Sign in with
Sign up | Sign in
Your question

Itels New 2 Billion Transistor CPU with 30MB of cache !

Last response: in CPUs
Share
April 2, 2008 1:21:08 PM

http://www.tomshardware.com/2008/04/02/intel_shows_off_...

Quote:
Gelsinger did show off Intel's upcoming Tukwila processor which is a monster 2-billion transistor quad-core chip with 30 megabytes of cache. Due to ship in late 2008, the Tukwila will have dual integrated memory controllers and Quickpath interconnects something which Intel says is faster than AMD's HyerTransport technology. The silicon seems to be progressing as the world's first demo of a Tukwila server running RedHat 5.1 was shown to the public.



Sorry if someone already posted about this, but does anyone know anymore about this?

Whats up with this ?

a c 127 à CPUs
April 2, 2008 1:40:58 PM

Basically according to Intel its a giant can of whoop ass.

I can't wait to see though cuz if it is it will be interesting to see how fast AMD gets something out to compete with it.
April 2, 2008 1:41:47 PM

The two billion transistor count is a bit misleading if you're looking at that as a way of measuring its processing power since most of those transistors are probably part of that massive chunk of cache in the thing.
a c 127 à CPUs
April 2, 2008 1:46:06 PM

rodney_ws said:
The two billion transistor count is a bit misleading if you're looking at that as a way of measuring its processing power since most of those transistors are probably part of that massive chunk of cache in the thing.


Well the reference to giant was due to size but if what Intel states is true it will be a can of whoop ass. I need to find the details but I read a report on their quickpath and the bandwidth available to it. It was insane.

But as I said I can't wait to see it in action just to see AMD's response.
April 2, 2008 1:48:45 PM

How does this compare to Nahelem ? How much cache will it have. I know desktops and server cpu's differer , just wondering.
April 2, 2008 2:00:23 PM

From what I remember when Intel compared QPI to Hypertransport, they were comparing it to HT2.0. Just means AMD will need to release chips that run HT3.0 at full speed, and probably need to come out with a faster version for AM3.
April 2, 2008 2:05:38 PM

Well. My transmission going out in my car will force me to wait until Helmmie comes out.. So then I can render my standard def videos faster than real time! If I had one of the current quads I probably could ?

I read a post about Nahelem earlier and someone had cpuz pics I think it showed cache. That could have been a fake though.
April 2, 2008 2:27:11 PM

rodney_ws said:
The two billion transistor count is a bit misleading if you're looking at that as a way of measuring its processing power since most of those transistors are probably part of that massive chunk of cache in the thing.


This is pretty much true.

Quote:
It operates at 2GHz and uses a large number of transistors for memory. It also contains lot of onboard memory and registers.




This is still a good step in the right direction IMO for Intel.

April 2, 2008 2:29:17 PM

I forget what exactly I read and where... but it said something to the effect of in 1950 there were 500 million transistors in the WORLD.

Now we're packing 4 times that amount into one CPU. Pretty amazing times we live in.
a b à CPUs
April 2, 2008 2:44:19 PM

It's mostly cache .. so the huge transistor count isn't a great leap forward in technology.

The design is old hat.

It's a clinker.

When software engineering (if ever) catches up and we actually have decent multithreaded code running ...

Alas ... the world doesn't seem to have enough smart people employed in the area.

Maybe we need to market programming as the profession for the world's smartest people ...

After all ... the hardware can't do much without them.


:) 
a b à CPUs
April 2, 2008 2:56:42 PM

rodney_ws said:
I forget what exactly I read and where... but it said something to the effect of in 1950 there were 500 million transistors in the WORLD.

Now we're packing 4 times that amount into one CPU. Pretty amazing times we live in.


Yeah, I think those were contact transistors which basically were a germanium rock with wires soldered to it :)  I'd like to see Intel put 2 billion of those puppies in a building, let alone a chip.
April 2, 2008 2:57:46 PM

^^

lets see you try and program for multi thread, its more of a limitation of the language than the programmer tbh... ive got some training in Multi-threaded Java programming, but i hate the language. prefer C+.
April 2, 2008 3:16:46 PM

Yeh the Itanium2 is crazy nuts. I think several hundred are used in one of NASA's supercomputer....
April 3, 2008 6:31:41 AM

Ah yes, Itanium.
They are also known as "the money pit", and Itanic.
Intel has spent about $100billion on that puppy over the last ten years.
Since it first went on sale about 6 years ago, they have sold thousands.
It is intended as an HPC server chip. Problem is, it's more efficient to go with multiple clusters, with IBM, opteron, and even xeon chips.
I'm sure that AMD will be happy to see Intel waste more $ and research on it.
a c 127 à CPUs
April 3, 2008 1:50:37 PM

endyen said:
Ah yes, Itanium.
They are also known as "the money pit", and Itanic.
Intel has spent about $100billion on that puppy over the last ten years.
Since it first went on sale about 6 years ago, they have sold thousands.
It is intended as an HPC server chip. Problem is, it's more efficient to go with multiple clusters, with IBM, opteron, and even xeon chips.
I'm sure that AMD will be happy to see Intel waste more $ and research on it.


