Carmack on Vista for games: "grasping at straws for reasons to upgrade"

[Babel-17]

http://www.dailytech.com/Article.aspx?newsid=5665

Swiped from the /. discussion.

http://games.slashdot.org/games/07/01/19/1426249.shtml


Carmack then said that he’s quite satisfied with Windows XP, going as far to say that Microsoft is ‘artificially’ forcing gamers to move to Windows Vista for DX10. “Nothing is going to help a new game by going to a new operating system. There were some clear wins going from Windows 95 to Windows XP for games, but there really aren’t any for Vista. They’re artificially doing that by tying DX10 so close it, which is really nothing about the OS ... They’re really grasping at straws for reasons to upgrade the operating system. I suspect I could run XP for a great many more years without having a problem with it,” he said.

Hehehe, speaking truth to power.

I can see giving DX10 to Vista first but denying it forever to XP users bespeaks of shoddy and highly suspect support for the XP OS.

[Drakkenfyre]

If not for the fact that they force you to upgrade to new OS's by making new features and programs useable only on new said OS's, hardly anyone would buy them

Back in Windows 95's day I said I would have to be dragged kicking and screaming into Windows 98, I finally got it so I could play WarCraft 3, then I said it would be forever before I got Windows XP, which I still have not used

this time, however, I think getting Vista right off the bat (errrrrrrr, or after an SP) is the smart thing to do simply because Microsoft will be Microsoft, resistance is futile when it comes to your OS

-Drakk )))

[opus512]

Nothing artificial about it, it's blatant. IMO there's an anti-trust suit hiding in there if someone in the administration ahd the balls to flesh it out.

[Ailuros]

I can see giving DX10 to Vista first but denying it forever to XP users bespeaks of shoddy and highly suspect support for the XP OS.

Artifically tying D3D10 so close to Vista doesn't mean that it can work under winXP, rather the contrary.

The way D3D10 has been written and how it correlates to the OS, according to all developers out there it seems impossible to port it over to XP, unless either the API or XP goes through a major revamp. Of course was it intentional from M$s behalf, but it doesn't change the fact that one shouldn't expect any miracles.

You have to understand Carmack's usual rambles a bit better; he most certainly has a point. A G80/R600 will look like today's IGPs in terms of performance when first real D3D10 games will arrive. Ignore the fancy stickers on any future game box until then. Games like Crysis, FS X and Lord knows what else are and WILL REMAIN pure D3D9 games. Yes there will be additional D3D10 paths down the line for performance increases yet nothing groundbreaking.

D3D10 game development started only a couple of months ago. W/o any real D3D10 HW around no one would bother to touch any D3D10 stuff anyway.

[Ailuros]

Nothing artificial about it, it's blatant. IMO there's an anti-trust suit hiding in there if someone in the administration ahd the balls to flesh it out.


Will you be able to use a DX9 GPU under Vista? Yes.

Would a DX9 GPU see any benefits from using D3D10 under XP? No.

It's not that the whole affair isn't annoying, yet as long as one can use a D3D10 GPU in XP w/o blinking much of an eye for the foreseeable future and anyone can work under Vista with a DX9.0 GPU, I don't see where the problem exactly is.

As for hypothetical anti trust suites try thinking of how much ground it would have legally when it comes to M$s own OS and own future APIs.

[sj_hurst]

Shouldn't it have some performance advantages when operating w/ 64-bit CPU & OS? Or is Carmack arguing that there's no bottleneck w/ 32-bit XP yet?

You see all these people w/ 32-bit XP OS and 64-bit CPU. Isn't the whole point of having a 64-bit CPU so that it can run in a 64-bit OS environment and get a performance boost?

[Babel-17]

Huh, so Carmack was wrong/misquoted/taken out of context?


http://games.slashdot.org/comments.pl?sid=217766&cid=17688582

"Carmack then said that he's quite satisfied with Windows XP, going as far to say that Microsoft is artificially forcing gamers to move to Windows Vista for DX10"

This is such FUD... There are REAL reasons that DX10 requires Vista. For example the WDDM model in Vista is needed for DX10 to do what it needs with the RAM usage and multi-tasking features of the GPU introduced in DX10.

These are NOT things that can be done on XP unless MS completely redesigned the Video Subsystem in WindowsXP, which would be a long process, and utterly insane.

People please do your own research and not buy into the ranting of a fool that thinks he has all the facts. DX10 has a lot of new features pushing the envelope on gaming graphics with larger textures and other pipeline optimization techniques, but many of the DX10 features would FAIL on XP or any other OS that did not have the WDDM driver model introduced in Vista, as no other OS can multi-task GPU operations or selectively share large portions of system RAM with the GPU RAM with no performance loss.

