Intel held an analyst day today

  • Thread starter Thread starter Yousuf Khan
  • Start date Start date
I've had packages back with no chips in them before, but that normally
meant we 'forgot' to test them. I've never managed to write a test
program that passed an empty socket, but I guess there's still time.
'Test escape' normally means you tested it incorrectly, not that your
program doesn't find the fault. So either you used the wrong program, or
bins got mixed up, or it wasn't tested at all.

Test holes, on the other hand, are where you missed some coverage which
the customer didn't. It happens occasionally, and usually equates to a
tiny dpm. So you just add in a test for it, if you can and it's not too
expensive, and move on.

probably old news already, but fwiw....
AMD has changed the screening process for rating the two product lines
as the chips come off the production line, Taylor said. As a result,
some chips that would have been rated with clock speeds of 2.8 MHz in
the past would be listed at 2.6 MHz, making them less likely to be used
in extreme computing environments.
http://www.informationweek.com/news/showArticle.jhtml?articleID=187001959
 
I've had packages back with no chips in them before, but that normally
meant we 'forgot' to test them. I've never managed to write a test
program that passed an empty socket, but I guess there's still time.

It's not hard to do. After this incident (at another site, BTW) we went
through our entire library looking for programs that would pass an empty
socket. We found a surprising number that did. They did test parts, but
also no parts. Oops! After that a macro was added to every test to trap
an emptu socket. This was all 20 years ago (I left that group in '88), so...
'Test escape' normally means you tested it incorrectly, not that your
program doesn't find the fault. So either you used the wrong program, or
bins got mixed up, or it wasn't tested at all.

You have a different dictionary. In mine it means that "we shoulda caught
it but didn't", for any number of reasons.
Test holes, on the other hand, are where you missed some coverage which
the customer didn't. It happens occasionally, and usually equates to a
tiny dpm. So you just add in a test for it, if you can and it's not too
expensive, and move on.

Ok. <shrug>
 
It's not hard to do. After this incident (at another site, BTW) we
went through our entire library looking for programs that would pass
an empty socket. We found a surprising number that did. They did
test parts, but also no parts. Oops! After that a macro was added to
every test to trap an emptu socket. This was all 20 years ago (I left
that group in '88), so...

I suppose it depends on what sort of device you're testing - and what
tester. For instance, I'd expect digital vectors to fail if there's no
chip. But on the other hand you can get very good leakage currents if
you aren't connected to anything. I've done that one myself.
You have a different dictionary. In mine it means that "we shoulda
caught it but didn't", for any number of reasons.

Fair enough. I was just trying to guess what AMD might have meant.
 
I suppose it depends on what sort of device you're testing - and what
tester. For instance, I'd expect digital vectors to fail if there's no
chip. But on the other hand you can get very good leakage currents if
you aren't connected to anything. I've done that one myself.

Intel Microprocessors (though incoming inspection not outgoing product),
so they were digital testers (Fairchild and MegaTest, in our case).
Fair enough. I was just trying to guess what AMD might have meant.

Sure. In another case our group used a bazillion motherboards to burn in
every processor for a few minutes because the test programs weren't good
enough at binning parts. Any number of things could have gone wrong,
gone wrong, gone wrong,...
 
If it is found helpful for marketing purpose,

If they're definign "microarchitecture" down to ink on a "roadmap" well,
they're further gone than even I suspected. Though reading some of the
stuff here I'm getting Deja Vu, all over again. Perhaps it is time to
short!
even different speed
grades of the same thing may be declared "new, massively improved"
architecture.

A gelding may be called a stallion too, but it won't help "him" much.
After all, the multiplier is different, and if Mr. Otellini decides
that the multiplier is what defines the microarchitecture, it will be
so.

....and if he calls his dog's tail a leg, how many feet does Mr. O have in
his mouth?
Most of the anal...ysts will take it gladly up their ...

Sadly so, but the good folks here should be the first ones to call him on
this shite!
 
Don't forget the recent manouver which scattered engineers out to
"solutions" units and mixed in newly hired "social scientists" to stir the
brew.:-) I tend to think Sculley:Apple == Otellini:Intel... no?