IDK. That may change with this chip. If it performs better tand cost less than a certain amount of clusters then it would be to the benefit of the buyer.

I of course think that tera-scale would be better but hey I don't control them.
April 3, 2008 2:18:28 PM

fazers_on_stun said:
Yeah, I think those were contact transistors which basically were a germanium rock with wires soldered to it :)  I'd like to see Intel put 2 billion of those puppies in a building, let alone a chip.


Oh come on... I wasn't implying they were same size of transistor.
a b à CPUs
April 3, 2008 3:00:57 PM

The iceberg hit it so many times and it refuses to sink ... so ... ???

Intel probably use the entire project as a means to avoid paying tax.

That is possibly the entire reason for it's existence.

On that basis alone it could be a fantastic success !!

I actually have seem an Itanium ... so they do exist.

A guy was walking out of a shop with one ... in a wheelbarrow. Two mad scientists with bald heads and facial ticks were at his side ... whispering some very long sentences in his ears.

He was unable to comprehend.
April 3, 2008 3:32:37 PM

Reynod said:
The iceberg hit it so many times and it refuses to sink ... so ... ???

Intel probably use the entire project as a means to avoid paying tax.

That is possibly the entire reason for it's existence.

On that basis alone it could be a fantastic success !!

I actually have seem an Itanium ... so they do exist.

A guy was walking out of a shop with one ... in a wheelbarrow. Two mad scientists with bald heads and facial ticks were at his side ... whispering some very long sentences in his ears.

He was unable to comprehend.

Go post on Sharikou's blog!
a b à CPUs
April 3, 2008 9:06:21 PM

rodney_ws said:
Oh come on... I wasn't implying they were same size of transistor.


Neither was I :) . However since you made the comparison, at least number-wise, I just took it one step further. Fun to imagine a CPU the size of an office building, running at maybe 100 Hz, with a 10 megawatt power draw. Then AMD could regain the computing power-per-watt crown :) ...
a c 127 à CPUs
April 3, 2008 9:14:05 PM

Ok lets look at the facts. 2billion transistors on a single die. Impressive? Very. Nice QPI links and uber fast with lots o memory bandwidth.

The chip looks promising. Now we need to wait and see its performance nubers when it hits. :D 
a b à CPUs
April 3, 2008 9:35:11 PM

You have to give credit to Intel for this, although we (most of us any way) won't be using those CPUs the technology and the knowledge that Intel learned from the creation of those CPUs will filter down in one way or another to the newer consumer CPUs, (ie. Quickpath in the future Nehalm)
April 4, 2008 1:17:52 AM

rodney_ws said:
Go post on Sharikou's blog!


Have a problem with humour? Maybe just a bad day?
FYI Itanium has been the laughing stock of the server market for longer than I can remember.
It has been called Itanic since 2K or earlier.
Just consider yourself lucky. Intel had planned to shift everyone to IA64 itanic chips, up to the point where AMD brought out x86-64.
It may still happen.
The price range for Itanics is $800 to $3000.
April 4, 2008 8:21:18 AM

Quote:
FYI Itanium has been the laughing stock of the server market for longer than I can remember.


I think that the Rejecteron (or Oopseron?!) takes the cake now. The Barcelona launch was far more botched (Launched last September... servers still MIA!) than any Itanium launch.
April 4, 2008 1:10:11 PM

endyen said:
Have a problem with humour? Maybe just a bad day?
FYI Itanium has been the laughing stock of the server market for longer than I can remember.
It has been called Itanic since 2K or earlier.
Just consider yourself lucky. Intel had planned to shift everyone to IA64 itanic chips, up to the point where AMD brought out x86-64.
It may still happen.
The price range for Itanics is $800 to $3000.


Same old crap from you Intel wanted to move us to IA64 there was no talk of EPIC in the mix.

Word, Playa.
a c 127 à CPUs
April 4, 2008 1:38:36 PM

endyen said:
Have a problem with humour? Maybe just a bad day?
FYI Itanium has been the laughing stock of the server market for longer than I can remember.
It has been called Itanic since 2K or earlier.
Just consider yourself lucky. Intel had planned to shift everyone to IA64 itanic chips, up to the point where AMD brought out x86-64.
It may still happen.
The price range for Itanics is $800 to $3000.


So in a way AMD killed some innovation for all users. Thats nice to know. x86 is old and IA64 has a lot of features with a lot of promise.

But for now we shall remain stuck with x86-64. Unless AMD suddenly decides to go with a IA64 type architecture.
April 4, 2008 1:43:01 PM

endyen said:
Ah yes, Itanium.
They are also known as "the money pit", and Itanic.
Intel has spent about $100billion on that puppy over the last ten years.
Since it first went on sale about 6 years ago, they have sold thousands.
It is intended as an HPC server chip. Problem is, it's more efficient to go with multiple clusters, with IBM, opteron, and even xeon chips.
I'm sure that AMD will be happy to see Intel waste more $ and research on it.

