Reverend At The Pulpit #14

[Reverend]

I just finished a review of a 6800GT. The review took more than 2 months to complete simply because I just didn't have the time and, to a certain extent, I have lost some considerable enthusiasm for reviewing since it has been such an extremely long time since I last reviewed any kind of hardware. For that, I have to apologize (and I have, privately) to Dave. Actually, there shouldn't have been much pressure to review it quickly since the card was one that Dave bought and sent to me. More urgent is the review I'm currently working on (Visiontek X800XT) which I have had for more than a month now... I'll need to apologize to Visiontek in some way. In any case, you will all see that the 6800GT review isn't a good one. By that, I don't mean that the card is bad but the quality of the review really doesn't live up to my own personal expectations -- there are "just" 10 pages and IMO most of the pages just aren't as informative as I would've preferred... very disturbingly is that this is the first time I wanted a review "to be done with". This has to do with the aforementioned lack of enthusiasm for reviewing overall. I need to quickly re-discover it and start being a investigative reviewer! Dave will publish the 6800GT review when he sees fit. BTW, notice what benchmark software (that B3D regularly use) was missing from Dave's recent X850XT review? Well, it's missing from my 6800GT review too... and this was something I decided back in September.

Recently, a developer had said that Beyond3D will be the only media outlet to have access to a game benchmark (prior to public release of the full game), not counting all the IHVs. The game is an anticipated title. After informing me that a NDA document should be forthcoming between mid and end November, things have been silent from the developer, even after two inquisitive emails from me. I just hope things eventually turn out good for us.

There are three games I really want to buy but just haven't found the time : Half Life 2 (PC, of course), Metal Gear Solid 3 and Ratchet&Clank:Up Your Arsenal (both PS2 of course). More interestingly is that I want to buy the two console games because I really want to play them but I am buying HL2 more because I can include it in reviews. Something's not right.

BTW, can anyone give me a link that provides sales stats for PC games and console games? I have no idea which is a bigger industry. And preferably a link that breaks down the console stats into the various platforms (PS2, XBOX, etc. ).

Oh, and this RATP edition really is the Carmack edition. So here goes, some (not every) correspondences between us in no particular order of "flow" :

Hi John,

Well, that certainly is true for both the major IHVs, especially ATI.

Can you say if the new engine will also be D3D? You've said at QuakeCon04 that you were tempted to use it for Doom, with the improvements made with DX. Have the Microsoft Xenon guys (Chris Donahue and XBOX2 team) been nicer?

It is currently still OGL.

John Carmack

Why do you have your aversion to DX, John? I've been talking to Chris (a friend) and he has expressed his frustration!!

Also, how's the voting going with regards to the fix for the RT with pbuffer nonsense?

I don't have any aversion to it. I have been working to consolidate and minimize the OpenGL code in preparation for Xenon port, but if we continue to support OpenGL for the mac and linux, it might remain the windows standard well. The Xenon code probably won't be directly compatible with windows,

ARB_framebuffer_object is expected to go to a vote next month.

John Carmack

The Xenon code probably won't be directly compatible with windows,

Can you tell me why that may be the case, John?

They plan on exposing all of the low level hardware capabilities, including some architectural things that aren't directly in line with the DX model. I consider that a good thing -- no reason to keep a lowest common denominator approach on a console. I don't expect it to be "far" off, but a bunch of progs will probably be slightly different.

There is a lot less API specific code in the new engine anyway, with most of the effort moved into vertex / fragment code.

ARB_framebuffer_object is expected to go to a vote next month.

Well, it better be positive -- like I said, the RT using the pbuffer is just horrible for performance!

Yes, lousy performance, and the multiple contexts make things suck even more. Pbuffers are part of the ugly legacy of X windows.

John Carmack

John,

First of all, the understanding is that we're talking NDA info here (of course!).

They plan on exposing all of the low level hardware capabilities, including some architectural things that aren't directly in line with the DX model. I consider that a good thing -- no reason to keep a lowest common denominator approach on a console. I don't expect it to be "far" off, but a bunch of progs will probably be slightly different.
There is a lot less API specific code in the new engine anyway, with most of the effort moved into vertex / fragment code.

Does this have anything to do with the 3-core Xenon CPU? Also, I suppose the fact that we're talking about Xenon's supersets of NT and DX9 also has to do with the difficulties regarding Xenon-to-PC "compatability"?

Anything you can tell me with regards to Xenon's compatability with the first Xbox in terms of software?

I'm working on limited NDA info regarding Xenon and, well, I'm not developing games for the consoles

Sorry, I'm not going to spill anything microsoft won't give you.

