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

[Sonicfan]

No one is saying that it would do a version exactly as the ps2 version, I think everyone agree or should, with that. The accumulative blur effect alone would have killed performance on the Dreamcast. But the fact that some early prototypes were first person shouldn't be taken as a fact of anything let alone performance.

Yep, that Neon 250 video pretty much shows that accumulation needs to be turned off. I personally turned it off on all the gta games. I hated it

 

[davis.anthony]

No one is saying that it would do a version exactly as the ps2 version, I think everyone agree or should, with that. The accumulative blur effect alone would have killed performance on the Dreamcast. But the fact that some early prototypes were first person shouldn't be taken as a fact of anything let alone performance.

I've always said in this thread that DC could do some form of a 3D GTA game, just not the PS2 version of GTA 3 we ultimately got.

 

[davis.anthony]

Important plot point:

You have different opinions from different developer. One saying something about what they expected doesn't make it true, or not.

And what's up with that formatting Tweet?

It is a important point, going by his other Tweets all they had running on DC was a primitive City with some pedestrians in it.

So they knew from a pure rendering point of view that DC could do a 3D GTA game.

It would also seem that they never got around to testing the systems they wanted (Traffic, physics, weather, world simulation)

So as Nesh said above, if DC wasn't killed off, Rockstar could have done further testing of more advanced systems, discovered that DC wouldn't be able to handle them and still moved to PS2 anyway.

 

[Karamazov]

was headhunter the closed we got to GTA3 on DC ?

 

[Sonicfan]

was headhunter the closed we got to GTA3 on DC ?

Nah, I'd say super runabout or crazy taxi

 

[Sonicfan]

was headhunter the closed we got to GTA3 on DC ?

 

[EsppiraK]

Running under Windows CE a true performance hogh

 

[Sonicfan]

For comparison sake, this is what the sequel looks like on ps2, 2 years later. Idk if it's an emulator, though, it looks really slowmo. Only thing i could see DC having a problem with if it got the sequel too is the hevier smoke and flame effects

 

[Sonicfan]

@EsppiraK Yep, that's what i was gonna say. The tools weren't there yet, unfortunately

 

[Cloofoofoo]

People are way too confrontational in this thread. That console war ended decades ago. Let go of your emotional attachments and just engage in the "what if?" discussion for fun.

In the end of the day its all in good fun. Comments may seem sharp/pointed but it isnt anything serious.

It maybe have gone off the rails slightly though from can dc do this effect which in turn leads yes nothing in the hardware stops but its best to replace with "b" effect which in turn leads to " aha so it cant handle it all thus it would have not even done and lead to abandonment". which kinda defeats the whole theoretical talk.

 

[Cloofoofoo]

It is a important point, going by his other Tweets all they had running on DC was a primitive City with some pedestrians in it.

So they knew from a pure rendering point of view that DC could do a 3D GTA game.

It would also seem that they never got around to testing the systems they wanted (Traffic, physics, weather, world simulation)

So as Nesh said above, if DC wasn't killed off, Rockstar could have done further testing of more advanced systems, discovered that DC wouldn't be able to handle them and still moved to PS2 anyway.

Thats the thing the comment you posted form the developer isnt saying that. He said they needed more test, and it was 4 MONTHS of work, you dont strictly graphical work for 4 months on these project you know that? each person has their work to fill( physics , gui , controls, memory management system). That argument is really and you shifted from your prior argument that " oh graphically it wouldnt be able to handle it" now that the lead tech designer said graphics is no issue. If there was money on it on dc they would followed through hardships and all simple as that.

Streaming the dreamcast is more robust than you think, as long as your important data is ON THE OUTER EDGE of the disc you will basically be hitting close to 1mb/s on the gdrom( look back on the video i posted on optimized BURNT CDR layout issues). Thats on them. ACtually the development team that ported Crazy taxi to psp found out the hardway psp disc wasnt fast enough to play music and stream files from the disc without causing slowdown unlike the dc version. The game is framelocked to 30 fps because of this !( half of what it runs on dc). They said if you play music from your memory stick duo it stabilizes the framerate. And this from a system that has 2 Grand theft auto games. Yeah you idea of dc cant handle the city stream is bogus.

