Video Card Destroyer...

[Lin Evans]

Don't even try this unless you have a very strong video card. This one is just too much because of the masking, for any except a top-end card. It was too much for an MP4 as well. It runs smoothly on my nVidia 8800 GT, but crashes on my wife's overclocked 8600 GT and all the rest of my video cards. It's really on the ragged edge for even the 8800 GT.

http://www.learntomakeslideshows.net/xmas/xmastreepc.zip (Windows)

http://www.learntomakeslideshows.net/xmas/xmastreemac.zip (Mac)

About 35 meg...

Lin

[Lin Evans]

Hey Tom,

The HD 4670 has a 784 speed rating compared to the 8800 GT's 957 so I would say to really play everything smoothly with zero hesitation, probably a rating of around 1000 or anything better would be ideal. I originally had snowfall in each of the Christmas Tree Ornaments simultaneously, but it was way too much for the 8800GT so I removed the animation from all but two at a time for the pans and that made it just tolerable to the 8800 GT. I would love to see what the limits were on a really top end card but don't have the means to get one yet.

Each Christmas Tree ornament has a mask so that really occupies the video card and doesn't leave a lot for the animation. I was trying to push the limits with this one to see where they were for my card. I think I've reached it - LOL.

Best regards,

Lin

Hi Lin,

It was near perfect on my medium performance ATI Radeon HD 4670 - 512 MB card. Slight hesitation during ornament panning in tree, but not bad at all. Just a few days remaining before winter solstice.

Thanks,

Tom

[fh1805]

Lin,

When I try to open the Windows version, WinZip v9 says it isn't a valid archive file.

regards,

Peter

[Lin Evans]

Hi Peter,

I just checked and the file is FUBAR - LOL. I am uploading it again now and this one is definitely O.K.

Try again in 10 minutes and it should be uploaded.

I had made a small change and rezipped and the archive only contained about 4 meg of the 37 meg file.

It's up and ready now - looks like it all there.....

Best regards,

Lin

Lin,

When I try to open the Windows version, WinZip v9 says it isn't a valid archive file.

regards,

Peter

[Ken Cox]

Lin

the original was fine until the zoom to top of tree then the movements were a bit ratty, but as Tom says acceptable

would have tested sooner but was on a streaming site

will try the new one now

03:48 am

rev 1

seems to be a bit smoother on the zoom function - still acceptable

ken

[Scorpion]

Frame rate varied from 60 fps down to 10 fps on my Nvidia 9500 GT.

Check your frame rate with "Fraps" available as a free download from - http://www.fraps.com/download.php

[Lin Evans]

On the nVidia 8800 GT it's almost all 60 fps with a couple quick 62 fps and for an instant a 42 fps.

The nVidia 9500 GT has a speed rating of 368 making it a bit more than 1/3 as powerful as the 8800 GT. It would be marginal for these heavy video graphics and a little under the 8600 GT which works well for many graphics intensive files but can't handle this one at all.

Best regards,

Lin

Frame rate varied from 60 fps down to 10 fps on my Nvidia 9500 GT.

Check your frame rate with "Fraps" available as a free download from - http://www.fraps.com/download.php

[fh1805]

Hi Lin,

The new upload solved my problem with WinZip. Thanks!

The show ran successfully on my nVidia 8400GS. The only aberrations were that the zoom from the globe to just the tree scene was very jerky (and, I thought, at too fast a pace; but that's an artistic call, not a technical one) and then each time after you settled on a new globe, there was some "juddering" of the image which slowly wore off. But the pans from globe to globe were as smooth as silk.

Thanks for sharing this with us.

regards,

Peter

[CorVdK]

the 8600 GT which works well for many graphics intensive files but can't handle this one at all.

Hello Lin,

I played your show 3 times. The only visual problem I see is a slight speed difference in the snow falling just after a zoom-in, but only for max. one second.

Config.:

HP core 2 quad Q6600 2,4 Ghz

Geforce 8600 GT 512 MB

So when the show crashed completely with an overclocked 8600 GT there must be something else involved I think.

Greetings,

Cor

[Lin Evans]

Hi Cor,

There could indeed be other variables (CPU power, available RAM, etc.) which, in borderline situations, could make a marginal card perform better or worse.

When I say it "crashed," I meant that the motion was way too jerky for a realistic animation. I have made a few changes since the original which might have ameliorated some of the problems. Originally, I had eight "ornaments" running simultaneously with snow animation inside the globes plus the snow animation on the main image. I changed the format to have more slides rather than so much going on in multiple object on fewer slides. In doing this, I only have masking animation happening in two of the eight globes simultaneously on each slide. This greatly reduces the GPU stress. Still, this one will not (at least yet) make a smooth mp4 h.264 so the "jury" is still out on whether it exceeds boundaries of propriety for "most" video cards. When I purchased my nVidia 8800GT card it was top-of-the-line. Now it's only a medium powerful card. I suspect on a really good video card there would be no problem even in the original configuration.

