removing heatsink with Arctic Silver Ceramique...?

[Richard Hopkins]

Decent LGA sockets aren't cheap.

There's no fundamental reason why this design of socket should be
fundamentally more expensive than the mPGA solution, as the requirements are
basically the same.

IIRC Foxconn charges something like four times as much for a Socket 775
assembly as they do for an mPGA478 at the moment, so in some ways you're
right. However, pretty much anything gets cheap if you make enough of 'em,
so as volumes ramp and production techniques improve, I'm sure the cost will
drop.

Huh? Have you looked at the Opteron stack?

Of course. However, how AMD do their stuff is AMD's business. This thread's
about LGA775, and why Intel chose the design they did as the successor to
mPGA478.

....but has no benefit to the end user, other than another place for the
board maker (the one making close to nothing) to screw up.

Don't forget the biggest manufacturer of motherboards for Intel processors
is... Intel. It's not in their interests to make a socket design that'll
come back to bite them, while it's also unlikely that they'd set out to make
a socket that'd annoy their chipset customers.

Again, there is good reason LGA's didn't make it to the PC space
before.

Sure, the reason was that ZIF sockets were a proven solution that met the
requirements of previous technologies. The PC industry doesn't stand still,
and from time to time solutions need to be rethought. This is just one of
those occasions. Did you react with equal skepticism when PCI replaced ISA
and VL-Bus?

This is a definite wait-n-see.

It's difficult not to see this as flat-earthism of a sort. Sure, the new
socket looks very different, but from a system integrity and electrical
standpoint it looks good. All the scaremongering that's been going on is
frankly ridiculous, and very difficult to understand.
--


Richard Hopkins
Cardiff, Wales, United Kingdom
(replace .nospam with .com in reply address)

The UK's leading technology reseller www.dabs.com
Get the most out of your digital photos www.dabsxpose.com

 

[Richard Hopkins]

"Rob Stow" wrote in message...

And if there are pins to be broken, I'd as soon have
them broken on a CPU.

The thing you're forgetting is that ZIF sockets are also full of little
fiddly components, many of which are effectively moving parts. The higher
the pin density gets, the more fiddly and fragile they get. If Intel had
stuck with the pinned CPU/ZIF design for the 775 conductor socket, there's
every likelihood that socket damage would have been *more* of an issue than
whatever results from Socket 775.

It takes a heck of a lot less time to replace a CPU than it takes for
a motherboard swap.

I can see where your concern is coming from, but bitching about what we got
while failing to consider what the alternative would have been seems a
little unrealistic. The pinned CPU/ZIF socket concept becomes increasingly
difficult to implement as the pin density rises.

In any case, have you actually broken an LGA775 assembly yet? While outright
ham-fistedness probably wouldn't do them any good, they're not chronically
fragile as some seem to be fearing/painting them as.

Breaking a pin off a CPU socket would be a real ... umm ...
annoyance after you told the customer that his server would
only be down for ten minutes while you upgraded the
processors and added some RAM.

Come again? You reckon you can take a server out of service, strip it,
upgrade two (for the sake of example) processors, add at least one DIMM,
reassemble the case and get it back into service within ten minutes? You
might get it done in the time if you rush, but if you're rushing, there's
more chance of screwing something up.

Provided you're not terminally clumsy or somewhat unlucky you won't break an
LGA775 socket. Surely the way to go would be to tell your client that the
server will be down for 20 minutes, and take the extra to make sure you get
everything right without having to rush. That way you'll probably take less
time than the estimate so leaving the customer smiling and impressed with
your apparently rapid handiwork.

You almost always have a spare CPU on hand, but even if you have
a spare motherboard on hand it can cost the customer a lot of money
if his server is down for an hour instead of the 10 minutes you
promised him. The price difference between a CPU and a
motherboard can be trivial in that situation.

If anything, the Socket 775 design is probably *more* robust than the
obvious alternative. By simplifying the design of the socket and
re-orienting the conductor "fingers" horizontally instead of vertically,
Intel have effectively freed up more space for the conductors, which allows
them to be larger than they otherwise would have been, and hence
stronger/more durable.