Streaming data off the UMD also proved to be a difficult issue for us to nail down. We were on a compressed schedule to try and finish CTFW as quickly as possible. The lead times to request and obtain testable UMDs burned from SCEA were prohibitively long, so we didn’t get any UMDs made until very late in the development cycle.

The theoretical assumptions we made early on in the design of streaming the data weren’t tested until just before we were preparing to submit to SCEA for approval. What we found were long load times and decreased frame rate that we had to adjust for. You’ll notice that if you use the custom music player instead of the in-game music, the performance of the game improves a bit. This is due to the fact that the custom music player streams the data from the Memory Stick and not the UMD, thus the seek times for the rest of the game data improved.

Physics code was on crazy taxi 1 and 2 was more complex than they thought and didnt edit because they didnt want to introduce bugs they would never find when adding 2 player adhoc mode .

The original source code layout for the collisions and physics of the cars was spread out into several areas, making the task even more technically challenging. Our engineers were perplexed by this layout and knew that it would take too long to try and completely track down all of the effects that a single change might make to either part of the code.

So overall it was decided that we would not attempt to change any part of this code, both to keep the original feel of the game as true as we could and also so we didn’t accidentally create bugs that we would then never be able to track down. This isn’t a complaint against SOJ’s Dreamcast code; it’s just a fact that the original code was never intended to support two players at the same time.

Hackers found if you unlock the framerate and play the games off your memory stick the framerate is closer to dreamcast framerate with some slow down , meaning you have to play it off solid state media!( this isnt even taking into account perhapds these unlocked psp might even have the cpu unlocked to 333 to reach this state).

Edit: another thing ct pace far faster than gta3 as well so the data it demands has to be loaded pretty fast to keep up with the players high speed.

 

[Nesh]

4 months of work is too early and too premature work. It doesn't say much. Of course some other aspects besides graphics like physics etc would have been in the testing stage or early implementation, but it doesn't say much on the state of the game.

 

[davis.anthony]

Thats the thing the comment you posted form the developer isnt saying that. He said they needed more test, and it was 4 MONTHS of work, you dont strictly graphical work for 4 months on these project you know that?

You seem to be taking this very personally, it's OK to say the DC isn't as powerful as PS2 and has different limitations.

each person has their work to fill( physics , gui , controls, memory management system).

Can you provide proof that all of the systems they wanted in the game were present and working in the 4 months they worked on the DC version.

That argument is really and you shifted from your prior argument that " oh graphically it wouldnt be able to handle it" now that the lead tech designer said graphics is no issue. If there was money on it on dc they would followed through hardships and all simple as that.

It's already been acknowledged by pretty much everyone but you that DC wouldn't be able to handle the game exactly like PS2 version.

It would handle a version of a 3D GTA game, and this is what Rockstar's tests were checking

Streaming the dreamcast is more robust than you think, as long as your important data is ON THE OUTER EDGE of the disc you will basically be hitting close to 1mb/s on the gdrom( look back on the video i posted on optimized BURNT CDR layout issues).

The developer stated that streaming was one of the biggest challenges on PS2, so imagine how bad it would have been on DC with only 1MB/s compared to PS2's 5.28MB/s.

Having 1/5 the streaming performance alone with half the system RAM of PS2 would have required substantial cut-backs to environment detail and complexity to keep the streaming requirements within the 1MB budget.

ACtually the development team that ported Crazy taxi to psp found out the hardway psp disc wasnt fast enough to play music and stream files from the disc without causing slowdown unlike the dc version. The game is framelocked to 30 fps because of this !( half of what it runs on dc). They said if you play music from your memory stick duo it stabilizes the framerate. And this from a system that has 2 Grand theft auto games. Yeah you idea of dc cant handle the city stream is bogus.

PSP has 32MB (Latter models have 64MB) of system ram compared to DC's 16MB which is a huge advantage over DC.

PSP's UMD (1.37MB/s) also has a slightly higher transfer rate than DC's GD-ROM so sounds more of a game problem, than an issue with PSP's hardware and transfer rate.

Edit: another thing ct pace far faster than gta3 as well so the data it demands has to be loaded pretty fast to keep up with the players high speed.

