Could Dreamcast et al handle this/that game/effect? *DC tech retrospective *spawn

[Shifty Geezer]

I wouldn't know about vc nor care. The gt3 section literally ran 23 to 30 fps . Including cutscenes. That simple. So it did constantly fall below 25 fps.

It's exactly what I said here that's been ignored and you and davis.anthony have dropped a personal crapper in this thread that's shameful.

You said "Falls below 25 fps all the time on PS2. That's the truth," without qualifying to counter this idea that there's a big disparity between the two versions' framerates. davis.anthony said "[GTA] holds 30fps very well".

I said people need to be specific and choose something like the 95% percentile, but you just stuck with your number and stretched the definition over and over to not admit your information wasn't appropriate. david.anthony tried to make sense of your statement and the only that did to him is you were counting VC's framerate, not GTAs.

Does GTA dip to 25 fps on PS2? Yes. Does it do it 'all the time'? What does 'all the time' mean? Watch the video you linked and most of the time during that video, the amount of time a framerate is shown, the framerate is 29-30, sometimes 28. It dips down to 25 in two instances for a fraction of a second. There's no way that can be categorised as "falls below 25 fps all the time", or if it is, you need to say the DC version drops to 5 fps all the time, or whatever fair comparison would be.

I said earlier this is just supposed to be fun thread. Anyone who can't engage in a fun technical discussion without an emotional attachment to the outcome - etiher side - will be banned from this thread.

To compare framerates between the DC GTA build and the PS2 build, use the 95% of time framerate. You can call PS2 29 fps if you want. Or do proper metrics and produce a box-plot of min-median-interquartiles-max framerates or something.

 

[Sonicfan]

 

[EsppiraK]

Seems there is people on a crusade to criticise anything that it is not the almighty PS2 hilarious.

 

[davis.anthony]

Looking at that video, what resolution is the game running at?

It looks really low, either that or they're using an RF video connection.

EDIT: It must be the video connection as it has much better image quality here, unless reducing image quality was one of things done to increase the frame rate?

 

[Cloofoofoo]

I know DF is doing an interview with the devs asap, and i can't wait to watch!

Df is gonna interview the homebrew devs? Well that was quick. I can only hope they do comparisons to existing versions . It would be most interesting.

 

[Shifty Geezer]

Seems there is people on a crusade to criticise anything that it is not the almighty PS2 hilarious.

There are people on both sides making unfair arguments. Everyone cut it out.

It's worth nothing that for the benefit of discussion, having a 'negative' party is good to challenge the 'positive' party. If you study debate, you'll experience this. The path to the truth requires asking questions and questions from a different perspective, trying to disprove rather than prove, are valuable in that search for truth.

If people asking questions to challenge the results of the experiment make you uncomfortable, leave the thread. If you enjoy the discussion and you are right and they are wrong, that'll come put in the conclusion and you can enjoy the warm glow of having been right all along.

 

[XboxKING]

But cd-roms did run at up to 72x speeds with the kenwood true-x 72x drive that split the laser into 7 beams to achieve a high read rate and if your using a virtual cd rom drive with an emulator you could have nvme ssd speeds

That sounds really heavy metal lol

 

[Cloofoofoo]

It's exactly what I said here that's been ignored and you and davis.anthony have dropped a personal crapper in this thread that's shameful.

You said "Falls below 25 fps all the time on PS2. That's the truth," without qualifying to counter this idea that there's a big disparity between the two versions' framerates. davis.anthony said "[GTA] holds 30fps very well".

I said people need to be specific and choose something like the 95% percentile, but you just stuck with your number and stretched the definition over and over to not admit your information wasn't appropriate. david.anthony tried to make sense of your statement and the only that did to him is you were counting VC's framerate, not GTAs.

Does GTA dip to 25 fps on PS2? Yes. Does it do it 'all the time'? What does 'all the time' mean? Watch the video you linked and most of the time during that video, the amount of time a framerate is shown, the framerate is 29-30, sometimes 28. It dips down to 25 in two instances for a fraction of a second. There's no way that can be categorised as "falls below 25 fps all the time", or if it is, you need to say the DC version drops to 5 fps all the time, or whatever fair comparison would be.

I said earlier this is just supposed to be fun thread. Anyone who can't engage in a fun technical discussion without an emotional attachment to the outcome - etiher side - will be banned from this thread.