Yes I know AMD stuck with the "old way" for their 939/940 pin designs, but
their CPU substrates are appreciably bigger than Intel's, and they also use
the full area of the socket for pins, something that either apparently
wasn't an option for Intel, or if it was, one they didn't feel was as good
as what they went with. As I said further up, just wait and see what AMD's
next package looks like...
--


Richard Hopkins
Cardiff, Wales, United Kingdom
(replace .nospam with .com in reply address)

The UK's leading technology reseller www.dabs.com
Get the most out of your digital photos www.dabsxpose.com

 

[keith]

There's no fundamental reason why this design of socket should be
fundamentally more expensive than the mPGA solution, as the requirements are
basically the same.

DO you have experience in this area? I have a little. LGAs sockets
aren't all that wonderful. They are expensive.

IIRC Foxconn charges something like four times as much for a Socket 775
assembly as they do for an mPGA478 at the moment, so in some ways you're
right. However, pretty much anything gets cheap if you make enough of
'em, so as volumes ramp and production techniques improve, I'm sure the
cost will drop.

Perhaps. I'll wait for the dummies to pay, thanks. LGA doesn't pay
unless you're going to bond 'em right to the board, in which case BGA is a
better solution, IMO. Unfortunately, the PC marketplace doesn't allow
this.

Of course. However, how AMD do their stuff is AMD's business. This
thread's about LGA775, and why Intel chose the design they did as the
successor to mPGA478.

Didn't you say the fansink was going to rip out the pins and bend 'em or
sonething equally silly? Hell, even the PII had better retention than
this. Or are you saying that Intel has completely lost their marbles?

My point is that there *are* ways of securing the stack, better than a
LGA. In fact the LGA package, by itself, solves nothing. The brackets
around it still have minimize the forces on the chip/socket interface.

Don't forget the biggest manufacturer of motherboards for Intel
processors is... Intel.

Wrong! Intel hasn't made a board in *years* (five or six, IIRC). They
rely on the same board makers as everyone else.

It's not in their interests to make a socket
design that'll come back to bite them, while it's also unlikely that
they'd set out to make a socket that'd annoy their chipset customers.

Their track record of "not getting bitten" isn't too great these days.

Sure, the reason was that ZIF sockets were a proven solution that met
the requirements of previous technologies. The PC industry doesn't stand
still, and from time to time solutions need to be rethought. This is
just one of those occasions. Did you react with equal skepticism when
PCI replaced ISA and VL-Bus?

Irrelevant posturing. LGAs are *not* new. Just because Intel now makes a
stink about them only makes it new stink.

It's difficult not to see this as flat-earthism of a sort. Sure, the new
socket looks very different, but from a system integrity and electrical
standpoint it looks good. All the scaremongering that's been going on is
frankly ridiculous, and very difficult to understand.

LGA is *NOT* new. Can't you get that through your mind?! I
worked with LGAs five-six years ago. I didn't like them then (only BGA
sockets were worse) and don't see how they're going to be magically
improved by Intel. ZIFs aren't all that wonderful, but large pinout ZIFs
are well known, and cheap.

 

[keith]

"Rob Stow" wrote in message...

The thing you're forgetting is that ZIF sockets are also full of little
fiddly components, many of which are effectively moving parts. The higher
the pin density gets, the more fiddly and fragile they get. If Intel had
stuck with the pinned CPU/ZIF design for the 775 conductor socket, there's
every likelihood that socket damage would have been *more* of an issue than
whatever results from Socket 775.

....and we had 1800 pin ZIFs 30 years ago. They were a tad on the
expensive side, but they worked fine. They still work and are cheap.

The 940 pin Opteron package is quite robust. I don't see a couple of
hundred more pins to be any big problem. Keeping that many pins planar,
I see to be a real problem though.

I can see where your concern is coming from, but bitching about what we
got while failing to consider what the alternative would have been seems
a little unrealistic. The pinned CPU/ZIF socket concept becomes
increasingly difficult to implement as the pin density rises.

No more so than a pinned socket. Something has to make up for the
difference in planarity between the board and the chip. A pin in a clip
is proven to work very well.

In any case, have you actually broken an LGA775 assembly yet? While
outright ham-fistedness probably wouldn't do them any good, they're not
chronically fragile as some seem to be fearing/painting them as.

