Intel CC: Ivy Bridge began volume 22nm production in Q3

[Ika]

"which means that Intel will be begin shipping final products to its customers in the second half of the quarter"
"According to Otellini, first Ivy Bridge systems should become available in Spring 2012"

What's the difference between these two statements? Does "customers" mean OEMs and distributors? In other words, a consumer wouldn't be able to buy a retail Ivy Bridge CPU until Spring 2012, which is basically when it's been slated for for a while, right?

 

[zlejedi]

Yep. They probably want to give OEMs time to sell Sandy Bridge stuff during Christmas.

 

[Rvenger]

Are we going to be able to use our existing 1.5v memory though? It seems that when the processes get smaller, the memory controller gets more and more sensitive.

 

[Ika]

Are we going to be able to use our existing 1.5v memory though? It seems that when the processes get smaller, the memory controller gets more and more sensitive.

Sandy Bridge was supposed to be incompatible with 1.65v memory, and yet everyone uses it without issues. 1.5v will probably be okay on IB, especially considering its ubiquitousness and the relative unavailability of lower-voltage RAM.

 

[Idontcare]

"which means that Intel will be begin shipping final products to its customers in the second half of the quarter"
"According to Otellini, first Ivy Bridge systems should become available in Spring 2012"

What's the difference between these two statements? Does "customers" mean OEMs and distributors? In other words, a consumer wouldn't be able to buy a retail Ivy Bridge CPU until Spring 2012, which is basically when it's been slated for for a while, right?

Intel's customers are the resellers from whom we all buy our chips. We don't buy them from Intel.

Intel's customers need time to verify platform readiness before they release the platform/system to the end-user (you and me).

 

[Rvenger]

Are we going to be able to use our existing 1.5v memory though? It seems that when the processes get smaller, the memory controller gets more and more sensitive.

My problem is I actually have 1.65v memory but undervolted to 1.5v. I am afraid my board will POST with the new chip at 1.65v which may damage it. So I might have to consider new memory but well see. If I was smart, I'd just run this setup for a couple years.

 

[Idontcare]

My problem is I actually have 1.65v memory but undervolted to 1.5v. I am afraid my board will POST with the new chip at 1.65v which may damage it. So I might have to consider new memory but well see. If I was smart, I'd just run this setup for a couple years.

That's OK, I have 1.5V rated DDR3 which my mobo (MIVE-Z) insisted on defaulting to 1.65V until I went into the BIOS and manually set them to 1.5V.

No matter how hard you try, someone in the loop is going to screw you.

 

[ElFenix]

so will IB be out before SB-E?

 

[Idontcare]

so will IB be out before SB-E?

You know, its unfortunate but Intel is making a total spaghetti mess of their processor lines and release timelines.

It used to be just Itanium was on some weird uber-delayed timelag relative to leading edge node launches.

Now we've got Itanium way out in left field, and the even the xeon server/workstation line (including the enthusiast extreme chips) is coming up on being nearly a full year out of sync with the leading edge release of the base family architecture.

And then you've got Atom, the very chip that needs to be expedited to newer nodes faster than any other chip...and 22nm is nearly here but we've yet to see 32nm atom reviews.

 

[Dadofamunky]

You know, its unfortunate but Intel is making a total spaghetti mess of their processor lines and release timelines.

It used to be just Itanium was on some weird uber-delayed timelag relative to leading edge node launches.

Now we've got Itanium way out in left field, and the even the xeon server/workstation line (including the enthusiast extreme chips) is coming up on being nearly a full year out of sync with the leading edge release of the base family architecture.

And then you've got Atom, the very chip that needs to be expedited to newer nodes faster than any other chip...and 22nm is nearly here but we've yet to see 32nm atom reviews.

That's a good point, since Intel's mobile strategy is conspicuously... absent.

One thing Intel hasn't made mistakes with lately is their fabbing processes. As someone else alluded to, everyone else is fumbling around in the weeds while Intel is slamming out 22nm product. Of course it remains to be seen what the final results will be, but I'm sure looking forward to that drop-in 22nm upgrade.

 

[Dadofamunky]

My problem is I actually have 1.65v memory but undervolted to 1.5v. I am afraid my board will POST with the new chip at 1.65v which may damage it. So I might have to consider new memory but well see. If I was smart, I'd just run this setup for a couple years.

Shoot, keep your box and overclock it. I don't see how Ivy will give you a huge marginal increase in performance. Of course there would be some. But the 2600K is a Mack Daddy.

 

[IntelUser2000]

And then you've got Atom, the very chip that needs to be expedited to newer nodes faster than any other chip...and 22nm is nearly here but we've yet to see 32nm atom reviews.

Well, Intel has said Atom's lag will be addressed. Now for the Xeon chip, you are right.

For the Xeons, it may be like the following:

Intel arguably had a gap between the regular Xeon DP chips and the Xeon MP "RISC replacement chips". If you wanted the benefits of the DP platform(lower platform costs, lower watts, memory controller optimized for more compute intensive apps vs scability and database optimized for MP) but with 4 chips, there were no choice.

Then the Opteron 6100 came. Opteron 6100 managed to find itself a niche. It fits exactly between the Xeon 5500/5600 and the Xeon 7500/E7. No matter how good the cores in the 5600 chip might perform, it doesn't beat 4P Opteron 6100(6x2 vs 12x4 cores) with only 2 chips. And if you can't justify the RAS and scability features of the Xeon MP, its good too.

Right now: Xeon 5600 2P->Opteron 6100 4P-->Xeon MP

With Sandy Bridge EP, there's a chip that will directly address the niche Opteron 6100 has with the 4P option.

I assume there might have been a delay to get the EP more like the MP. Sure, the delay isn't nice, but now they'll have all bases covered.

 

[aviat72]

Sandy Bridge was supposed to be incompatible with 1.65v memory, and yet everyone uses it without issues. 1.5v will probably be okay on IB, especially considering its ubiquitousness and the relative unavailability of lower-voltage RAM.

The issues, if they arise, will occur over time; these are reliability related.

IB is going down another node in size to 22nm so lower voltages become even more important.

People are still running 1.65V because that is what they had. Intel was quite clear right from the word go that 1.5V is what the specs are.

 

[IntelUser2000]

People are still running 1.65V because that is what they had. Intel was quite clear right from the word go that 1.5V is what the specs are.

Exactly the same thing was said when 45nm chips came out. Then articles came out saying the real limit was actually 1.65V. Of course, people take that as "oh that's the official maximum" and put it even higher. But there's more to that.

http://www.anandtech.com/show/2636

http://www.bit-tech.net/hardware/cpus/2008/11/03/intel-core-i7-nehalem-architecture-dive/3

You can push the memory voltage, but Intel performance guru François Piednoel was keen to stress that the CPU voltage must be kept within a 0.5V potential difference to memory.

By that rationale, AMD's Athlon 64s suffered the same fate with DDR2

It has to be related to having the memory controller integrated into the CPU.

 

[VirtualLarry]

IB next spring? Hmm, decisions, decisions. SB unlocked quad-core was cheap (relatively speaking), will IB quad-cores come in at the same price (and discontinue SB), or will IB quad-cores slide into the pricing schedule at something higher than SB, and both continue on the market for a while?

 

[Edrick]

IB next spring? Hmm, decisions, decisions. SB unlocked quad-core was cheap (relatively speaking), will IB quad-cores come in at the same price (and discontinue SB), or will IB quad-cores slide into the pricing schedule at something higher than SB, and both continue on the market for a while?

IB LGA1155 should replace SB LGA1155 at very similar price points. IB will then coexist in the market with SB-E for awhile until IB-E replaces that later in the year.