AI, but thats way too much to ask of a hardware reviewer). Since I too believe that disabling either of those altogether would be a bad choice, his method was the best that I can think of.As for image quality, I couldn't agree more with Big, you and Michael. ATI's AA and AF, in *this* round of designs and drivers, have always been better. As for your next purchase, I'd most certainly wait to see the cards out at the time you make that decision first.Despite my title, this really isn't an ATI vs NVidia comparison (as in, the companies): this is a current ATI card vs current NVidia card comparison (as in specific hardware and drivers in a slice of time). Each design of the respective companies will vary widely in its capabilities, strengths and weaknesses. In this long lasting round (NV3x vs R3xx), ATI most certainly has the IQ locked up in my opinion. But again, I'd wait to see the next round, as always, before making a purchasing decision based on those future cards and drivers.Take care,Elrond

This is an interesting thread--thanks for the post. It's interesting to see how the big performers rate side by side, and this was among the better comparisons I've seen since the FS2004 benchmark was well spelled out and there were images to compare side by side...One thing in regards to image quality....I don't believe LOD adjustments were made, or did I miss that? I know the 30.82 driver set had to be patched with Rivatuner before I could dive into negative LOD's, but negative LOD's go a long way, along with AF, to improve image quality. Once my drivers were patched and my LOD set in a negative range, my image quality went way up on my 4200 and seems the equal of the ATI shots in the comparison. That makes me wonder though whether ATI adjusts the LOD so that the "normal" or default setting is actually close to -1 or less with the Nvidia cards... Can anyone who has compared ATI and Nvidia cards of a similar quality comment?Sorry I'm in the forum less these days, BTW... I was out of the country until Tuesday, and had to start planning my next business trip as soon as I set my foot in the door. I haven't had much time for our hobby, although I did get a two hour COF session in Tuesday afternoon...-John

Elrond,Just a few thoughts to share:It is good to see FS again being used for benchmarking and long over due.Having seen HardOCP's review a few things seem a bit odd as per fs9...At first I couldn't fathom the FPS scores for the 5950 vs. the 9800 as per my own experience using a 5900 overclocked higher than the 5950 there is just no way that it is faster than then the 9800XT...but then when you look at the IQ shots the results speak for themselves... They have got something very wrong with thier setup as 8xAF as well as all other AF modes work just fine and are supported in the new 52.16 drivers and it is plainly obvious that not even 4xAF was used in those IQ test and as a result you can't let those FPS score fool you.As for the IQ of the new 52.16 it is a notch backward in many ways but hopefully when dx9.1 is released that will go back up, also it seems the menu problem is NOT solved as many have claimed.