I havent' touched a 775, and am never likely to do so. However Iv'e seen
a fair share of LGAs wasted half that size. ...adn they were *expensife*
sockets ($50 each). We spend upwards of $3K each for good
low-inductance BGA/LGA sockets.

Yes I know AMD stuck with the "old way" for their 939/940 pin designs,
but their CPU substrates are appreciably bigger than Intel's, and they
also use the full area of the socket for pins, something that either
apparently wasn't an option for Intel, or if it was, one they didn't
feel was as good as what they went with. As I said further up, just wait
and see what AMD's next package looks like...

Ah, there you have it. The "new" way (which isn't) is *always* better
than the "old" way. ...that works perfectly well. Thre is a reason LGA
didn't take the world by storm 5-10 years ago.

 

[Rob Stow]

"Rob Stow" wrote in message...



The thing you're forgetting is that ZIF sockets are also full of little
fiddly components, many of which are effectively moving parts. The
higher the pin density gets, the more fiddly and fragile they get. If
Intel had stuck with the pinned CPU/ZIF design for the 775 conductor
socket, there's every likelihood that socket damage would have been
*more* of an issue than whatever results from Socket 775.



I can see where your concern is coming from, but bitching about what we
got while failing to consider what the alternative would have been seems
a little unrealistic. The pinned CPU/ZIF socket concept becomes
increasingly difficult to implement as the pin density rises.

In any case, have you actually broken an LGA775 assembly yet? While
outright ham-fistedness probably wouldn't do them any good, they're not
chronically fragile as some seem to be fearing/painting them as.

I've been getting my paranoia from the review sites.

The very few LGA775 systems I've seen are so brand
spanking new that they are still a year away from
cpu upgrades.

Come again? You reckon you can take a server out of service, strip it,
upgrade two (for the sake of example) processors, add at least one DIMM,
reassemble the case and get it back into service within ten minutes? You
might get it done in the time if you rush, but if you're rushing,
there's more chance of screwing something up.

I have my tools, new parts, and the work area ready
before the server is shut down. It *is* rushing -
but I don't tell the customer I can do it that fast
unless he is pushing for a rush job and he signs
off on the risk.

Usually I can convince them that if the business is
so dependent on that one server that a saving a few
minutes is such a big deal, then maybe they need a
more robust solution.

Provided you're not terminally clumsy or somewhat unlucky you won't
break an LGA775 socket. Surely the way to go would be to tell your
client that the server will be down for 20 minutes, and take the extra
to make sure you get everything right without having to rush. That way
you'll probably take less time than the estimate so leaving the customer
smiling and impressed with your apparently rapid handiwork.



If anything, the Socket 775 design is probably *more* robust than the
obvious alternative. By simplifying the design of the socket and
re-orienting the conductor "fingers" horizontally instead of vertically,
Intel have effectively freed up more space for the conductors, which
allows them to be larger than they otherwise would have been, and hence
stronger/more durable.

Yes I know AMD stuck with the "old way" for their 939/940 pin designs,
but their CPU substrates are appreciably bigger than Intel's, and they
also use the full area of the socket for pins, something that either
apparently wasn't an option for Intel, or if it was, one they didn't
feel was as good as what they went with. As I said further up, just wait
and see what AMD's next package looks like...

I've been awaiting it with dread. I've heard that it is
going to have 1270 pins.

 

[George Macdonald]

in message...

They're not really "pins", they're fingers. IIRC the "Socket T" nickname
originally came from the fact that the hole in the top of the CPU retainer
is sorta "T"-shaped.


Yes and no. Intel's transition to this design is nothing more than a logical
side effect of increasing pin counts and decreasing CPU package sizes. More
connections in a smaller package means higher pin density, so you have to
cram the pin, insulation and (if you keep the ZIF socket) all the contacts
and sliding stuff into ever decreasing spaces. Intel obviously decided the
time was right to re-evaluate things, and design something that'll work for
the current and a number of future generations of CPU's.

So yes, as AMD's CPU's are also transitioning to smaller packages and larger
pin counts, it's only a matter of time before they adopt a similar design.
Probably sooner rather than later, seeing as Intel have now done all the
R&D, and taken the initial lumps for daring to reinvent the wheel.

