Consoles that provide(d) the best performance increase from 1stgen to last gen titles

[Lazy8s]

A deferred renderer like PowerVR captures the scene and processes the vertex workload of a frame while processing the pixel workload of the previous frame, decoupling the two workloads from typical dependency.

PS2 having only around 10 MB more RAM than a console over a year older, much less expensive, and sold at a better margin, and having only comparable RAM space to Gamecube, a console sold for a lower price and with much, much better margins is just more glaring evidence of the high cost of its RDRAM.

More performance was important to Sega when selecting their next arcade board in 1999. The system that could afford cheaper and therefore more plentiful RAM utilized that advantage in delivering higher performance.

A little 300 MHz MIPS was disappointing for a system which packed significantly more processor silicon than NAOMI2 while performing worse in-game.

Silicon costs of the memory chips were of course factored in to the higher selling prices of DDR DRAM.

The die area used to offer unrealistic peaks, like the GS's triangle set-up, should've been used to its raise real world performance.

A general purpose processor is simply inherently better for the flexibility of an algorithm, so CPU processing like on an SH-4 could offer more efficient implementations of more advanced geometry shading algorithms than a DirectX processor, too.

 

[swaaye]

I'm curious as to what developers think about the 360 in terms of what we are going to see in improvements from this point on graphically. My feeing is that we've pretty much seen what we are going to see and there will only be subtle increases.Definately not the dramtic increases that we saw on PS2.
Reason being the system is easy to develop for to begin with,and the devs working on the 360 are used to creating very advanced graphics already.Devs like Epic,Monolith,Valve and Bioware are PC devs and used to dealing with high resolution graphics and advanced effects and graphics in general. They didn't have to learn that stuff,just the system.
Not only that but we are already seeing the best looking games with unstable framerates and games being made at lower than 720p.

Considering most games are based on middleware, and that the middleware is often very popular (and thus gets a lot of tweaking from a range of programmers), I'd say 360 is basically at its limits. PS2 didn't really have middleware on any sort of similar scale. The hardware was a lot less understood, because of its quirks and workarounds that were developed over years of toiling. I think other than the multi-core CPU aspect, that the current consoles have 3D hardware that is a lot more efficient and understood than older machines. PS3 in particular is a known quantity in its 3D hardware; I wouldn't expect any surprises from it at all.

For example: Mass Effect really seems to push the limits of streaming, in particular, to get its high-resolution texturing in. It really struggles to swap textures in and out and stutters a lot. But is it really that much prettier than Gears? Not really.

Forza 2 and DIRT push the 360 to its limits in other ways. Forza makes all sorts of sacrifices to get 60 fps, while DIRT uses 4X AA and AF while stumbling at 20-30fps often enough.

 

[fearsomepirate]

PS2 having only around 10 MB more RAM than a console over a year older, much less expensive,

10MB is nothing to sneeze at when you're measuring memory in double digits. The impact of a spec increase is based on the proportion of the increase. 10MB is a 38% increase.

having only comparable RAM space to Gamecube, a console sold for a lower price and with much, much better margins

A lot happened in the year between Gamecube and PS2. Also, Gamecube had only 24 MB of main RAM vs PS2's 32 MB. Its 16MB of "A-RAM" was the cheapest memory money could buy. In addition, Cube had a number of cost-cutting measures over the PS2, so it's not really a good comparison.

A little 300 MHz MIPS was disappointing for a system which packed significantly more processor silicon than NAOMI2 while performing worse in-game.

PS2 haters always talk about its CPU as though VU0 and VU1 weren't there. Without the vector units, sure, PS2's CPU would have been totally outclassed by an SH-4. Just like a Dodge Viper minus four cylinders would be outclassed by a Mustang GT.

 

[Fox5]

I doubt 360 is at its max. Developers have barely even begun to make use of its geometry shader capabilities, so even if they have to cut back in other areas, I think 360 games will overall continue to look better year after year.

BTW, I've never understood the fascination with Sega arcade machines. Many hold Model 3 in high regard over Dreamcast, but looking at the games made for both, Dreamcast had the better looking games imo. In other words, Virtua Fighter 3 !> all.
And likewise for Naomi 2, Virtua Fighter 4 is a good looking game but I wouldn't say it beats everything on the ps2. And naomi 2 didn't even last very long, with sega soon replacing it with chihiro and triforce (mostly chihiro) and later their geforce 6 based arcade machine, and most of the rest of arcade vendors going for playstation 2 derived hardware.