I do think it will be possible to write code that basically just compiles for xenon and windows, but there will be some options to go a lot closer to the actual hardware on xenon if you want to.

ARB_framebuffer_object is expected to go to a vote next month.

Well, it better be positive -- like I said, the RT using the pbuffer is just horrible for performance!

Yes, lousy performance, and the multiple contexts make things suck even more. Pbuffers are part of the ugly legacy of X windows.

You mentioned allowing shaders to write back to vertex buffers. I assume this is render to vertex array.

Yes, that was what I was talking about.

Is it possible to allow writeback from the vertex shaders... in future 3D hardware? Coz the current ones don't allow this.

There is some discussion about it, and I'm not convinced adding the full random memory system of a CPU is a good idea.

John Carmack

John,

Is it possible to allow writeback from the vertex shaders... in future 3D hardware? Coz the current ones don't allow this.

There is some discussion about it, and I'm not convinced adding the full random memory system of a CPU is a good idea.

Hmm... well, as far as I know, the pros are that the GPU would then be a full-fledged computing device rather than just a game accelerator.

The con is that the hardware complexity increases, because random-access writes on a parallel computing architecture are only practical if there exists synchronization mechanisms similar to those that CPU programs use to synchronize threads in a multithreaded environment.

I'm not sure which is a more important consideration... what are your thoughts (or perhaps you'll elaborate on why you're not convinced)?

No reply from John

Hopefully some of the above information are interesting to you guys. Let me know if I'm an arrogant SOB that craves attention by publicly posting my email correspondences with John which to some folks probably means arrogance in itself. Just remember that I correspond with a number of other game developers and I feel no need to publicly post such correspondences... do you know why this is the case? If you say I'm being arrogant again by saying so, well...

 

[Ostsol]

Ooh! Cool info-type stuff. . .

It's nice to know that the frame buffer extension is coming along and will see the light of day. . .

 

[Guden Oden]

I think you post your conversations with John because you know people are interested in what you two are talking about.

Please consider posting your correspondance with other game devs as well.

Thank you!

 

[PeterAce]

Rev,

First of all thanks for posting up your JC discussions (I had some questions after quakecon04, and you've answered some of them here).

Secondly if reviewing hardware is becoming boring to you, why not start writing articles that you DO enjoy inverstigating.

For example somthing like 'alogrithm overview' articles for Beyond3D. What I mean is articles that cover 'shadowing algorithms like; stensil, shadow buffer, PSM Etc - but importnatly provide a 'test app' example to demonstrate.

One of the really useful things for me was the abillity to run the test program provided with Dean Calver's 'Photo-realistic Deferred Lighting' article.

So if 'testing hardware' has lost it's interest for you, make a change and write articles (that are 3D related) but DO interest you.

 

[sytaylor]

... and maybe get B3D forum members to write "sample" reviews to which you can consider and perhaps even post? I'm sure there are plenty of budding/talented reviewrs somewhere further down the food chain? Worth a shot I guess

 

[Rodéric]

Interesting infos indeed Reverend, thanks.

 

[Reverend]

It's nice to know that the frame buffer extension is coming along and will see the light of day. . .

So, you already know that the ARB will vote positively for this?

 

[Zeross]

It's nice to know that the frame buffer extension is coming along and will see the light of day. . .

So, you already know that the ARB will vote positively for this?

Given that :
-this extension was previously known as EXT_render_target and was jointly written by 3DLabs and nVidia
-the spec was modified with inputs from ATI
-the final spec was written by both ATI and nVidia
-OpenGL coders desperately need a RTT extension to definitely drop the whole PBuffer mess

I think we can say yes

If a member of the ARB vote against this extension he better run fast because he will have to face a lot of angry OpenGL coders

More seriously, nVidia has some kind of support for EXT_frame_buffer_object in its drivers and it seems that some software company has already access to this functionality (given that some bugs with this extension were corrected in 66.93 I've seen it in release notes). I guess that like GLSL there is some key in the registry to expose the extension.

 

[pascal]

Hey Rev,

Good you keep us updated with news from the tech frontier.
When we will see some game review/insights like you used to do in the past?

 

[Tim Murray]

Recently, a developer had said that Beyond3D will be the only media outlet to have access to a game benchmark (prior to public release of the full game), not counting all the IHVs. The game is an anticipated title. After informing me that a NDA document should be forthcoming between mid and end November, things have been silent from the developer, even after two inquisitive emails from me. I just hope things eventually turn out good for us.

To those of us who have posted here for a while, I'm guessing it should be immediately obvious what game he's talking about...

 

[John Reynolds]

To those of us who have posted here for a while, I'm guessing it should be immediately obvious what game he's talking about...

Your theory is correct.