[maarek99]

Your situation sounded more like some kind of freak virtual memory swapping thing was going on eventhough you had enough memory. But it doesn't have anything to do with the videocards memory. It can't. Videocards don't store information like that in their memory.

 

[mono]

I dont' think it can be any kind of "freak" thing, as I've observed it multiple times on different systems, both with systems running weathered-in OS and clean installs, systems that worked fine in every way, benchmarked fine, just this situation. I have to wonder if some readers were editing very large images back when cards didn't have more than 16-32MB of memory, that the problem didnt' disappear, rather less noticable the more video memory there was, which of course is what would happen.

Video cards dont' store information like what, exactly?
ANY kind of data can be written to a video card's memory, this much I am 100% certain of. It doesn't even have to be remotely related to video.

What is it you claim it can't do? Store a virtual screen larger than it actually outputs to the monitor? nVidia's Detonator drivers allow zooming the overlay per the Display Properties page, which would be one example of that happening.

To clarify something, I was not suggesting that the video card nor it's memory speeds up processing, rather that it speeds up the video output, the feedback to the user when user is manipulating large blocks of pixels. I am not alone in observing this happening.

... or maybe I'm wrong... it happens BUT still I find no other explaination for why the difference.

 

[AdamWill]

I can't explain what happened to mono either, but I also don't buy his explanation. AFAIK there simply isn't an API which would allow you to store 2D image data outside the currently displayed buffer in the video card's memory. There wouldn't be a reason to, either, as system RAM is certainly fast enough for the purpose. So I don't know what caused the problem, mono, but I don't think it's that the memory on your video card is being used for offscreen data in Photoshop and doing it faster than system RAM would.

 

[mono]

AdamWill,
No, system memory is certainly not fast enough. I'm not quite sure why you would believe it is, as integrated video quite easily demonstrates that framerates drop horribly when the resolution is very high, like it would be with a virtual (zoomed) resolution. Remember that what we're talking about is large images, not small ones. The prior example a 1280 x 960 image zoomed 4X, would be a virtual image of 5120 x 3840. If system memory were fast enough, then even with dependence on an aperture or (anything but the hard drive) there would be zero stuttering. It is easy enough to see there was no HDD activity. Recall my prior detail that the system benchmarks normally for CPU, memory, HDD, etc. All indicators of system performance are within normal, expected parameters.

Let's dig deeper. I just fired up Paint Shop Pro on a different system with a 64MB card in it. Sorry Photoshop fans but I'm not going to buy Photoshop for an old 2nd hand testbed. Regardless, PSP was just loaded and waiting idle, I see a memory Commit Charge of 264,196. Merely using the File->Open dialog before actually opening the image it jumps to 264,552. After image is opened, Commit Charge is 268,432. It is a 3,601KB image.

We can clearly see that the amount of memory used to store this image is roughly coinciding to it's file size. When I then ZOOM in on the image, clearly the image, picture displayed on the screen is "different", but the allocated memory is NOT, in only changed by about a dozen K. WHERE is this zoomed image bitmap being stored? It can't be stored in system memory, because there was no allocation for it. When it's zoomed 4X, it can't merely be writing system memory to the video card memory, because there are now 3 new unique pixels per each one original (since it's a 4X zoom). Where are these "new" pixels coming from? How is this data being stored right at this moment while i look at it?

Now I have a 2nd test. I closed PSP. Commit Charge now 253,504. I closed it to be sure PSP wasn't allocating additional buffers which hid the memory usage. I reopen PSP. Commit Charge now 262,640. As a process, PSP is showing 13,212K memory usage. Next, I opened a bigger bitmap file, this one is 49.4MB I check PSP's memory usage again, and it went up to 67,508. It may be creating buffers, but only ~ 3.5MB worth of them... clearly not even remotely close to large enough to hold multiple times a (zoomed) 49.4MB image. So now I zoom in on it. PSP's now using 67,500. Where is this larger bitmap i'm looking at, being stored? System memory is definitely only storing the original bitmap, the total Commit Charge is now 315,856. Maybe I am wrong, again I concede it's possible but what accounts for the preceeding memory figures?

