Xbox 2 coming in Nov-Dec 2005 - Revolution could be stronger

[Qroach]

The Flipper's clockspeed was dropped by 40.5mhz. You made it sound like an "extreme" downgrade, & made no mention of Gekko's 80mhz upgrade initially when they were both were in fact altered at the same time & the upgrade was not done "later."

I didn't make it sounds like anything more than it was, as there's was no need to. Yes, I only said it was downgraded since I didn't need to talk about the CPU because I was referring to the performance states in the IGN article which was strictly referring to the GPU.

I wasn't referring to the GC's 8 standard hw lights, but "a custom" lighting method. Something akin to Light Scattering & Bumped Specularity comes to mind. (any technique that involves the global lights being fetched from a texture, or any other location instead of being computed by the lighting hardware)

I have no idea where you’re getting this idea from but, you certainly wouldn’t have any idea what the performance on a custom lighting scheme would be like. If it’s not computed by the hardware they why would you have posted that IGN article? I’m getting the impression you forgot what we were arguing about. I’m sure you recall what Teasy wrote regarding his thoughts on the gamecube being the most efficient console because it (in his opinion) drops performance the least from the specs released.

You seemed to want to continue his argument on, or try to further support his opinion, but what you just wrote above makes little sense when you what to argue efficiency. It clearly wouldn’t be all that efficient to not utilize a section of the hardware because it’s not flexible and perform those tasks on a CPU since the main CPU’s on the other consoles would be sitting idle in most cases. Taken the argument we had in mind, how can you argue for a custom lighting scheme and turn around and tell me this is an advantage of the game cube with regards to efficiency without having any idea what the performance would be like?? For instance, what if Xbox could perform custom lighting in vertex shaders? That would more than likely be faster than performing the same task on the CPU.

This is exactly what I said before, that you wouldn’t have any idea how efficient each console is unless you code on all of them.

Basically any vertex shader operation can be done on the CPU *OR* the GPU, but if it's done on the CPU it obviously uses up processor time and adds the overhead of sending the resultant vertex data to the GPU. On the GC, this must be performed on the Gekko. So the increase in speed would prove beneficial here also.

Well yes of course you can transform and light on the CPU. That’s old school knowledge. That does only depend if the CPU could perform the task faster than you could on the PS2 CPU or Xbox in a vertex shader program (or CPU) to be an advantage. I’d be willing to bet from what I’ve seen of performance on the Xbox, that a vertex shader on the GPU will be head and shoulders faster than any vertex program running on Gekko.

So once again, how is this an advantage in efficiency? How does this support Teasy’s comments that gamecube is the most efficient console, because it (in his opinion) drops performance the least from the specs released.

Neither have you, so you can place your unaltering faith in them if you choose. But then any game that exceeds these benchmarks numerically would mean what? A more procient & efficient code was created & manipulated in a real-time gaming scenario correct?

Um, yes I actually have worked on products for all the consoles currently on the market. I’ve worked closely with really experienced game developers on each platform, which is how I’m able to form an opinion not based on rumor or IGN/Anandtech. I personally have a very good relationship with Sony, MS and Nintendo developer support thanks. So don’t try and tell me I haven’t done anything when you’d clearly not have a clue!

…and yes, I much rather put my faith in something EA does as they have a lot more credibility regarding console development than you do ( I don’t mean for that to sound rude but there’s no real way to put it). It’s is a bit odd to see you taking a hypothetical situation, and using that to support your argument, when you really don’t know one way or the other. So what If a few other games get better performance while they are doing something completely different than EA. Typically those games would be exclusive titles to the platform obviously.

Let me put it this way. When you want to argue which console is the most efficient, then you have to try and perform similar tests on each platform. Yes I’m sure they can all be optimized in different ways, but in the case of EA they did some synthetic tests and then tried to do tests that left the CPU idle as they know there’s going to be AI, physics and other areas of code spending cpu cycles.

Strange that their cross-platform games generally aren't a reflection of these tests then isn't it? Or did you completely ignore my Madden example? I could name quite a number of other EA games if you like. (ROTK, The Sims, etc, etc.)

You can say they aren’t a reflection of console performance if you want, but you really don’t have anything to back up that opinion. I think the performance in EA’s games generally is a reflection of the performance achieved in tests. That being, Xbox is usually the top performer, with gamecube toping out around the same level as PS2 in real world multiplatform game performance. Just because you don’t believe it to be so, doesn’t mean it’s not what’s true.

You still fail to mention where the majority of their coding is now done. Or where they've gained most of their expertise. How large is the japanese PC gaming market again? Within in a university level programming environment, yes. But I was referring strictly to one region, & you still made a blanket statement nonetheless.

It fail to realize what you are saying doesn’t matter for the point I was making. I mentioned that almost every programmer I’ve met is familiar with the Intel style CPU’s from having performed development on it before. Game haven’t been very big for mac computers, so very few game developers have ever tried coding on those until they were used in a console like gamecube. Either way, I mentioned the only reason I mentioned this was to prevent us form going down a route of CPU verses CPU.

Yeah it’s a blanket statement, but it’s true. Name one CPU processor used in as many computers with information on how it works widely as available. I have no idea how you could refer to one region, when talking about this. However even in Japan I’m sure you wouldn’t find one game developer that doesn’t know their way around Intel style processors even if all they’ve coded for 5 years has been consoles.