For further reading, go check out NVidia or ATI's site for information on DX10, or even be as bold to check on MS's site on DX10 and why the WDDM driver model in Vista is EXPECTED to be there and IS USED by the DX10 framework to provide the many of the new features of DX10.





http://games.slashdot.org/comments.pl?sid=217766&cid=17681378

DX10 by itself doesn't require Vista, but they decided to get rid of the legacy cruft and re-wrote the entire graphics system. This allows neat things like multi-tasking and virtual memory handling for GPUs, but requires completely new drivers. This also supposedly enables a much higher performance, a game running on DX10 should be a lot faster than the same game running on DX9, assuming that they both use the exact same shaders. So yes, they could port DX10 to XP, but this would require two different kinds of DX10, with different features (no virtual GPU memory on XP = lame), and with different levels of performance. IMO the clean break is a good thing, but the HDCP bloat makes me hesitate to upgrade.

[Drakkenfyre]

You see all these people w/ 32-bit XP OS and 64-bit CPU. Isn't the whole point of having a 64-bit CPU so that it can run in a 64-bit OS environment and get a performance boost?

If you are talking strictly about Windows, you have to remember that Windows XP 64-bit requires 64-bit drivers, it will not work with 32-bit drivers, so the fact that not every hardware manufacturer out there is providing decent 64-bit drivers (if any at all) combined with the fact that Windows XP 64-bit was never released at retail, and that explains why there are not more people with 64-bit CPU's using it

-Drakk )))

[msan_msw]

Looks like Carmack got owned.

[sj_hurst]

If you are talking strictly about Windows, you have to remember that Windows XP 64-bit requires 64-bit drivers, it will not work with 32-bit drivers, so the fact that not every hardware manufacturer out there is providing decent 64-bit drivers (if any at all) combined with the fact that Windows XP 64-bit was never released at retail, and that explains why there are not more people with 64-bit CPU's using it

-Drakk )))

I know that. But therein lies the conundrum.. how to get to those advantages of 64-bit. You have the shiny new 64-bit CPU, but you (i.e. the consumer) purposefully hold back on 64-bit OS cuz other hardware manufactures don't support it (w/ 64-bit drivers). Upgrading to 64-bit OS doesn't make sense with few manufacturers making 64-bit drivers.

Many people would upgrade to 64-bit OS if 64-bit drivers were available. Companies don't make 64-bit drivers because the user base for 64-bit OS is too small. Both forces work in concert to keep 64-bit from happening. It takes a big bully like MS to step in and break these chains. It will be MS's doing that the consumer PC market finally makes the move to true 64-bit once & for all.

[Babel-17]

Looks like Carmack got owned.

Sure does. Unless he's starting the controversy to pave the way for an announcement of an update to the OGL API?

But even that hypothetical would lack the geekiest features that are mentioned in the posts I quoted. Unless it too was designed for Vista ....... which would put us back to square one.

I guess we'll be hearing more about this.

[Drakkenfyre]

I know that. But therein lies the conundrum.. how to get to those advantages of 64-bit. You have the shiny new 64-bit CPU, but you (i.e. the consumer) purposefully hold back on 64-bit OS cuz other hardware manufactures don't support it (w/ 64-bit drivers). Upgrading to 64-bit OS doesn't make sense with few manufacturers making 64-bit drivers.

Many people would upgrade to 64-bit OS if 64-bit drivers were available. Companies don't make 64-bit drivers because the user base for 64-bit OS is too small. Both forces work in concert to keep 64-bit from happening. It takes a big bully like MS to step in and break these chains. It will be MS's doing that the consumer PC market finally makes the move to true 64-bit once & for all.

That would constitute what I would call a cycle of suck

I do not remember and too lazy to look it up right now, but isn't Vista 64-bit to begin with, or only the "ultimate" versions?

-Drakk )))

[Babel-17]

Huh

http://tomshardware.co.uk/2007/01/17/gameplay-only-gets-worse-with-vista_uk/page4.html

I had the opportunity to speak with OpenGL architecture review board Chairman Barthold Lichtenbelt. He specified that if there is not a manufacturer ICD present then the API wrap for D3D will operate. This means that OpenGL 1.4 will run in software on D3D. The best implementation will utilize the OpenGL driver from your hardware manufacturer.

I asked Barthold about the future of OpenGL on Vista and other developments. While the breadth of that conversation does not entirely pertain to this article, he made it clear that OpenGL has a home on Vista and touted that he predicts that OpenGL will have greater performance than that of D3D under upcoming standard changes. As OpenGL is 14 years old, Lichtenbelt stated that there are two scheduled specification releases due out this year.

"We are revamping OpenGL," he said, to make it easier for developer to "write applications, faster, more stable and with less bugs, which is a good benefit."

"[The ARB] is very aggressively pushing OpenGL out this year," Lichtenbelt said.

