AMD and Intel General Discussion (not for getting help)

[Reynod]

18 months ago they showcased an 80 core CPU.

I imagine they have 4 or more variations (not the 80 core though) they can go into production with, depending on the global economy, and AMD's alternatives.

The economy isn't strong enough to warrant ticktocking at the speed they have been lately.

Plus the perfomance increase from the 45nm core2 line to i7 hasn't exactly been stellar ... only in a narrow range is there a real performance boost.

An overclocked 45nm core2 quad still kicks an i7 on most gaming ... and it runs cooler and draws less power.

Tech sites have overlooked that fact ...

It is primarily a server cpu though ... no criticism intended.

Overclocking a core2 quad is going to give you better bang for the buck on a home PC ... unless your into video production work etc ... where the OMC really helps.

 

[Upendra09]

u do know that was a total false piece of info right?

 

[Reynod]

What ... the 80 core thing they showed or my outrageous post designed to incite a riot?

http://techresearch.intel.com/articles/Tera-Scale/1449.htm

 

[Upendra09]

no my post about the i11

 

[Reynod]

Yeah well I think that one is a given.

We are pretty sure the i8 got eaten tho ...

 

[jj463rd]

i13 will be bad luck ouch.

 

[randomizer]

I will get i13 purely because 13 is my favourite number.

 

[jj463rd]

I'm always lucky with the 666 number myself.

 

[Reynod]

http://sch.ci.lexington.ma.us/~jfrost/HWR/10digitnumber.pdf

Get working on it geeks and report back ...

 

[archibael]

There's only 576 possible numbers. How hard could it be to beat it by brute force?

 

[Reynod]

Digits Kernel
2 None
3 495
4 6174
5 None
6 549945, 631764
7 None
8 63317664, 97508421
9 554999445, 864197532
10 6333176664, 9753086421, 9975084201

 

[archibael]

9876541230
9276541830
9632581470
9216543870
7896543210
9816543270
9876543210
9276543810
9876543210
9216547830
3816547290
9816547230
9216547830
9632587410
7412589630


Did I do something wrong here?

 

[Upendra09]

just multiply all the numbers together from 1-10

 

[Upendra09]

nvr mind i didin't read the whoole thing

 

[Reynod]

http://www.fudzilla.com/content/view/15706/1/

Good news ... the pushpins on the Intel HSF's are going with i9.

my thumbs and index fingers are quivering with excitement !!!

Can anyone find out who the Wally was who designed those push pins so we can sue them for nerve damage already done?

Any system assembler will be very happy I imagine.

 

[amdfangirl]

Sweet! I hate pushpins!

 

[Reynod]

I have 4 of those damned things now ... yeah call me an Intelfanboi eh?

The A64 and Phenom are sweet in comparison.

I am sure Intel never realised how many times the HSF gets taken off and put back on again.

I already have one dead HSF due to stripped pins ...

They are also harder to fit to better quality mobos ... which are thicker .. hence the pins need to go a mm or so further.

 

[1st duke of marlborough]

Instead of six or eight core processors, I would rather see programmers actually utilize four cores. Seems to me that programs right now hardly have the multi-threading necessary to make any use of more than one or two cores, thus four really isn't needed. Thus, six cores are just overkill, since we can't even make full use of four yet. Hopefully the prevalence of multi-threaded programs will drastically increase as the next year or two progress.

 

[JAYDEEJOHN]

And thats the problem, as even if they get their apps down to the best MT that can be done, theres still too many areas that are single threaded in each app, no matter what they do.
So, the reasoning is, to go back to what works, speed and IPC.
Aince its somewhat deadended with cpus, thats why were seeing LRB and the G300, and theres tons of potential "fastness" there, and the cpu will only be directing alot of things, and not running everything, as itll be too slow.
So, LRB, fusion and CUDA/G300 type units are hopefully the future, and anyone thats buying hex and octo cores for the "future" is waaaaaay behind, because you dont need massive cores, just alot of fast ones

 

[randomizer]

If any of you have tried to make a simple, multithreaded application that works 100% you'll see that it's not a simple task. One thread is dead easy, more threads can be relatively easy or extremely difficult. If you simply offload some ad-hoc work to another thread that works independently of the rest of the program it's not a big task, but to split the ongoing workload over multiple threads which need to access the same memory locations and depend on the outputs of each other means controlling thread execution precisely or errors will occur. Magnify this to the scale of a 5GB game and you'll see that it's not a menial task.

 

[Upendra09]

sounds hard

 

[Dekasav]

Multithread programming is incredibly difficult, and a vast number of programmers didn't start learning how to use it until a few years ago.

Patience.

 

[Upendra09]

is a virtue

 

[randomizer]

But hard is not impossible. It's a bit of a jump from a procedural to an object-oriented programming model as well, but it's been done, so the "it's too hard" excuse can only be used for so long.

 

[amdfangirl]

< Doesn't know scrap about programming

Yeh, random can do it.