To compare framerates between the DC GTA build and the PS2 build, use the 95% of time framerate. You can call PS2 29 fps if you want. Or do proper metrics and produce a box-plot of min-median-interquartiles-max framerates or something.

I guess only way to tell is wait to see and do 1 to 1 comparison scene wise after they take the dc version as far as they are willing to. Though according to someone here df might just do that, convenient.

As for the personal crapper thing , I said as clear as possible what the video was for, didn't care for the other game in the video at least 3 times while the other guy just went on a rampage about vc. Feel free to go an delete the unrelated posts.

I guess you want something like the screen shot below? I only have the metrics left over , actually lost the link to this video/info. Not far off from what df was showing though or what I said.

 

[Shifty Geezer]

I guess only way to tell is wait to see and do 1 to 1 comparison scene wise after they take the dc version as far as they are willing to. Though according to someone here df might just do that, convenient.

It doesn't need a wait-and-see. It just needs the ongoing discussion to be fair. People want to compare current framerate with final. That makes sense because if current framerate is same as PS2, that's not a problem. And if it isn't, we can wait and see whether the framerate gets there or not while speculating on whether it will or won't.

The problem is people are being very wishy-washy with their use of framerate, and different people are using different measures which favour their position. If you want to take PS2's lowest at 25 fps, then compare the port's current lowest. If you want to use a median, then compare DC's current median framerate to PS2, which is 30 fps using the DF video as source. The measurements have be fair and the same thing.

As for the personal crapper thing , I said as clear as possible what the video was for

The video included two games. One was mostly 29 fps, the other sub 25 fps. As you claimed GTA fell below 25 fps all the time, davis.anthony felt you must have been referring to the VC part of the video as that game, not GTA, constantly ran below 25 fps. Then the silly back-and-forth with you two not understanding that the other guy couldn't understand your point.

Neither of you respected the other enough to question their own position and make an effort to get to know exactly where they were coming from.

I'm going to ask for clarification on this statement you made because it confuses me:

Here's digital foundry Real analysis of the actual frame rate. Falls below 25 fps all the time on PS2.

I see about 80 seconds of GTA footage. With the framerate changing once per second, there are I think 2 seconds of below 25 fps, and the vast majority of framerate counter is at 28 or 29 fps. That's 1/40th of the time below 25 fps where you asserted it falls below 25 fps all the time. It we measure framerate at or below 25 fps, it's still only about 5 seconds out of 80. I cannot reconcile the evidence you presented with your interpretation of that evidence.

Can you please explain your interpretation of the video or in some other way revisit your statement so we can get on to the same page in regards PS2's framerate so we can fairly compare it to the DC port's?

 

[Liandry]

Interesting, game industry now is in such low point what most popular theme in recent days on best game and computer technology forum in the world is theme about fan made port of GTA 3 for Dreamcast.

 

[sheathx013]

I highly suspect these videos with the graph overlays unless the graph is merely used as a general "where to look for dips" and frame time stalls. If they are a similar to a Fraps Text file turned into a video graph I don't think they are synced with the videos all of the time. If they are something else, well, the process is undisclosed so I don't trust it as absolute truth.

I also remember GTA 3 dropping below 20FPS frequently, especially during heavy driving scenes while the police are after you. But even for just driving, I don't see "28-30FPS" average, especially whenever the car turns more than 45degrees. Check out this video from 44:22-44:28, 1:04:46-1:04:49 , 105:49-105:51, 6:23:44-6:23:47, 6:44-6:45 and so on. It is basically only a stable framerate while the car is driving straight and anything but stable whenever you have to turn.

Also of note, those newspapers pop up right in front of the car, and there are typically less than four pedestrians and two other cars on screen in any given moment. I'm curious to see what the PowerVR 2 Tile Based Renderer can do with the denser segments, especially since Vogons claims GTA 3 is an overdraw hog.

 

[davis.anthony]

The game suffers with terrible frame pacing which would feel like a lower frame rate.

So trying to draw performance conclusions from a video with no frame rate or frame time graph is impossible.

I also had a flick through and easily found scenes with more cars and pedestrians.

The 'rampage' chapter of the video shows a high amount of cars and pedestrians depending on what the rampage task is.

One rampage shows ~10 pedestrians on screen at once.

