Stereoscopic 3D Benchmarking: DDD, iZ3D, Nvidia

[PressBot]

By Neil Schneider

While traditional gaming media has every benchmark under the sun, it’s very rare to find ratings for modern stereoscopic 3D drivers like DDD, iZ3D, and Nvidia’s GeForce 3D Vision.  This is completely understandable because it’s much more time consuming, the drivers don’t work equally well with the available measurement tools (e.g. FRAPS), and game setting expectations are different from one solution to the next.

We’ve been getting our share of graphics cards for game testing, and it got us thinking.  What’s stopping us from doing some benchmarking on MTBS?  Could we share information that other sites don’t?  Let’s find out!

Read full article...

[skabais]

Very nice review!
I'm especially pleased that finally there is an objective proof that the widespread myth about stereoscopic gaming cutting framerate to half is a myth!
I had a feeling that 50% drop in performance is more an exception than a rule and im glad to see that tests are proving that.

[tritosine5G]

thats not a myth, and this is no evidence ( averaging over averaging).


For evidence, you should look at the GPU itself with monitoring program like rivatuner or evga precision. There you have the gpu usage percent. I use it to determine if AA is worth the effort or not ( currently UE3 games without decent AA suport eat away 50% perfomance when 4xAA is forced. AMD's MLAA is an answer to this.)



Also metro2033 is acting funny at every resolution, especially outside.

[Neil]

The benchmarks were 100% identical and not an approximation. To do otherwise would make the effort useless.

I'm thinking of updating this again with an Nvidia GTX460 instead of the GTX285. Or just adding it to the list. This was more of an experiment than a diverse benchmarking exercise. It's a starting point for something much bigger.

I don't have a 460 yet, so don't hold it against me!

Regards,
Neil

[Chiefwinston]

Thanks Neil,
I'm currently researching for my next stereoscopic 3D gaming PC. My last one burned up in '06. 3D gaming is not cheap. I've spent thousands of dolllars over the years on constant upgrades. You have zeroed in on the performance area that I'm interested in. Also, any time I hear about turning shadowing or other features off just to play. well I kinda sigh. These are features that seperate old games from new games. I'm glad your up front and clear about these issues. Any chance you'll be including SLI or Crossfire performance numbers in the future. I'm building my newest system from the graphics card on down. HDMI 1.4 support is also, paramount in my selection. So thank you for your well done article. I'll be looking forward to future additions.
cheers

[Neil]

Phew! I'm glad you liked it.

Yes, SLI and CrossFireX are fair game when properly implemented ('m not sure if all drivers support this yet).

I need to get our benchmarking system updated and standardized. This article was intended as an experiment to get a measure of whether or not it's worthwhile to develop further.

Regards,
Neil

[cybereality]

Very nice job.

[Zerofool]

Finally someone made a decent performance testing . For a long time I was trying to find someone with similar videocard to the one I have, so we can do some ATI vs nVidia tests. Some time ago a friend of mine had GTS 250 (which is neck to neck with my HD 4850 in 2D) but he sold it few days before we were about to do the tests, what a douche .

Anyways, I have some concerns about the validity of some of the tests you did. Firs of all, in RE5 the 2D performance is almost identical in all modes (close to 60fps) which brings me to the thought that you maybe forgot to disable the FPS limiter and Vsync (in game's options menu). If this is the case, this will render the percentage graphs incorrect.

Another thing is the fact that Metro 2033 is DX11 game. Are you sure that the game runs in DX9 or DX10 mode on the AMD card? Because GTX 285 is DX10 card, and we would be comparing apples to oranges if it runs in DX11 on the AMD...
The same thing is valid for BattleForge, it's also an DX11 game (after a certain patch).

I hope in the future you'll also test in 1680x1050 resolution because it's still widely used mode by many S3D gamers. And also add some more pure-DX9 benchmarks like FarCry 2 and GTA IV for example. And maybe some DX10 benchies like S.T.A.L.K.E.R. Clear Sky, at least until you get DX11 Nvidia card (you may want to wait for GTX 560 in Q1 2011, if not sooner; whatever you chose, avoid 768MB models).

