Intel to segregate Mainstream and Enthusiast Platforms even more with Sandy Bridge

[SHAQ]

It looks like they missed on their tick-tock cycle. The high end should have been out late this year. Are they going to launch any 32nm quads?

 

[yh125d]

It looks like they missed on their tick-tock cycle. The high end should have been out late this year. Are they going to launch any 32nm quads?

No they are on track with their tick/tock. I forget which is tick and which is tock but

new arch old process 65nm = Conroe
old arch new process 45nm = penryn
new arch old process 45nm = nehalem
old arch new process 32nm = westmere
new arch old process 32nm = upcoming sandy bridge


And we wont see 32nm quads til SB unless intel changes their minds (unlikely this late in the game

 

[beginner99]

It looks like they missed on their tick-tock cycle. The high end should have been out late this year. Are they going to launch any 32nm quads?

There are already 32nm quads but all of them are xeons and you won't see any no-xeon variants of these. but they do run in x58 mobos.
But useless (overpriced) if you don't plan on overclocking (or need the encryption feature).

 

[Idontcare]

It looks like they missed on their tick-tock cycle. The high end should have been out late this year. Are they going to launch any 32nm quads?

tick/tock correlates process-node (65/45/32nm) iterations with microarchitecture (Core2/Nehalem/Sandy) iterations...tick tock is not intended to correlate either string of iterations with an absolute timeline.

To "miss" a tick/tock would mean Intel cancelled either a process-node or a microarchitecture. That hasn't happened.

 

[mv2devnull]

Tick-tock separates two unrelated tasks: process-node shrink and microarchitecture development. Good management strategy, particularly if you can hire a specialist team to each task.

 

[SHAQ]

But a tick-tock cycle is supposed to be two years. i7 was out in December '08. December 2010 should have been the new architecture. And this time they are releasing the low-end version first which is a complete reversal over past launches. A few months late is no big deal however.

 

[mv2devnull]

But a tick-tock cycle is supposed to be two years.

Is that an official policy published by Intel?

 

[yh125d]

Tick-tock has nothing to do with time shaq. It's all about the cadence, the timing is separate

 

[Idontcare]

But a tick-tock cycle is supposed to be two years. i7 was out in December '08. December 2010 should have been the new architecture. And this time they are releasing the low-end version first which is a complete reversal over past launches. A few months late is no big deal however.

It's really no different than what they did with Westmere. Mainstream released first, then after months went but they released the high-end LGA1366 enthusiast variant.

 

[Acanthus]

X58 will have been king from November 2008 through Q3 2011. Who says intel is shifting platforms too quickly?

I for one, think its rather the opposite. CPU upgrades are slowing down. Drastically. Of course, they're also needed less than ever before. I think my Q8200 is gonna last through to a LGA1155 chip. Now if only I could afford an arrandale laptop (that satisfies my dock requirement) Id be on schedule.

ugh, this is what happens when the gorilla muscles everyone else out (Intel did not allow Nvidia / AMD / etc. to make a competitor to X58.)

They can just sit on their hands and leave us with 2 year old crap.

 

[IntelUser2000]

But a tick-tock cycle is supposed to be two years. i7 was out in December '08. December 2010 should have been the new architecture. And this time they are releasing the low-end version first which is a complete reversal over past launches. A few months late is no big deal however.

Yes that is true, but Gulftown didn't release until mid-late March this year. It's likely client Socket 1366 successor will be close to H1 of 2011 than H2, similar to Gulftown. Bunch of sites are giving out mixed messages out now, but there's truth in there somewhere.

And I doubt they want to do 4 channel DDR3 version for client other reasons than have it as a successor to the V8 platform.

It's likely that with Nehalem they desperately wanted server lead back from AMD so the Bloomfield launched first. On Sandy Bridge though, they might want the client side first again.

 

[Soleron]

Aren't there three SB sockets? 1155, 1356 and 2011?

http://pc.watch.impress.co.jp/docs/column/kaigai/20100428_364200.html

Look at the increasing segmentation on that towards 2012 and 2013:

http://pc.watch.impress.co.jp/img/pcw/docs/364/200/html/04.jpg.html

And 6- and 8-core SB are 2H '11, not 1H. An opportunity for AMD, if the 8-core Bulldozer is 1H.