AMD's pin density has been higher for some time - we'll see if they follow
Intel or not -- and you've already been told that Intel did not invent
anything. I don't see much R&D here.

Just another bullshit scare story that stupid people like to spread. This
all comes from Intel specifying that the design should be capable of at
least 20 successful insert/extract procedures. It doesn't mean that your
socket will self-destruct the 21st time you pull the chip out.

The thing the stupid people forget to mention is that Socket 478, Socket
370, Socket 7, basically every ZIF socket Intel ever designed was also only
ever specified for 20 operations, so in this respect it's no worse any any
of its predecessors. Provided you're not ham-fisted or simply unlucky,
there's no reason to expect you'd have problems with it.

Anyone can tell at a glance that the mechanism is more fragile and prone to
damage... though it does make reqts. for handling & packaging of the CPU
less rigorous.

Richard Hopkins
Cardiff, Wales, United Kingdom
(replace .nospam with .com in reply address)

The UK's leading technology reseller www.dabs.com
Get the most out of your digital photos www.dabsxpose.com

I'm not sure how other people feel about direct advertising in this
newsgroup but I'm sure there are a number of other people -- not me -- who
would like to benefit from the privilege if accorded. If it comes to a
question of all or none I'd think the choice would have to be the latter...
but maybe others can comment??

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

 

[keith]

On Mon, 03 Jan 2005 02:42:19 -0500, George Macdonald wrote:

I'm not sure how other people feel about direct advertising in this
newsgroup but I'm sure there are a number of other people -- not me -- who
would like to benefit from the privilege if accorded. If it comes to a
question of all or none I'd think the choice would have to be the latter...
but maybe others can comment??

I think the usual Usenet rules are to allow a short advert in the siggy.

Rgds, George Macdonald

Though a siggy delimiter (dash-dash-space) would be ,umm nice George! ;-)

 

[George Macdonald]

On Mon, 03 Jan 2005 02:42:19 -0500, George Macdonald wrote:




I think the usual Usenet rules are to allow a short advert in the siggy.

DABS is probably the largest computer parts e-tailer in the U.K. I don't
think we need this... or expect NewEgg, Monarch, Dell<gawp>,
Overclockers.xxxx et.al. to follow. I can't find a newsgroup charter and
the guidelines are: if you can't find a FAQ, don't do it.

Though a siggy delimiter (dash-dash-space) would be ,umm nice George! ;-)

I've been thinking about dropping mine for a bit now - probably given the
poor folks searching on the keywords enough bother.

Rgds, George Macdonald

 

[Rob Stow]

DABS is probably the largest computer parts e-tailer in the U.K. I don't
think we need this... or expect NewEgg, Monarch, Dell<gawp>,
Overclockers.xxxx et.al. to follow.

It appears to be generally accepted that people
can use their sig file to /briefly/ promote businesses
that the own or work for.

I can't find a newsgroup charter and
the guidelines are: if you can't find a FAQ, don't do it.

Police states work on "Everything is forbidden
unless explicitly stated otherwise." Here in the
"free" world it is supposed to be the other way
around.


In any case, the charter can be found at
http://www.landfield.com/ftp/pub/usenet/control/comp/comp.sys.ibm.pc.hardware.chips

Once the headers are stripped out it is amazingly brief:

Description: Discussion of processor, cache & memory chips for the ibm-pc.

This newsgroup will be for the discussion of those topics directly related to
computer chips and technologies. "Chips" is a broad term to include memory
chips (such as SIMMs) as well as new system chips from Intel (such as the
Pentium) or other chip manufacturers (such as AMD). Topics may include chip
pricing & availability, clock speeds (eg.- overclocking), and future chip
releases.


Charters in general are easy to find by using Yahoo to search for
newsgroup charter "newsgroup.name.goes.here"

 

[Robert Redelmeier]

It appears to be generally accepted that people can use
their sig file to /briefly/ promote businesses that the
own or work for.

General netiquette (RFC?) is to keep USENET sig files down
to 4 lines. No limit on content.

Police states work on "Everything is forbidden unless explicitly
stated otherwise." Here in the "free" world it is supposed to
be the other way around.