If you are curious how much vram is used in the different modes, you can use (as I stated in my S3D Vram usage test) GPU-Z and as far as I know also EVGA Precision/MSI Afterburner (on Nvidia cards only). In Crysis it's definitely around or more than 1GB in 1080p (in 2D, not to mention in S3D).

Oh, and which self-respecting gamer/enthusiast uses AMD platform nowadays? Intel is the real deal! No offence to anybody still using AMD . And no overclocking whatsoever? This first-gen Phenom is bottlenecking the cards quite much.

Anyway, kudos for the effort, and I'm looking forward to your next benchmark run.

[Neil]

Hi Zerofool,

Thanks for your kind words!

Just to set your mind at ease:

Resident Evil 5 was benchmarked correctly with VSYNC off and the FPS limiter was also off. Metro 2033 definitely supports DirectX9, and it was only tested in DX9. DX10 and DX11 was not included in any of the benchmark tests because it isn't supported by all three driver developers equally. Battleforge also supports DX11, but in all cases, was manually set to DX9 in the game's config file. In fact, when playing on DDD, it tells you what DX version is being used (to be sure).

I'm equally concerned about bottlenecking, and I'm looking at options for updating the benchmarking system. This was just an experiment to start with.

Thanks again!

Regards,
Neil

[tritosine5G]

Oh, and which self-respecting gamer/enthusiast uses AMD platform nowadays? Intel is the real deal!

me with rebadged g92 gpu SLI, a ph2. Also have a ph1 in parts. Only my bro bought intel cpu.. I had 5x86, duron and all that.

Intel?

Intel wanted to be big in LCOS. Instead, china 's going to be big in LCOS . You call yourself self respecting gamers with intel cpus and lcd screens?

http://www.splendidoptronics.com/" onclick="window.open(this.href);return false;
http://translate.googleusercontent.com/ ... xYYAnk3xSg" onclick="window.open(this.href);return false;
300.000 china cities get many many hundred 70-100" 3xLCOS displays with 10.000 hour bulbs

[Neil]

Language, language!

Regards,
Neil

[Zerofool]

Hello Neil,

Thanks for the clarification.
You are welcome, keep it up!

I have never used DDD's driver, but I know for sure that iZ3D driver can also show which DirecX version is used. Just turn the "Show FPS" feature on from the profiles menu of the driver panel. But use it just for checking, I suspect it slows things down a bit (as does the laser sight ). And another observation - FRAPS works with iZ3D in all games, when using anaglyph output (which is good because all 3D drivers support anaglyph output).


tritosine, I don't want to offend you but ... wake up man, stop living in the good old Athlon days. I also used many AMD CPUs throughout the years (actually before my current Intel machine, my previous Intel one was based on Pentium MMX 233MHz (OCed to 266MHz )) but since s754/939, AMD is always behind Intel (when Core2 architecture was released). Yes, in terms of Performance/$ they are much better, indisputably, but in terms of peak performance Intel is in the lead. Just check out some clock-to-clock comparisons in SuperPi, wPrime, Cinebench, Sisoft Sandra, x264 compression, WinRAR/7-Zip benchmarks, BOINC Benchmark, PPD in SETI@home, Folding@Home and so on... Now factor in the fact that Intel CPUs overclock better and there you have it. Oh, there's something for the extreme users that Intel has, and AMD doesn't - 24 frickin' threads (google "SR-2"), this baby is unbeatable in crunching/rendering/transcoding (but it has it's price).

In games though, the difference is often very small to justify the extra money for Intel system, but most power users rarely use their PCs for gaming only (that's what consoles are for (and come to think of it, none of the consoles use AMD CPUs )). Here, we're talking about benchmarking of video cards, and therefore, we must eliminate all other possible bottlenecks. For the past few years Nvidia cards have been more CPU-dependent than ATI/AMD cards. That's why to eliminate all doubts, the usage of LGA-1366 Core i7 overclocked @4GHz at least, is recommended (using i7 980X for this purpose would be overkill though).

Don't get me wrong, I'm not some stuped fanboy, I always love good competition because we all benefit from it. That's why I have high hopes on the upcoming AMD Bulldozer architecture, but I'm little skeptical about it beating the LGA-2011 Sandy Bridge (which probably won't be that much faster than LGA-1366 Nehalem/i7, clock-to-clock).

