Article: Parallelism at HotPar 2010
By: koby m. (kobyDOT_DOUBLE_M.delete@this.gmail.com), August 7, 2010 1:06 am
Room: Moderated Discussions
Linus Torvalds (torvalds@linux-foundation.org) on 8/6/10 wrote:
---------------------------
>Jouni Osmala (josmala@cc.hut.fi) on 8/5/10 wrote:
>>
>>Only problem is that their integrated processor is going
>>on worse process than the competitors.
>
>Problem, yes. "Only problem"? No.
>
>Their even bigger problem is that Intel is just damn good.
>
>There was a much better window back when Intel had its
>head up its ass, and was in denial about the problems with
>Netburst, and holding x86 back due to pushing IPF.
>
>AMD took advantage of that weak period for Intel (and
>others tried ;)
>
>These days? Unless Intel stumbles, I would not want to
>compete against them as a newcomer. Especially not if I
>were to try to sell higher-end graphics for gamers as a
>package. Because that means that you have to compete with
>Intel where they are strongest - you need good performance
>from the CPU and multi-threading too.
>
>[...]
>
>Linus
As long as everyone are comparing their crystal balls...
In some parallel universe (which is, of course, very similar to ours), NVIDIA
has already sorted out what it need to do. And that is, simply put, offer the
main component (and maybe overall design) of "the best gaming laptop ever,
(it's even better then Xbox/PS3)" (according to NVIDIA marketing
department at least).
Now, this main component is really not all that special.
All NVIDIA really did was take the most powerful ARM design they could find
(and buy). Beefed it up so it is powerful enough to run whatever non GPU code
that is required by current and upcoming tripleA games titles. Surrounded it
a heck of a lot of GPU, the kind that doesn't completely ignore power and
thermal considerations (cough*fermi*cough), and that isn’t encumbered by the need
to offer certain facilities to the HPC crowd. Last, embed any/most other
main facilities (as long as you have the die area) so that main component
basically defines the system it is a part of.
Or IOW, a fully fledge gaming console in a laptop form factor, one that
has a 15 inch or larger screen size with a decent resolution("fit for the
hardcore gamer" said the marketing department), input devices, and all the
latest and greatest laptop bells and whistles included.
It's the perfect choice.
First, instead of going head to head with Intel (any path that involves x86)
in NVIDIA main market (games), they went around then. Second, laptops is a market that
can still inject fresh USD oxygen to the company and on top of that, since
NVIDIA is now also the CPU (and chipset/other) provider they are also getting
the slice that those vendors took until now (Intel, meet can of woopass).
Right now, NVIDIA is mostly busy helping game developers port their games to
the new "platform"...
In this universe however (which appear to posses a fondness of placing thing
between a hammer and a hard place) windows7 doesn't run on ARM, although it
appears that Tegra2 is sufficient to run high profile game engine*
(didn’t bother to check the resolution in which it was running).
OTOH, is this proposition that preposterous in this universe?
(probably yes)
Sure, now NVIDIA needs to reinvent a good handful worth of software wheels,
but they are already familiar with most of these wheels. And sure, the cost
will be in 6-7 figures ballpark (or more), but so is the cost of playing against
Intel (in its own backyard no less). And it's not like they have to start from
scratch, they already have one design running Linux, there are already
some software efforts being made in that department (wine for example, even
if as a crud stating point if nothing else), and so on and so forth.
Convincing games studios to port their software will not necessarily be such
a huge challenge, most tripleA games are already multiplatform (although some
wooing will still be necessary at first at least). And a good emulator should(?)
take care of all the older games.
But overall, in contrast to competing successfully with Intel (and AMD) on
x86, is the possibility of developing a platform that can claims superiority in
high-end gaming in the mobile sector with ARM sound all that ludicrous?
(not too crazy to be considered, and just crazy enough to work...?)
*) http://www.geek.com/articles/mobile/tegra-2-runs-unreal-engine-3-puts-high-end-console-3d-graphics-in-your-pocket-2010018/
koby
---------------------------
>Jouni Osmala (josmala@cc.hut.fi) on 8/5/10 wrote:
>>
>>Only problem is that their integrated processor is going
>>on worse process than the competitors.
>
>Problem, yes. "Only problem"? No.
>
>Their even bigger problem is that Intel is just damn good.
>
>There was a much better window back when Intel had its
>head up its ass, and was in denial about the problems with
>Netburst, and holding x86 back due to pushing IPF.
>
>AMD took advantage of that weak period for Intel (and
>others tried ;)
>
>These days? Unless Intel stumbles, I would not want to
>compete against them as a newcomer. Especially not if I
>were to try to sell higher-end graphics for gamers as a
>package. Because that means that you have to compete with
>Intel where they are strongest - you need good performance
>from the CPU and multi-threading too.
>
>[...]
>
>Linus
As long as everyone are comparing their crystal balls...
In some parallel universe (which is, of course, very similar to ours), NVIDIA
has already sorted out what it need to do. And that is, simply put, offer the
main component (and maybe overall design) of "the best gaming laptop ever,
(it's even better then Xbox/PS3)" (according to NVIDIA marketing
department at least).
Now, this main component is really not all that special.
All NVIDIA really did was take the most powerful ARM design they could find
(and buy). Beefed it up so it is powerful enough to run whatever non GPU code
that is required by current and upcoming tripleA games titles. Surrounded it
a heck of a lot of GPU, the kind that doesn't completely ignore power and
thermal considerations (cough*fermi*cough), and that isn’t encumbered by the need
to offer certain facilities to the HPC crowd. Last, embed any/most other
main facilities (as long as you have the die area) so that main component
basically defines the system it is a part of.
Or IOW, a fully fledge gaming console in a laptop form factor, one that
has a 15 inch or larger screen size with a decent resolution("fit for the
hardcore gamer" said the marketing department), input devices, and all the
latest and greatest laptop bells and whistles included.
It's the perfect choice.
First, instead of going head to head with Intel (any path that involves x86)
in NVIDIA main market (games), they went around then. Second, laptops is a market that
can still inject fresh USD oxygen to the company and on top of that, since
NVIDIA is now also the CPU (and chipset/other) provider they are also getting
the slice that those vendors took until now (Intel, meet can of woopass).
Right now, NVIDIA is mostly busy helping game developers port their games to
the new "platform"...
In this universe however (which appear to posses a fondness of placing thing
between a hammer and a hard place) windows7 doesn't run on ARM, although it
appears that Tegra2 is sufficient to run high profile game engine*
(didn’t bother to check the resolution in which it was running).
OTOH, is this proposition that preposterous in this universe?
(probably yes)
Sure, now NVIDIA needs to reinvent a good handful worth of software wheels,
but they are already familiar with most of these wheels. And sure, the cost
will be in 6-7 figures ballpark (or more), but so is the cost of playing against
Intel (in its own backyard no less). And it's not like they have to start from
scratch, they already have one design running Linux, there are already
some software efforts being made in that department (wine for example, even
if as a crud stating point if nothing else), and so on and so forth.
Convincing games studios to port their software will not necessarily be such
a huge challenge, most tripleA games are already multiplatform (although some
wooing will still be necessary at first at least). And a good emulator should(?)
take care of all the older games.
But overall, in contrast to competing successfully with Intel (and AMD) on
x86, is the possibility of developing a platform that can claims superiority in
high-end gaming in the mobile sector with ARM sound all that ludicrous?
(not too crazy to be considered, and just crazy enough to work...?)
*) http://www.geek.com/articles/mobile/tegra-2-runs-unreal-engine-3-puts-high-end-console-3d-graphics-in-your-pocket-2010018/
koby