I think it's more a matter of mindset -- the old joke:

In England, everything is permitted unless forbidden.
In Germany, everything is forbidden unless permitted.
In Russia, everything is forbidden even the permitted.
In France, everything is permitted even the forbidden.

-- Robert

 

[The little lost angel]

I'm not sure how other people feel about direct advertising in this
newsgroup but I'm sure there are a number of other people -- not me -- who
would like to benefit from the privilege if accorded. If it comes to a
question of all or none I'd think the choice would have to be the latter...
but maybe others can comment??

It's not really a bother right, I remembered being told that it's ok
to advertise your business/service. As long as the sig is kept short,
the content doesn't matter. Plus, I think it's good to know when the
poster has a direct interest in something when getting
advice/information from them. e.g.
"John Kenbi-aniwan
Visit Just Kool Online
Largest U.S. Distributor for AMD processors" ;PpPpP




--
L.Angel: I'm looking for web design work.
If you need basic to med complexity webpages at affordable rates, email me
Standard HTML, SHTML, MySQL + PHP or ASP, Javascript.
If you really want, FrontPage & DreamWeaver too.
But keep in mind you pay extra bandwidth for their bloated code

 

[Richard Hopkins]

and you've already been told that Intel did not invent anything.

My point is that Intel have initiated the use of this style of socket in the
PC segment, and they've certainly had to bear the brunt of the flak for
doing so.

I don't see much R&D here.

You think this socket happened by itself? What about the electrical design
(and by implication motherboard design) that's gone into it? If you don't
see much R&D here George, you need thicker glasses.

Anyone can tell at a glance that the mechanism is more fragile and prone
to damage...

It looks more fragile than Socket 478 because there's no cover plate on the
socket, and because there are more, smaller contact fingers. However,
whether it *is* more fragile than Socket 478 is something you can't tell
just by glancing.

I'm not sure how other people feel about direct advertising in this
newsgroup

Lol. It's two lines, it's not HTML, it's at the bottom of my signature, and
it's not prohibited by the newsgroup's charter. If you don't like it, don't
scroll down. Simple.
--


Richard Hopkins
Cardiff, Wales, United Kingdom
(replace .nospam with .com in reply address)

The UK's leading technology reseller www.dabs.com
Get the most out of your digital photos www.dabsxpose.com

 

[keith]

DABS is probably the largest computer parts e-tailer in the U.K. I don't
think we need this... or expect NewEgg, Monarch, Dell<gawp>,
Overclockers.xxxx et.al. to follow. I can't find a newsgroup charter and
the guidelines are: if you can't find a FAQ, don't do it.


I've been thinking about dropping mine for a bit now - probably given the
poor folks searching on the keywords enough bother.

I wasn't talking about your sig, though it is getting nigh on six years
*old*. I was referring to your lack of the signature delimiter, the lack
of which is just short of top-posting on the Usenet agravation list. ;-)

 

[George Macdonald]

It appears to be generally accepted that people
can use their sig file to /briefly/ promote businesses
that the own or work for.

It's sorta moot here but sorry I don't see that for a large retail
enterprise, who might even be paying individuals (per hit ?) who do the
planting. How would you feel about someone touting the Home Shopping
Network with ads for some computer they sell? I suggest you find out a bit
about DABS.

Police states work on "Everything is forbidden
unless explicitly stated otherwise." Here in the
"free" world it is supposed to be the other way
around.

Shove your "police state" comment back where you pulled it from - the "free
world" has alternative means to restrict your behavior... they're called
lawyers and their elite ranks are the ones who make up the rules for *all*
of us. What do you not understand about "guidelines"? Go check out some
of the "FAQs" at the newsfeed services - that is precisely what I read at a
couple, about refraining if there was no stated policy for a group.

In any case, the charter can be found at
http://www.landfield.com/ftp/pub/usenet/control/comp/comp.sys.ibm.pc.hardware.chips

Once the headers are stripped out it is amazingly brief:

Description: Discussion of processor, cache & memory chips for the ibm-pc.

