Texture cache of the N64 vs. that of Playstation

[phoenix_chipset]

But to address your previous angry post The 64 was state of the art in 95 when Nintendo finished its design, the only held it back to 96 because of mario 64 not being ready. Any argument that Nintendo used year old hardware due to cost or incompetence is bogus.

 

[swaaye]

Boss Game Studios did custom microcode. If you haven't tried World Driver Championship, you might want to look that up.

His Gamecube opinion seemed to be that it was easy to get max performance out of, but it could be outperformed by a team that put major effort into leveraging PS2 to the limit. Because PS2 was a pain in the ass to max out. Gamecube visuals would look cleaner though.

 

[function]

I'm not interested in a biased opponent with regards to the gamecube, it's an undeniable fact it was more capable than ps2 in everything but alphas and dynamic geometry.

1) That's incorrect; 2) viewing people who see the GC hardware differently as "opponents" is interesting to say the least and 3) dismissing ERP's opinions because you've decided he's biased is shitposting in technical forums and unhelpful. My opinions of hardware are heavily influenced by real world experiences of people who've worked on it.

And I rate the GC above the PS2, but that doesn't mean any of the things ERP measured, he lied about.

I'm not calling erp a liar, but it's one man's assessment : from the perspective of someone who didn't ship a game on it, let alone solely focus on it for an exclusive game with a large budget.

I'm interested in which GC game you shipped, after hand tuning it to perform beyond the limits of the T&L hardware, to be able to dismiss his findings.

I'm especially interested to know why it's irrelevant that ERP's findings sit nicely with the talented, AAA, Burnout crew's assessment of the hardware's relative merits.

You and the rest of you old boys here can cherrypick which data is reliable and agree on it here, but the agreement will be isolated to the forum.

Okay, so what we discuss here doesn't matter, and all the people outside will decide on what they think is true in isolation of the acutal merits and capability of the hardware.

Cool. That's a big win.

I'm esp. not interested in discussing with you in particular since you already assumed I was ignorant on what hardware was available in 96. Move on if i'm just a ninty fanboy yeh?

This isn't a discussion. You don't want one. But don't assume that you can post outrageously nonsense subjective opinions in a tech forum and not have them challenged. And don't assume you can shitpost about a highly experienced developer here - with no counter evidence of your own to present - and not have it challenged.

But to address your previous angry post The 64 was state of the art in 95 when Nintendo finished its design, the only held it back to 96 because of mario 64 not being ready. Any argument that Nintendo used year old hardware due to cost or incompetence is bogus.

Nintendo are one of the most competent companies that have ever been involved in the games industry. But the N64 architecture was dated on release. A more cutting edge solution could have been made, but would have been more expensive and would have been contrary to Nintendo's philosophy. Cost was certainly a factor, they bought an already designed but dated architecture, like they did with GC, Wii, WiiU, Switch.

That's right. That what they do. They buy none bleeding edge hardware at a reduced rate and ship it when they're ready.

It. Is. What. They. Do.

And it works for them.

 

[phoenix_chipset]

What was more advanced than 64 hardware in 95? Besides a high clocked pentium perhaps, but nothing graphics side was more advanced. You're full of it. You don't choose hardware to manufacture, start making a game (mario 64) then scrap your old hardware design so you have to start the software from scratch. You make it sound like they choose a dated design at the START of development - they didn't, not with 64 or GC. Wii, Wii U and switch, yes. When Ninty signed with art x for the cube's gpu. It was state of the art.

Like I already said? GC was weaker in terms of transforming geometry which is what erp and burnout devs spoke to. You're the blowhard trying to win arguments for the sake. I'm just saying - if your boy says GC isn't as capable as ps2 in terms of detail, that's plainly false. I point to the myriad of games with per pixel lighting, high poly counts and basically any effect xbox does sans normal maps. If anything was a dated design, it was the ps2 with its n64 esque design in the era of shaders. Nobody said erp is biased I said YOU and the mods were biased for accepting one mans opinion as scripture. I don't doubt what he experienced in development : doesn't mean other devs hit the same limitations.

The sync with erp and burnout speaks only about the cube's skinning deficiency.

 

[swaaye]

You could probably find other developer comments here. I don't remember if Fafalada wrote about N64. It's unfortunate that you weren't here a decade ago to chat with these people.

 

[phoenix_chipset]

You could probably find other developer comments here. I don't remember if Fafalada wrote about N64. It's unfortunate that you weren't here a decade ago to chat with these people.

In a perfect scenario i'd like to chat with the high profile developers of exclusives games for each machine, compare their experiences with other developers on the same machine, what they could achieve, what the bottlenecks are and compare that data. Who here would think all conclusions would be identical?

I'm bowing out of the convo for now, maybe i'll pick up world driver 64.

