Will Direct3D 10 ever come to Windows XP?

[swaaye]

Yes, yes, I agree that Vista is most likely the future. Unless the current Mac initiative actually goes somewhere (who knows?)

But, people are too short-sighted for you, I think. XP was slower than 9x and, thus, many people stuck with it for years even after XP's performance was improved. XP was slower though partially because it is simply a heavier OS with more overhead than 9x. Stability was gained by reducing the ways things could bring down the OS. This didn't really help performance, however. Of course there were driver writer challenges that were a major challenge to overcome too.

I really wonder if XP ever did reach performance parity with 9x? Most people just forgot about the OS. And I'm not saying that's a bad thing because 9x was definitely not the best OS in any way imaginable.

Are we going to see the same with Vista? Does the OS strike anyone as significantly streamlined? I'm pretty sure it's not a featherweight compared to XP. Direct3D 10 certainly is looking to be the roughest Direct3D update yet. Bad drivers and non-optimal hardware from the looks of things. And MS is killing the adoption rate by not having it backwards compatible with XP at the least. IMO this spells out that DX9 will be here for years to come and that DX10 is probably a year away from being useful for gamers.

Anyway, I think the question is: Does Vista currently offer a tangible improvement for gaming over XP? Nope. Will it in the future? It might. Could I install DirectX 9 on Windows 98? Yes. Did MS cripple XP's future intentionally? Hmmmm..... Honestly I think that goes down the "yes" path without asking. Does it really matter? Only in that it would cost me a few hundred bucks to pick up the currently-of-questionable-worth-for-gamers Vista OS.

BTW, Win3.1 was never much of a gaming platform. DOS was the choice till Win95. DirectX was created to make 95 an option for game devs after MS failed so horribly with 3.x.

 

[Geo]

Yes, yes, I agree that Vista is most likely the future. Unless the current Mac initiative actually goes somewhere (who knows?)

But, people are too short-sighted for you, I think. XP was slower than 9x and, thus, many people stuck with it for years even after XP's performance was improved. XP was slower though partially because it is simply a heavier OS with more overhead than 9x. Stability was gained by reducing the ways things could bring down the OS. This didn't really help performance, however. Of course there were driver writer challenges that were a major challenge to overcome too.

I really wonder if XP ever did reach performance parity with 9x? Most people just forgot about the OS. And I'm not saying that's a bad thing because 9x was definitely not the best OS in any way imaginable.

I think it did, actually. Might take some spelunking at [H] or THG or Anandtech to prove it tho.

Are we going to see the same with Vista? Does the OS strike anyone as significantly streamlined? I'm pretty sure it's not a featherweight compared to XP. Direct3D 10 certainly is looking to be the roughest Direct3D update yet. Bad drivers and non-optimal hardware from the looks of things. And MS is killing the adoption rate by not having it backwards compatible with XP at the least. IMO this spells out that DX9 will be here for years to come and that DX10 is probably a year away from being useful for gamers.

I hold MS responsible for this to a degree, and the IHV's to a degree as well. For not communicating very well in the two years or so before Vista launched. Or possibly communicating one true thing that will eventually be true while neglecting to communicate another true thing that is immediately true.

The immediately true thing that didn't get communicated until Vista was actually launching was that moving the dx interface out of the kernel was going to have an inescapable performance cost of a few percent vs XP. When you see benchies that are just a touch slower on Vista, that's what you're seeing. When you see benchies that are comparable, you're seeing performance improvements in the Vista drivers overcoming that inescapable deficit. At least that's my understanding now, and I can tell you that even tho I was deeply interested in this subject for a couple years before Vista launched, I didn't gain that understanding until the end of January in a phone interview with Andrew Dodd of AMD. It seemed like it was pretty obvious to him, but I couldn't remember any time prior to that when AMD (or NV or MS) had communicated to that effect to the mass audience (maybe the hardcore dx techies reading this are going "Well, duh!").

The longer term "eventually true" thing that's going to require software support, for the most part, to reap the benefit of, is the overhead issue that everyone talks about. My understanding is that ISVs understood that issue pretty well on XP and thus architected to not bump into it for the most part. So a straight move over to Vista isn't going to give you a performance advantage on those current titles. However, now that they know they have that increased headroom they'll start taking advantage of it in future titles. As they do take advantage of it for DX10, most likely this means that DX9 paths of their games are either going to be slower on XP as they bump that overhead wall, or less fully featured in some visible fashion as they once again maneuver to avoid it on their DX9 path.

Maybe JHoxley or Demirug can correct my understanding in the above three paras if they don't agree with it.

