It's rumored that Intel hammered out both preliminary specs and also final
release specs while enjoying the fishing on Kaby Lake. That's the reason
for the new Kaby Lake name chosen for the new chip. It's my opinion that
we've reached the point where we're not going to increase processor or
memory speed. Somewhere back there we had
3.2 Ghz and then 3.4 Ghz chips. Intel backed off to 3.2 Ghz and never
increased CPU speed since. It's also my opinion that the 3.2 Ghz speed is
not the speed that your programs run at. That's the speed that the micro
processor runs at. The micro processor is what determines how the machine
instructions run. The machine instructions run at the speed of the memory.
Another fallicy is memory speed. In the days of core memory the memory ran
at the clock speed. You could access a memory location for read or write in
one machine cycle. Part of that single machine cycle included a refresh cycle
for the location you just accessed. When you did a write to that location you
just used the new data instead of restoring the old data. You could run memory
without tremendous losses of time because core memory was permanent, or nonvolatile.
To explore nonvolatile memory I used to run diagnostics on a CPU with core memory.
If I wanted to swap a card I shut down the system, swapped the card, powered up and
restarted the system without rebooting. The data stayed in memory and was fine
even thru a shutdown. I never did this but if I had unplugged
memory stack, plugged it back in and
powered back up I could have have restarted the program without problems.
We didn't do that with customer programs of course.
Today there's all those delays so that the entire memory can be
restored before the data is lost. The delays are known as CAS and RAS delays. The truth
is that todays memory runs at 100-150 Mhz, not in Ghz speeds. Let's say
one memory has a CAS delay of one. That's core memory. A memory with a CAS
delay of 2 runs at half the speed of a CAS 1 memory. This is because the CAS 2
memory runs every other clock cycle. A CAS 3 memory runs at 1/3 the speed, or on
every third clock cycle. The horror is that todays memory is horrible, look at
this one from today: DDR3-1866 11-11-11. You can also buy a: DDR3-1333 9-9-9.
This one runs at 1333 instead of 1866, but with all the extra delays, 11-11-11,
the only difference is the price!
It's also my opinion that multiple cpu's, or cores, on a chip are only as
useful as the number of paths you have to memory. If you have 4 cpus, cores, each needs
its own path to memory. And each needs it's own memory. Each cpu needs a path to
access instructions.... machine
instructions. And it also has to access the data the instruction requires. To
summarize each core needs a memory and a path to that memory. Once a core has to wait
while another core is accessing memory it becomes useless. The PC is said to have
multiple memories; or slots. It's said to have multiple paths, two. But if you look
closely at a block diagram you'll find only one real path.
Another thing to ponder; every time a program is interrupted to allow another program to run,
you've got to store the data from all the registers the program was using
back to memory and then load all the registers with the data from memory for
that program. That's known as a PSW swap.
Software solutions for todays issues!
This page presents the software developed by uServe.
If you wish to create free virtual buttons for your web page check out the uServe Button Site.
They give you the ability to create web buttons in any appropriate size, containing any text you wish, and any color combination.
They let you create rectangular oval or round buttons and amazingly they even give you gradient capabilities.
Imagine a button bleeding from two colors that you pick.
Their round and oval buttons even have transparant backgrounds unlike some of the other sites. And All For FREE!
And no pop ups.