Best regards,

Lin

Hello Lin,

I played your show 3 times. The only visual problem I see is a slight speed difference in the snow falling just after a zoom-in, but only for max. one second.

Config.:

HP core 2 quad Q6600 2,4 Ghz

Geforce 8600 GT 512 MB

So when the show crashed completely with an overclocked 8600 GT there must be something else involved I think.

Greetings,

Cor

[Lin Evans]

Hi Peter,

That's absolutely amazing! The 8400GS is really quite a slow card with very marginal, even low speed rating of 138 in benchmarks. This tells me that perhaps the loading is being shared somehow with the CPU contributing more than I'm expecting to the performance. If you get a chance, run the FRAPS test and let me know what kind of fps values you are seeing.

I had the zoom faster because it actually works better with marginal cards than a slow zoom, but I have since slowed it down for a better aesthetic appearance. The "juddering" could be caused by the transition to another slide using identical settings and a cross-fade. This is very smooth on my 8800GT, but it's running 957 benchmark speed to 138 for the 8400GS card.

Best regards,

Lin

Hi Lin,

The new upload solved my problem with WinZip. Thanks!

The show ran successfully on my nVidia 8400GS. The only aberrations were that the zoom from the globe to just the tree scene was very jerky (and, I thought, at too fast a pace; but that's an artistic call, not a technical one) and then each time after you settled on a new globe, there was some "juddering" of the image which slowly wore off. But the pans from globe to globe were as smooth as silk.

Thanks for sharing this with us.

regards,

Peter

[Lin Evans]

Hi Ken,

I'm learning some interesting tidbits with this one. There seems to be more variables at work than just the video card. Apparently, a powerful video card can compensate for other system inadequacies, but a marginal card can have a quite different result depending on other system resources. My wife's overclocked 8600 GT has problems running this show smoothly while Cor's regular 8600 GT card performs much better. I'm betting that he has a more powerful CPU and better system resources than she has. Her system has a Pentium 4 running at 3 GHz but only 1 meg (oops, 1 Gigabyte) of system RAM.

Best regards,

Lin

Lin

the original was fine until the zoom to top of tree then the movements were a bit ratty, but as Tom says acceptable

would have tested sooner but was on a streaming site

will try the new one now

03:48 am

rev 1

seems to be a bit smoother on the zoom function - still acceptable

ken

[Ken Cox]

Hi Ken,

Oops, actually I just checked and it only has 1 "gigabyte" of system RAM - yep, I'm still living in 1984 - LOL.

Paging may indeed help the streaming. I'm not certain how her system is set up, I'll have to check it.

Lin

[Gérard de Lux]

Hi Lin,

I didn't vave any problems, everything ran pefectly smooth.

Config: Intel Quad Q9450 @ 2.66 GHz; RAM 4 GB; Cache L2 12 288 Ko; NVidia GeForce 8800 GT, 512 MB; 85 Hz

Thanks for sharing.

[CorVdK]

My wife's overclocked 8600 GT has problems running this show smoothly while Cor's regular 8600 GT card performs much better. I'm betting that he has a more powerful CPU and better system resources than she has. Her system has a Pentium 4 running at 3 GHz but only 1 meg (oops, 1 Gigabyte) of system RAM.

Hello Lin,

Just for info, I have:

HP core 2 quad Q6600 2,4 Ghz

with 3 Gig of system RAM

Greetings,

Cor

[davegee]

I thought that we had decided long ago that the video card was not the only "variable" when it comes to performance?

DG

[Lin Evans]

Hi David,

It depends on what you mean by "performance." The GPU is the most "important" variable in animation smoothness, but only a part of the puzzle. A really powerful video card can compensate for lots of system weakness with PTE animations, but a really powerful CPU, plenty of RAM, proper caching, etc., seem to be able to help a mediocre video card perform at its best. When using hardware rendering, it's the GPU which gets the lion's share of the work so what one has in the way of video card is extremely important when trying to play back complex animations.

The mystery is in determining just which animations and which "constructs" are most problematic and what the best and most effective ways of building a PTE show with animations are. What theDom and others have discovered is that masks seem to be particularly problematic. I'm discovering that zooms and rotates are not much of an issue, but pans, especially with larger file sizes, can tax a system.

I've discovered that pans cause issues with MP4 h.264 smoothness regardless of resolution. What I would like to know is how much the GPU contributes, if any, to this issue and whether the contribution as far as them being problematic is due to problems in the "creation" of the MP4 or in the playback of the MP4. Players seem to be optimized for widescreen and none of them actually seem to be able to play a 1600x1200 mp4 h.264 correctly.