 

[Shifty Geezer]

I highly suspect these videos with the graph overlays unless the graph is merely used as a general "where to look for dips" and frame time stalls.

it's Digital Foundry. They have a complete timeline of frame times as well as their average score. It's as good as we're going to get. And far, far more useful than simply eyeballing and marking out some times in an hours long video where the framerate is low. It's utterly impossible to look at your evidence and derive a performance metric that can be used in a sensible discussion. If we're going to compare framerates, we need reasonably accurate measurements.

All your qualitative analysis shows is there were times when GTA on PS2 had low framerates. You only mark a few spots so can't quantify that into how much time the game spent at or around its 30 fps target.

 

[sheathx013]

The video is at 60FPS on youtube. It would be easier if we could frame advance, but at .25 speed it should be displaying at 15 frames per second. Even if we were frame advancing, if the game was 30FPS we would not see objects jump entirely across the screen in a single frame. You can plainly see while watching the background in virtually any turn that it is not blurring across the screen (30FPS), it is jumping across the screen.

I fail to see how this is a qualitative statement on my part. The game demonstrably stutters during turns, and that gets even more dramatic when the action gets chaotic (which this video rarely shows). I think it is also clear from this video that pedestrians and cars tend to be higher in count while on foot than while driving. Other graphical details change based on the camera view in car, or in the flying segment as well. These differences would be even more difficult to quantify.

 

[Shifty Geezer]

The video is at 60FPS on youtube. It would be easier if we could frame advance, but at .25 speed it should be displaying at 15 frames per second. Even if we were frame advancing, if the game was 30FPS we would not see objects jump entirely across the screen in a single frame. You can plainly see while watching the background in virtually any turn that it is not blurring across the screen (30FPS), it is jumping across the screen.

Are you saying DF isn't to be trusted in their framerates?

I fail to see how this is a qualitative statement on my part. The game demonstrably stutters during turns, and that gets even more dramatic when the action gets chaotic (which this video rarely shows).

Yes it stutters. That doesn't give us a metric to compare. The DC port also stutters.

I think it is also clear from this video that pedestrians and cars tend to be higher in count while on foot than while driving. Other graphical details change based on the camera view in car, or in the flying segment as well. These differences would be even more difficult to quantify.

We can only look at one aspect at a time. The issue was raised over framerate. How does DC's framerate compare to PS2's? It's lower? Okay, how much? Then we talk about reasons why, which might include the DC doing more working, and optimisation, etc. But to be able to compare the two different hardwares, we need objective points of comparison. We need well-sources metrics.

 

[Mittens0407]

Looking at that video, what resolution is the game running at?

It looks really low, either that or they're using an RF video connection.

EDIT: It must be the video connection as it has much better image quality here, unless reducing image quality was one of things done to increase the frame rate?

Resolution is unchanged between the two builds.

 

[sheathx013]

Are you saying DF isn't to be trusted in their framerates?

Yes it stutters. That doesn't give us a metric to compare. The DC port also stutters.

We can only look at one aspect at a time. The issue was raised over framerate. How does DC's framerate compare to PS2's? It's lower? Okay, how much? Then we talk about reasons why, which might include the DC doing more working, and optimisation, etc. But to be able to compare the two different hardwares, we need objective points of comparison. We need well-sources metrics.

I'll always say that playing the game while looking for dips, and then clearly describing the when's and where's of the performance issue is better than watching a video. Trusting an undisclosed analog to digital conversion frame counter setup over my own eyes seeing the game dip when the car turns doesn't make any sense to me. I think we both know that a software frame counter can "see" a frame as different when the scene didn't change, etc. Saying "5% low is 11FPS" doesn't mean as much to me as saying "when it gets over five cars on the screen and while panning it's below 20FPS." I'm pretty sure at this point after I record the game bottoming out during action and post a video that'll be more suspect than DF's frame counter though.

I brought up the other "modes" of GTA III because the mostly on foot "Rampage" segment was already being directly compared to the driving sequences I was discussing. Making an objective comparison between the PS2 original and the Dreamcast port is going to be extremely complex. Especially if people are going to be squinting at a moving graph and jumping on highs/lows as constants instead of watching the game without anything to draw their attention to a particular scene, or playing the game themselves.

 

[TapamN]

What I actually said is that PSP's 32MB vs 16MB system RAM is a huge advantage for PSP, which it is.

Developers didn't have access to 32MB, 8MB was reserved, so saying it's 32MB vs 16MB is misleading. It's 24MB vs 16MB of main RAM.