Anyway, I think the question is: Does Vista currently offer a tangible improvement for gaming over XP? Nope. Will it in the future? It might. Could I install DirectX 9 on Windows 98? Yes. Did MS cripple XP's future intentionally? Hmmmm..... Honestly I think that goes down the "yes" path without asking. Does it really matter? Only in that it would cost me a few hundred bucks to pick up the currently-of-questionable-worth-for-gamers Vista OS.

Currently? Nope. No DX10 content to speak of, so how could it? Neither did XP at first, and my memory is that the comparative gaming performance situation for XP vs Win9x was much worse at the same point in the OS transition. Anyone who lived thru both disagree with that?

 

[Demirug]

Were there any significant framerate differences in those comparisons too?

I must confess that I didn’t measure the framerate. As long as both results are not visual comparable it doesn’t make much sense IMHO.

Additional this tutorial samples are not very complex on the GPU side. Therefore I expect that we have a CPU limit there. This is naturally a bad starting point for a wrapper solution.

 

[Tim Murray]

After seeing that the Alky project is giving away Shadowrun and Halo 2 "modifications" that allow them to run on XP if you order it from Amazon using their referrer link, I have to wonder how much of it is a scam. It seems remarkably fishy to me.

 

[Demirug]

The longer term "eventually true" thing that's going to require software support, for the most part, to reap the benefit of, is the overhead issue that everyone talks about. My understanding is that ISVs understood that issue pretty well on XP and thus architected to not bump into it for the most part. So a straight move over to Vista isn't going to give you a performance advantage on those current titles. However, now that they know they have that increased headroom they'll start taking advantage of it in future titles. As they do take advantage of it for DX10, most likely this means that DX9 paths of their games are either going to be slower on XP as they bump that overhead wall, or less fully featured in some visible fashion as they once again maneuver to avoid it on their DX9 path.

Well, the “Draw Call Overhead Problemâ€￾ is well known since years by core engine developers out there. There is even a magic number of 500 draw calls per frame per 1 GHz CPU power. This is not necessarily the full true as you need to consider additional calls needed to support a draw call (state changes, etc.).

Therefore the number of objects becomes a more critical component then the number of polys. You may remember the effect of instancing in games like FarCry. It allows to increase the number of plants on the screen. There were even some bad jokes from the OpenGL side. â€￾We don’t need instancing as we don’t have the high draw call overhead from D3D.â€￾ That’s right as a D3D draw call costs round about two to three times more than an OpenGL draw call. But if we talk about instancing 10 or more projects this is not enough and therefore even OpenGL gets it’s instancing extension.

 

[Demirug]

After seeing that the Alky project is giving away Shadowrun and Halo 2 "modifications" that allow them to run on XP if you order it from Amazon using their referrer link, I have to wonder how much of it is a scam. It seems remarkably fishy to me.

Both games don’t need Direct3D 10 at all. I am currently not sure which parts of the new Vista API they need. If we talk about only a few calls it would be possible to add them with common system tweaking tricks to an XP. Therefore it is not impossible but these guys don’t seem to have much luck with this at the moment. The aim for late July maybe the will hit this point.

 

[JHoxley]

IMO this spells out that DX9 will be here for years to come and that DX10 is probably a year away from being useful for gamers..<snip>..it would cost me a few hundred bucks to pick up the currently-of-questionable-worth-for-gamers Vista OS.

From what you've posted I get the impression you're not an early adopter. Which is absolutely fine - for people interested in playing games there isn't really any financial sense in upgrading to Vista.

I see it as being the same as HDTV - everyone agrees it's better but currently there is limited to no HDTV content in a lot of places, which means many consumers aren't going to fork out £/$/€ 1000's for a shiny new TV they can't make much use of. Once there's more HDTV content readily available i'm sure sales will rocket...

I also don't see why people don't get the business angle.

Just be glad most of them aren't here. Reading other forums is painful at times.

I Beyond3D.

Jack

 

[Davros]

I dont need dx10 I can easilly reduce my frame rate by 70% in xp

 

[digitalwanderer]

So just to make certain; aside from it being technically impossible it's illegal to do and if someone DOES manage to make a wrapper that emulates dx10 on XP they're gonna get their balls sued off.

To answer the question, no way in hell.

 

[Demirug]

The legal status of such a thing seems more complicated. But I am not a lawyer therefore I have include this aspect. I know for sure that you can’t use the names DirectX, Direct3D 10 without telling the people that Microsoft is the owner of these trademarks. But reimplementing one of Microsoft’s APIs seems not illegal at all. Look at Transgaming. There whole business model is based on reimplementing windows APIs.

 

[JHoxley]

My interpretation of the legal aspect (again, IANAL) is that it boils down to one of two things:

1. "stealing" intellectual property. If someone threatens the R&D or patents (etc..) involved in developing Vista/WDDM/D3D10 through reverse engineering then you can expect the relevant companies to defend their property. If it's a free wrapper/patch then maybe not, but if someone intends to make money from it then they'd better be *very* careful!