The streaming load of one game doesn't represent the streaming load of another, so please stop the argument of trying use what other games are doing.

 

[Karamazov]

true crime games on PS2 were more complex technically than GTA.

 

[Cloofoofoo]

You seem to be taking this very personally, it's OK to say the DC isn't as powerful as PS2 and has different limitations.



Can you provide proof that all of the systems they wanted in the game were present and working in the 4 months they worked on the DC version.



It's already been acknowledged by pretty much everyone but you that DC wouldn't be able to handle the game exactly like PS2 version.

It would handle a version of a 3D GTA game, and this is what Rockstar's tests were checking



The developer stated that streaming was one of the biggest challenges on PS2, so imagine how bad it would have been on DC with only 1MB/s compared to PS2's 5.28MB/s.

Having 1/5 the streaming performance alone with half the system RAM of PS2 would have required substantial cut-backs to environment detail and complexity to keep the streaming requirements within the 1MB budget.



PSP has 32MB (Latter models have 64MB) of system ram compared to DC's 16MB which is a huge advantage over DC.

PSP's UMD (1.37MB/s) also has a slightly higher transfer rate than DC's GD-ROM so sounds more of a game problem, than an issue with PSP's hardware and transfer rate.



The streaming load of one game doesn't represent the streaming load of another, so please stop the argument of trying use what other games are doing.

I am not taking this personal like I said above it's all in good fun. But by your same logic you can't prove or disprove they they did not test it /even slight either despite your answering with such finality( oh they WILL abandon because it CANT DO IT because I said so with no proof or even try anecdotal proof). So basically I by the same coin I can say yeah it can.

Also your wrong about the PSP. The PSP always has slow drive problems which is why they offered installs to improve overall performance and lower loading times. Even gta games on the PSP have the main complaint of being jittery when played at the umd users usually ask how do I fix this? Answer by most are " hey bud overclock to 333 and install on memory stick duo ", basically rely on solid state media. Let's not mention other Dreamcast to palsp ports. Power Stone 1 is as low poly as it gets least resource intensive for dc( 1vs 1 with characters being 1.5k to 3k and same for the stages). What's the result low quality texture and missing lighting and complaints of long load times. There's more to dc than you give credit for . Same for crazy taxi streaming, despite being faster on paper its umd could not cope with the load during the actual game.

Of course the DC isn't as powerful as the PS2. Things would have to be done somewhat differently to leverage its advantages. Issue is you don't accept those answer , if it isn't the PS2 way it's not valid.

True it's the not same load but it doesn't mean what it's doing is tiny.

 

[Sonicfan]

I am not taking this personal like I said above it's all in good fun. But by your same logic you can't prove or disprove they they did not test it /even slight either despite your answering with such finality( oh they WILL abandon because it CANT DO IT because I said so with no proof or even try anecdotal proof). So basically I by the same coin I can say yeah it can.

Also your wrong about the PSP. The PSP always has slow drive problems which is why they offered installs to improve overall performance and lower loading times. Even gta games on the PSP have the main complaint of being jittery when played at the umd users usually ask how do I fix this? Answer by most are " hey bud overclock to 333 and install on memory stick duo ", basically rely on solid state media. Let's not mention other Dreamcast to palsp ports. Power Stone 1 is as low poly as it gets least resource intensive for dc( 1vs 1 with characters being 1.5k to 3k and same for the stages). What's the result low quality texture and missing lighting and complaints of long load times. There's more to dc than you give credit for . Same for crazy taxi streaming, despite being faster on paper its umd could not cope with the load during the actual game.

Of course the DC isn't as powerful as the PS2. Things would have to be done somewhat differently to leverage its advantages. Issue is you don't accept those answer , if it isn't the PS2 way it's not valid.

True it's the not same load but it doesn't mean what it's doing is tiny.

Preach!

 

[function]

The developer stated that streaming was one of the biggest challenges on PS2, so imagine how bad it would have been on DC with only 1MB/s compared to PS2's 5.28MB/s.

Having 1/5 the streaming performance alone with half the system RAM of PS2 would have required substantial cut-backs to environment detail and complexity to keep the streaming requirements within the 1MB budget.