The RAM difference is not a major advantage. It's like how the Xbox 360 did not have a huge memory advantage over the PS3, despite the 360 having 512MB of main RAM compared to the PS3's 256MB. Both consoles provide the same total amount of RAM to the developer, once you factor in video and sound RAM. You can develop pathological cases where one system couldn't handle the other's workload (for example, the DC, with it's higher texture compression ratio, could load more textures than the PSP could handle), but the fact that the PSP's RAM is unified is not be a major factor in most cases, and can generally be worked around.

I worked on optimizing Gens4All, a Genesis emulator for the Dreamcast. It uses some very large lookup tables to accelerate emulation, so RAM ends up pretty tight. Some hacks and homebrew games have large ROMs (up to 12MB), and can't fit in main RAM in Gens4All's current state. One idea I had (but never implemented) was to store parts of larger ROMs in video or sound RAM, and use the SH4's MMU to help automate swapping things into main RAM when needed (directly accessing video or sound RAM with the CPU would be too slow). On the PSP this wouldn't be needed, so it does have an advantage, but the DC can still work around it, so I think calling it a huge advantage is wrong.

That is an uncommon case, where the main CPU needs direct access a lot of RAM. On most 3D or 2D games, a lot of RAM would be dedicated to textures/sprites and sound effects, which could be keep permanently in video and sound RAM on the DC, so there would be less main RAM pressure and no need for similar tricks. As I said, it's like PS3 vs 360. The RAM advantage the PSP has over the DC is minor overall.

 

[davis.anthony]

Developers didn't have access to 32MB.

I've never said they did.

It's 24MB vs 16MB of main RAM.

Which is a very large difference given the small amount of RAM in total.

The RAM difference is not a major advantage.

Yes it is.

It's like how the Xbox 360 did not have a huge memory advantage over the PS3, despite the 360 having 512MB of main RAM compared to the PS3's 256MB.

360 did have a large memory advantage over PS3, especially in the first couple years of development.

Both consoles provide the same total amount of RAM to the developer, once you factor in video and sound RAM.

No they don't, the OS on 360 requires less RAM, so 360 has more usable RAM for developers.

Not to mention the 10MB EDRAM on 360 that PS3 doesn't even have.

You can develop pathological cases where one system couldn't handle the other's workload (for example, the DC, with it's higher texture compression ratio, could load more textures than the PSP could handle),

PSP has a compression ratio just as good as DC does it not?

But we're massively off topic now, so back to DC.

 

[TapamN]

360 did have a large memory advantage over PS3, especially in the first couple years of development.

I suppose this depends on what your definition of a large advantage is. I would consider stuff like the PS2 port of Deus Ex, where you have to go, break up, and redesign levels to make the game work, to be something that indicates a major advantage, whereas 360 vs PS3 seems like a minor difference to me. This is really a matter of opinion.

No they don't, the OS on 360 requires less RAM, so 360 has more usable RAM for developers.

The DC and PSP have the same amount of RAM available. The PSP has 24MB of main RAM available for games, and 2MB of video RAM, totaling 26MB. The DC has 16MB of main RAM, 8 MB of video RAM, and 2 MB of sound RAM, totaling 26MB. Games on both systems can use 26MB.

PSP has a compression ratio just as good as DC does it not?

No. As I've said before, the PSP's S3TC compresses to 4 bits per pixel at best, while the DC's VQ compresses to 2 bits per pixel (plus some extra over head) so it's close to half the size.

An S3TC 256x256 16bpp mipmapped texture would be about 42.7 KB, while VQ would be about 23.3 KB. For a similar 128x128 texture, the difference would be smaller due to the 2KB overhead, with 10.7 KB for S3TC and 7.3 KB for VQ. but VQ could be reduced somewhat (typically by 0.5-1.5KB) at the cost of quality.

If you want S3TC with 4-bit alpha, it doubles in size to 8BPP, while the DC's VQ remains unchanged. However, S3TC would have much better quality. IIRC, it's basically ARGB4565, while the DC would be using ARGB4444 and have worse compression artifacts. 1-bit alpha S3TC is the same size as RGB565 S3TC.

But we're massively off topic now, so back to DC.

The thread title is, "Could Dreamcast et al handle this/that game/effect? *DC tech retrospective *spawn." Memory factors in to whether or not the DC could handle a game. I even provided an example of how the DC could work around its memory limitations to do something PSP could do easier. It seems on topic to me.