 

[mpg1]

There is some nostalgia here for me but I preferred the aesthetic of N64 graphics over PS. I remember even at the time saying N64 games "looked smoother". I guess you could use the comparison of N64 having an "analogue look" and PS having a "digital look." I think overall N64 games hold up better for what I consider for simple low-end 3D graphics.

Also this game destroys everything from that time period imo:

 

[Lalaland]

It was a fun game (except for the protect the hacker bit which can die in a fire) but it had some very funky issues such as the shoulders pinching out of existence when the arms were held at certain angles. Just for accuracy I'll post the N64 capture version here as even back in the day I hated the filtering that was applied to every N64 texture and the emulated footage cleans that up.

 

[phoenix_chipset]

Just for giggles I took a second look at what graphics hardware was available from 1993 onward, and for 1995 the PC alternatives to silicon graphics hardware was the ridiculously weak NV1 which used quads, and had a low 12 MP/s a second. Likewise the ati rage 1, is built on the same 500nm process (older than 64s 350 nm process), didn't launch until spring of 1996, and had peak fillrate of 26 mp\s. Maybe there was some clockspeed variance on cards back then but it couldn't have been much. That's lower than 64's rating of 32.5 with z enabled (which sounds like it might not take into account custom micro code? Not sure.)

So, unless Nintendo could've used magic to shove a sega model 2 or another arcade board in a home console, (which model 2 is 1993 so saturn and ps1 was dated as well with this logic, shit even voodoo 1) N64 was indeed top of the heap when its design was conceived.

With regards to the cube, the only things better available were the ati r200 (again, too late as art x design was finished and chosen at this point of October 2001), and the xbox gpu. Nvidia didn't even have something as good for the pc market. Even with its top tier graphics and faster cpu, xbox had corners cut that weren't there on the cube. Lower memory bandwidth for nearly everything but for e.g. many alpha textures which couldn't be stored in cube's eDRAM, and a slower fsb than gc made even worse given xbox's higher clocks.

My main point is this pervasive idea that Nintendo cheaped out, or didn't make the best designs they could back in those days is nothing but intellectually dishonesty.

 

[Roly]

Much interesting discussion here (and some intriguing on a Spanish n64 sub-forum I recently encountered). In terms of the fill-rate numbers, I think many years ago, when this was discussed (c. October 2002), there was a fairly well supported observation that the fill-rate was, in practical terms, a very long distance from the 62.5Mpx, with corresponding diminution in the multiple cycle mode rates. I very much like the aesthetic of the console (amongst a handful I view most fondly), and having spent a good portion of 2017 holiday with WDC, I found it very impressive (no obvious z-errors, possibly one subtle one with a fence priority in a replay). There was some discussion on another forum, which I have lost track of, as to the possibility of using part of the instruction memory space for texturing, but it seemed very speculative hypothesis; and another on some exotic use of the colour combiner.

 

[Squeak]

ERP talked a lot about the cache and some other aspects years ago.

Yeah, I read those posts with great interest back then, and before this thread. But he doesn't say anything explicitly about whether the CLUT is stored in the texture buffer though.

For 64x64x4 textures plus MIPmaps it would still fit, but it would rule out anything larger, while using the build in MIPmapping (you'd have to vertex alphablend between near and far geometry or ignore the artifacts).

As said, it doesn't appear that most games are in fact using 64x64 textures. It's actually a healthy size of texture for the target resolution and the time.
If all textures where that size N64 games would have looked a lot different.

Edit:
A breakthrough here I think: http://fgfc.ddns.net/PerfectGold/Textures.htm

Color-Indexed Textures

These are some restrictions on the size and placement of CI texture maps within the TMEM. The TMEM is actually partitioned into two halves. Four texels are sampled from the first bank and fed into the second bank for texture/color/index table lookup (TLUT).

8-bit CI textures all require 2 KB (256 x 64 bits per entry) second half banks to hold the TLUT, while 4-bit CI texture can have up to 16 separate TLUTs.

Since the Upper half is reserved for TLUTS all tiles must reside below 2Kb.
This means that a 64x64x4 CI texture does not have room for Mip-Map Tiles unless they are loaded seperatly.

The Largest single Automatically Mip-Mapping (3Level) Texture that can be loaded is:
64x48 (32x24 + 16x12) 4Bit = 1536+(384+96)=2016byte

An example taken from PD is the washing Mashine at 64x56x4. This can have 1 more Mip-Map Tile at 32x16 to take it up to exactly 2048.

However refer to wrapping requirements on both these examples.