 

[DrMrLordX]

I haven't heard much in official Intel releases about LGA1356 lately. This thread almost makes it seem like LGA1356 has been canned in favor of LGA2011 for the desktop (putting the high-end enthusiast sector on the same socket as Intel's workstation/server sector). I could be very wrong about that though.

 

[aigomorla]

I haven't heard much in official Intel releases about LGA1356 lately. This thread almost makes it seem like LGA1356 has been canned in favor of LGA2011 for the desktop (putting the high-end enthusiast sector on the same socket as Intel's workstation/server sector). I could be very wrong about that though.

if u guys ask me..

and i think they messed up or got a memo wrong.

But LGA1366 is not going away. And its not gonna go away for a while.
Because enterprise sector will be on it, and its gonna stay for a while.

Once again.. our X58 chipset is based off the enterprise sector 5500.
The 5500 got a small upgrade to a 5520 chipset.
We are waiting on a revision to that chipset, then we shall see the X68.

If your going die shrinkage.. it would make more sense to go down on pin count and not go up.

Aren't there three SB sockets? 1155, 1356 and 2011?

http://pc.watch.impress.co.jp/docs/column/kaigai/20100428_364200.html

Look at the increasing segmentation on that towards 2012 and 2013:

http://pc.watch.impress.co.jp/img/pcw/docs/364/200/html/04.jpg.html

And 6- and 8-core SB are 2H '11, not 1H. An opportunity for AMD, if the 8-core Bulldozer is 1H.

Socket 2011 shows its a 2xqpi socket.

Theres one more which should be a 4xqpi socket for beckton and its 4 way controller. Im guessing that was mislabled as the "3xqpi?"

2011 is NO WHERE NEAR MAINSTREAM with 2xqpi.

2xqpi = DUAL CPU... when has dualys ever been considered mainstream consumer?

4xqpi = 4 CPU's.


Edit: after seeing these ima have to go back to my friend at intel and ask wtf is going on...
1356??? 1155??
it seems too much like a typo then a real change... and intel has been very quiet lately.

 

[yh125d]

With 4x QPI links and 4 channel IMC, they won't have any room for the actual cores!

 

[aigomorla]

With 4x QPI links and 4 channel IMC, they won't have any room for the actual cores!

lol.... which is why the beckton is not doing so great.
They need more cache is what everyone is saying, and what im hearing.

But the problem is i hard also in the 4way qpi.

The QPI link is how the cpu's talk to each other.

So a dual setup will have 2xqpi, meaning 1 for each cpu in a 2 cpu setup.
A 4 way qpi would mean a link to all 4 processors.

Anyhow as soon as i get more info about this i'll fill you guys in.

 

[IntelUser2000]

lol.... which is why the beckton is not doing so great.
They need more cache is what everyone is saying, and what im hearing.

Right. That's why if you stick them on database apps like OLTP and TPC-C, it'll perform AND scale better than Gulftown. *rollseyes*

I do have to admit I was a bit disappointed they weren't completely dominant across everything, but for the target market, it'll crush Gulftown hands down.

 

[dac7nco]

I swear to God, if LGA2011 is a server socket, I'll posts some pictures. I'll sell my dual 5580s in a TJ-07 on this forum to some red-haired cheap bastard... thats USA dollars.

Daimon

 

[DrMrLordX]

Edit: after seeing these ima have to go back to my friend at intel and ask wtf is going on...
1356??? 1155??
it seems too much like a typo then a real change... and intel has been very quiet lately.

http://en.wikipedia.org/wiki/Intel_Sandy_Bridge_(microarchitecture)

Check that out, particularly when it talks about future products. Not guaranteed to be accurate, but if you're wondering why people have been talking about LGA1356, well, there you go (Sandy Bridge EN, for example).

That being said, that's the only place I've seen LGA1356 mentioned, and Intel may have moved all their future LGA1356 procs onto LGA2011.

LGA1155 is mentioned in the OP's link and post.