AMD vs Intel for video format conversions and editing

[Peabody]

I'm close to springing for a new computer with an Athlon 64 3500+
processor. I don't do any gaming, but do a good bit of video
format conversion and video editing.

I read in a forum somewhere that gamers like AMD, but video editors
like Intel (something to do with hyperthreading). So I wondered if
the experts here have any thoughts on that.

Well, here's the machine I'm looking at, at Circuit City:

eMachines T6522 - full system for $629 after rebates.

AMD Athlon 64 3500+
1GB ram
200 GB HD - parallel ATA
DVD burner
Integrated 128MB ATI Radeon Xpress 200 (display model has 400 chip)
TV tuner with remote
17" CRT
Canon photo printer
Windows Media Center
Three PCI slots, one PCI express slot.

 

[YKhan]

Video editing seems to be the Pentium 4's last remaining advantage over
Athlon 64; everything else is faster on the Athlon. I don't really
think Pentium's advantage is on all video encoders, just a few select
ones like Divx, but I think Xvid is faster on Athlon.

Part of the reason might even be as a result of the fact that Divx uses
the Intel compiler, while Xvid uses the Microsoft one.

Yousuf Khan

 

[Carlos Moreno]

Video editing seems to be the Pentium 4's last remaining advantage over
Athlon 64; everything else is faster on the Athlon. I don't really
think Pentium's advantage is on all video encoders, just a few select
ones like Divx, but I think Xvid is faster on Athlon.

Part of the reason might even be as a result of the fact that Divx uses
the Intel compiler, while Xvid uses the Microsoft one.

Whoa! That would be one of those really-hard-to-believe occasions
in which Microsoft apparently does the right thing?! ;-) That
really doesn't happen often!

Carlos
--

 

[Yousuf Khan]

Whoa! That would be one of those really-hard-to-believe occasions
in which Microsoft apparently does the right thing?! ;-) That
really doesn't happen often!

Not really, look at the targets. Microsoft is only out to screw other
software makers, it doesn't care who's chips it runs on. Intel on the
other hand will try to screw other semiconductor makers.

Whereas Intel might put code in its compilers to generate suboptimal
code paths when run on non-Intel processors, Microsoft might try to sell
its competitors less capable compilers whereas it keeps its better
work for its own inhouse development.

Yousuf Khan

 

[Carlos Moreno]

Not really, look at the targets. Microsoft is only out to screw other
software makers, it doesn't care who's chips it runs on. Intel on the
other hand will try to screw other semiconductor makers.

Whereas Intel might put code in its compilers to generate suboptimal
code paths when run on non-Intel processors, Microsoft might try to sell
its competitors less capable compilers whereas it keeps its better work
for its own inhouse development.

Oh no, of course! It was perfectly clear to me the logic behind
all this (both on the Intel and the Microsoft sides).

Just thought I'd comment on the humor of the situation -- Microsoft's
compiler is doing its job reflecting what really should be happening
(that AMD processors run faster). A Microsoft product *does its
job correctly*, a job that Intel's compiler doesn't. When you look
at the "summary" without digging into the reasons why, you *have* to
admit that there's a certain component of humor in this

Carlos
--

 

[Peabody]

YKhan says...

Video editing seems to be the Pentium 4's last remaining
advantage over Athlon 64; everything else is faster on
the Athlon. I don't really think Pentium's advantage is
on all video encoders, just a few select ones like Divx,
but I think Xvid is faster on Athlon.

Part of the reason might even be as a result of the fact
that Divx uses the Intel compiler, while Xvid uses the
Microsoft one.

Thanks very much. Could it have anything to do with the
instruction sets? I think both have SSE, but could there be
differences in the instruction sets that allow for
Intel-friendly optimization?

 

[Yousuf Khan]

YKhan says...

Thanks very much. Could it have anything to do with the
instruction sets? I think both have SSE, but could there be
differences in the instruction sets that allow for
Intel-friendly optimization?

No actually, it's not quite so innocent as a mere technical difference.
It's got more to do with marketing. Intel's compilers tend to enable
features only if they're running on an Intel processor, even if the
other processor supports those features. So even if the AMD processor
supports SSE2, the compiler will ignore it, and just go with basic
instruction sets. This has been proven when people have manually removed
the "Intel" processor check and let the program run. The program ran
faster than it did before on the Athlon, and it ran just as fast as it
did previously on Pentium. And yes, it did run faster on the Athlon than
on the Pentium, even using the Intel compiler, as long as the processor
identity check was removed.

Unfortunately, there's quite a few well-known applications that are
compiled with the Intel compiler. They basically act as free advertising
for Intel.

Yousuf Khan

 

[George Macdonald]

Whoa! That would be one of those really-hard-to-believe occasions
in which Microsoft apparently does the right thing?! ;-) That
really doesn't happen often!

Depends what you mean by "right thing". It could also be that Intel has
managed to get optimizations from the compiler which are particularly
applicable to their CPU design; it could also be, if you believe AMD's
anti-trust complaint, that Intel has special code in their compiler to
penalize Athlon64.

 

[Carlos Moreno]

Depends what you mean by "right thing". It could also be that Intel has
managed to get optimizations from the compiler which are particularly
applicable to their CPU design; it could also be, if you believe AMD's
anti-trust complaint, that Intel has special code in their compiler to
penalize Athlon64.

Hmm, it seems like my humor is a bit too different from what people
are used to?

Yes, as you can see from my other message, I'm clear on those
possibilities -- they're more or less direct conclusions from what
Yousuf said in his OP, and was just pointing out the component of
humor in it.