Below is a list of types and the total number of Texels that will fit in TMEM

• 
  
  4-bit(I,IA) 8,192
  4-bit Color Index 4,096 (plus 16 palettes)
  8-bit(I,IA) 4,096
  8-bit Color Index 2,048 (plus 256-entry LUT)
  16-bit RGBA 2,048
  16-bit IA 2,048
  16-bit YUV 2,048 Y's, 1,024 UV pairs
  32-bit RGBA 1,024

The problem I see is that Nintendo Waisted TMEM in CI Opperation. There is NO benifit in 8bit over 16bit in TMEM. In fact 16bit has a slight advantage since 2 texures can co-exist while 8bit can only have 1 TLUT.

4bit CI should have loaded TLUTS back from end of TMEM and dispenced with mirroring 4 times the colours.


Now, what Id love to know is how GoldenEye's manual Mip-maps work.
I can see a 128x47 and 64x16 fitting together but not a 128x47 and 64x24...
Also, Rare used resolutions lower than 16 wide which is pretty pointless.
eg

Oops, My Mistake, I miscalculated... They do fit
((128*47)+(64*24)+(16*13)+(16*7)+(16*4)+(16*3))/2=3992 (<4096=True)

Trev​

 

[Squeak]

Much interesting discussion here (and some intriguing on a Spanish n64 sub-forum I recently encountered). In terms of the fill-rate numbers, I think many years ago, when this was discussed (c. October 2002), there was a fairly well supported observation that the fill-rate was, in practical terms, a very long distance from the 62.5Mpx, with corresponding diminution in the multiple cycle mode rates. I very much like the aesthetic of the console (amongst a handful I view most fondly), and having spent a good portion of 2017 holiday with WDC, I found it very impressive (no obvious z-errors, possibly one subtle one with a fence priority in a replay). There was some discussion on another forum, which I have lost track of, as to the possibility of using part of the instruction memory space for texturing, but it seemed very speculative hypothesis; and another on some exotic use of the colour combiner.

Actually, it appears Rocket Robot on Wheels is using the combiner in one rare instance, to do some bump mapping like stuff on a single piece of geometry in the Tim Burtonesque level. I remember being quite surprised by it at the time and explored with the first person camera to be sure.
Someone at Sucker Punch had a bit of time off and felt adventurous perhaps.

There is a weird misconception that it is bad that fillrate drops when it is used. That is the whole point. To use it. Just use it on the right things.
With good culling overdraw for plain surfaces has been around three for decades.

 

[phoenix_chipset]

Hmm yes I saw Glenn plants review on rocket and the textures seem pretty nice though geometry is simple.

Also I decided I'm mad enough to buy an ultra HDMI modded 64 lol.

 

[Squeak]

Yeah, I read those posts with great interest back then, and before this thread. But he doesn't say anything explicitly about whether the CLUT is stored in the texture buffer though.

For 64x64x4 textures plus MIPmaps it would still fit, but it would rule out anything larger, while using the build in MIPmapping (you'd have to vertex alphablend between near and far geometry or ignore the artifacts).

As said, it doesn't appear that most games are in fact using 64x64 textures. It's actually a healthy size of texture for the target resolution and the time.
If all textures where that size N64 games would have looked a lot different.

Edit:
A breakthrough here I think: http://fgfc.ddns.net/PerfectGold/Textures.htm

Color-Indexed Textures

These are some restrictions on the size and placement of CI texture maps within the TMEM. The TMEM is actually partitioned into two halves. Four texels are sampled from the first bank and fed into the second bank for texture/color/index table lookup (TLUT).

8-bit CI textures all require 2 KB (256 x 64 bits per entry) second half banks to hold the TLUT, while 4-bit CI texture can have up to 16 separate TLUTs.

Since the Upper half is reserved for TLUTS all tiles must reside below 2Kb.
This means that a 64x64x4 CI texture does not have room for Mip-Map Tiles unless they are loaded seperatly.

The Largest single Automatically Mip-Mapping (3Level) Texture that can be loaded is:
64x48 (32x24 + 16x12) 4Bit = 1536+(384+96)=2016byte

An example taken from PD is the washing Mashine at 64x56x4. This can have 1 more Mip-Map Tile at 32x16 to take it up to exactly 2048.

However refer to wrapping requirements on both these examples.


Below is a list of types and the total number of Texels that will fit in TMEM

• 
  
  4-bit(I,IA) 8,192
  4-bit Color Index 4,096 (plus 16 palettes)
  8-bit(I,IA) 4,096
  8-bit Color Index 2,048 (plus 256-entry LUT)
  16-bit RGBA 2,048
  16-bit IA 2,048
  16-bit YUV 2,048 Y's, 1,024 UV pairs
  32-bit RGBA 1,024

The problem I see is that Nintendo Waisted TMEM in CI Opperation. There is NO benifit in 8bit over 16bit in TMEM. In fact 16bit has a slight advantage since 2 texures can co-exist while 8bit can only have 1 TLUT.

