Talking about grapefruits and watermelons

A CPU is a CPU is a CPU is a CPU.

https://cdn.arstechnica.net/wp-conte...5/epyc_amd.jpg

And that's a CPU.

Now imagine an after-market CPU cooler/fan(s), e.g. fins and attached
heat-wick to a pad to fit it. Most go on a presumption for a base
marginally to exceed the CPU X-Y/upper axis dimension. One, to say,
such as might be aftermarket manufacturer from Cooler Master and their
big market representation in supplementary cooling with the Hyper
series. That's one serious fan, and I'm still somewhat in shock over
its sizing from being a Cooler Master supporter ever since they came
out with the design. So long ago it was called something else I can't
offhand recall. I've even gotten into the habit of calling it one of
either -- 1) a hefty grapefruit (to palm), if not simply, however more
aptly a 2) monstrosity. Qualifiedly and simply fantastic for overkill
cooling, though, which is as much acknowledgement of a reasonable
price-structuring, among its H-series coolers, for a contemporary
budget solution.

That CPU, however -- AMD's Epyc: encompassing 32 cores for 64
threads, 128 PCIe 3.0 paths, and eight memory channels covering 16
DDR4 channels for 32 physically socketed DIMMs -- please to note, is
a slightly bigger-built CPU, almost as if an afterthought in all
that's technologically off-kilter, due, of course, to Murphy's Law and
leaps and bounds.

A toast, then, to comparably scaled coolers (in my favorite design,
Cooler Master Hypers). One which would take, fully, half the space
above what space is presently occupied on a conventional mini-MB
factor -- if not to negate altogether any prior expectation of what
coolers have or will become, based over the past decade of
advancements.

Flasherly wrote:
A CPU is a CPU is a CPU is a CPU.

https://cdn.arstechnica.net/wp-conte...5/epyc_amd.jpg

And that's a CPU.

Now imagine an after-market CPU cooler/fan(s), e.g. fins and attached
heat-wick to a pad to fit it. Most go on a presumption for a base
marginally to exceed the CPU X-Y/upper axis dimension. One, to say,
such as might be aftermarket manufacturer from Cooler Master and their
big market representation in supplementary cooling with the Hyper
series. That's one serious fan, and I'm still somewhat in shock over
its sizing from being a Cooler Master supporter ever since they came
out with the design. So long ago it was called something else I can't
offhand recall. I've even gotten into the habit of calling it one of
either -- 1) a hefty grapefruit (to palm), if not simply, however more
aptly a 2) monstrosity. Qualifiedly and simply fantastic for overkill
cooling, though, which is as much acknowledgement of a reasonable
price-structuring, among its H-series coolers, for a contemporary
budget solution.

That CPU, however -- AMD's Epyc: encompassing 32 cores for 64
threads, 128 PCIe 3.0 paths, and eight memory channels covering 16
DDR4 channels for 32 physically socketed DIMMs -- please to note, is
a slightly bigger-built CPU, almost as if an afterthought in all
that's technologically off-kilter, due, of course, to Murphy's Law and
leaps and bounds.

A toast, then, to comparably scaled coolers (in my favorite design,
Cooler Master Hypers). One which would take, fully, half the space
above what space is presently occupied on a conventional mini-MB
factor -- if not to negate altogether any prior expectation of what
coolers have or will become, based over the past decade of
advancements.

Anyone who can afford one of these setups, is going to
use something more exotic than air cooling.

I've seen no discussion yet about the reliability of that
many contacts on an LGA socket. I guess we'll find out
soon, whether the mechanical engineers did a good job or not.
It would take one hell of a lot of normal force ("down" force)
to compress 4000 springs. Notice they don't use levers for it.
(That's if you can trust any of the pictures shown so far.
At least one site "fudged" their socket picture...)

The record for socket contacts, used to be a 6000 contact
interface, but it used "fuzz buttons", which is similar
to neither of the concepts around today (LGA, PGA).

The base of that processor, is going to have to be pretty
exotic. So it won't be bent by all the various forces
applied to it.

