Is Itanium the first 64-bit casualty?

Brett Davis wrote:

The big problem for both these firms is OutOfOrder design and getting
all the details right, which took AMD a decade. Also the portfolio of
patents needed to actually implement OoO for x86 in a competitive way.

VIAs latest core is OoO.

--
Mvh./Regards, Niels Jørgen Kruse, Vanløse, Denmark

Brett Davis wrote:
In article , Yousuf Khan
wrote:

Brett Davis wrote:
The only question I have about NVDA is who is going to buy the remains
when the company craters in two years.
Nvidia still has the cash to buy VIA many times over. VIA owns an x86
core that can be made to compete in more markets with some adequate R&D
budget.

Lots of people have the money to buy VIA, its just an engineering firm
with 2,000 people. VIA is years and billions away from being a real
threat to Intel, and Intel having learned its lesson from letting AMD
live is not about to make that mistake twice.
Intel is using Engineering (Atom and 45nm) and marketing/bribes/threats
to VIAs supply chain shut VIA down. I do not see how VIA survives as an
independent.
The firm that buys VIA might just want the engineers, and would get a
under the table bonus from Intel for shutting down x86 development.
Otherwise Intel will just keep the pressure up with Atom, etc.

VIA has an x86 license, which is something Nvidia doesn't have. I'm sure
Nvidia could build their own x86 core if they wanted to, but they
wouldn't be allowed to sell it.

There is also the Renesas RX series CPUs which is already almost an
exact copy down to including string opcodes. The opcode encoding is
completely different, but that is just a detail.

http://documentation.renesas.com/eng...b0435_rxsm.pdf

Are you saying this processor is x86 compatible without them claiming to
be x86 compatible?

The big problem for both these firms is OutOfOrder design and getting
all the details right, which took AMD a decade. Also the portfolio of
patents needed to actually implement OoO for x86 in a competitive way.

VIA now has the out-of-order execution, with its latest core.

VIA Nano™ Processor - VIA Technologies, Inc.
"The superscalar, speculative out-of-order 'Isaiah, architecture of the
VIA Nano processor family supports a full 64-bit instruction set and
provides for macro-fusion and micro-fusion functionality, and
sophisticated branch prediction for greater processor efficiency and
performance. "
http://www.via.com.tw/en/products/processors/nano/

Yousuf Khan

Brett Davis wrote:
I could argue that the recent run up in NVidea was Wall Street in a
panic putting the screws on Intel to make a deal now, before the price
goes "higher". When everyone knows the price is going down. Look at the
top ten holders of NVidea, these big Wall Street firms cannot sell
without collapsing the price, so they are pushing hard for someone to
buy NVidea, or they will end up with big losses.
http://finance.yahoo.com/q/mh?s=NVDA


There would be considerable anti-trust implications for Intel if they
bought Nvidia. They already own the majority of the overall graphics
market share, albeit with non-cutting-edge just basic graphics. If they
bought Nvidia, they'd have control over the cutting-edge graphics too.

Yousuf Khan

In article ,
wrote:

In article ,
(Brett Davis) wrote:

There is also the Renesas RX series CPUs which is already almost an
exact copy down to including string opcodes. The opcode encoding is
completely different, but that is just a detail.

No it isn't. It means that you can't break out of the embedded market
into the high-volume desktop/laptop market, because you won't have
Windows. When Intel first planned EM64T, they were going to have a
different encoding to AMD64, so that 64-bit software would have to be
built separately for Intel and AMD. Microsoft just said "no".

The embedded market is the high volume market, not low end x86.
Once the volumes are up and the bugs worked out, then you worry about
changing to x86 encoding. Changing the instruction decoder ROM is easy,
especially if you have planned for it.

Now you have volume sales protected from Intels interference, flip a
switch for x86 compatibility and start selling into the low end x86
market.

Intel sees this coming, which is why Atom exists, to turn the low end
x86 market into a profit wasteland.

They weren't prepared to support a different encoding for the sake of
artificial market segmentation, even when Intel asked nicely, so they
seem quite unlikely to do so when a small CPU company with far less
market muscle asks them.

