Intel Is Not Cutting Off The Prescott

[Guest]

Archived from groups: alt.comp.hardware.overclocking (More info?)

The day in the Life of the Socket 478 mPGA
Is almost over, Intel is going to make a few more, but they couldn't tell
me when.
They could only tell me that they will be stopping making the Socket 478.
They also acted surprised when I asked.
" I heard that you were soon going to cut off making the Prescott? "

Intel.. "No, were not. We will be stopping making the mPGA 478 Pin though
But for the time, we are still making them. "

"How about the 775 LGA?"

Intel "Why would we change that? It's only been out for a few months."

I don't remember how I asked the question about it's core, I do remember
that Intel
did say that the Prescott core will be around for a while.

Think about it for yourself, the 478 pin has been out for a while.
First the 400 Bus then 533/800
That is a pretty long life don't you think?

Denny. ;-) 🙂

 

[Guest]

Archived from groups: alt.comp.hardware.overclocking (More info?)

"Dennis E Strausser Jr" <dstrausser33@comcast.net> wrote in message
news:41e6249c$1_3@alt.athenanews.com...
> The day in the Life of the Socket 478 mPGA
> Is almost over, Intel is going to make a few more, but they couldn't tell
> me when.
> They could only tell me that they will be stopping making the Socket 478.
> They also acted surprised when I asked.
> " I heard that you were soon going to cut off making the Prescott? "
>
> Intel.. "No, were not. We will be stopping making the mPGA 478 Pin though
> But for the time, we are still making them. "

Of course they're going to say that. They don't have anything else ready and
the last thing they want is confirmation to sway everyone over to AMD.

 

[Guest]

Archived from groups: alt.comp.hardware.overclocking (More info?)

> Think about it for yourself, the 478 pin has been out for a while.
> First the 400 Bus then 533/800
> That is a pretty long life don't you think?

Slot 1 didn't stick around.

Socket 775 isn't mechanically sound and I would't trust it. There is no good
reason to pull the pins off the CPU and replace it with hundreds of tiny,
moving parts on the mainboard except to save Intel a few cents and seriously
effect reliability.

I predict we'll see socket 775 abandoned in short time.

Personally, they could take socket XXX and leave the pins on the CPU and put
pads on top of the chip for power and use the clamping mechanism to connect
power to the CPU. This allows for large pads for the high amps that could be
used by the CPU and reduces the number of pins on the underside of the CPU.

 

[Guest]

Archived from groups: alt.comp.hardware.overclocking (More info?)

"Noozer" wrote in message...
> Socket 775 isn't mechanically sound and I would't trust it.

It's probably not as mechanically robust as Socket 478, but it seems to work
fine in practice. It's certainly better than the obvious advantage. When you
say that you wouldn't "trust" it, do you mean you've actually assembled a
few boards using it?

> There is no good reason to pull the pins off the CPU and replace
> it with hundreds of tiny, moving parts on the mainboard

That comment is based on a fundamental misunderstanding. You're incorrect
in thinking that LGA775 has more "moving parts" than 478, the reverse is
true. With ZIF, every pin on the CPU and every socket side connector is
effectively a moving part, as the CPU and top plate slides to enable the
contacts when the latch is closed. With LGA775, the CPU drops into place in
the socket and is held from above by the clamp, it doesn't slide within the
socket during closure.

This simplifies the socket design, the elimination of the sliding contacts
allowing each trace/connector/finger (etc.) to be *bigger* than it otherwise
would have been. The only reason it looks more complicated is because there
are more pins, and because it doesn't have the cover plate on it.

You also have to consider how thin and spindly the pins on the CPU and the
ZIF socket components would be if you tried to maintain the "old" design
within the 775 pin/35 square millimetre package. It's a given that Intel and
the socket manufacturers will have tried this solution and found it lacking,
otherwise they wouldn't have gone to the trouble of bringing something new
to market.

> except to save Intel a few cents and seriously effect reliability.

The "reliability" argument just doesn't hold water. The biggest distributor
of motherboards using LGA775 sockets is Intel themselves, so they're hardly
going to design something that'll come round and bite them on the bum when
the motherboard warranty claims start to come in. I've put together a few
systems using this socket now, and have yet to see any evidence that it's
fundamentally fragile or deficient in any way. As for saving Intel a few
cents,

> I predict we'll see socket 775 abandoned in short time.

That's two questions. Will the LGA775 pin design hang around for extended
periods of time? Probably not, as Intel's move to dual cores will likely
require an even larger number of traces sooner rather than later. However,
if you're expecting a move back to pinned CPU's, that's not even remotely
likely. The LGA design is fundamentally better suited to high pin density
solutions. That's why Intel went with it in the first place, why it'll make
even more sense in future, and why AMD will almost certainly go the same way
at the next convenient juncture.

