AMD/Linux vs Intel/Microsoft

  • Thread starter Thread starter E
  • Start date Start date
["Followup-To:" header set to comp.os.linux.advocacy.]
Datagram from Jim Richardson incoming on netlink socket
I use gcc 3.x (whatever is the latest in Sid at the time I compile it)
and you can supposedly use icc (the intel compiler) to compile (for
intel cpu's only) Additionally, icc is supposed to have better
optimizations for intel cpus, probably at least in part, because they
don't have to support so many other architectures.

I use GCC 3.2.3 to compile both userland and kernel components. Intel
has tried to make ICC be able to compile kernel, and thus have added
support for some needed GCC extensions.

And GCC never has had speed as a goal. GCC is designed to be highly
portable.

-Ilari
--
When it comes to security of computer attached to the Internet, you
have preceisely two options: To be paranoid or to be cracked. --
Ilari Liusvaara
Linux LK_Perkele_IV9 2.4.23-selinux1 #2 Mon Jan 5 20:12:55 EET 2004 i686 unknown
4:21pm up 4 days, 19:58, 5 users, load average: 0.07, 0.06, 0.01
 
What am I missing? And even if my time estimates are unrealistic, it's
sure a hell of a lot easier for them to do it, ONCE for each distro/XFree,
rather than asking thousands of end-users to make the individual effort.

I'm the customer, ATI. I'm the guy with the money that YOU want. Make
some effort to help me out!

ATI's sales depend primarily on OEMs. If, once we see significant
rollout of Linux boxes from OEMs, ATI will be more likely to offer
Linux drivers.

They've never put enough effort towards drivers, even for WINDOWS,
IMHO.
 
Kevin said:
If a manufacturer chooses to not supply drivers for your chosen OS then
what is the point in 'spitting into the wind' building a machine to run
that OS using parts lacking in decent driver support ?

Exactly. I needed a new photo printer and have used canon in the past. But
when I started using linux I found the driver support for canon printers
sucks and this is canon's fault IMHO. Epson and HP provide the info for
good drivers to be written or actively support linux themselves so guess
whose printer I bought? A hint, it wasn't Canon. Maybe they don't care that
I bought someone else's $400+ printer just because they don't want to be
involved with linux?
 
Fearing a spontaneous XP reboot, stacey mumbled this incantation:
Exactly. I needed a new photo printer and have used canon in the past. But
when I started using linux I found the driver support for canon printers
sucks and this is canon's fault IMHO. Epson and HP provide the info for
good drivers to be written or actively support linux themselves so guess
whose printer I bought? A hint, it wasn't Canon. Maybe they don't care that
I bought someone else's $400+ printer just because they don't want to be
involved with linux?

I have an additional consideration. A company may not see enough of a
market to make Linux support worth the cost. Fine. But a good company
will have developers who are keenly interested in getting their
company's product supported on Linux. They will be motivated to write
the code themselves, and push it through the process to at least have
"officially unsupported" drivers posted at the corporate download site.

An motivated coders are quite often good coders, as shown by the whole
Open/Free software community.
 
If MS (or anyone else) had really wanted RISC editions of NT to
succeed, they could have done worse than given away zillions of
x86->RISC cross-compilers. Sure, it wouldn't have been properly
tested on the new platform, but most software isn't properly
tested on the developer's platform either. :o)

That may be true of the vendors that you deal with, but most application
vendors take testing pretty seriously because it has a direct effect on
customer satisfaction (revenue) and support costs (expense).

The commercial vendors that I'm familiar with would (and did) consider
a cross compilation environment with no opportunity for testing a
non-starter.

Heck, the CAD vendors only certify and support specific versions
of the graphics card drivers!

Tom
 
Tony Hill said:
there. At various times NT was reported to have been running on
PowerPC and MIPS in addition to the Alpha, i386 and IA-64 instruction
sets that it was officially released for. Combined with the upcoming

My first WinNT box was a MIPSpower (MIPS cpu) machine running 3.5.1.
We used it to run a cheaper (than unix) version of Pro/Engineer
(parametric CAD software).