2. (mis-)using the brand names or images. As Ralf commented - things like using "Direct3D" or making any sort of assertion or implication that Microsoft or the IHV's endorse the wrapper would be a no-no. Big companies like to protect their reputation and branding.


Jack

 

[digitalwanderer]

My non-lawyer legal analysis is based on the theory that letting something like that out in the wild would mess up M$'s strategy for forcing people to upgrade to Vista, so if there is even the slightest of angles for a legal challenge M$ is gonna tie 'em up for years with it.

 

[Geo]

Oh, you don't actually need to think you can win in a lawsuit that goes all the way to the jury in order to sue. Particularly if the other guy is a whole lot smaller than you are. You just wear him down with lawyers and your deep pockets.

 

[Silent_Buddha]

Currently? Nope. No DX10 content to speak of, so how could it? Neither did XP at first, and my memory is that the comparative gaming performance situation for XP vs Win9x was much worse at the same point in the OS transition. Anyone who lived thru both disagree with that?

Win 9x games fared considerably worse at Win XP's inception than Win XP games are faring on Vista.

I see no noticable difference between the games I currently play on Vista 64 vs running the same games on Win XP.

Win XP also had similarly bad and oft times worse drivers during the transition period than I'm currently seeing with Vista. Some of the culprits are the same. Creative Labs and HP were notorious for bad or non-existant drivers in XP and that appears to be the case in Vista also.

However, for every company with bad drivers there's companies with superlative drivers. Realtek and Samsung (for counterpoints to Creative and HP) have quite stable and well performing drivers for both Vista 32 and 64.

As with any previous Windows transition it pays to do a little research ahead of time to see if you are going to have potential problems with your hardware as it concerns drivers.

Assuming you have hardware from companies with nimble/good driver teams, you'll have a quite pleasant experience.

For me I noticed the benefits of Vista WRT gaming almost immediately as I tend to play certain types of games (MMORPG/RTS) games windowed whenever possible. And the fact that the windows driver model supports full 3D and video acceleration across multiple displays is a very nice boon for me.

Enough so that I simply cannot stand to use Windows XP on a multi-display machine. Likewise the increased usefulness of Explorer has completely drawn me in. I'm now constantly finding stuff that I'd forgotten I had.

I can only imagine that people in general are resistant to change. /shrug. It's been the same repeating pattern ever since I can remember. If OS's were good enough as is, we'd still be using DOS. After all people complained about the need to go from DOS to Windows. And even resisted the move from DOS to Win 95. Pretty much the exact same situation as we have now with people resisting the move from XP to Vista.

Regards,
SB

 

[swaaye]

Resistant to early adoption, more like it. Experience from prior instances of things not working right makes people not want to deal with it again unless there is a significant advantage to mitigate that risk. Let's call it learned resistance to frothing at hyped technology.

By 2008, Vista will be the only way to go I imagine. Primarily because by then it will be almost everywhere and companies will either have support for it or they will be done for.

But, I bet my scanner is junk once I go to Vista. HP has planned obsolescence in full swing. I imagine my 15 year old Laserjet 4P will continue to function, though, ironically. Probably only because PCL5 is so popular. Windows 3.1 - Vista. A printer that has been around.

 

[Frank]

I would like it MUCH better, if Windows wasn't so incredibly stupidly bloated with endless amounts of crap. I don't even use any of the "freebies": I use better apps.

What in hell does the GPU driver/model have to do with the OS? Nothing, that's what.

So, I have to buy all those endless amounts of bloatware and eye-candy that I don't need or want for ~ 500 Euro, simply to be able to use DX10? No, thanks. I pass.

"No, you have it wrong! Windows Media Player / Internet Explorer / Direct3D / whatever are tightly integrated into the OS, so it's impossible to take them out, or make them available for old versions of the OS."

Yeah, right. How many times have we heard that crap? If you believe that, I have a nice bridge to sell to you.


I don't use an operating system to write letters, or browse the internet. I use applications for that.

 

[Frank]

But, I bet my scanner is junk once I go to Vista. HP has planned obsolescence in full swing. I imagine my 15 year old Laserjet 4P will continue to function, though, ironically. Probably only because PCL5 is so popular. Windows 3.1 - Vista. A printer that has been around.

Simply tell whatever OS it's a LaserJet II. It works just as well.

Some people like to buy everything new every few years, others simply want to use the stuff they bought as long as it works well. I do as well.



It's like your garage telling you that you need to buy a new car, because the new firmware doesn't support the old one anymore. And you really need the new firmware, because all the replacement parts, like tyres, require that to function.