[opus512]

Carmack's a twat, but Microsoft still sucks large ones. And it's still bullshit tyng DX10 to Vista. Did they have to design it that way? Maybe so, but it still sucks. It's a defacto standard, and you don't have to be a monopoly to wield monopoly influence over a market. That was the whole basis of the suit against Microsoft under Clinton that Bush dropped when he got into office. There ws a solid case, MS was going to lose.And Bush dropped the case.

Rationalize it all you want on so-called technical grounds, but it's still bullshit.

[Ailuros]

Sure does. Unless he's starting the controversy to pave the way for an announcement of an update to the OGL API?

But even that hypothetical would lack the geekiest features that are mentioned in the posts I quoted. Unless it too was designed for Vista ....... which would put us back to square one.

I guess we'll be hearing more about this.

I read that Carmack interview a couple of days ago and somehow I did understand what he wanted to say without taking it out of context. Granted it's phrased a tad weird and can mislead, but I cannot believe or accept that an as experienced developer like JC doesn't know what is and what isn't possible within XP. In a weird sense M$ artificially excluded D3D10 from XP; the twist of the story being that the "artificial" solution was the only sensible one.

JC isn't the type of guy that rambles nonsense like Sweeney at times ;) LOL

[Ailuros]

Rationalize it all you want on so-called technical grounds, but it's still bullshit.

Rationalizing either software or hardware on non technical grounds would be the most genuine bullshit ever.

[Ailuros]

Huh

http://tomshardware.co.uk/2007/01/17/gameplay-only-gets-worse-with-vista_uk/page4.html

Things are truly moving more rapidly lately with OpenGL. I'm not sure if the PDA/mobile/handheld market and it's huge upswing and the closely tied OpenGL_ES have anything to do with it, yet the signs are truly promising.

[opus512]

Yeah well, you know what opinions are like.

[Sharkfood]

Wow, marking my calendar.. I'm actually agreeing with something John Carmack has spoken.. lol.

I take a totally different outlook on his (unmolested/untranslated) words though.

The whole basis of his diatribe is:
" “Personally, I wouldn’t jump at something like DX10 right now. I would let things settle out a little bit and wait until there’s a really strong need for it.”

So from that standpoint, I totally agree. We dont NEED DX10 right now.. hell, 65%+ of the gaming user base isn't even up to PCI-Express nor DX9.0. Integrated graphics are still the lionshare and those aren't even up to DX8.1 level shaders.

The high-end, discrete graphics market not only is in the minority, but the current hardware is fine as-is. The "improvements" in DX10 are the same as the "improvements" in PCI-Express, which are, in practice, irrelevant. The industry is pushing numbers and white papers that mob-like mentality sheep of consumers are lapping up, when at the end of the day, it comes down to making those few pry open their wallets and dump more cash into a blackhole that wont even be *partially* actualized for another 2-4 years. I doubt even the current sockets or PCI-Express will even live that long, which will in turn force even more, newer standards into play.

I truly do not like the "pushes" the past 2 years have seen as they have no practical or real value to gamers. I especially don't like the pushes from Microsoft to force the world to upgrade (yet again) but with no SLA or accountable value. We have only perceived values being pushed and a marketplace that falls in-line to hurry up and lay down the cash with little to no delivery. Current solutions have value... what is down the pipe has NO real value for years, just perceived and artificial.

[Babel-17]

Huh, so, is DX10 too much, too soon?

In other words, would we the consumers be better off, long and short term, if more focus was paid to getting more out of DX9 software and to leveraging more performance out of DX9 hardware?

No one is happy about transistors going towards features that aren't going to deliver much in a reasonable frame of time but we've come to accept, somewhat, that new features have to start somewhere.

I'm curious, I've read about the flexibility of the new hardware and I wondered if all of their power comes into play when running a DX9 game. (Assuming the gpu isn't waiting on the cpu)

Or are there transistors that sit idle?

Should I be looking at these DX10 cards as actually being next generation DX9 parts that have the bonus of doing DX10?

Sorry for the painful ignorance.

[Ailuros]

So from that standpoint, I totally agree. We dont NEED DX10 right now.. hell, 65%+ of the gaming user base isn't even up to PCI-Express nor DX9.0. Integrated graphics are still the lionshare and those aren't even up to DX8.1 level shaders.

There's a difference between fancy word processors (being IGPs) and a GPU which isn't allergic to anything that moves on the screen. Hell when analysts categorize IGPs they shouldn't dump it into the low-end segment of the market, but rather a separate category on the bottom of the chain named "dog slow low-end".