They did work and were fairly fast at the time.
 
E said:
Thats interesting, so Intel might one day have an IA-64 and an AMD x86-64
aimed at desktop PCs?


Yes, I read this same rumor. But I don't understand why if Microsoft already
has a version of XP for the Itanium, they would drop support for it in
favor of a version of Windows for x86-64, and tell Intel to follow suite
with there desktop CPU line. I guess its the built in 32 bit compatiblity
of the AMD x86-64.


My own observation is that it comes down to politics more then
technology.
Microsoft promised support to AMD for x64 when it came out and pulled
back at the last minute for two reasons.

1) Microsoft still needs Intel for sales as Intel still has a
veritable monopoly on consumer systems although they would very much
like this not to be the case very soon. So a bone was kicked in hte
direction on Intel to allow them to claim 64 bit Windows on Itanium.

2) Microsoft is in the process off revamping everything around Rights
Denial
( NGSCB or Palladium if you will ), Microsoft DRM, etc. and really
wants this so that it can pander to the likes of the music and movie
monopoly slobs at the expense of it's customer base who you get the
feeling that they think are stupid enough to believe that it's for
their benefit. You could see the logic in this if you accept that you
are a monopoly leveraging your OS product to sell you productivity and
back room apps and you want to seal up the vast majority of systems
from your competition that you know that you have no other way of
competing against. I think it's interesting to note that both AMD64
based chips and Itanium and all future Intel processors support the NX
( No execute ) register. This is one way to allow NGSCB to lock out
"illegal" applications and for Microsoft and others to control the
list of legal applications that can run in the processor. Do you
suppose this was the price of admission the Microsoft wanted from the
processor people? I think the Intel got the better end of the deal in
that case.

THX
 
+---------------
| ... but all that may have to be re-done... AGAIN. Linux doesn't
| suffer nearly so much from this stupidity (but it does, to some extent,
| as many manufacturers distribute binary-only drivers for Linux.
| That's not the fault of Linux, although many regard binary-only
| drivers as pure evil.)
+---------------

In some cases the manufacturer simply has no other choice. A good case
in point is for wireless devices (e.g., 802-11.{b,a,g}) in which some
aspect of the transmission (frequency, power, or encoding, say) is
dynamically controlled directly by the driver software [so-called
"software-defined radios"]. The FCC has ruled that the end-user *must*
not be able to alter those characteristics of the driver, since it
might result in the device no longer meeting its emissions requirements.
See "Section C: Software Modifications" in the following:

<URL:http://ftp.fcc.gov/Bureaus/Engineering_Technology/Orders/
2001/fcc01264.pdf>

Sam Leffler, in particular, has been working to create a standard for
an API to the FCC-mandated binary parts of wireless drivers (as well as
parts considered vendor-proprietary) for open-source operating systems
(sort of like the old MS-DOS "NDIS" driver API for networking cards),
starting initially with the Atheros chipset and the Linux & FreeBSD
operating systems, to encourage more wireless manufacturers to support
open-source operating systems. See:

<URL:http://cvs.sourceforge.net/viewcvs.py/madwifi/madwifi/
README?rev=1.17>
<URL:http://www.atheros.com/news/linux.html>
<URL:http://wifinetnews.com/archives/002545.html>
<URL:https://sourceforge.net/projects/madwifi/>
<URL:http://ndiswrapper.sourceforge.net/>


-Rob
 
**** 'em. nVidia make some good cards and drivers.

Yeah, I may have to go that way with my next card. I've never liked
nVidia or it's products all that much, but if their card has better Linux
support...

These companies must know that there's a lot people who are going to be
checking-out Linux, and they should be ready. Tux Racer @ 1fps is NO
GOOD! 8)
 
In comp.arch chrisv said:
Yeah, I may have to go that way with my next card. I've never liked
nVidia or it's products all that much, but if their card has better Linux
support...

I don't buy bleeding edge cards, but I've had very good results with X and
the ATI cards I've got (a mobile version of the Radeon 7500 on several
laptops I support, and a desktop Radeon 8500 on my home Linux box.)