The big problem for both these firms is OutOfOrder design and getting
all the details right, which took AMD a decade. Also the portfolio of
patents needed to actually implement OoO for x86 in a competitive way.

OK, so they're doomed to uncompetitive performance for the desktop and
laptop market too. Definitely not going to break out of the embedded
market. AMD have significant patents that they license to Intel and
vice-versa. Until those patents expire, anyone else trying to get into
the high-end x86 chip business is going to have to have some major
inventions and be prepared to cross-license them, or have very deep
pockets to buy their way in.

In article , Yousuf Khan
wrote:

Brett Davis wrote:
In article , Yousuf Khan
wrote:

Brett Davis wrote:
The only question I have about NVDA is who is going to buy the remains
when the company craters in two years.
Nvidia still has the cash to buy VIA many times over. VIA owns an x86
core that can be made to compete in more markets with some adequate R&D
budget.

Lots of people have the money to buy VIA, its just an engineering firm
with 2,000 people. VIA is years and billions away from being a real
threat to Intel, and Intel having learned its lesson from letting AMD
live is not about to make that mistake twice.
Intel is using Engineering (Atom and 45nm) and marketing/bribes/threats
to VIAs supply chain shut VIA down. I do not see how VIA survives as an
independent.
The firm that buys VIA might just want the engineers, and would get a
under the table bonus from Intel for shutting down x86 development.
Otherwise Intel will just keep the pressure up with Atom, etc.

VIA has an x86 license, which is something Nvidia doesn't have. I'm sure
Nvidia could build their own x86 core if they wanted to, but they
wouldn't be allowed to sell it.

Just like with AMD, that license goes away on corporate change of
control. NVidea would only get the design and the engineers, and would
face lawsuits over the VIA chips.

There is also the Renesas RX series CPUs which is already almost an
exact copy down to including string opcodes. The opcode encoding is
completely different, but that is just a detail.

http://documentation.renesas.com/eng...b0435_rxsm.pdf

Are you saying this processor is x86 compatible without them claiming to
be x86 compatible?

No, I am saying it is easy to change the decoder ROM to add x86
compatibility. I am assuming that some key Intel patents are soon to
expire. Another reason for Intel to plant its stake in the low end
market first with Atom.

The big problem for both these firms is OutOfOrder design and getting
all the details right, which took AMD a decade. Also the portfolio of
patents needed to actually implement OoO for x86 in a competitive way.

VIA now has the out-of-order execution, with its latest core.

VIA Nano� Processor - VIA Technologies, Inc.
"The superscalar, speculative out-of-order 'Isaiah, architecture of the
VIA Nano processor family supports a full 64-bit instruction set and
provides for macro-fusion and micro-fusion functionality, and
sophisticated branch prediction for greater processor efficiency and
performance. "
http://www.via.com.tw/en/products/processors/nano/

Yousuf Khan

In article ,
(Brett Davis) wrote:
wrote:
No it isn't. It means that you can't break out of the embedded
market into the high-volume desktop/laptop market, because you
won't have Windows. When Intel first planned EM64T, they were going
to have a different encoding to AMD64, so that 64-bit software
would have to be built separately for Intel and AMD. Microsoft just
said "no".

The embedded market is the high volume market, not low end x86.

Ah, I wasn't clear. I'm perfectly well aware that the embedded market
has much higher volume, although at lower unit prices. What I failed to
do clearly was distinguish the non-x86 part of the desktop/laptop market
(which is very small) from the x86 part.

Once the volumes are up and the bugs worked out, then you worry about
changing to x86 encoding. Changing the instruction decoder ROM is
easy, especially if you have planned for it.

Now you have volume sales protected from Intels interference, flip a
switch for x86 compatibility and start selling into the low end x86
market.

If you can evade the patents, or they've expired. Otherwise you're
definitely committing commercial suicide. It's a pretty risky business
anyway; how many x86 challengers have failed to establish themselves so
far? VIA acquired the remains of at least two engineering teams, and
while they have a niche, they don't seem to be growing fast.