Why in gods name are you back?
a b à CPUs
April 4, 2008 2:40:15 PM

Just how does "The Barcelona launch was far more botched (Launched last September... servers still MIA!) than any Itanium launch" fit reality Mandrake??

Barcelona has already outsold Itanium ... with the bugged ones they shifted first.

Barcelona high end servers are rapidly taking back market share at the 4 and 8 socket end ... Intel has no competition because their current FSB based products just don't scale as well.

Till nehalem.


Itanium is the laughing stock of the server industry ... an absolute joke.

They can stick 60Mb of cache on it and nobody will care.

Rodney ... I am shocked ... did you have to spell check that and get some help??
April 4, 2008 3:05:09 PM

jimmysmitty said:
So in a way AMD killed some innovation for all users. Thats nice to know. x86 is old and IA64 has a lot of features with a lot of promise.

But for now we shall remain stuck with x86-64. Unless AMD suddenly decides to go with a IA64 type architecture.


You're very wrong about that. Read up on why IA64 was dumped by the industry and x86-64 was welcomed and adopted before you make a statement like that.
a c 127 à CPUs
April 4, 2008 3:18:32 PM

wingless said:
You're very wrong about that. Read up on why IA64 was dumped by the industry and x86-64 was welcomed and adopted before you make a statement like that.


Or you could take it as a joke like it was meant to be.

I know why IA64 was not accepted and x86-64 was. IA64 had no direct backwards support for 32bit and x86-64 was basically 32bit with 64bit extensions.

Although IA64 was and still is a more efficient architecture from what I have read, it would take too much to fully drop 32bit support and switch everyone to only 64bit support.

Man people here need to get a sense of humor.
a b à CPUs
April 4, 2008 3:42:12 PM

Well, Barcelona is out now. You can grab one off Newegg if you'd like.
April 4, 2008 4:54:47 PM

EXT64 said:
Well, Barcelona is out now. You can grab one off Newegg if you'd like.

I'll pass :) 
a c 127 à CPUs
April 4, 2008 4:58:29 PM

EXT64 said:
Well, Barcelona is out now. You can grab one off Newegg if you'd like.


IDK. I kinda prefer a Double Whopper w/cheese to a quarter pounder without cheese...
April 5, 2008 6:14:48 AM

mmmmm.....cheese
April 5, 2008 7:57:31 AM

jimmysmitty said:
So in a way AMD killed some innovation for all users. Thats nice to know. x86 is old and IA64 has a lot of features with a lot of promise.

But for now we shall remain stuck with x86-64. Unless AMD suddenly decides to go with a IA64 type architecture.

In case you didn't know, Itanic is what's called a VLIW, or very large instruction word chip.
The problem with this type of architecture, is the compiling requirements. A bad compile can cost big time.
It is also not very fault tolerent. A small misque can have severe consequences. While most OSes were ported to IA64, performance costs were problematic.
It doesn't help that it's huge size/cost also makes reasonable speeds almost impossible.
April 5, 2008 8:01:57 AM

Ycon said:
Why in gods name are you back?

Back?
When you say anything that is even remotely inteligent, I will listen (and be very amazed). Until then, I will think of you as an idiot troll, and not really give a rats rectum what you say.
a b à CPUs
April 5, 2008 9:53:57 AM

rats rectum ... LOL

a b à CPUs
April 11, 2008 1:55:35 PM

From Wikipedia...



The chip will have 4 processor cores per die and 30MB of cache[u said:
[3]; it may only be able to run at full rated clock with some cores deactivated. In this way it can be configured for highest multithreaded performance or highest single thread performance, while staying within its thermal limits.
]

The chip will have 4 processor cores per die and 30MB of cache[3]; it may only be able to run at full rated clock with some cores deactivated. In this way it can be configured for highest multithreaded performance or highest single thread performance, while staying within its thermal limits.
[/u]

So wait a minute...it may only be able to run at rated speeds if one or more of the cores is deactivated? Is this Intel's version of tri-core?

Itanium is/was great in theory but never really delivered any benefits in the real world. I'm surprised they even keep the project alive.


jimmysmitty wrote:


But as I said I can't wait to see it in action just to see AMD's response. said:


But as I said I can't wait to see it in action just to see AMD's response.

Don't hold your breath. I doubt AMD will even respond to it.
a c 127 à CPUs
April 11, 2008 2:05:15 PM

chunkymonster said:
From Wikipedia...



So wait a minute...it may only be able to run at rated speeds if one or more of the cores is deactivated? Is this Intel's version of tri-core?

Itanium is/was great in theory but never really delivered any benefits in the real world. I'm surprised they even keep the project alive.


jimmysmitty wrote:

Don't hold your breath. I doubt AMD will even respond to it.


One problem here. Its Wiki. Can't fully trust it as a reliable source. From what I gather it will do the same thing that was meant to be in Nehalem. Where, for single threaded apps, it will downclock the cores not being used and OC like crazy the core that is processing the thread. This will result in two things. Faster single threaded performance and it can easily stay in the same thermal envelope as when it had all cores at the same clock.
!