I've had good results with the nvidia drivers on old TNT cards, but I can't
get DRI to work properly (~10fps in small glxgears) with the one GeForce
card I've had to set up.
 
In comp.arch Mainframe Linux said:
linux doesn't need fancy 3d drivers
it uses the cpu for x rendering

DRI helps for more than 3D... X on a raw framebuffer is fine with a simple
window manager and no truetype or other complex fonts, but it's an utter dog
when running relatively modern applications, window managers and fonts
without a proper accelerated driver.
 
| On Sat, 10 Jan 2004 14:39:02 +1300, ~misfit~ wrote:
|
|| chrisv wrote:
|||
||| I'm the customer, ATI. I'm the guy with the money that YOU want.
||| Make some effort to help me out!
||
|| **** 'em. nVidia make some good cards and drivers.
|
| Yeah, I may have to go that way with my next card. I've never liked
| nVidia or it's products all that much, but if their card has better
| Linux support...
|
| These companies must know that there's a lot people who are going to
| be checking-out Linux, and they should be ready. Tux Racer @ 1fps is
| NO GOOD! 8)

Exactly - so vote with your wallet ! ;-)
This brings me back to using Matrox video cards in systems which run a
non-Micro$oft OS - often Linux, maybe something else (BeOS ?).
The high-spec nVidia and ATI cards show their best performance ratings
used on Windoze systems - and the battle for top position seems to change
monthly. This isn't much of an indication of how the card might perform
under, say, Linux as the system for displaying complex 3D graphics is a bit
different. Those top-ranking cards rely on Windoze drivers which have been
carefully written to exploit their features. In a Linux box they might not
seem quite so clever.
Matrox have the good sense, decency and end-user committment to offer
Linux drivers for most of their cards - that is one of the reason why I use
them and recommend them. The other two reasons are image quality and the
ability to drive two monitors (dual head). I even have a system running BeOS
giving a dual-head display from its Matrox card (G400).
To me, these things are more important than being able to show about 70
fps in the latest games. My eyes posses a feature called 'persistance of
vision' which prevents such speed being necessary - and ceratinly not worth
paying for.
Kevin.
 
My own observation is that it comes down to politics more then
technology.
Microsoft promised support to AMD for x64 when it came out and pulled
back at the last minute for two reasons.

They haven't really pulled back so much as delayed. That is hardly
anything new for Microsoft, I can't even remember the last time they
shipped something on time! The 6-9 month delay is fairly standard for
their products, no need for any conspiracy theories there!
1) Microsoft still needs Intel for sales as Intel still has a
veritable monopoly on consumer systems although they would very much
like this not to be the case very soon. So a bone was kicked in hte
direction on Intel to allow them to claim 64 bit Windows on Itanium.

Intel needs Microsoft much more than Microsoft needs Intel.
2) Microsoft is in the process off revamping everything around Rights
Denial
( NGSCB or Palladium if you will ), Microsoft DRM, etc. and really
wants this so that it can pander to the likes of the music and movie
monopoly slobs at the expense of it's customer base who you get the
feeling that they think are stupid enough to believe that it's for
their benefit.

Minor FWIW, but NGSCB/Palladium/La Grande/name-of-day actually DOES
have some very beneficial aspects to it. The DRM stuff is really only
one small portion of the whole deal, but of course it's the ONLY part
that ever gets any coverage in the geek-sream media (I don't think any
of it gets coverage in main-stream media :> ). The ability to run a
service in it's own 'sandbox' of a sorts is a very good thing from a
sever security standpoint.

Of course, the goal of the DRM stuff is most certainly not to help the
users, but rather to attract the media companies to release their
wares using Microsoft DRM. What better way to keep people using your
product than to REQUIRE the use of your product to view certain
content?
competing against. I think it's interesting to note that both AMD64
based chips and Itanium and all future Intel processors support the NX
( No execute ) register. This is one way to allow NGSCB to lock out
"illegal" applications and for Microsoft and others to control the
list of legal applications that can run in the processor. Do you

What in the hell are you talking about? First off, the NX bit is just
that, a bit in the page table. It has ABSOLUTELY NOTHING to do with
applications! It does exactly ONE thing and one thing alone, it marks
pages in the page table as executable or non-executable. This is a
feature that is implemented in many (most?) other architectures out
there but has been sadly lacking in x86 (err, it's implemented in
IA32, but only in a rather ass-backwards manner that can not be easily
used by most operating systems).