John

Brett Davis wrote:
In article , Yousuf Khan
wrote:
VIA has an x86 license, which is something Nvidia doesn't have. I'm sure
Nvidia could build their own x86 core if they wanted to, but they
wouldn't be allowed to sell it.

Just like with AMD, that license goes away on corporate change of
control. NVidea would only get the design and the engineers, and would
face lawsuits over the VIA chips.

Not necessarily. I've not heard that about VIA's x86 license. My
understanding is that VIA is just a client licensee, and that as long as
they simply keep paying the royalty fees, they can continue to use it no
matter what their ownership status. On the other hand, the AMD x86
license is an actual cross-license where AMD gets to use certain Intel
patents for free, and vice-versa, under certain circumstances. The
circumstances for AMD are of course AMD's ownership status, and its
internal manufacturing status.

There is also the Renesas RX series CPUs which is already almost an
exact copy down to including string opcodes. The opcode encoding is
completely different, but that is just a detail.

http://documentation.renesas.com/eng...b0435_rxsm.pdf
Are you saying this processor is x86 compatible without them claiming to
be x86 compatible?

No, I am saying it is easy to change the decoder ROM to add x86
compatibility. I am assuming that some key Intel patents are soon to
expire. Another reason for Intel to plant its stake in the low end
market first with Atom.

Transmeta went that way, and the results were none-too-good for performance.

Yousuf Khan

Brett Davis wrote:
The embedded market is the high volume market, not low end x86.
Once the volumes are up and the bugs worked out, then you worry about
changing to x86 encoding. Changing the instruction decoder ROM is easy,
especially if you have planned for it.

Now you have volume sales protected from Intels interference, flip a
switch for x86 compatibility and start selling into the low end x86
market.

Intel sees this coming, which is why Atom exists, to turn the low end
x86 market into a profit wasteland.


This magic x86 decoder ROM that you keep talking about, doesn't exist.
The decoding is all done internally in the processor, and the opcodes
are built-into the processor. Transmeta is the only one that ever tried
fully reprogrammable opcodes, and it wasn't a great performer.

Yousuf Khan

On Jun 6, 9:56*pm, Yousuf Khan wrote:


There would be considerable anti-trust implications for Intel if they
bought Nvidia. They already own the majority of the overall graphics
market share, albeit with non-cutting-edge just basic graphics. If they
bought Nvidia, they'd have control over the cutting-edge graphics too.

So, the best anti-trust solution is an AMD/ATI monopoly.

So far as the presumed death of Nvidia is concerned and the recent
stock price runup, Tegra might figure into it all somewhe

http://techvideoblog.com/computex/nv...bile-software/

http://www.arm.com/iqonline/news/partnernews/25208.html

and over a hundred others from google news.

Moving more and more action onto the GPU with a fairly wimpy CPU looks
like a serious threat to x86 to me. The wimpy CPU can handle legacy
software. Modern software will make maximum use of GPU-like
processors. Even Microsoft will have to respond.

Robert.

In article , Yousuf Khan
wrote:

Brett Davis wrote:
The embedded market is the high volume market, not low end x86.
Once the volumes are up and the bugs worked out, then you worry about
changing to x86 encoding. Changing the instruction decoder ROM is easy,
especially if you have planned for it.

Now you have volume sales protected from Intels interference, flip a
switch for x86 compatibility and start selling into the low end x86
market.

Intel sees this coming, which is why Atom exists, to turn the low end
x86 market into a profit wasteland.


This magic x86 decoder ROM that you keep talking about, doesn't exist.
The decoding is all done internally in the processor, and the opcodes
are built-into the processor. Transmeta is the only one that ever tried
fully reprogrammable opcodes, and it wasn't a great performer.

There is a tiny part of the CPU that maps a read opcode byte to the
instruction used. The 6502 CPU for example has only 4,000 transistors
for everything and implements a pretty complex opcode scheme.

Changing an ADD so it maps to hex 5A instead of hex B6 is not that hard.
Even after adding all the x86 garbage, its still just not a big deal.

The hard parts are the millions of transistors that implement the
OutOfOrder part of the CPU. And the patents you will run into.

Brett