Previews - Intel introduces CPUs with GPUs on 32nm Westmere: Nehalem axed

[HEXUS]

New (official) roadmap shows CPU-and-GPU chips coming sooner than you think.

Read more.

[matty-hodgson]

how fast is a 32nm i7 gonna be?!
i wish i could afford all the newest tech when it comes out!

and what use will a graphics core and cpu core on the same chip be?
HEAT?

[TAKTAK]

and what use will a graphics core and cpu core on the same chip be?

Thats kind of the answer...

[matty-hodgson]

but GFX on the CPU isn't going to be very powerful at all.. so what exactly would the gains be?
gaming on less powerful PC's?
but still, there's the problem of heat again.

[AdamAnubis]

Which is why it will only come in on 32nm. I'm guessing that's why anyway. Less power cos of the smaller die size = less heat?

I'm wondering now, whether it's worth upgrading to i7 (months off yet anyway for me) or to wait until the next wave of CPU's come in?

[magneticman]

gulftown oh baby.

[Infinite]

Which is why it will only come in on 32nm. I'm guessing that's why anyway. Less power cos of the smaller die size = less heat?

I'm wondering now, whether it's worth upgrading to i7 (months off yet anyway for me) or to wait until the next wave of CPU's come in?

If you can wait months then you might aswell wait til christmas. Well thats what im gona do

[amdavies]

Simple rule to upgrading:-
If you need it now, then buy it; if you don't, then wait.

[TAKTAK]

but GFX on the CPU isn't going to be very powerful at all.. so what exactly would the gains be?
gaming on less powerful PC's?
but still, there's the problem of heat again.

Because putting them on one chip reduces the manufacturing cost, which (eventually) will be reflected in the price, and as it is incorporating a die shrink too, it will use less power and spread the heat over a larger area, thus meaning that cooling isn't a problem.

Also, it allows the basic customers (enthusiasts really are a minority segment remember), to have a better graphics experience without a discrete card, but also has benefits for the enthusiast, in that the GPU on the die can be linked up with a discrete card in SLi/Crossfire/Whatever arrangements, and therefore give increased graphics potential, or, possibly (i don't know about feasibility) but one presumes that in furture incarnations with nVidia onboard, that the on-die GPU could be used as a physX renderer leaving the main discrete card to take the brunt of the effort and effectivly reduce the load put onto the discrete card. (Obviously it is still early days, and i have added mere speculation )

In essence it allows lower end components to have more features, (e.g. the ever increasing netbook market), without impacting on size (too much ), still a WIP but any progress is (generally) good progress for the consumer market, especially in times such as these, as i would have expected development to have slowed down not increase, but obviously i was incorrect (again )

If you want to look more into it, a good place to look is at 360's as Valhalla is expected to have a combined chip, which could reduce the console size by upto 30%, and there is alot of information floating around about the benefits of such an action (if it takes place). But CPU + GPU on one chip doesn't necessarily mean rubbish...

[kalniel]

So Clarkdale doesn't have an intergrated memory controller in the CPU? I wasn't expecting that change - I would have guessed the IGP to move to a seperate piece of silicon and stay at 45nm, but taking the IMC out of the CPU and sticking it with the IGP at 45nm is very odd. Might help differentiate lynnfield performance I guess if that is going to have a 'native' IMC albeit it all at 45nm.

Makes me wonder if Clarkdale actually doesn't have all the much in common with Nehalem after all, instead is more similar to penryn +QPI, for cost reasons.

edit: holy cow.. anandtech point out something - if you take the roadmap at face value then 45nm lynnfield will be the only 4core (8ht) performance mainstream chip for a while (ie not LGA 1366/i58 chipset) - it's being replaced with only a 2C/4HC part at 32nm.

I can't quite believe that Intel would just roll over and let AMD take the performance mainstream market, but it looks like that's what they're doing. Unless they're trying a bit of FUD...

[Tarinder]

edit: holy cow.. anandtech point out something - if you take the roadmap at face value then 45nm lynnfield will be the only 4core (8ht) performance mainstream chip for a while (ie not LGA 1366/i58 chipset) - it's being replaced with only a 2C/4HC part at 32nm.

But there's nothing stopping Intel having, say, six speed-grades with Lynnfield, with differing cache sizes as well.

In effect, I see Lynnfield take over from Core 2 Quad on a like-for-like basis. Expect to see a glut of processors, just as we do now, so nothing's really changed?

[kalniel]

But there's nothing stopping Intel having, say, six speed-grades with Lynnfield, with differing cache sizes as well.

In effect, I see Lynnfield take over from Core 2 Quad on a like-for-like basis. Expect to see a glut of processors, just as we do now, so nothing's really changed?

I can see an initial spread of speeds, but then what? If Intel are pushing forward the drop to 32nm for economic reasons (makes sense) then it seems the lynnefield line will be a relatively short run - possibly constrained in order to avoid competing with i7 - without even a penryn equivalent to follow up.

[Noxvayl]

Because putting them on one chip reduces the manufacturing cost, which (eventually) will be reflected in the price, and as it is incorporating a die shrink too, it will use less power and spread the heat over a larger area, thus meaning that cooling isn't a problem.

Also, it allows the basic customers (enthusiasts really are a minority segment remember), to have a better graphics experience without a discrete card, but also has benefits for the enthusiast, in that the GPU on the die can be linked up with a discrete card in SLi/Crossfire/Whatever arrangements, and therefore give increased graphics potential, or, possibly (i don't know about feasibility) but one presumes that in furture incarnations with nVidia onboard, that the on-die GPU could be used as a physX renderer leaving the main discrete card to take the brunt of the effort and effectivly reduce the load put onto the discrete card. (Obviously it is still early days, and i have added mere speculation )

Sorry I just couldn't help myself, had to correct you on your proposed speculation.

Intel own Havok physics outfit and AMD have joined the fold supporting Havok physics. I'd have to say that a more realistic speculation would be that the on die GPU from Intel would support Havok physics instead of the nVidia PhysX. The feud between nVidia and Intel also should lead you to believe Intel have no intention of helping nVidia in any way.

With Havok Physics being a free tool to be used by gaming developers and the likes of the Source graphics engine utilising Havok physics you have to favour Havok for the future.

Again its speculation but you'd expect hell to freeze over before Intel and nVidia have a partnership which diminishes a company Intel owns.

[TAKTAK]

Sorry I just couldn't help myself, had to correct you on your proposed speculation.

Intel own Havok physics outfit and AMD have joined the fold supporting Havok physics. I'd have to say that a more realistic speculation would be that the on die GPU from Intel would support Havok physics instead of the nVidia PhysX. The feud between nVidia and Intel also should lead you to believe Intel have no intention of helping nVidia in any way.

With Havok Physics being a free tool to be used by gaming developers and the likes of the Source graphics engine utilising Havok physics you have to favour Havok for the future.

Again its speculation but you'd expect hell to freeze over before Intel and nVidia have a partnership which diminishes a company Intel owns.

Ahh, but if (specutivly [sp?]) VIA and nVid teamed up and created and integrated solution (fat chance, with via and S3...) then physX would be on the cards (and grossly irrelevant as the thread is about Intels plans...)

But you are correct

It would have been better to use Intel + Larrabee as the example, or AMD + ATi, but either way you look at it, nVid get the cold shoulder in the end...

[UseItNow]

Looks like a new build is on it's way in Q3.....time to bring out the wallet.

[cobhc]

Unless Gulftown is at a decent price, then I think I'm gonna be better off with a current generation Nehalem, I don't need IGP and the 2 core 4 thread cpu's are a step back performance wise.