It's quite unclear whether Intel will support this feature in their
upcoming P4 "Prescott" chips or not. One recent third-party report
was quoted as saying that they will, however several other recent
reports have Intel saying that they the Prescott does not support the
NX bit and they have no plans on doing so.
suppose this was the price of admission the Microsoft wanted from the
processor people? I think the Intel got the better end of the deal in
that case.

Both AMD and Intel have included the hardware required for Microsoft's
NGSCB in their new or upcoming processors. Nothing in ANY of this
hardware has any connection at all to locking out applications or
operating systems.
 
Tony said:
On 12 Jan 2004 06:44:30 -0800, (e-mail address removed) (THX1138) wrote:

Intel needs Microsoft much more than Microsoft needs Intel.


I wouldn' be so sure. Intel just jumped on the anit-SCO bandwagon, bet MS
isn't too happy about that after they dumped ~50million into SCO's pockets
to squash linux. Maybe Intel is getting tired of being strong armed by MS?
 
Intel needs Microsoft much more than Microsoft needs Intel.


Why do you say that? It isn't as though Microsoft can suddenly drop support
for Intel CPUs and support only AMD. The cross licensing between Intel & AMD
goes both ways, so AMD can't create incompatible instructions that give them
any more of an advantage than MMX/SSE/SSE2 gave Intel when they were briefly
the only ones who offered it. Microsoft surely can't start supporting say
PowerPC and get anywhere with it today. In the early 90s the ACE platform
was more of a worry when the PC didn't have much home market penetration.

OK, Microsoft can quit supporting IA64, but IMHO the ship has pretty much
sailed on that one being a mass market for Intel in the Windows world
anyway thanks to AMD forcing Intel to do 64 bit x86.

Intel could certainly retaliate by doing things Microsoft doesn't want and
effectively prevents them from doing, like doing their own software (media
players, etc.) giving x86 full virtualization capability, or throwing 100%
support behind Linux instead of the half assed support they give now. (If
anyone argues it is 100% now, where are the Centrino drivers a year after
the intro of that platform?)
 
Douglas said:
or throwing 100%
support behind Linux instead of the half assed support they give now. (If
anyone argues it is 100% now, where are the Centrino drivers a year after
the intro of that platform?)

Yep that is sad but Intel does seem to be getting behind linux of late? At
least they are anti-SCO which can't please MS.
 
Tony Hill said:
They haven't really pulled back so much as delayed. That is hardly
anything new for Microsoft, I can't even remember the last time they
shipped something on time! The 6-9 month delay is fairly standard for
their products, no need for any conspiracy theories there!

Still, Don't you find it interesting that MS was able to come up with
a 64 bit implimentation of Windows that happens to only run on
Itanium?
Intel needs Microsoft much more than Microsoft needs Intel.


Minor FWIW, but NGSCB/Palladium/La Grande/name-of-day actually DOES
have some very beneficial aspects to it. The DRM stuff is really only
one small portion of the whole deal, but of course it's the ONLY part
that ever gets any coverage in the geek-sream media (I don't think any
of it gets coverage in main-stream media :> ). The ability to run a
service in it's own 'sandbox' of a sorts is a very good thing from a
sever security standpoint.

This is very true but you know you are being obtuse if you think that
is what is driving this.

Intel ( AMD ) and especially MS don't give a wet fart about server
security because if they did they would have had at least 10 years to
fix the problem. They only care about money and security wasn't part
of the deal.
Of course, the goal of the DRM stuff is most certainly not to help the
users, but rather to attract the media companies to release their
wares using Microsoft DRM. What better way to keep people using your
product than to REQUIRE the use of your product to view certain
content?