> Personally, they could take socket XXX and leave the pins on the
> CPU

....no they can't. Once the connector density increases beyond a certain
point, the ZIF socket design simply gets too small to implement. Contrary to
your belief, the ZIF design is *more* fiddly and contains more small parts
than the LGA alternative. As connector densities are only going one way, ZIF
becomes less and less practical as time goes on.

> and put pads on top of the chip for power and use the clamping
> mechanism to connect power to the CPU. This allows for large pads
> for the high amps that could be used by the CPU

That's a fantasy solution that simply doesn't make sense when you try and
implement it. It appears you're proposing some sort of busbar/mains type
power distribution system integrated into the CPU package. How would you
segregate/subdivide this "topside" power distribution and inject the power
into the right spots on the CPU, bearing in mind that the top of the die
must be left free for heatspreader/heatsink, and how would you do it all in
a context that could be mass-produced cost effectively?

Your suggestion implies a fiendishly complex "sandwich" CPU packaging system
that would cost an absolute fortune to produce, and that would require a
complicated (hence expensive) socket on the motherboard. It might be okay
for small volume mainframes and supercomputers, but as a high volume
solution it just doesn't work. You were complaining earlier about LGA775
having too many "moving parts". In what way would your system, with traces
on upper and lower socket planes, be better?

> and reduces the number of pins on the underside of the CPU.

....So what? You're creating a solution to a problem that doesn't exists, and
in the process increasing complexity and production cost. Not going to
happen.

Intel's electrical design for LGA775 is far neater than your proposal. It
injects power in lots of different places using a comparatively large number
of traces (thus reducing the current going through each trace) while keeping
everything on one side of the die. This is by far the best/easiest/most
efficient way of doing things.

Mind you, what this has to do with the idiot's fairy story about "talking"
to Intel about the future of Prescott I'm not sure! ;-) As you say in your
post above, Intel will continue making this core until they have a viable
alternative to replace it. The point that Dennis has misquoted (again) was
that Intel have stopped *developing* Prescott, and in this respect, nothing
has changed. They stopped developing it last year...

 

[Guest]

Archived from groups: alt.comp.hardware.overclocking (More info?)

"Noozer" wrote in message...
> Socket 775 isn't mechanically sound and I would't trust it.

It's probably not as mechanically robust as Socket 478, but it seems to work
fine in practice. It's certainly better than the obvious advantage. When you
say that you wouldn't "trust" it, do you mean you've actually assembled a
few boards using it?

> There is no good reason to pull the pins off the CPU and replace
> it with hundreds of tiny, moving parts on the mainboard

That comment is based on a fundamental misunderstanding. You're incorrect
in thinking that LGA775 has more "moving parts" than 478, the reverse is
true. With ZIF, every pin on the CPU and every socket side connector is
effectively a moving part, as the CPU and top plate slides to enable the
contacts when the latch is closed. With LGA775, the CPU drops into place in
the socket and is held from the sides by the socket moulding and from above
by the clamp, it doesn't slide within the socket during closure.

This simplifies the socket design, the elimination of the sliding contacts
allowing each trace/connector/finger (etc.) to be *bigger* than it otherwise
would have been. The only reason it looks more complicated is because there
are more "pins", and because it doesn't have the cover plate on it.

You also have to consider how thin and spindly the pins on the CPU and the
ZIF socket components would be if you tried to maintain the "old" design
within the 775 pin/35 square millimetre package. It's eminently possible
that either pin breakage, or pin bending during socket closure, or both,
would have become significant issues with this design.

It's a given that Intel and the socket manufacturers will have tried this
solution and found it lacking, otherwise they wouldn't have gone to the
trouble of bringing something new to market.

> except to save Intel a few cents and seriously effect reliability.

The "reliability" argument just doesn't hold water. The biggest distributor
of motherboards using LGA775 sockets is Intel themselves, so they're hardly
going to design something that'll come round and bite them on the bum when
the motherboard warranty claims start to come in. I've put together a few
systems using this socket now, and have yet to see any evidence that it's
fundamentally fragile or deficient in any way. As for saving Intel a few
cents, in what way is that fundamentally a bad thing?

> I predict we'll see socket 775 abandoned in short time.

That's two questions. Will the LGA775 design hang around for extended
periods of time? Probably not, as Intel's move to dual cores will likely
require an even larger number of traces sooner rather than later. However,
if you're expecting a move back to pinned CPU's, that's not even remotely
likely. The LGA design is fundamentally better suited to high pin density
solutions. That's why Intel went with it in the first place, why it'll make
even more sense in future, and why AMD will almost certainly go the same way
at the next convenient juncture.

