Larrabee delayed to 2011 ?

[Voxilla]

To me it seems that switching from x86 only makes sense for LRB if you're 100% convinced that these cores won't at some day in the future end up running your operating system kernel. That's not where LRB is right now of course, nor will it be for a few years hence, but in the longer term they may well end up doing so. Sticking with x86 makes options for those sorts of changes a much simpler transition for the end-users, and hence simpler to get into the market.

Like the x86 ISA would be the only one capable of running an operating system.
At the contrary Larrabee having a decent instruction set would enable getting rid of x86 completely in the long term, given that Larrabee cores would be integrated side by side with legacy x86 cores on the same die.

 

[nutball]

Like the x86 ISA would be the only one capable of running an operating system.

That's not the point though. Right now Windows dominates the marketplace, and x86 dominates Windows. Microsoft have been persuaded to port Windows to other ISAs in the past, ports which have proven to be a total failure and waste of money. If I were them I've be extremely reticent to go through all that again.

I'm not arguing that any of this is technical infeasible. Apple have changed ISAs underneath their OS, but they did that because they basically had to to survive. MS don't, and Intel won't get very far trying to push them *again*.

My point is that if you look at the cost implications to Intel, MS and the end-users of trying to migrate 90% of the worlds computers away from x86 to another ISA, you have to have a pretty bloody good reason to go that route beyond a few aesthetic whimsies and the saving of 15% on your CPU transistor budget.

 

[pcchen]

Another possible reason why Larrabee uses x86 is that Intel may want to incoporate Larrabee into their CPU in the future. That is, they may have a CPU with 4 big cores and many small cores. Of course, in this case having the same ISA for both big and small cores would be convenient.

 

[Voxilla]

My point is that if you look at the cost implications to Intel, MS and the end-users of trying to migrate 90% of the worlds computers away from x86 to another ISA, you have to have a pretty bloody good reason to go that route beyond a few aesthetic whimsies and the saving of 15% on your CPU transistor budget.

As you tell, Apple has already done it twice before and end users base only has grown, so it's quite feasible and desirable if it is for the better.

On the other hand now at least we have some competition between two x86 suppliers AMD and Intel, else we might end up with a complete monopoly.

Infact there already is a better ISA available in all modern x86 processors (thanks to AMD), being x64. Having twice the amount of registers makes a very real difference, I can tell.

 

[Andrew Lauritzen]

As you tell, Apple has already done it twice before and end users base only has grown, so it's quite feasible and desirable if it is for the better.

Well but the irony here is that it grew the most when they transitioned TO x86, where the effect of "now I can dual-boot Windows on my Mac!" is not to be under-estimated.

 

[nAo]

Well but the irony here is that it grew the most when they transitioned TO x86, where the effect of "now I can dual-boot Windows on my Mac!" is not to be under-estimated.

This is actually how I became a mac user. I bought 3 years ago my first mac(book) because I wanted to learn how to use a new OS, thinking that I had windows as fall back option anyway.
It turns out I never installed windows on it

 

[Voxilla]

Well but the irony here is that it grew the most when they transitioned TO x86, where the effect of "now I can dual-boot Windows on my Mac!" is not to be under-estimated.

Obviously there were other reasons for Apple to make that move (or call it sacrifice), such as PowerPC not doing too well in the mobile segment and it also enabled Apple to benefit from other widely available PC components such as chipsets and GPUs.

I wouldn't be too surprised if they switched again for some of their upcoming gear to the ARM processor, in fact they already have sort of with the iPhone/Pod.

 

[Voxilla]

This is actually how I became a mac user. I bought 3 years ago my first mac(book) because I wanted to learn how to use a new OS, thinking that I had windows as fall back option anyway.
It turns out I never installed windows on it

Yeah, I also bought one, be it a mac mini, only because it was relatively cheap and cute, never thought of putting windows on it though.

 

[nAo]

in fact they already have sort of with the iPhone/Pod.

Well, it's not like they had any option (no x86 CPU for that market segment)

 

[Panajev2001a]

This is actually how I became a mac user. I bought 3 years ago my first mac(book) because I wanted to learn how to use a new OS, thinking that I had windows as fall back option anyway.
It turns out I never installed windows on it

You should really try Windows 7 out, the new taskbar and polished-up UI is just THAT good... did I mention that the new taskbar now finally usable when placed vertically (and window management) is really good ?

 

[nAo]

You should really try Windows 7 out, the new taskbar and polished-up UI is just THAT good... did I mention that the new taskbar now finally usable when placed vertically (and window management) is really good ?

I am actually a Win7 user as well

 

[Megadrive1988]

Another possible reason why Larrabee uses x86 is that Intel may want to incoporate Larrabee into their CPU in the future. That is, they may have a CPU with 4 big cores and many small cores. Of course, in this case having the same ISA for both big and small cores would be convenient.

Agreed.

That to me sounds a lot like Intel's "Platform 2015" a whitepaper from 2005, before Larrabee was known outside of Intel.

 

[Groo The Wanderer]

Charlie of Semiaccurate fame claims he spoke with an Atom engineer who stated that core was 15-20% larger because it was x86.

No, Charlie of SemiAccurate fame was told by Pat Gelsinger directly that the *POWER* overhead of x86 on the IALP cores was between 0 and 50%. If you are going to quote me, at least be close. KTHXBYE.

-Charlie

 

[Karoshi]

No, Charlie of SemiAccurate fame was told by Pat Gelsinger directly that the *POWER* overhead of x86 on the IALP cores was between 0 and 50%. If you are going to quote me, at least be close. KTHXBYE.

-Charlie

I suppose that's 0% overhead at deep sleep and 50% overhead when doing something.

SCNR

 

[3dilettante]

No, Charlie of SemiAccurate fame was told by Pat Gelsinger directly that the *POWER* overhead of x86 on the IALP cores was between 0 and 50%. If you are going to quote me, at least be close. KTHXBYE.

My apologies for my inaccuracy.
I thought I recalled you talking about this on either the RWT or aceshardware forums, and I didn't recall you dropping Pat Gelsinger's name.

 

[Groo The Wanderer]

NP, probably because I didn't want to say it because of potential annoyances it might cause him. Not a problem now.

To be more specific, it was more a talk about the ARM ISA vs x86 as implemented in Atom, IE you put the same level of work into an intel ARM chip.

-Charlie

 

[dkanter]

NP, probably because I didn't want to say it because of potential annoyances it might cause him. Not a problem now.

To be more specific, it was more a talk about the ARM ISA vs x86 as implemented in Atom, IE you put the same level of work into an intel ARM chip.

-Charlie

That really depends on the microarchitecture of the design.

David

 

[Cookie Monster]

link
Did anyone post this regarding GPGPU "viruses".

Pretty much confirms LRB isnt ready in a long while. Funny how they are downplaying GPGPU because of it.

 

[Neb]

From link above:

(in fact, certain combinations of shaders can damage graphics chips that are not properly cooled down),

Is that a punch at ATI reference power circuit solution for 48xx series?

 

[compres]

Ohh man and I was hopping to have one of these soon.