I've always been curious about why powervr hardware didn't catch on more though. Did it not scale well? Did the design make it difficult to increase complexity/shrink the die/raise clock? PowerVR hardware, as far as I know, has always offered best in its class performance for the die size/memory cost, but but outside of its early efforts it's never been positioned against high end solutions. Were there complications that kept IMGtech from doubling die size or adjusting latencies for better clock speeds that kept PowerVR tech out of the high end single chip market?

 

[Crazyace]

A deferred renderer like PowerVR captures the scene and processes the vertex workload of a frame while processing the pixel workload of the previous frame, decoupling the two workloads from typical dependency.

That has nothing to do with deferred rendering ... I think you are confusing general vertex processing with the binning of triangles required for deferred rendering. Maybe the elan chip buffered a frame of transformed geometry - but I'm not sure that would be wanted for a low latency game like VF4.

PS2 having only around 10 MB more RAM than a console over a year older, much less expensive, and sold at a better margin, and having only comparable RAM space to Gamecube, a console sold for a lower price and with much, much better margins is just more glaring evidence of the high cost of its RDRAM.

As dreamcast failed in the market, and was dumped - I dont think you can talk about better margins. ( Wasn't it 26MB(16+8+2) vs. 40MB(32+4+2+2) anyway )
Gamecube was later - and used it's own non standard expensive ram as well

More performance was important to Sega when selecting their next arcade board in 1999. The system that could afford cheaper and therefore more plentiful RAM utilized that advantage in delivering higher performance.

I guess that's why Naomi2 was soon replaced was Xbox, and Geforce based hardware.

A little 300 MHz MIPS was disappointing for a system which packed significantly more processor silicon than NAOMI2 while performing worse in-game.

Silicon costs of the memory chips were of course factored in to the higher selling prices of DDR DRAM.

The die area used to offer unrealistic peaks, like the GS's triangle set-up, should've been used to its raise real world performance.

in your opinion I guess... and why do you keep confusing DDR ram with rambus ram?

A general purpose processor is simply inherently better for the flexibility of an algorithm, so CPU processing like on an SH-4 could offer more efficient implementations of more advanced geometry shading algorithms than a DirectX processor, too.

So are you suggesting now that the Elan processor was a complete waste of space? Please make up your mind - What is a 'directX processor'?
VU1 was forward thinking in the way it could act as a 'geometry shader' as well as simple transform and lighting. It was incredibly more efficient than a SH-4 at this due to it's wide datapaths from memory and it's direct connection to the GS, and for algorithms that needed more flexibility the EE+VU0 combination was also available.

 

[Crazyace]

I think 360 has a lot of headroom left ( I even think the Wii has a few surprises )
PC's are leading the graphics race though, and more effort will be spent 'behind the scenes' to ensure parity with what people expect.

I expect that as more titles on the PC used multilayering, and transparencies - along with more effective Z culling techniques - that the advantages of the PowerVR deferred rendering showed up less. However the point is moot, as PowerVR moved away from the PC space to their current markets

 

[Lazy8s]

While an SH-4 also is more functional than a specialized shader or T&L unit, it would've been no substitute for Elan at accelerating VF4, either. Even reduced to three-quarters, 75 MHz speed, Elan would've made NAOMI2 significantly faster than PS2 at VF4.

For almost two years, NAOMI2 was SEGA's premiere arcade board, and it continued to receive high profile releases, like new VF4s, Initial Ds, and World Club Champion Footballs for a while afterward, the typical lifespan of a SEGA arcade board.

 

[Crazyace]

While an SH-4 also is more functional than a specialized shader or T&L unit, it would've been no substitute for Elan at accelerating VF4, either. Even reduced to three-quarters, 75 MHz speed, Elan would've made NAOMI2 significantly faster than PS2 at VF4.

SH4 had functionality ( it was a general purpose CPU after all ), but it didn't match the performance of a vertex shader, or VU unit. ( The EE+FPU on it's own would have been able to match the SH-4 in terms of real world graphics transformation, due to the better DMA paths to the GS, and also the increased number of registers )
And do you have any real justification for comparing a Elan at 75MHz? In my opinion memory was the major limitation for the conversion of VF4 to PS2, but I would be interested in facts that show otherwise.

For almost two years, NAOMI2 was SEGA's premiere arcade board, and it continued to receive high profile releases, like new VF4s, Initial Ds, and World Club Champion Footballs for a while afterward, the typical lifespan of a SEGA arcade board.