Maybe it's not that there is a virtual image at the zoom resolution, I'm still chewing on that... but "something" is indeed causing stutters based on the size of the image, even when less than 1/3 of the system's memory is being used. In this new example system, the video card is a Geforce4 TI4200, which meets all expected performance criteria in benchmarks, as the other system did per it's ATI card.

The systems are fairly well matched except the video card. Again I want to relearn this if there is a flaw in my theory, but nobody else is doing more than claiming limited theories... I am the only one actually testing anything and do see a difference. If it'll make anyone happy, I"m about to do a clean OS install on one of the systems, if that makes a difference I'll report it but I see no reason why it should matter, as I mentioned previously the system meets expected memory, CPU, video performance per benchmarks.

Regardless of the theories some have posed, there is another element being ignored, because there is an obvious difference in smoothness of image manipulation with large, zoomed images, depending on the card. If it's not what I mentioned then it is something no one has mentioned yet.

 

[vagarach]

The only reason for a new video card in the case of image editing is for better image quality, and in this regard, onboard video is not usually of the finest quality. 64MB of Video RAM is plenty for image editing. Photoshop doesn't actually use the screen buffer, but rather system RAM. Of course, with the upcoming OSX Tiger release, this will all change due to CoreImage, but this will only affect Macintosh versions of Photoshop

As an aside, it is very easy to directly write to the screen buffer, without using API's.

 

[PFRfan]

mono, have you had a chance to swap in your G200 yet? Also, I may have missed it, but what 2D card are you using?

PSP System Requirements & Recommendations

Notice that it only adresses the need for a certain resolution at a certain color depth. No mention of video memory is made EVEN in the recommended system. Adobe does not post recommendations either. In fact, in doing multiple searches for information on the topic, it seems that nobody feels strongly enough that more video RAM (above ~16MB) increases the speed of image manipulation to actually post such a recommendation.

If your G200 has more than 8MB RAM, and it is slower than whatever you're using now, I'll be blown away. Let me know, though, because I may have to research more (i.e. test).

 

[devwild]

I do understand high resolution image editting and have done so for years. Photoshop, or any Windows photo-editting application that I know of, does not use video ram for image storage. Nor is any additional memory used by being zoomed in on an image-- the image and the pixels being processed have not changed, only their portrayal on screen. Photoshop stores the image in memory, along with all layers, channels, effects, history, etc, amounting to hundreds of megs of data when working on images for print. The more complex the project, the more heavy the swapping, and the faster the slowdown.

Until you can present an article from an authoritative source that suggests otherwise, I suggest dropping the topic before this thread becomes nothing but a flame war.

Because windows and effects are processed in pseudo 3D, VRAM does make a difference, even in the current version of OSX, and it will in Longhorn as well as the Avalon addon for XP. But currently Windows uses little more than accellerated versions of the framebuffer methods that have been around for ages.

 

[mono]

Nope, I still dont' know where the G200 is, unfortunately HUMANS here keep trying to vie for attention over my 'pooters. It's looking like I won't find the G200 today, maybe several days. I do mean to find it though, as with all treasure hunts one finds other things buried in the process.

I realize PSP doesn't have hefty system requirements, but on the other hand, who'd try to edit large images on a system with those specs? Needing IE6 for PSP? I won't even get into a rant about that, LOL.

Anyway, sure, PSP runs, it does it's job. That doesn't mean I'd be satisifed with the lag seen on the box with the 8MB card, if I were going to use it for an image editing station. I'd liken it to Half-Life 2 being able to run on a GF2MX... it will run, but it ain't pretty.

On the other hand, I could edit with the lag, live with it if I had to, which I don't. It could depend a lot on the types of editing done, in addition to the image size.

Putting the above ramblings aside, it still doesn't explain how there is an observable difference depending on the video card. I appreciate the comments from others and to a certain extent I'd have wholeheartedly agreed if our positions were reversed, except that the lag IS there, the proof is right in front of me. Frankly at this point I'm apathetic about it, since the typical person isn't going to be pairing an old 8MB video card with a more modern system, the usefulness of any tests would be quite limited. Even so the memory allocation issue is another curious thing, but better to have only Y amount of memory used and wonder why-not-more, than have 3Y memory used and wonder why-so-much.