When the mp4's are uploaded to Youtube, they are converted to Flash and seem to run properly even at 1600x1200. The downside is the buffering necessary to get continuous smooth playback. It seems that Youtube doesn't incorporate Flash flv preload very well. Apparently, only those with extremely fast broadband can play back large HD quality files on either Vimeo or Youtube without waiting for full buffering before beginning playback.

The difficulty is that we have limited resources for testing. Most of us have at least two systems to test with (I'm fortunate that I have about seven - all with different video cards) but to gather facts, we need a systematic approach. Some time ago I posted a link on the "Equipment" forum to a video and system benchmarks site where they run systematic testing without so many variables. One can look up their video card or CPU and determine where it might rank in performance, but we still need to know which specific animation and/or features of PTE cause the most problems not only with executable playback but also with video creation.

My purpose in putting up these snow animations in various configurations is primarily an attempt to decypher the variables and find out what works and what doesn't work well with the majority of systems. There are apparently variables which we still don't understand very well. For example, Ken "often" gets really good and smooth performance out of what "should" be a rather mediocre performance video card as far as animations are concerned. On the other hand, with this last "video card killer" demo, he gets better and smoother performance on his wife's laptop. In general, laptops are far less likely to produce good performance on complex animations because even the best of the "M" cards (mobility) are magnitudes less powerful than even medium performance desktop GPU's. Even the top-end Dell (purchased from Alienware) gaming laptops costing upwards of $6,000 rarely have video cards available much better than a mid range desktop card such as my nVidia 8800 GT.

Cor's 8600 GT performs better in his system than my wife's overclocked 8600 GT does in her system. The main difference is the amount of system RAM and perhaps configurations. So "apparently," system RAM and perhaps caching can somehow "help" the video card perform optimally. What we need to know are specifics on how this works so we can design test shows which better allow users who intend to run complex animated shows to test systems before purchasing.

Best regards,

Lin

I thought that we had decided long ago that the video card was not the only "variable" when it comes to performance?

DG

[davegee]

P.S. Runs smooth as silk on my laptop (Dual Core T8100 2.1Ghz - 3Gb Ram - with nVidia 8400GT and 256Graphics Ram).

I will try it on my desktop later after I have dug myself out of the snow.

DG

[davegee]

Lin,

I have now had a chance to try this demo on my Desktop and also read your post #18.

I must admit to a little confusion since your Title for this Thread (and post #1) and the contents of post #18 seem to be suggesting different things, but I understand.

Anyway, my desktop played the demo equally well (Dual Core 3Ghz Media PC; 2Gb RAM; nVidia 7600 GS with 512RAM).

DG

[Ken Cox]

looks like the 512 mb card is the secret

ken

[Scorpion]

Hi Lin

Some additional information, to help, in your enquiry.

Windows XP Sp3

System - Intel dual core E 8440 3.0 GHz with 4Gb ram.

Graphics - Nvidia 9500 GT with 1Gb ram.

Running the "PassMark" 2D video test on my system gave an average score of 382 which compares with the rated assessment of 368.

The opening credits and the moonrise with falling snow ran at 60 fps, speed then fell to an average of 20 fps occasionally dipping to 10 fps.

The show was perfectly watchable, a casual observer would not have noticed any problems, only the frame rate indicator suggested otherwise.

Regards, Ken

[fh1805]

Lin,

Is the as yet unconfirmed performance factor related to the fact that some (all?) video cards will, if needed, take some main RAM memory to supplement their onboard memory? My nVidia 8400GS has 256MB on board but will share up to a further 768MB.

So, if it does take some RAM, then the speed of the RAM, the speed of the internal bus, the other activity (if any) in the PC which also wants RAM and possibly even the speed of the main processor could all come into play. And if the system is busy enough and RAM-constrained enough to be very actively paging, then the speed of the paging disk and the data bus it uses will also be factors.

As I'm sure you are aware, PC performance is a minefield because everything interacts together. The only golden rule is that it cannot go faster than its slowest component.

regards,

Peter

[davegee]

Peter,

To what extent does Anti-Virus etc play a part during the playing of a PTE EXE file (if at all).

DG

[fh1805]

Dave,

Depending upon the settings of your Anti-Virus software, it may scan the exe file as soon as you double-click on it to launch it. In this situation you would see a delay (perhaps 2-10 seconds) before the exe started running. The other possibility, of course, is that it could be performing a scheduled scan or a scheduled update of the detection files. Neither of these would directly affect the launch of the exe but they would consume PC resources which might be required by the exe as it runs.

regards,

Peter

[Lmoreels]

Hi Lin,

I played the MAC version on my iMAC 2,7 Hhz 4 G ram ATI radion with 256 mB video.

In the zoom effects the snow falls in short jumps.

For the rest it played fairly well

Greetings

Luc