4bit CI should have loaded TLUTS back from end of TMEM and dispenced with mirroring 4 times the colours.


Now, what Id love to know is how GoldenEye's manual Mip-maps work.
I can see a 128x47 and 64x16 fitting together but not a 128x47 and 64x24...
Also, Rare used resolutions lower than 16 wide which is pretty pointless.
eg

Oops, My Mistake, I miscalculated... They do fit
((128*47)+(64*24)+(16*13)+(16*7)+(16*4)+(16*3))/2=3992 (<4096=True)

Trev​

Also: http://fgfc.ddns.net/PerfectGold/SetTile.htm

​

 

[Squeak]

Hmm yes I saw Glenn plants review on rocket and the textures seem pretty nice though geometry is simple.

Also I decided I'm mad enough to buy an ultra HDMI modded 64 lol.

Get it! it's a great game. Best platformer on the machine after Mario 64 IMO.
It also has some of the best and most technically impressive music on the system.

 

[vipa899]

N64 released in summer 1996 in japan, for sure it didnt best a high end pc in anything by then. Not often ive seen someone saying that lol.

PS2 being more capable then GC seems strange but its something i can believe.

 

[phoenix_chipset]

N64 released in summer 1996 in japan, for sure it didnt best a high end pc in anything by then. Not often ive seen someone saying that lol.

PS2 being more capable then GC seems strange but its something i can believe.

Like I posted above N64 was ahead of PC in 95 and almost all of 96. Look up what was available for yourself. Voodoo 1 wasn't out until October.

Ps2 was trounced in everything but geometric manipulation and alphas. Which, it beats xbox in these regards as well. With polygon counts, lighting, textures and effects gamecube was next level.

There's a reason gamecube and xbox were seperated from ps2 in digital foundry's latest h20 videos, go watch them if you haven't.

 

[phoenix_chipset]

I think a lot of people look at PC through today's lens and sometimes assume it was always home to the most advanced graphics hardware, but that definitely wasn't the case in the 90's. Dreamcast as well, beat the pants off pc hardware in 98 and still ahead 99 I do believe. Truly in terms of tech there's an argument to be made that DC was the most state of the art console ever released. IIRC, there was even a pc gpu based on DC in 99 but it wasn't as good.

Believe it or not, Snes and genesis were much better at 2d graphics than PC at the time they released and for much time after.

The 90's was the arcade's time to rule with the ultimate in graphics technology, and this didn't change until 2001.

 

[Nesh]

N64 released in summer 1996 in japan, for sure it didnt best a high end pc in anything by then. Not often ive seen someone saying that lol.

PS2 being more capable then GC seems strange but its something i can believe.

Is it? I think it was in some respects better and in some other worse. Overall I d say the GC was better. Resi 4 was a case that demonstrated this

Edit: unless you meant at the time of their respected launch PS2 more capable compared to what was available

 

[vipa899]

Like I posted above N64 was ahead of PC in 95 and almost all of 96. Look up what was available for yourself. Voodoo 1 wasn't out until October.

N64 ahead of PC in 1995 how? The machine was released to the public at its earliest in june 1996, thats japan release, october 1996 for usa, and we in EU got it spring 1997.
Voodoo 1 is ahead of N64 in most ways, then you have CPU, 200mhz pentium beats the N64 too for most things, PC's had much more ram aswell as bigger storage medium.
Not only that but 3d GPU hardware was moving fast during this time era.

I always saw the N64 as equal to the PSX somehow, just that the PSX was a much more successfull console and the better one imo.

There's a reason gamecube and xbox were seperated from ps2 in digital foundry's latest h20 videos, go watch them if you haven't.

Im very well aware the xbox was about a generation ahead of both GC/PS2, it wasnt that much of a succes/not that much exploited as PS2. Hower, i think the GC was one of, if not Nintendos best console so far. It was small, yet powerfull and a competitive price, cheaper then PS2/xbox.

Is it? I think it was in some respects better and in some other worse. Overall I d say the GC was better. Resi 4 was a case that demonstrated this

Edit: unless you meant at the time of their respected launch PS2 more capable compared to what was available

I meant that, i myself see the GC as the more powerfull/capable console overall, but if ERP/a developer says its the other way around for most things, it wouldnt strike me as something unbelieveble, just that it didnt really show. I put the GC closer to PS2 then Xbox atleast, i know many dont, but i think its closer to the truth.

https://www.neogaf.com/threads/the-...asts-throughout-console-history.240021/page-5

Scroll down and read dark10x's post, he sums it upp nicely about pc vs N64, spot on, and how i experienced the pc vs n64. Btw, isnt dark10x one of DF's people?