Streaming performance - in terms of sustained linear read - peaks at 1.8 MB/s on the outer edge of the disk on DC, so it's a little above 1/3 of the PS2's peak read rate.

Still a sizeable difference, but not bad from the DC.

The biggest challenge for streaming in a open world game isn't sustained linear read rate, but transfer rates when access times are accounted for. This is more complicated than just a single figure, where things like how fast the laser head can move (and perhaps re-focus), and even how fast the disk is physically rotating become important factors.

I've got some of the DC developer docs which have some info on the GD-Rom drive, but I've yet to find the equivalent for PS2, so I can't really compare them in this regard yet. I expect they'd be fairly similar, with maybe a negligibly lower rotational latency for the Dreamcast (~3000 rpm for DC vs ~2900 rpm for PS2) but a practical advantage for PS2 coming from data being more densely packed.

Edit: and with BluRay being so much higher capacity you could potentially duplicate key assets across the disk to reduce access times, but you'd need to carefully think out how you would do this.

PSP has 32MB (Latter models have 64MB) of system ram compared to DC's 16MB which is a huge advantage over DC.

PSP stored textures in main ram, so a fairer comparison would be 32MB vs 16MB + ~6MB, so 32MB vs about 22MB. Still an advantage for the PSP, but I think PSP used CLUT textures like PS2. In practice, this would have closed at least some of the 32MB vs ~22MB gap, potentially quite a bit.

PSP's UMD (1.37MB/s) also has a slightly higher transfer rate than DC's GD-ROM so sounds more of a game problem, than an issue with PSP's hardware and transfer rate.

Peak transfer rate of the GD-Rom was 1.8 MB/s, so a bit faster than PSP's UMD. Transfer rates for the GD-ROM were actually in the range between 900KB/s on the inner edge of the high density area (the part where games were stored) to 1.8MB/s on the outer. So an x2 difference between fastest and slowest - typically it's larger than this for an optical drive, but GD-Rom uses the slowest innermost section of the disk for none game purposes.

But I suspect a bigger problem for UMD than linear read rate are its reportedly awful access times. If you needed to read from somewhere else on the disk, it was slow to get there and start reading again and that represents dead time that slows down loading. The more often you have to move between different locations, the more you suffer. In the case of Crazy Taxi on PSP struggling when also streaming music off the UMD, I think there's a good chance that's primarily because of poor access times as the head moves repeatedly between game data and music data.

But I suppose that makes sense. The DC's GD-Rom drive was physically much bigger, could use vastly more power, could spin the disk much much faster (greatly lowering rotational latency), and almost certainly could move its read head across the surface of the disk faster.

For anyone that's interested, here is the GD-Rom outline page from Dreamcast developer docs:

 

[davis.anthony]

But by your same logic you can't prove or disprove they they did not test it /even slight either despite your answering with such finality

There's more evidence to support my theory than yours.

( oh they WILL abandon because it CANT DO IT because I said so with no proof or even try anecdotal proof).

Never said that.

So basically I by the same coin I can say yeah it can.

So you're saying that you have looked at the performance specs of both CPU's, and have determined DC's CPU can do the same work as PS2's CPU?

Also your wrong about the PSP. The PSP always has slow drive problems which is why they offered installs to improve overall performance and lower loading times. Even gta games on the PSP have the main complaint of being jittery when played at the umd users usually ask how do I fix this? Answer by most are " hey bud overclock to 333 and install on memory stick duo ", basically rely on solid state media.

So what we can take from this is that DC would have also struggled with GTA3 due to streaming.

Let's not mention other Dreamcast to palsp ports. Power Stone 1 is as low poly as it gets least resource intensive for dc( 1vs 1 with characters being 1.5k to 3k and same for the stages). What's the result low quality texture and missing lighting and complaints of long load times.

PSP has nothing to do with this thread.

There's more to dc than you give credit for.

Such as?

Same for crazy taxi streaming, despite being faster on paper its umd could not cope with the load during the actual game.

There's absolutely nothing about Crazy Taxi on PSP that looks taxing from a streaming point of view, but again, this isn't a PSP thread.

Of course the DC isn't as powerful as the PS2. Things would have to be done somewhat differently to leverage its advantages. Issue is you don't accept those answer , if it isn't the PS2 way it's not valid.