I haven't seen any coolers for that thing announced yet,
but there is still time. I would want water cooling, not
because I like water, but because it would leave more room
inside the PC to work. I have a giant air cooler now, and
it makes it difficult to do anything inside the PC. (I had to
put "legs" on the air cooler to hold it up, so it doesn't
stress the motherboard.) And that's one area, where a water
cooler would be better. You have just a water block to hold up.

Paul

On Tue, 13 Jun 2017 02:38:19 -0400, Paul
wrote:

Anyone who can afford one of these setups, is going to
use something more exotic than air cooling.

Believe I saw a mention of Epyc being a 150-watt driven processor.
Modest or all the right goodies in die fabrication technology, at
least for a AMD of hefty residuals and 125-watt Bulldozer packaging
until of recent. Then again, at 95-watts in an upper conservative
E(nergy efficient)-series Bulldozer, stats of a 300-watt draw capacity
might tend send the uninitiated quickly scurrying to H2O territory.

I've seen no discussion yet about the reliability of that
many contacts on an LGA socket. I guess we'll find out
soon, whether the mechanical engineers did a good job or not.
It would take one hell of a lot of normal force ("down" force)
to compress 4000 springs. Notice they don't use levers for it.
(That's if you can trust any of the pictures shown so far.
At least one site "fudged" their socket picture...)

A few bent springs, I repaired (under two magnifying glasses), on a
Intel 775 LGA does tend to impress its own unique nightmarish quality,
as again opposed to a recent departure from Bulldozer tech. (Just
having checked a "normal" Ryzen for normal pins. . .) a picture of the
above-Ryzen or Epyc/Threadripper socket...
https://www.techpowerup.com/img/17-0...da29debfdc.jpg
Force equals...ready, purchased and got your inexpensive Chinese
toolkit with all oddball screw-tips (inexpensive kits, couple dozen
tips and a driver socket for taking apart a likes of tablets and
handheld phones)?

The record for socket contacts, used to be a 6000 contact
interface, but it used "fuzz buttons", which is similar
to neither of the concepts around today (LGA, PGA).

AMD supposedly uses them for breadboard testing.
http://www.custominterconnects.com/fuzzbuttons.html
Pretty wild stuff. Not as wild, nearly so, as a design factor in the
actual facilities incorporated when building AMD research and
manufacturing plants in Germany. A layer of literal stability first
layered beneath the buildings (to isolate them) was as staggering to
me to read, perhaps as a naked and painted Gaul's glimpse at the
effects of the concrete, used first in Roman architecture.


The base of that processor, is going to have to be pretty
exotic. So it won't be bent by all the various forces
applied to it.

And then some -- one highly likely candidate for custom, 4-CPU
socketed industrial, military and scientific applications.


I haven't seen any coolers for that thing announced yet,
but there is still time. I would want water cooling, not
because I like water, but because it would leave more room
inside the PC to work. I have a giant air cooler now, and
it makes it difficult to do anything inside the PC. (I had to
put "legs" on the air cooler to hold it up, so it doesn't
stress the motherboard.) And that's one area, where a water
cooler would be better. You have just a water block to hold up.

Hm...custom structured flying buttresses to Paulie's Tower of Power.
You've definitely got some Italian genotypes floating around in there.
....just kidding, I've filed down the neoprene spacers, conversely,
distributed for standard accessory/gear with MBs and cases, to attack
a similar problem from the reverse side.

I've gotten the water bug a couple times, only didn't quite bite.
Probably because air cooling was so affordable to be at least
adequate;- more so, I'm strongly questioning now, where higher temps
have shifted to the VRM power modules supplying (AMD) CPUs.

I can be sitting at a cool ambient 105F ambient temperature, drop a
couple dozen audio files for simultaneous distribution across 8 cores,
and there's not a 500millisecond's elapse before temps register a 130F
start state, a base ramp to further escalating to 140F, within a few
seconds, and a continuing but slower rise to beyond. (Dunno where,
how far or long that can go on. I don't especially enjoy watching it
and shut down whatever the cause from processes running. ...Stuff to
do and figure on another rainy day.)