What in the hell are you talking about? First off, the NX bit is just
that, a bit in the page table. It has ABSOLUTELY NOTHING to do with
applications! It does exactly ONE thing and one thing alone, it marks
pages in the page table as executable or non-executable. This is a
feature that is implemented in many (most?) other architectures out
there but has been sadly lacking in x86 (err, it's implemented in
IA32, but only in a rather ass-backwards manner that can not be easily
used by most operating systems).

It's quite unclear whether Intel will support this feature in their
upcoming P4 "Prescott" chips or not. One recent third-party report
was quoted as saying that they will, however several other recent
reports have Intel saying that they the Prescott does not support the
NX bit and they have no plans on doing so.


Both AMD and Intel have included the hardware required for Microsoft's
NGSCB in their new or upcoming processors. Nothing in ANY of this
hardware has any connection at all to locking out applications or
operating systems.

http://www.theregister.co.uk/content/4/33729.html
MS to intro hardware-linked security for AMD64, Itanium, future CPUs
By John Lettice

If I find it later there is an article on AMD's site that states that
NX IS for DRM...It's a sales pitch for corp types that was link
public.

THX
 
THX1138 said:
http://www.theregister.co.uk/content/4/33729.html
MS to intro hardware-linked security for AMD64, Itanium, future CPUs
By John Lettice

Even that article concedes that the NX page protection isn't DRM.

The article hypothesises that because NX breaks some apps, it may
soften up customers for DRM that breaks rather more. You are free
to buy that theory if you like, but I don't. (NX doesn't break
much, for one thing.)
 
| On Sat, 10 Jan 2004 14:39:02 +1300, ~misfit~ wrote:
|
|| chrisv wrote:
|||
||| I'm the customer, ATI. I'm the guy with the money that YOU want.
||| Make some effort to help me out!
||
|| **** 'em. nVidia make some good cards and drivers.
|
| Yeah, I may have to go that way with my next card. I've never liked
| nVidia or it's products all that much, but if their card has better
| Linux support...
|
| These companies must know that there's a lot people who are going to
| be checking-out Linux, and they should be ready. Tux Racer @ 1fps is
| NO GOOD! 8)

Exactly - so vote with your wallet ! ;-)
This brings me back to using Matrox video cards in systems which run a
non-Micro$oft OS - often Linux, maybe something else (BeOS ?).
The high-spec nVidia and ATI cards show their best performance ratings
used on Windoze systems - and the battle for top position seems to change
monthly. This isn't much of an indication of how the card might perform
under, say, Linux as the system for displaying complex 3D graphics is a bit
different. Those top-ranking cards rely on Windoze drivers which have been
carefully written to exploit their features. In a Linux box they might not
seem quite so clever.

The Nvidia supplied drivers for Linux are quite good. I don't have any
recent references but I seem to remember that some of their cards work
*better* in Linux than in Windows.
Matrox have the good sense, decency and end-user committment to offer
Linux drivers for most of their cards - that is one of the reason why I
use them and recommend them. The other two reasons are image quality and
the ability to drive two monitors (dual head). I even have a system
running BeOS giving a dual-head display from its Matrox card (G400).

Nvidia supplies drivers for Linux for ALL their currently shipping x86
chipsets.

I have systems with Nvidia graphics that are dual head (dvi and vga outs
to a flat panel and tube monitor respectively). I also have systems with
both an Nvidia card and a Matrox card. My dual head Nvidia setup... even
though using one card still does OpenGL perfectly. I get around 2500FPS
from an older GeForce2 GTS card using the Nvidia drivers.
To me, these things are more important than being able to show about
70
fps in the latest games. My eyes posses a feature called 'persistance of
vision' which prevents such speed being necessary - and ceratinly not
worth paying for.

One can get a very decent card from Nvidia for under US$100 and one that
is decent for under US$50. While it would be nice if we could get the
specs from Nvidia or they would completely open-source their drivers I
have no problem with the current ones. The stock XFree86 "nv" driver works
just fine in 2D.

-DU-...etc...
 
Back
Top