This newsgroup will be for the discussion of those topics directly related to
computer chips and technologies. "Chips" is a broad term to include memory
chips (such as SIMMs) as well as new system chips from Intel (such as the
Pentium) or other chip manufacturers (such as AMD). Topics may include chip
pricing & availability, clock speeds (eg.- overclocking), and future chip
releases.


Charters in general are easy to find by using Yahoo to search for
newsgroup charter "newsgroup.name.goes.here"

And the FAQ can also be found at www.faqs.org:

"Q) 1.3 Is it ok to (sell/buy/job-offer/advertise) things here?

No, none of the above fit within the charter of the
comp.sys.ibm.pc.hardware.* hierarchy, therefore such posts are
considered unacceptable. For buying/selling things, use groups with
the words 'wanted' or 'forsale', and for job offers, use groups with
the words 'jobs'. All of these can be found in the misc.* hierarchy.
For commercial advertisements, use only the biz.* hierarchy as per the
guidelines of USENET. (refer to the news.* groups for more
information)."

Now whether the Usenet netiquette allowance for sig "adverts" overrides the
above, is up for discussion - the "How To" articles on sig.ads particularly
suggest just pointing people to a Web site, without any hype.... hardly
seems to fit the bill exactly here.

Rgds, George Macdonald

 

[George Macdonald]

in message...

My point is that Intel have initiated the use of this style of socket in the
PC segment, and they've certainly had to bear the brunt of the flak for
doing so.

No, your "point" was perfectly clear... and don't project any "flak" on to
me please.

You think this socket happened by itself? What about the electrical design
(and by implication motherboard design) that's gone into it? If you don't
see much R&D here George, you need thicker glasses.

No Richard all you have to do is remove the iBlinkers. Any necessary
"electrical design" is quite divorced from any issue of mechanical design
of contacts and retention mechanism.

It looks more fragile than Socket 478 because there's no cover plate on the
socket, and because there are more, smaller contact fingers. However,
whether it *is* more fragile than Socket 478 is something you can't tell
just by glancing.

Seems a wee bit contradictory to me. Well you can glance... and then
confirm by looking at the warnings and consequences in Intel's
documentation
(http://www.intel.com/cd/channel/res...top/proc_dsk_p4/technical_reference/99345.htm):
5 different "failure types" and some illustrative pics. Apparently they
were only able to quell the uproar at the mbrd mfrs by proving that they
had undertaken an extensive training program for system builders.

Lol. It's two lines, it's not HTML, it's at the bottom of my signature, and
it's not prohibited by the newsgroup's charter. If you don't like it, don't
scroll down. Simple.

One can only hope then that it fulfills Korpela's Law 17.: "Advertizing
works on Usenet; but the effect is opposite to the intended one."... and
the newsgroup FAQ *does* say no to advertising.

Rgds, George Macdonald

 

[George Macdonald]

I wasn't talking about your sig, though it is getting nigh on six years
*old*. I was referring to your lack of the signature delimiter, the lack
of which is just short of top-posting on the Usenet agravation list. ;-)

That bad?... tsk...tsk - doesn't seem to be well respected or even
documented but there are so many broken links in the Usenet docs now. The
sig extension is gone anyway.

 

[keith]

That bad?... tsk...tsk - doesn't seem to be well respected or even
documented but there are so many broken links in the Usenet docs now. The
sig extension is gone anyway.

Have you ever noticed that my sig, as humble as it is, dissapears when you
reply? That's the sig-delimiter (dash-dash-space) at work. See how
yours just went poof? ...t'wasn't me!

 

[George Macdonald]

Have you ever noticed that my sig, as humble as it is, dissapears when you
reply? That's the sig-delimiter (dash-dash-space) at work. See how
yours just went poof? ...t'wasn't me!

Hmmm, Free Agent doesn't care... apparently. I keep meaning to buy the
damned thing but now with that and no quote level limiting, I dunno.

 

[keith]

Hmmm, Free Agent doesn't care... apparently. I keep meaning to buy the
damned thing but now with that and no quote level limiting, I dunno.

Strange. There was supposed to be an Earth shattering kaboom!
(Free) Agent used to work. I was going to buy Agent, but decided to try
Gravity first. Though Gravity is not perfect I never went back to Agent,
free or otherwise.