Oh, &diety! This reminds me of 1990 with Akers and "platform managers"
and again in the mid '90s and "towers". Get your forks ready! The meat
is on the table.
 
Yousuf Khan said:
Intel held its annual analyst day today. Lots of news coverage about it.
Everybody is asking what's going wrong with Intel and what Intel intends
to do about it.

This article says that Intel will from now on start introducing new
microarchitectures every two years instead of every 4-6 years:

Intel steps up chip cadence | CNET News.com
http://news.com.com/Intel+steps+up+chip+cadence/2100-1006_3-6065925.html

Also Intel has announced that it will start restructuring its business.
With possible job cuts and pairing of underperforming business groups:

"Otellini: Intel plans broad restructuring"
http://wireservice.wired.com/wired/story.asp?section=Business&storyId=1517387

It's also announced a bunch of new platforms in the last little while,
such as Viiv from a few months ago, and now a business desktop platform
called Vpro:

Intel's vPro technology gets smarter with security - TechSpot News
http://www.techspot.com/news/21452-Intels-vPro-technology-gets-smarter-with-security.html

*Yawn*

Lets see...Intel still has no On-Die memory controller. Intel still does
not have a true Dual Core processor. Intel procs do not support Direct
Connect Architecture. Intel procs do not support HyperTransport. Intel has
not made a truly new processor in 8 years.

It is sad, really. Intel used to be an innovator, but became complacent.
Now they are happy to just dream up a bunch of new stickers, a rebranding,
and add more confusion to the landscape.

Can anyone honestly say that *ViiV* has brought anything new to the table?

Intel used to have a clue. Now they just depend on smoke and mirrors.

Bobby
 
Lets see...Intel still has no On-Die memory controller. Intel still does
not have a true Dual Core processor. Intel procs do not support Direct
Connect Architecture. Intel procs do not support HyperTransport. Intel has
not made a truly new processor in 8 years.
It is sad, really. Intel used to be an innovator, but became complacent.
Now they are happy to just dream up a bunch of new stickers, a rebranding,
and add more confusion to the landscape.

Can anyone honestly say that *ViiV* has brought anything new to the table?

Intel used to have a clue. Now they just depend on smoke and mirrors.

Despite all that, the signs are their new Conroe, nevermind if it's
"new" or just "evolved", is significantly better than the competition.
Hardly just smoke and mirrors there :P
 
That depends on the spin they put on it. It was sufficient to
repackage PPro in slot 1 cartridge to call it PII - new
microarchitecture, isn't it?

Don't forget dropping the cache to half speed...
 
Ed said:
They intend to sabotage some Opterons so AMD gets some bad press. :)
http://www.amd.com/us-en/0,,3715_13965,00.html?redir=CORPR01

I'm not sure who "they" are here, are you implying that Intel somehow
forced defects into the AMD QA? Or that they planted a phony web page on
AMD and there's really no problem? I read this as "when they get really
hot they work funny," which applies to most CPUs. AMD seems to be
handling this better than Intel handled the f00f bug, but I don't recall
a vendor having to replace a whole generation of CPUs before.
 
Intel procs do not support Direct Connect Architecture. Intel procs do not support HyperTransport.
Isn't Direct Connect Architecture and HyperTransport the same thing?
Intel has not made a truly new processor in 8 years.
Maybe because they don't need to? (yet)

Ed
 
Ed said:
Isn't Direct Connect Architecture and HyperTransport the same thing?



Maybe because they don't need to? (yet)

Ed
Sure they did. You guys just don't like it. (Itanium)

:-)
Crospost trimmed, spruikers.
 
Sure they did. You guys just don't like it. (Itanium)

:-)

Hmm, is that a new processor that came out in the last 8 yrs or just a
old processor from 10+ years ago that finally got kinda fixed in the
last 8yrs? :P
 
I'm not sure who "they" are here, are you implying that Intel somehow
forced defects into the AMD QA?