You call yourself self respecting gamers with intel cpus and lcd screens?

Please, don't insult me. I would never use LCD monitor, ever. I'm a proud CRT (professional 22" one) owner and my next display probably will be 2011 Plasma 3DTV. LCDs can't get even close to these black levels (even the LED-backlit ones, with their blooming) and response times, not to mention viewing angles and color reproduction (especially on the most popular TN panels).

Peace (and excuse me for the off-topic)

[tritosine5G]

ofc im not serious You know, tablets and ARM won this war already.

Our first PC (mother's accountant machine at home):
http://www.redhill.net.au/c/c-4.html" onclick="window.open(this.href);return false;

This was the last and greatest 386, and AMD made it in enormous quantities. The DX-40 was vastly cheaper than a 486SX, which usually could not out-perform it anyway, and easily faster than Intel's best 386, the DX-33.

hahah!

The only thing I really use my x86 for, that really makes it struggle is high quality resizing, denoising, deblurring , motion interpolating videos thru AVISYNTH in realtime.. I can always find a good compromise, I dont really overclock, 3.2ghz is enough. Im just reluctant to spend money on cooler, I have the PC in another room, noise is no concern.

[Zerofool]

ofc im not serious You know, tablets and ARM won this war already.

Aw, silly me! How could have I forgotten that these can run Crysis in 3D @1080p at solid 60fps...

The only thing I really use my x86 for, that really makes it struggle is high quality resizing, denoising, deblurring , motion interpolating videos thru AVISYNTH in realtime..

Believe me, real-time high quality motion estimation & compensation of HD video is hard task for any CPU (well, except for that monster I described, and only in case the script/plugin(s) is(are) well multi-threaded). I'm telling this from my experience, I've worked with such scripts in which the average speed is measured in seconds per frame, not the other way around .

I dont really overclock, 3.2ghz is enough. Im just reluctant to spend money on cooler, I have the PC in another room, noise is no concern.

It's personal choice, but in my book "self-respecting gamer/enthusiast" and running your CPU at stock speeds and with the box cooler don't fit together. But that's just my interpretation of that term. And it's not about money, it's about philosophy... or state of mind if you will. My PC components were one of the cheapest in their class, and still offer great overclockability and features, you just need to know what you're doing when shopping.
I will disagree that 3.2GHz is enough for whatever nvidia cards in SLI you have (2x 8800 GT/9800 GT(X) or the GX2 I guess, because the newer 55nm-shrink chips are called g92b). Just borrow a nice air cooler from a friend, clock the CPU at 3.8GHz and you'll see the difference in games.

[Neil]

Doesn't anyone buy Cyrix anymore?

Regards,
Neil

[Zerofool]

Now serious - VIA bought them a while ago, and they only make low-power/mobile platforms now, which are quite competitive in their class to be honest.

To continue the hilarious tone..
Neil, why don't you also test any of the S3 Chrome cards, they are DX10.1 compatible .

[tritosine5G]

Tablets going to win, seriously. PC is a dyin breed....

Aw, silly me! How could have I forgotten that these can run Crysis in 3D @1080p at solid 60fps...

heard about cloud computing? Nvidia sez , every gfx intensive task goin to run on cloud by 2014
So rather think about real time raytracing , & avatar quality gfx on tablet& stuff


...nV integrates ARM processors inside GPU, just go , read about nvidia echelon ...

I play with 3 games these days:

Mass Effect2 - 120fps with 4xAA (forced 4xAA costs 50% fps alone, consolitis)
Borderlands - 120fps only here I have fps drop with 0AA, when huuuge geometry is present on the screen. Tried OC , 3.8ghz, not CPU limit. Its something else
Men of War - 120fps , RTS with destructible terrain, and 3d crosshair not bad.
...moreover expect to play DNF and crysis2 at 120fps
I'll upgrade when 28nm GPU's come out. 40nm won't be enough for this (preview, not dx11 yet), no joke, wanna bet we see gtx580 on knees?:
[youtube]http://www.youtube.com/watch?v=dRzRLwSb ... re=related[/youtube]

also , metro2033 runs very good. But it needs an ultra high gain screen