Cheers,

Carlos
--

 

[gaffo]

Yousuf Khan wrote:

This has been proven when people have manually removed
the "Intel" processor check and let the program run. The program ran
faster than it did before on the Athlon, and it ran just as fast as it
did previously on Pentium. And yes, it did run faster on the Athlon than
on the Pentium, even using the Intel compiler, as long as the processor
identity check was removed.

Unfortunately, there's quite a few well-known applications that are
compiled with the Intel compiler. They basically act as free advertising
for Intel.

Yousuf Khan

WOW I've never heard this. do you have a site I can read about this?

--

Jurors should acquit, even against the judge's instruction...if
exercising their judgement with discretion and honesty they have a clear
conviction that the charge of the court is wrong.

Alexander Hamilton, 1804

 

[Yousuf Khan]

Yousuf Khan wrote:




WOW I've never heard this. do you have a site I can read about this?

Here's one:

Your Processor, Your Compiler, and You: The Case of the Secret CPUID String
http://www.devx.com/amd/Article/28001

There might be others here:

http://tinyurl.com/blk7j

Yousuf Khan

 

[Peabody]

Yousuf Khan says...

Here's one:

Your Processor, Your Compiler, and You: The Case of the
Secret CPUID String
http://www.devx.com/amd/Article/28001

There might be others here:

http://tinyurl.com/blk7j

As the OP on this thread, I really appreciate having this
information. I guess the question is whether major video
editing and conversion software authors pretty much already
know about this and avoid using Intel compilers, or will do
so in future versions. Can you think of a forum where I
might ask that question and get an informed response from
the multimedia gurus?

I understand that this Intel compiler behavior is one of the
anti-competitive complaints in AMD's antitrust suit against
Intel. I wish them luck in their suit.

 

[YKhan]

As the OP on this thread, I really appreciate having this
information. I guess the question is whether major video
editing and conversion software authors pretty much already
know about this and avoid using Intel compilers, or will do
so in future versions. Can you think of a forum where I
might ask that question and get an informed response from
the multimedia gurus?

I wouldn't worry too much about it. My understanding is that most
companies actually compile their code with the Microsoft (or GCC in
Linux) compiler rather than the Intel one. Intel sort of markets
itself as the Lexus to the Microsoft Toyota: almost the same, just a
little better. There are other companies out there marketing themselves
as the luxury compiler too, like Pathscale and PGI, so Intel isn't the
only luxury compiler game in town.

I understand that this Intel compiler behavior is one of the
anti-competitive complaints in AMD's antitrust suit against
Intel. I wish them luck in their suit.

What's amazing is that despite all of the sabotage, the AMD processors
manage to hold their own against Intel processors, even while executing
basic x86 code rather than the fancy high-performance extended x86
code. It's been said that AMD has come to expect this from Intel so
they've spent a lot of effort in getting their basic x86 instructions
working extremely well, and not relying on instruction extensions to
get them running fast. So they seem to have infused some of the basic
x86 features with resources from the extended x86 features, so some of
the optimizations that are available to the extended instructions may
also be valid in the basic instructions for AMD processors.

Yousuf Khan

 

[gaffo]

Here's one:

Your Processor, Your Compiler, and You: The Case of the Secret CPUID String
http://www.devx.com/amd/Article/28001

There might be others here:

http://tinyurl.com/blk7j

Yousuf Khan

Thank you Sir!!
a true gentleman!
peace to you...



--


If the jury feels the law is unjust, we recognize the undisputed power
of the jury to acquit, even if its verdict is contrary to the law as
given by a judge, and contrary to the evidence...If the jury feels that
the law under which the defendant is accused is unjust, or that exigent
circumstances justified the actions of the accused, or for any reason
which appeals to their logic or passion, the jury has the power to
acquit, and the courts must abide by that decision.

4th Circuit Court of Appeals, United States v. Moylan, 1969

 

[gaffo]

Here's one:

Your Processor, Your Compiler, and You: The Case of the Secret CPUID String
http://www.devx.com/amd/Article/28001

There might be others here:

http://tinyurl.com/blk7j

Yousuf Khan

dumb question............

how did you get your "tinyurl" link to google "intel compiler cpuid
amd"?????????????????


nice trick.




--

Jurors should acquit, even against the judge's instruction...if
exercising their judgement with discretion and honesty they have a clear
conviction that the charge of the court is wrong.

Alexander Hamilton, 1804

 

[Yousuf Khan]

dumb question............

how did you get your "tinyurl" link to google "intel compiler cpuid
amd"?????????????????


nice trick.

I used the Firefox Tinyurl extension. I'm sure you can do it even
without the extension, just by copying the address bar and pasting it
into the Tinyurl website.

Yousuf Khan

 

[EdG]

I used the Firefox Tinyurl extension. I'm sure you can do it even
without the extension, just by copying the address bar and pasting it
into the Tinyurl website.

Yousuf Khan

Sort of funny ,weren't you one of the people who wouldn't even click on
those tiny links I was posting here not so long ago? If not sorry got
you mixed up with one of the other nerds.

Ed

 

[daytripper]

Sort of funny ,weren't you one of the people who wouldn't even click on
those tiny links I was posting here not so long ago? If not sorry got
you mixed up with one of the other nerds.

Ed

ahahahahahahaha!

 

[George Macdonald]

Sort of funny ,weren't you one of the people who wouldn't even click on
those tiny links I was posting here not so long ago? If not sorry got
you mixed up with one of the other nerds.

Nah it was some guy who posted as RusH from Poland who complained to
Yousuf: [email protected]

 

[YKhan]

No, actually it's the opposite way. People have told me not to post
those shortened links, as they were too scared to click on them.

Yousuf Khan