While I don't disagree one bit that the userbase doesn't and won't need for quite some time to come D3D10, IGPs and in extension Intel got served well with D3D10, since it's more than painful for each developer to have to either write fallback modes or disregard the work he wants to do because Intel wants to produce only ultra cheap and fairly useless crap (for anything 3D that is). Up until now Intel didn't even bother to implement any sort of geometry unit; while one would had argued that in 2000 it wouldn't had been necessary at all, nowadays it's fairly ridiculous to dump all the geometry on the CPU. With the G965 all of the sudden Intel was forced to aim for at least SM3.0 and on board geometry units (irrelevant if it can get "upgraded" via drivers to D3D10 or not as they claim) in order to get the Vista compliance seal.

It might sound too strict but at least now there's finally a lowest common denominator and frankly no IGP user really cares if 3D performance stinks a little bit more or a little bit less. It just stinks and that's the end of the story.

The high-end, discrete graphics market not only is in the minority, but the current hardware is fine as-is. The "improvements" in DX10 are the same as the "improvements" in PCI-Express, which are, in practice, irrelevant. The industry is pushing numbers and white papers that mob-like mentality sheep of consumers are lapping up, when at the end of the day, it comes down to making those few pry open their wallets and dump more cash into a blackhole that wont even be *partially* actualized for another 2-4 years. I doubt even the current sockets or PCI-Express will even live that long, which will in turn force even more, newer standards into play.

PCI-e could have made a significant difference if the AGP userbase wouldn't still be as large. How and why would one expect a developer to take advantage of the PCIe backchannel with all the AGP system users screaming for unfair treatment.

I truly do not like the "pushes" the past 2 years have seen as they have no practical or real value to gamers. I especially don't like the pushes from Microsoft to force the world to upgrade (yet again) but with no SLA or accountable value. We have only perceived values being pushed and a marketplace that falls in-line to hurry up and lay down the cash with little to no delivery. Current solutions have value... what is down the pipe has NO real value for years, just perceived and artificial.

Since major semiconductors don't seem to be willing to jump on the Vista bandwagon yet due to extreme cost issues, I wouldn't be that much worried Vista making a major breakthrough that soon.

Any user can still stay on XP as long as the OS continues to be supported. I've no intention myself to enter the bug ridden hell of Vista just yet. A couple of service packs down the line and I'll think about it.

[Ailuros]

Huh, so, is DX10 too much, too soon?

In other words, would we the consumers be better off, long and short term, if more focus was paid to getting more out of DX9 software and to leveraging more performance out of DX9 hardware?

Don't you think that half a decade is a mighty long time for an API to last? DX9 lasted times longer than any former version of D3D, despite it being broken up into three stages of compliance (SM2.0, SM2.0_extendend, SM3.0).

In D3D10 something similar will happen, namely D3D10, D3D10.1 and D3D10.2, lasting all together over 4-5 years down the line.

IMHLO SM4.0 is an evolutionary step compared to SM3.0 and despite the unification of shader units is not obligatory in D3D10, a unified shader core is far more better for many tasks including minimal shader texturing latency, load balancing and what not.

Whether now virtualizing RAM under 10 was necessary or not is up to anyone's perspective. But when you support up to 8096*8096 2D textures you have to have a sollution for any professional to get to work with it, even if it's just for the less foreseeable future or for pure experimental reasons.

No one is happy about transistors going towards features that aren't going to deliver much in a reasonable frame of time but we've come to accept, somewhat, that new features have to start somewhere.

Most added capabilities of purely new generation hardware are mostly of developer interest. If you run somewhere in mid summer to a store to buy let's say a midrange D3D10 GPU, a logical reason to do so is to be able to run a game like Crysis and there's no reason yet to either jump on Vista /D3D10 yet for that one, since a DX9.0 game will run perfectly fine on a D3D10 GPU under XP. It would be the added horsepower of such a GPU that could drive you to such a decision and less the fancy D3D10 sticker on the box.

I'm curious, I've read about the flexibility of the new hardware and I wondered if all of their power comes into play when running a DX9 game. (Assuming the gpu isn't waiting on the cpu)

Or are there transistors that sit idle?

You've more chances on a DX9.0 GPU for transistors to sit idle than on a pure USC/D3D10 GPU. When geometry shaders come into modest play in the foreseeable future, D3D10 GPU power consumption might actually slightly decrease, due to it being highly relevant to I/O. The higher the I/O the relatively smaller the power consumption; and yes before anyone says it when massive shader calls will appear in future games, todays D3D10 GPUs will be just a tad more powerful than future IGPs ;)

Should I be looking at these DX10 cards as actually being next generation DX9 parts that have the bonus of doing DX10?

It's the exact opposite. With the slight difference being that if we'd have real D3D10 games today, any D3D10 GPU would be way too slow for it. But that's a pretty normal trend since the dawn of 3D. The R300 was an outstanding first D3D9.0 GPU. Try running Oblivion on it today or Lord help upcoming Crysis or UT2k7 as simple examples.