> Personally, they could take socket XXX and leave the pins on the
> CPU

....no they can't. Once the connector density increases beyond a certain
point, the ZIF socket design simply gets too small to implement. Contrary to
your belief, the ZIF design is *more* fiddly and contains more small parts
than the LGA alternative. As connector densities are only going one way, ZIF
becomes less and less practical as time goes on.

> and put pads on top of the chip for power and use the clamping
> mechanism to connect power to the CPU. This allows for large pads
> for the high amps that could be used by the CPU

That's a fantasy solution that simply doesn't make sense when you try and
implement it. It appears you're proposing some sort of busbar/mains type
power distribution system integrated into the CPU package. How would you
segregate/subdivide this "topside" power distribution and inject the power
into the right spots on the CPU, bearing in mind that the top of the die
must be left free for heatspreader/heatsink, and how would you do it all in
a context that could be mass produced cost-effectively?

Your suggestion implies a fiendishly complex "sandwich" CPU packaging system
that would be very expensive to produce, allied with a complicated (hence
expensive) socket on the motherboard, presumably with ribbon cables or some
similar technology to get the power into a clamp/top cover on the socket,
but still using the ZIF base.

It doesn't make sense, and as a high volume, economic solution it just
doesn't work. You were complaining earlier about LGA775 having too many
"moving parts". In what way would your system, with traces
on upper and lower socket planes, and power connections into the top, be any
better?

> and reduces the number of pins on the underside of the CPU.

....So what? You're creating a solution to a problem that doesn't exist, and
in the process greatly increasing complexity and production cost. Not going
to happen.

Intel's electrical design for LGA775 is far neater than your proposal. It
injects power in lots of different places using a comparatively large number
of traces (thus reducing the current going through each trace), improving
power delivery and reducing electrical noise while keeping everything on one
side of the die. This is by far the best/easiest/most efficient way of doing
things.

Mind you, what this has to do with the Dennis the idiot's fairy story about
"talking" to Intel about the future of Prescott I'm not sure! ;-)

The point that Dennis has misquoted (again) was that Intel have stopped
*developing* Prescott, and in this respect, nothing has changed. They
stopped developing it last year, and, as you say in your post, they'll stop
making it as soon as they have a viable replacement.
--


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

The UK's leading technology reseller www.dabs.com

 

[Guest]

Archived from groups: alt.comp.hardware.overclocking (More info?)

Richard Hopkins wrote:
> "Noozer" wrote in message...
[...]
>> and put pads on top of the chip for power and use the clamping
>> mechanism to connect power to the CPU. This allows for large pads
>> for the high amps that could be used by the CPU
>
> That's a fantasy solution that simply doesn't make sense when you try
> and implement it. It appears you're proposing some sort of
> busbar/mains type power distribution system integrated into the CPU
> package. How would you segregate/subdivide this "topside" power
> distribution and inject the power into the right spots on the CPU,
> bearing in mind that the top of the die must be left free for
> heatspreader/heatsink, and how would you do it all in a context that
> could be mass-produced cost effectively?

While the current CPU paradigm of having heatspreaders effectively prevents
the use of a top-based power supply without massively enlarging the base, I
don't see any reason why a K7-style chip could not have a top-based power
supply. The K7 is already sitting on a 6 layer PCB IIRC, but I obviously
have no idea how the routing requirements would change if you drew power
through the top as opposed to the bottom. The biggest problem IMO would be
the socket design to get power from the motherboard to the bar that sits on
the top of the CPU. It obviously has to be removable, which makes having a
good connection (due to the current) between the two somewhat more complex.

I don't think the idea can be dismissed too easily, given that neither LGA
nor ZIF sockets are particularily easy to manufacture either. That said, I
agree that ZIF (or rather PGA sockets in general) are on the way out, and
LGA appears to be the easiest way forward.

[...]

--
Michael Brown
www.emboss.co.nz : OOS/RSI software and more
Add michael@ to emboss.co.nz ---+--- My inbox is always open

 

[Dave]

Archived from groups: alt.comp.hardware.overclocking (More info?)

"Richard Hopkins" <richh@dsl.nospam.com> wrote in message
news:41e92b50$0$19157$cc9e4d1f@news-text.dial.pipex.com...
> "Dave" wrote in message...

Snip

> > AMDs socket 939 points to one way of using more pins on the bottom
>> a CPU PCB in ZIF format by filling the whole bottom area with pins,
>> unlike sockets 7/370/462/478/754 and also 775 with the little "void"
>> areas in the middle of the socket.
>
> ...But filling the entire underside of the package with pins/traces/lands
> etc. causes its own problems, in particular more complex routing of the
> traces within the package PCB and a consequent increase in the number of
> layers required. As with all such things, there are a number of issues to
> be juggled, and the result is invariably a compromise. Industrial design
> always is.
>
>> Weather PGA type sockets are on the way out is hard to predict
>> but given that old technology is refined to the limit before being
>> discarded I have my doubts that will be soon.
>
> Intel's move to the LGA design tends to indicate that, as far as they are
> concerned, PGA already has reached its practical limits. If you take the
> top cover off a Socket 478 assembly and compare it with a Socket 775, you
> can see where LGA has advantages, given the context of where the socket
> goes, and what it does.
>
>> To put you all in perspective:-
>
> The examples you're quoting don't mean what you seem to think they do.