I think that while it may have been the premiere Sega arcade board, it was completely outclassed by the home consoles. GT3 looked way better than initial Ds ( and GT4 went on to support HD and progressive displays ) , and the Xbox and Gamecube also eclipsed it.

 

[Lazy8s]

The vertex lighting reduction on top of the halved geometry in PS2's VF4 shows that its deficiency to NAOMI2 was also in T&L performance. Elan was calculated to be faster than even the GeForce 4 era chips at T&L speed.

Only separate HD and proscan was supported by GT4 (as in, HD or proscan but not HD and proscan) because field rendering was its method of supporting its above-NTSC-but-not-even-HD mode in the first place. Field rendering may be clever, but it's not such a technical feat.

Subjectively, I've never liked the graphics in the PS2 Gran Turismos because of the harsh shimmering; Metropolis Street Racers graphics have my distate for the same reason.

 

[Crazyace]

The proscan output was very nice - and with a 1080i TV ( as was more common in Japan ) the interlaced 1080i ( by 640 ) output gave a really clear view of the road - the lack of horizontal res wasn't as important as the extra perceived vertical resolution to me.
( I still think the vertex lighting reduction was more a programmer related issue, but we can agree to disagree on that )

 

[Fafalada]

The vertex lighting reduction on top of the halved geometry in PS2's VF4 shows that its deficiency to NAOMI2 was also in T&L performance.

The fact that halved geometry characters actually Fixed all the terrible skinning problems of arcade version, also shows that Naomi2 had its own deficiencies in T&L, most likely a result of Elan fixed pipeline limitations.

I tend to side with CrazyAce on the lighting though - local lights require writting a few extra shaders, which may very well have been out of the timeline scope for the port.

 

[Lazy8s]

The bounty of extra features, not the engine, of the PS2 version was the bound on its development time.

Dengeki PS2: When did development actually begin on the PS2 version?

Kataoka-san: We did preliminary research on the PlayStation 2 more than a year ago, before the arcade version of Virtua Fighter 4. Actual development began once the arcade version was finished.

Dengeki PS2: So the translation took about six months then?

Kataoka-san: About that [smiles..] If we simply wanted a direct port of the arcade version we could have released it in mid-December, it was technically finished then. Since then we've focused on adding new features to the PlayStation 2 version, making sure it's 120% done.

Suzuki-san: There was another arcade project planned, but the staff decided this was our highest priority.

 

[Nano]

Especially considering the challenges it presented early on, there were some amazing results on PS2. Jak and Daxter was one such title that pushed the envelope with such large, seamless environments, great effects and animation at 60fps, and to think games got even more impressive. With the way titles utilized the vector units capabilities over a period of time, I would say the PS2 made the best performance increase from first to last gen. I think 360 and PS3 will be an even more exciting prospect.

 

[Statix]

The difference between a Halo 1 and Splinter Cell: Chaos Theory or DOOM 3 on the original Xbox is really pretty drastic to me.

 

[Crazyace]

wreckless ( or double steal ) actually showed off most of the graphical effects as a launch title ( no comments about the gameplay though )

 

[ShaidarHaran]

I think people are getting into the "why" too much instead of just answering the question at hand, to which the obvious answer is the PS2.

 

[Shifty Geezer]

Huh? The 'why' is essential understanding to answer the question. You need ot know what improved to know which improved the most. Just stating an 'obvious' answer without any explanation isn't real discussion now, is it?

 

[function]

The 'obvious' answer isn't always the correct one!

 

[ShaidarHaran]

Huh? The 'why' is essential understanding to answer the question. You need ot know what improved to know which improved the most. Just stating an 'obvious' answer without any explanation isn't real discussion now, is it?

All you have to do is look at some screenshots of first gen games compared to last gen games. Anyone that's played a game or two (and even some that haven't) should instantly be able to distinguish massive differences in graphical quality, especially for the PS2. Every console has natural progression during its lifespan, but no other console had as big of a difference, unless you want to talk about a single comparison of a pair of titles. PS2's launch games looked like slightly-improved N64 games, last gen titles were on-par with Xbox games of the time.

 

[Crazyace]

Maybe the reverse should apply,
In some ways the DreamCast and the Gamecube showed the least increase from 1stgen to lastgen
Rogue Squadron pretty much had everything, and Luigi's mansion had lots of technical effects ( it just wasn't that good compared a game )
Soul Calibur and DOA2 were pretty much launch titles for DC, and things didn't get much better than those 2.