 

[Albuquerque]

I would like it MUCH better, if Windows wasn't so incredibly stupidly bloated with endless amounts of crap. I don't even use any of the "freebies": I use better apps.

Agreed, except...

What in hell does the GPU driver/model have to do with the OS? Nothing, that's what.

I can't believe you'd even say this, as it's so incredibly wrong that it borders on ignorance -- and I know you're smarter than that. Unless of course you're talking about the DOS days when applications were directly coded to the hardware on the machine. Or maybe you're talking about a console device? Yeah, didn't think so.

If there's a driver, that means there's dependance on the OS kernel in some way, shape or fashion. Hell, even in DOS the drivers were hooked into the (very, very simplistic in today's terms) OS kernel. And while it might be argued that an API could be written directly to the driver, you still have modern programming paradigms to consider regarding protected access to physical hardware.

Really, why do you keep with this thought process? XP could have certainly been updated for SM4, that's not arguable. But to have it completely upgraded to a full and true D3D10 implementation, complete with all the performance enhancements? You're smart enough to know the changes that would have to be made, and you're smart enough to know about the huge compatibility issues you'd suddenly surface -- just like the ones you see in Vista.

Sorry, but you're wrong. Driver model is very much dependant on the OS kernel, and making drastic changes to the driver model means making similar changes to the kernel interface. And making those kinds of changes to an OS kernel is essentially grounds for creating a new OS around it.

I'm not here to argue the merits of cost vs benefit, I'm here to argue common sense as to why D3D10 would never, ever make it to XP and nor should it have.

EDIT
Just for clarification and stating-of-the-obvious, your little example of IE / WMP / DirectX is incredibly flawed too.

IE = no hardware interface.
WMP = no hardware interface.
DirectX = nothing but hardware interface. Your argument holds water on the first two, but not the last.

 

[Frank]

Agreed, except...

I can't believe you'd even say this, as it's so incredibly wrong that it borders on ignorance -- and I know you're smarter than that. Unless of course you're talking about the DOS days when applications were directly coded to the hardware on the machine. Or maybe you're talking about a console device? Yeah, didn't think so.

If there's a driver, that means there's dependance on the OS kernel in some way, shape or fashion. Hell, even in DOS the drivers were hooked into the (very, very simplistic in today's terms) OS kernel. And while it might be argued that an API could be written directly to the driver, you still have modern programming paradigms to consider regarding protected access to physical hardware.

Really, why do you keep with this thought process? XP could have certainly been updated for SM4, that's not arguable. But to have it completely upgraded to a full and true D3D10 implementation, complete with all the performance enhancements? You're smart enough to know the changes that would have to be made, and you're smart enough to know about the huge compatibility issues you'd suddenly surface -- just like the ones you see in Vista.

Sorry, but you're wrong. Driver model is very much dependant on the OS kernel, and making drastic changes to the driver model means making similar changes to the kernel interface. And making those kinds of changes to an OS kernel is essentially grounds for creating a new OS around it.

I'm not here to argue the merits of cost vs benefit, I'm here to argue common sense as to why D3D10 would never, ever make it to XP and nor should it have.

EDIT
Just for clarification and stating-of-the-obvious, your little example of IE / WMP / DirectX is incredibly flawed too.

IE = no hardware interface.
WMP = no hardware interface.
DirectX = nothing but hardware interface. Your argument holds water on the first two, but not the last.

Albuquerque, are you saying that MS cannot make a DX10 subsystem for XP that performs just as well as the Vista variant, simply because they don't have the source code for the OS that is needed to see what they have to do to have it perform in exactly the same way? Or that even MS cannot make a DX10 subsystem that runs in user space, because nobody can access hardware in user space, simply because they don't have the knowledge of how to place the required hooks? Or simply that the managers of the different MS departments don't see eye to eye?

The hard parts of a DX10 driver are about making the GPU memory part of the virtual memory pool and being able to batch draw calls. The first is so the OS can do the memory management and makes the driver part of the OS, while the latter is a real improvement.

Only, the latter can be done even by old versions of OpenGL, while the first is really weird, as they have moved the driver outside kernel space.


It's only marketing. There is nothing technical allowing or preventing whatever. Especially not for MS. It's like why they moved the Presentation Manager to kernel space in NT4: to speed up the pretty graphics. IT managers complained, that the server would go down when the graphical subsystem or UI crashed. Which it did. And who cared about a nice GUI on a server at that time?

I like my servers simple, accessible and especially dependable. Command-line for the win.

 

[Geo]

The peasants with pitchforks are still at least one year, and probably two, too late. It's just that simple. What might have been done then and what is reasonable to do now are two very different things.