Yes they do:- Just because it is old does not mean it is useless and should
be discarded because somebody has invented something and says "This is the
next greatest thing, everybody should use it".

>> Edge connectors have been around for a lot longer than PGA and
>> are still in use.
>
> Your point being? The reason edge connectors are still commonly used is
> because they work well in those *specific* applications. Candles are still
> very useful in certain circumstances, but that hasn't stopped them being
> superceded by electric lighting in the vast majority of situations.

Point being? See above. I still use candles ocassionally.

>> Look at you RAM DIMMs and/or ISA/PCI/AGP cards.
>
> What is ISA? ;-)
>
> As for why DIMMs and add-in cards use SEC, the reason is nothing more than
> that this method is best suited to these particular components, given the
> necessary pin counts, physical component layout and modular nature of
> these designs.
>
> The reasons Intel went to SECC for Slot1 and Slot 2 were obvious at the
> time, as were the reasons why they went back to a smaller, socketed
> package as soon as they possibly could.
>
> What works best for a memory stick or a graphics card is not, by
> inference, necessarily the best solution for a CPU. Modern CPUs are
> monolithic silicon dies produced on very small processes, with relatively
> high power requirements, relatively high numbers of logic traces, a
> requirement for short trace lengths, a requirement to be non-permanently
> connected to the host circuit, with constrained structural requirements
> and with a fixed set of commercial design ideals. These are considerably
> different requirements to memory or add-in cards.
>
>> Look at your motherboards and check out the ICs. How many of
>> them, other than NB/SB, are GA chips? Close to zero.
>
> Again this is a largely irrelevant proposition. The reason why there are
> normally less than a handful of grid array ICs on a typical desktop
> motherboard is because there are normally only a handful of IC's with
> sufficient numbers of traces/level of integration to require this sort of
> connection. It's no coincidence that the two BGA chips found on most
> motherboards (the northbridge and southbridge) are the ones that do the
> overwhelming majority of the work.

Not irrelevant, at all. Most of the support computer component ICs have
mainly shrunk and not vastly increased in complexity, with obvious
exceptions.

> If you wanted to build a system within a much smaller form factor than
> current motherboard standards while offering equivalent levels of
> performance, it's a given that you would see increased levels of
> integration and component/connector complexity. Just look at a typical
> modern mobile phone/PDA/notebook PC.

Which have been made more practical mainly due component shrinkage. SMD
devices continue to shrink. Computers continue to hundreds/thousand of SMD
resistors and caps. Sure, some items continue to be more complex internally
(CPUs, bridges, etc) but that has not really reduced the overall component
count. Typical computer boards are getting more crowded as extra features
are added but component shrinkage has been mainly responsible for this, not
any HUGE scales of intergration.

>
>> and most of the rest have pins around all 4 sides (its late, can't
>> remember what they are called).
>
> LQFP in most cases.
>
>> ZIF sockets have been around for quite a while and will continue to
>> be used until other technology is proven better.
>
> Think that point is pretty much here. It is unlikely that AMD will be able
> to squeeze their PGA density any higher than they have done with Socket
> 939/940, so whatever they do next will be significant.

It could be as simple as a socket size increase. Add a few extra
columns/rows and there's your extra pin count. Maybe they won't even bother
with DDRII system RAM modules and will wait for DDRIII and DDRII will go the
way of RD-RAM. If their road-maps are to be believed then their current
aims seem to be process shrinkage and CPU speed increases and some other
specialty chips. Crystal ball gazing, anyone? 🙂

>> LGA is newer and will have to prove its worth over the long term.
>
> As a design concept, land grid array isn't "new", it's been around for
> some time. The only thing that's new about LGA775 is the introduction of
> this concept to desktop CPU packaging.
>
>> Intel has introduced some "duds" in its time so just wait and see.
>
> It appears that LGA775 as a physical package has been well received by the
> OEM assemblers (whose opinions are the only ones which matter), so the die
> is already largely cast.

LGA is still a grid array and is not a huge leap forward over PGA. While
its internals may be simpler, it still has to be attached to the motherboard
and this would not vary greatly compared to PGA - ie stuck on and soldered
from the back.

Hopefully, since LGA has been around for a while elsewhere, that they have
the bugs ironed out for the desktop scene so Joe Average does not destroy
the socket/CPU during installation. All I care about is that it should be
robust and reliable and be tool free for mounting the CPU.

Dave.