Maybe I'm overlooking the obvious here, the card I'm looking at right now IS an ATI card, and their drivers were routinely regarded as crap, much moreso before the Radeons than after. I guess I'm looking for an excuse to admit I'm wrong- The funny thing is I'd just as soon concede I'm wrong and (anyone) is right rather than do more testing... I must be getting old.

Or just overworked...

devwild, you seem to have completely missed the central issue I was trying to get across. It is not "image storage" for an image editing application that I was referring to. "Storage" implies retrieval, not a one-way output. Photoshop (or any other image editing program) wouldn't actually have anything to do with this issue, EXCEPT that they happen to be able to generate such large virtual resolutions, and thus, the effect would be seen due to this manipulation of something with this resolution. Understand that what you see on the screen is NOT an indicator of what the system, OS, or an appilcation is doing... it can (and does) serve this pupose but above all it's what the video card is doing, as an output device.

I went about my argument all wrong though, instead of doing testing I should've just challenged anyone to come up with a test that would pinpoint a deficiency that causes the lag with one card but not the other. Lesson learned.

 

[devwild]

Quote:

Originally Posted by mono devwild, you seem to have completely missed the central issue I was trying to get across. It is not "image storage" for an image editing application that I was referring to. "Storage" implies retrieval, not a one-way output.

I do not misunderstand the issue, you are making assumptions based on crude testing and a fundamentally flawed understanding of computer graphics (and, um, memory is for storage, busses are for directional data flow), with no concrete backing for your theory. Plenty of folks have already tried to present you with content about the subject, and this content is not theory as you state, it is computer science.

I will not comment further on the topic, but I beg anyone reading this thread to please read the facts before jumping to conclusions. Ask Adobe, JASC, Microsoft, ATI, nVidia if you really want the hardcore facts.

 

[Falqon]

Quote:

Originally Posted by mono I never claimed one shouldn't have sufficient memory or fast hard drive. I ask you same question as maarek99, how do you account for the performance difference in SAME SYSTEM when only change was video card?

You probably don't have the right drivers installed, my nice video card doesn't perform very well 2d or 3d until the correct drivers are installed.

Bottom line, nobody cares about the differences between your two video cards, the test isn't in front of me, as you admitted, you could be missing something obvious. Go and find any proof that your .tif or .jpg or whatever in the heck your 4x zoomed image is, is being stored in your VRAM and then come back. Until then, stop beating a dead horse.

 

[devwild]

oops that was supposed to be an edit to the last comment - Wrong button.

Edit: last comment, the difference in speed in 2D between cards is dependent on GPU, ramdac, drivers, and other nuances of the cards. As of the last 5 years, all cards have come up to aproximately the same level in this respect. (as mentioned early on in this thread)

 

[mono]

Quote:

Originally Posted by devwild I do not misunderstand the issue, you are making assumptions based on crude testing and a fundamentally flawed understanding of computer graphics (and, um, memory is for storage, busses are for directional data flow), with no concrete backing for your theory. Plenty of folks have already tried to present you with content about the subject, and this content is not theory as you state, it is computer science. I will not comment further on the topic, but I beg anyone reading this thread to please read the facts before jumping to conclusions. Ask Adobe, JASC, Microsoft, ATI, nVidia if you really want the hardcore facts.

I asked for a test suggestion. There is nothing crude about seeing a stuttering, nor merely noting memory allocation. It is good to make preliminary observations before going off on a wild tangent. it is as real-world and appropriate a test as it gets when actually running an image editing app, and is reproducible. You call a test crude but don't have a better test in mind. It's certainly not an ideal test BUT so far I'm the only one that even acknowledges that a different video card can make a difference for (whatever) reason.

I have done other benchmarking that showed no anomolies but you're not really interested in that because you don't know what the problem is nor how to resolve it, right?

I admit my assessment may not be 100% accurate. The issue though is that all the feeback has been akin to "no that's not it". Well, then what IS it? I've tried different drivers now too. So far we have a lot of theories but when all is said and done, the fact remains that changing the video card made a difference. We have several people mentioning the obvious, that per a given resolution, it takes "X" amount of memory for the frame buffer. I've never disagreed with that, BUT, that minimum requirement for a frame buffer, is not the same thing as a maximal requirement for all video processing done by the video card.