If you agree with Teasy without really knowing, or having worked on all three consoles then you’re supporting a blanket statement made by Teasy (that gamecube is the most efficient console). So, don’t chastise me for making blanket statements when you’re attempting to support someone that did the same thing!

First off, do you know what the words "Computationally for free mean?" The fixed T&L the GC utilizes causes real-world performance scenarios to degrade at a much slower rate vs. the other consoles iirc. You disregard suistanable main memory latency figures, the embedded 2mb DRAM on the Flipper which eliminates memory bandwidth intensive Z-buffer accesses, as well as seperate ops being collapsable into one pass. I guess ERP or Faf could clarify this for us?

First off, don’t get sarcastic with me, when you’re the one that wants to start throwing attitude around. Second yes the gamecube has an advantage with simple fixed function lighting. However it will quickly loose its advantage in other areas of game performance, which is the point I’ve been making OVER and OVER again, yet you keep ignoring it.

How many times have you seen me write that ERP has already previously stated in this forum (not only him but other developers) that you’ll always find specific cases where one console can out perform another console at a specific task? How many times have you seen me write that? Yet you go on making up hypothetical situations where you claim Gamecube the winner despite those situations not taking into account everything that a game does. Your spec numbers mean nothing by themselves.

This again is why I refer back to what EA has done as a more sensible benchmark.

And somehow this makes them irrelevant features? I was also highlighting these design aspects for those who are doubting the Revolution's potential to be on par with, or more powerful than the Xenon. (a design philosophy which will follow into the Revolution)

Yes, it really is irrelevant if it’s a feature hardly anyone is using. How could it be used in an argument for being the most efficient console if nobody is using EMBM? I’ve heard that EMBM is terrible for cache efficiency and that’s why it’s never been widely used. Also not everyone wants their bump maps to appear to shine.

Once again I say this isn’t relevant when you’re getting down to specific features where you’ll always find specific cases where one console can out perform another console at a specific task. For instance, more games (at least on Xbox and PC) use dot3 bump mapping compared to EMBM, but that can’t be used in an argument of which console is the most efficient since it’s a single feature and not the console as a whole.

I never implied that there were. (infinite methods) You helped verify my point indirectly, a few ways was all I was attempting to establish.

…and you just helped my point. What’s to say they aren’t doing things in the most optimal manor already? If there’s only a few different ways to get different performance levels, what makes you believe they aren’t already using the optimal method? A gut feeling? And educated guess? What exactly?

Just like I do not accept EA's benchmark claims.

The difference is EA has and always will developer games for all platforms, where as Factor 5 have only developed GAMES on gamecube. Factor 5 has consistently exaggerated in things they have stated without telling the full story (such as the FSAA and 16bit color thing on GC. I forget where I heard about that limitation, but it may have been form ERP or someone in these forums.

(I really don't think it's a question of who's technically more gifted here, as that's fairly obvious.) Although at least they are aware of what the X-Box's console is capable of, seeing as how they designed a DivX SDK toolkit for the console. RS3, RL, Crimson Skies, Star Wars: Jedi Starfighter, & Star Wars: Starfighter can be grouped into the same genre correct? static non-animating objects, primarily no characters or skinning) Which game is accomplishing the most from a technical standpoint? Despite the fixed T&L.

Unless it’s the same game how can you exactly compare them? Poly budgets are quite different depending on the game, not to mention what’s under the hood in the engine drawing everything. Look at it this way, if it’s just a contest of drawing non bone animated 3D objects, then look at the EA performance test for the same scene, and see which console cones out on top. Xbox had twice the performance in tests like that. If the games aren’t the same, then it’s down to opinion and not on a technical level. I realyl don't want to get into opinion since we'd end up just posting sceen shots to support our opinion and that's useless/pointless. Like you could compare RS3 and crimson skies on different levels. For instance, the physics in RS3 simply are more simple than what you have for flight physics in crimson skies.

Here you are assuming I know no programmers, nor do I have any friends within the gaming media. You're incorrect, but this is irrelevant. Are our arguements both now validated since we both have sources?

Hold on, I said that in regards to the comments about Japanese companies. That one is debatable though as not all companies are the same. However, you’re clearly not basing anything you saying about the gamecube hardware being the most efficient console off anything you’ve heard from any developers you know. If you did you probably wouldn’t be arguing most of the points you've been attmepting to…

 

[Qroach]

Did it ever occur to you that their best platform centric optimal code may not match another developers' best? Just a thought.

1. Did it ever occur to you don't have anyone else to gather multi console performance numbers to work from? Specifically when you won't trust what someone says as good enough reason.

2. Did it ever occur to you that the only real way to claim one console is more efficient then the other is to run the same test on both?

3.Did it ever occur to you that we have this same argument with teasy for what seems like every 6 months and we still come to the same conclusion. That there's always cases where one console can out perform another at speccif tasks, and in the majority of synthetic tests GC comes otu performance wise around the PS2 level with xbox nearly dobling it. Even still real world tests are much lower. (go back and search for it if you choose.)

You could easily make a case that the console that has the highest performance with the least amount of optimizing is the most efficient. Just a thought...