D'oh - did you miss the ":)"?
Or that they planted a phony web page on
AMD and there's really no problem? I read this as "when they get really
hot they work funny," which applies to most CPUs. AMD seems to be
handling this better than Intel handled the f00f bug, but I don't recall
a vendor having to replace a whole generation of CPUs before.

What "whole generation"? It was a "test escape" on a "limited number" of
certain models of Opteron. They got the timing margins wrong on supposedly
up to 3,000 chips during qualification; while heat may have been a
side-issue of that, it was not the primary fault.
 
George said:
D'oh - did you miss the ":)"?


What "whole generation"? It was a "test escape" on a "limited number" of
certain models of Opteron. They got the timing margins wrong on supposedly
up to 3,000 chips during qualification; while heat may have been a
side-issue of that, it was not the primary fault.
Sorry, I meant that Intel had to replace, or at least offer to replace,
all of the Pentia with the floating point bug, which was the first time
(AFAIK) a mass market CPU was caught that way. Intel managed a
workaround for the f00f bug, luckily.
 
Sorry, I meant that Intel had to replace, or at least offer to replace,
all of the Pentia with the floating point bug, which was the first time
(AFAIK) a mass market CPU was caught that way. Intel managed a
workaround for the f00f bug, luckily.

It's impossible for me to ever forget about the Intel FPU bug,
I program in VB6 and one of the compiler options is, "Remove Safe
Pentium(tm) FDIV Checks". :)
http://img288.imageshack.us/img288/8051/opteronheat1hi.gif
Ed
 
It's impossible for me to ever forget about the Intel FPU bug,
I program in VB6 and one of the compiler options is, "Remove Safe
Pentium(tm) FDIV Checks". :)
http://img288.imageshack.us/img288/8051/opteronheat1hi.gif
Ed

You forgot to mention "favor PPro" option - BTW, if any AMD lawyers
read this, here is another proof of Intel (ab)using its market
dominance. ;-)
But seriously I thought VB6 is near-extinct now that VB.NET (aka VB7)
is in its 3rd iteration. It's quite a surprise that somebody is still
actively using it.

NNN
 
You forgot to mention "favor PPro" option - BTW, if any AMD lawyers
read this, here is another proof of Intel (ab)using its market
dominance. ;-)
But seriously I thought VB6 is near-extinct now that VB.NET (aka VB7)
is in its 3rd iteration. It's quite a surprise that somebody is still
actively using it.

NNN

There are still a lot of folks using VB6 and want to keep using it.
http://classicvb.org/petition/

I don't like .Net, most of my current code and years of saved routines
will no longer work or requires a ton of fixes/rewrites. I basically use
calls to the Win API so my programs run on WinXP without the user
needing to install anything extra, all they have to do is run the EXE. I
don't even have .net framework on my PCs.

When VB6 is finally dead on Windows that'll probably be the day I dump
Windows all together. :)

Ed
 
There are still a lot of folks using VB6 and want to keep using it.
http://classicvb.org/petition/

I don't like .Net, most of my current code and years of saved routines
will no longer work or requires a ton of fixes/rewrites. I basically use
calls to the Win API so my programs run on WinXP without the user
needing to install anything extra, all they have to do is run the EXE. I
don't even have .net framework on my PCs.

When VB6 is finally dead on Windows that'll probably be the day I dump
Windows all together. :)

Ed

If your code is heavy on WinAPI then yes you'll have to do a ton of
fixes to make it work under .NET. Also all COM/COM+/DCOM, while can
be made to work under .NET (that's what COM interop is for), will work
much better if recoded to .NET. Other than that, old VB6 code can be
just recompiled in VB.NET (OK, with some minor fixes like () after the
functions, and such). But for new development .NET is just more
powerful, has more and better options, is more future-proof, and
simply more convenient to use. When I switched from VS6 to VS.NET,
the feeling was similar to that as when I switched from old Geo to a
new (well, slightly used) Volvo - so much more power and comfort. And
you DO have .NET framework, unless you use older Windows - it comes
with WinXP standard.

NNN
 
Back
Top