As for your misunderstanding - To be precise, you wrote "Photoshop... does not use video ram for image storage". The misunderstanding would be that I never wrote nor implied that PS used the video memory. Rather, that PS is producing output, but the OS and video card itself has control over the memory, not PS. Do you now see the distinction between what you suggested and what I meant? Perhaps you completely understood what YOU meant, but that wasn't what you wrote. Perhaps I didn't elaborate enough to make my point clear. Either way, there was no sync between your conveyance of your grasp of what I meant, and what I actually meant.

So in the end, what is the conclusion beyond this:

1) Frame buffers need to be big enough to hold an entire frame, based upon resolution and bit depth. If someone doesn't know of any other use for video memory, then the only use they know about, MUST be the only use there is. It might be the only use, but what is going to limit this?

2) Nobody knows why one video card stutters, and nobody has any idea of how to determine why.

3) Nobody really cares - I dont' either, it's an old 8MB video card! However, it IS significant that there was a performance difference, which was the whole issue to begin with, no? So, whether anyone admits it or not, I do have proof that a video card can make a difference... it's just that nobody knows why. It also happens that this video card has less memory but can display anything else in 2D at normal speeds. Even in 3D, it can exceed 40FPS in 3DMark99, which is nothing to stand up and cheer about, but nonetheless is a sign that GPU speed and memory throughput are not the limiters, in fact the ONLY time the card has such an issue is working with large image-editing. At least, no other common PC use has shown any problem or unexpected performance issues but if ANYONE has an idea, I'm all ears! DX and driver changes have been tried.

Do not misunderstand the situation here, after I'm done playing with this 8MB card, in all likelihood it'll be put into a box and never touched again till it gets thrown away. However, this is of potential interest, determining WHY it can make a difference. I suspect that as in the past, once one further increases the sizes of their images, they too will reach an upper limit, beyond which they'll get stuttering.

 

[Falqon]

Quote:

Originally Posted by mono 1) Frame buffers need to be big enough to hold an entire frame, based upon resolution and bit depth. If someone doesn't know of any other use for video memory, then the only use they know about, MUST be the only use there is. It might be the only use, but what is going to limit this? 2) Nobody knows why one video card stutters, and nobody has any idea of how to determine why.

Enough with the dumb posts, open mouth, take out foot. If you would have read page 2, and the link that was supplied you would find that

"If you're running your PC in 1600 by 1200 resolution, with 32 bit colour, then that's a total of 61,440,000 bits for the whole display. Which is 7,680,000 bytes, which is 7,500 kilobytes, which is about 7.32 megabytes. It's not 7.5 megabytes, "

It doesn't matter how many times you zoom or play with your stupid picture, all the video card does is render the 1600x1200 screen in 32 bit color. Your video card is using 7.32 megabytes NO MATTER WHAT YOU ZOOM.

And 2, I told you, my money is on drivers improperly uninstalled.

 

[mono]

Quote:

Originally Posted by Falqon You probably don't have the right drivers installed, my nice video card doesn't perform very well 2d or 3d until the correct drivers are installed. Bottom line, nobody cares about the differences between your two video cards, the test isn't in front of me, as you admitted, you could be missing something obvious. Go and find any proof that your .tif or .jpg or whatever in the heck your 4x zoomed image is, is being stored in your VRAM and then come back. Until then, stop beating a dead horse.

Without the right drivers installed it would not run 3D mode and high color, higher resolutions, not to mention that it simply wouldn't accept the driver at all. It is quite clearly the right driver and should've been obvious based upon the facts presented. I could indeed be missing something, but it's not even remotely obvious... to myself or anyone who's replied, and the card does benchmark as working fine too.

 

[mono]

I apologize for taking up everyone's time, I can see by some posts that a few are prone to get all emotional when they are faced with unresolvable problems. There was no driver issue, the cards perform perfectly except one specific situation, zooming & panning during very large image editing. Regardless of whether you agree with WHY it happens, it does, even with a video card working perfectly at ever other task. Since there is no new/useful info to add, enough of my back-and-forth about it, I'll post back anything new that sheds light on this but nothing more.