1. I have said multiple times now, a DC version of GTA3 would be different to the PS2 version of GTA3
2. You might want to read the thread title again.... "Could Dreamcast et al handle this/that game/effect?

That game, being GTA3 on PS2.

And we can all agree that DC can't handle the PS2 version of GTA3, but it would handle some alternate version of GTA3.

True it's the not same load but it doesn't mean what it's doing is tiny.

You bought that game up, not me.

 

[davis.anthony]

The biggest challenge for streaming in a open world game isn't sustained linear read rate, but transfer rates when access times are accounted for. This is more complicated than just a single figure, where things like how fast the laser head can move (and perhaps re-focus), and even how fast the disk is physically rotating become important factors.

GTA3 on PS2 factored all of this in to the game files and they were arranged on the DVD disk for maximum speed.

PSP stored textures in main ram, so a fairer comparison would be 32MB vs 16MB + ~6MB, so 32MB vs about 22MB. Still an advantage for the PSP, but I think PSP used CLUT textures like PS2. In practice, this would have closed at least some of the 32MB vs ~22MB gap, potentially quite a bit.

Not a PSP thread so not going to get caught up in talking about it, but it did use S3TC which was a nice upgrade over PS2's GS.

Also important to remember that in some ways, PSP was more powerful than DC.

Peak transfer rate of the GD-Rom was 1.8 MB/s, so a bit faster than PSP's UMD. Transfer rates for the GD-ROM were actually in the range between 900KB/s on the inner edge of the high density area (the part where games were stored) to 1.8MB/s on the outer. So an x2 difference between fastest and slowest - typically it's larger than this for an optical drive, but GD-Rom uses the slowest innermost section of the disk for none game purposes.

They wouldn't build the game for peak though with GD-ROM, likely aim for a lower average, 0.9MB/s maybe?

 

[function]

Not a PSP thread so not going to get caught up in talking about it, but it did use S3TC which was a nice upgrade over PS2's GS.

Also important to remember that in some ways, PSP was more powerful than DC.

I'd forgotten about PSP supporting S3TC to be honest. They still won't have anything as small as 2bpp VQ, but they'll be higher quality than VQ while still being as small as CLUT textures. Pretty cool for a 2005 handheld console.

I'm not sure what GTA games on the PSP are using, but it would seem odd to use 4 or 8 bpp CLUTs when 4 and 8 bpp S3TC was available.

They wouldn't build the game for peak though with GD-ROM, likely aim for a lower average, 0.9MB/s maybe?

Hmm. Crazy Taxi was originally built for Naomi and ran from carts and into a system with 2x the main ram and video ram, and 4x the audio ram. They may have been keeping a DC port in mind when they were making it though - that was after all the point of Naomi and DC sharing an architecture. Probably also why the DC had a fast, robust optical drive (at least by 1998 standards), because they knew they weren't going to be shipping arcade ports on carts .....

To fit the game on DC they talked about using higher levels of texture compression, and I'm assuming they had to also stream more heavily from disk than they did from the Naomi cartridge. I think the arcade cartridge was fairly small, something like ~160MB. On GD-Rom it was even smaller iirc. That means it would have lived in the area of the disk that was 1.6 MB/s or greater, with very short distances between textures and audio tracks.

But without a dev kit or an emulator that pretty accurately replicates the characteristics of the disk drive it's very hard to know what the average actual streaming rate was. Large, infrequent, carefully ordered seeks could see a relatively high throughput from the drive, where as small, frequent, poorly ordered reads would see it drop to be very low. Small random reads on a mechanical drive can easily see throughput drop to only a few percent of peak.

Given that Crazy Taxi can run pirated from a CD - when propperly padded out to use the outermost edge of the CD - where that is still only half as fast as when on a legit double density GD-Rom, I'm more inclined than ever to think that poor access times are behind the PSP version's issues.

I can well imagine that the GTA3 developers had to put a lot of work into the streaming system and the analytical tools used to work out how to best move world data from disk and into memory. There are so many factors - chunk size, access time, ram allocation, priority of access, statistically more likely access patterns, polygon budgets, distribution of polygons and unique textures throughout the world.... and probably a hundred things I can't even think of.