tail wagging the dog - "A Multicore Tejas? . . ."

[vgrigor]

Had read the articel:
"A Multicore Tejas? . . ."
http://www.overclockers.com/tips00508/

my opinion:

The problem is in understanding what is important in speed perception:
The most demanding speed people will buy most fast system almost at any price- not even but of course not
reasonable - but just by image of selves of having fastest, or due to accustom to tend to such a peak,
or sense displacement.

So if there is possibility to buy 10% faster for 100% price add - top (same as demanding this software) people
will buy all that is possible.

So:
If there will be ability to have common based OS and with your program runs faster twice,
this will be the only choice between competing programs.
So any critical program writer must immediately rewrite program to be not looser to competitor,
due to practically declared on every www site very wide "only top interesting" pattern.
The only need to designate hardware system as COMMON, like when it is one processor, but two core (the main base point -
ALL people use single proc systems, by default.).

So the industry custom will be decisive for immediate acceptance of multi-core processors, to rewrite software like in UT 2003 case they did. Even amd-64 processor is still unusual.
once they can be "just designated" in market as common system.

_______________________
Smoothness -
in the field of noise of processors coolers -first people was excited in only speed,
than when became more intelligent and work oriented -the quiet pc became more important widely.
Same can be made with image smoothness upon common speed, as I think, - that this is full and correct parallels.
(noise vs speed)


________________


"tail wagging the dog":
So the observers is self moving for their own tail mainly. (industry, that just satisfy in that case.
The only but the very game point -that 's for industry it is very benefitable to overclock observers and that entuthisasts to this accustom. So they are "vote for" and help this.).

tail - observers accustom, made by them. For only that system of parallels. (not all for a game.)
So it can be moving reason sometimes.

[Richard]

I think Dual Core CPUs will not be quite as bad as he thinks.

WARNING: My own SWAG.

These new CPUs will handle data a little differently than two separate processors. Up until now SMP has meant the software has been the one handling how data is processed. The OS assigns a task to whichever CPU isn't processing. Basically, acting as a mediator.

My hypothesis is that dual core CPUs will handle the task of load balancing on their own. Working in parallel they'll act more as a singular processor than two. Of course, I think the real benefit will come from alternating the workload. The drawback to this is that the CPU won't be much faster, if at all, than a singular CPU but it will run cooler than previous generations. There are many possible avenues that Intel / AMD can go down with dual core.

It's going to be about working smarter, not harder.

[DaMouse]

I saw this in CustomPC (UK PC magazine directed at Overclockers, modders and general wackos :)) and they thought about the theory of Quad HT :) as in Desktop PCs with 4 logical blocks :)

-DaMouse

[vgrigor]

Can be presented many methods of using multu core:
thread smoothing like today's
- OS based
- new Instructions based (SSE-X for n-cpu's)
- risk processors sub cores way in addition to some existed risk
parts of processors x-86.
so on..

partial dimension number of cores - like today (1.5),
but more smarter...
(today's - is a only secondary add-on to trying to use more
effectively single cpu processing technique,
the results only two - sometimes games go smoother, speed
is impoved for 5-10%. first try - but surely not commonly bad.)
- good things.

I think it possible to invent even more...

The issue that this is quite a sure working effectively way,
proven much in some special cases for important tasks...
So this is - future accepted way, while it will be good to have some
importants task working much faster with not too much cost
as it it planned in that system.

Applications: smoothing, speed, ....

I think at standad price, for quiet PC by price, this and more applications are very good.

[man_utd]

Its RISC (reduced instruction set computer) not risk

[vgrigor]

The issue in that muti core can be realized by many ways,
as from hardware as from softeware as from alse technology
point of view- and some of them can became efficient not far in the future.

[cack01]

Software will be there in the future. I've have never seen Intel make a major move without having a firm base of support. Intel can get software support simply by spending money.
What worries me most about these multicore cpus is the heat everyone is projecting. Around 150W!! Of course its way to early to be quoting specs, I just can't imagine OEMs being able to cool anything thing close to that amount.

What I want to happen in the future is for SMP to go away, and instead of trying to process two threads at once, try to process one thread on two processors. I think if it becomes possible, multicores will become a huge step foward.

[vgrigor]

Today, two parts of proc, process same thread.

It is just kind of prediction of instruction flow bridjes.

In the future better way - to smarter preanalyze program to
divide it to independently working parts,
in run time (hardware) or with compile time.

The better way of course - if many processors can work as single one but fast.
but it is not simple enogh.

The article was about real two core, due to it can be done by just simple adding second same part, that is looking not too impressive without optimisation of interaction between them,
or without special recompiling of systems - but REAL.

So,
"Restricted by human brain (disabilities)" - not silicon.

[FizzledFiend]

man just think of this...4 CPU's all cooled by the same block acting as I chip man that would be something...less heat per core...only problem i would see is trying to get matched cores like we do now with ram....neat concept, but lets see the reality of it.

[vgrigor]

twice heat- half in price of cooling system.
Water- already overcooling that.

But cost of effective double speed up - 1000%.

So looks proportions for cooling.

Good effect/price of heat ratio ?

[XWRed1]

A multicore Tejas isn't going to act like one physical processor.

1) Current multi-core cpus don't act this way.
2) Performance wouldn't be as good.

[nukeboymd]

as with the added core we now have 150W+ more electricty now this will cause more headachs with mobo and psu companies as well as more heat issues wich means louder and more elabrate ways of moving heat . also the phisical size of the chips are expandind now its almose 1000 pins so it is clear that copper and transitors need a replacement fast. on a side note UV lithogrophy shouls be used if redy on tejas wich means sub .13 micron right.

[XWRed1]

Huh? Why does 1000 pins suddenly mean that transistors won't work any more? Or copper for that matter?

You are definately not going to move off of transistors just for this.

[vgrigor]

I understand "THE Idea" -
win here, who can write as more domb.

- With "But winner" stand on, what hides this from evidence.