Intel Tick Tock slows down: Sandy Bridge slips into 2011?

[ajaidev]

"Lack of competition hurts the market - AMD's continuous delays of their next-gen CPU architecture combined with Intel's ill-fated effort on Larrabee resulted in a slowdown of the Tick-Tock cadence. By now, we all know that AMD Bulldozer should only surface in 2011 - definitely no 2010 major core advances from them above the current cores seen in Phenom and Opteron spread available now. By itself, that's not exactly good news, but then again it wasn't unexpected either..."

http://www.brightsideofnews.com/news/2009/12/21/intel-tick-tock-slows-down-sandy-bridge-slips-into-2011.aspx

Too bad quick someone sue Intel for not releasing SD due to less competition...

[Etihtsarom]

Had a feeling that would happen.

[safan80]

This is what happens when AMD can't compete :shakes:. If AMD can get back into the game and offer performance better than Intel then the cpu wars will be back. The sad thing is intel would be offering 8 core cpus now if amd could keep up. At least AMD got their ATI division back into the game. With the profit they are making off of the 5xxx series I hope they can spend their money wisely and hire more devs and chip designers.

[Manicdan]

who wants to say this wouldnt be happening if intel didnt do all those illegal things a few years ago, which would have given amd the cashflow then to help get new tech out sooner?

[cky2k6]

Wasn't this the case for quite a while now? q1 2011 was what I heard for sb a good amount of time ago. Blaming amd for this is silly, who can expect a rival 1/10th the size of intel keep up with such manufacturing capabilities, even if amd's cpu division could design as quickly as intel.

[ajaidev]

Wasn't this the case for quite a while now? q1 2011 was what I heard for sb a good amount of time ago. Blaming amd for this is silly, who can expect a rival 1/10th the size of intel keep up with such manufacturing capabilities, even if amd's cpu division could design as quickly as intel.

Naa SB was for late 2010 "Q3-Q4" and Ivy bridge was for 2011 Q2-Q3.

The reason for the slip is not AMD but not the reason provided i have been told. :yepp:

[Nedjo]

"Lack of competition hurts the market - AMD's continuous delays of their next-gen CPU architecture combined with Intel's ill-fated effort on Larrabee resulted in a slowdown of the Tick-Tock cadence. By now, we all know that AMD Bulldozer should only surface in 2011 - definitely no 2010 major core advances from them above the current cores seen in Phenom and Opteron spread available now. By itself, that's not exactly good news, but then again it wasn't unexpected either..."

http://www.brightsideofnews.com/news/2009/12/21/intel-tick-tock-slows-down-sandy-bridge-slips-into-2011.aspx

Too bad quick someone sue Intel for not releasing SD due to less competition...

this is ridiculous POV! Year ago AMD acknowledged that Bulldozer will be out in 2011:

http://www.techpowerup.com/reviews/AMD/Analysts_Day/3.html

what delay does author of the article is talking about?

[IvanAndreevich]

this is ridiculous POV! Year ago AMD acknowledged that Bulldozer will be out in 2011:

http://www.techpowerup.com/reviews/AMD/Analysts_Day/3.html

what delay does author of the article is talking about?

I think it's BS too. Intel has been pushing their tick-tock for many years now. If they fall behind it will be seen as their failure. Why would they do that, instead of increasing their lead over AMD :shrug:

[Tech Dav]

Naa SB was for late 2010 "Q3-Q4" and Ivy bridge was for 2011 Q2-Q3.

The reason for the slip is not AMD but not the reason provided i have been told. :yepp:

How could they promise that when 45nm came out only in early 2008 just like 32nm will come out in early 2010 :shrug:

32nm=2010
22nm=2012
16nm=2014
11nm=2016
8nm=2018
6nm=2020
4nm=2022

[xVeinx]

As the process gets smaller with each generation, and new architectures are implemented, it's inevitable that some aspects will take longer to get right. Regardless of whether Intel can afford to invest billions more to any particular generation or design, its people who have to be creative and find the best ways to get a processor working at optimal levels and with a profitable yield. A lot of brilliant people are working hard at both Intel and AMD on these problems, so maybe we should cut them some slack? Any company would love to churn out amazing processors like clockwork, but there are some critical design choices that are being made right now that could have lasting impact on computing in the next decade, and I'd rather they made the best choice and the better product than trying to push things too fast.

PS: Do you really believe AMD is at fault for this? Does competition solve all technological problems? Does it always mean we get what we want when we want it?

[ajaidev]

How could they promise that when 45nm came out only in early 2008 just like 32nm will come out in early 2010 :shrug:

32nm=2010
22nm=2012
16nm=2014
11nm=2016
8nm=2018
6nm=2020
4nm=2022

http://www.canardpc.com/img/dossier/divers/idf2008/ticktock.jpg

http://www.electronista.com/articles/09/08/12/sandy.bridge.in.q4.2010/

http://www.xbitlabs.com/news/cpu/display/20080319161323_Intel_Discloses_Details_about_Sandy _Bridge_Microprocessors_Due_in_2010.html


As the process gets smaller with each generation, and new architectures are implemented, it's inevitable that some aspects will take longer to get right. Regardless of whether Intel can afford to invest billions more to any particular generation or design, its people who have to be creative and find the best ways to get a processor working at optimal levels and with a profitable yield. A lot of brilliant people are working hard at both Intel and AMD on these problems, so maybe we should cut them some slack? Any company would love to churn out amazing processors like clockwork, but there are some critical design choices that are being made right now that could have lasting impact on computing in the next decade, and I'd rather they made the best choice and the better product than trying to push things too fast.

PS: Do you really believe AMD is at fault for this? Does competition solve all technological problems? Does it always mean we get what we want when we want it?

+1

But i have heard it is partly AMD's fault but its not that AMD did not offer any competition but something much more devious than that. Intel has peaty good 32nm process the i3/i9 will be made on 32nm i am sure that its capable of putting out a SB in 32nm...

[terrace215]

This is all based on BSON, looking at a desktop roadmap covering 2010, and concluding that any new product coming in 2010 (say Q4), would necessarily be disclosed on said roadmap at this time.

Did I get that right?

The same outfit that claims Apple will not use Arrandale, and is instead waiting for Sandybridge, right?
:ROTF:

And this caption is completely wrong:

Intel's Tick-Tock cadence lasted from Conroe to Nehalem... then it fell apart. Westmere is a year late, Sandy Bridge isn't on the product roadmaps for 2010. Don't ask about the Haswell, CPU+LRB part.

Westmere is hardly a year late. It is, in fact, already shipping (in 2009), and will launch (the first week of 2010) 1 year and 1 quarter after Nehalem, making it 1 quarter beyond "a year", or 1 QUARTER "late". But hey, let's pretend 3 months is one year, and try to up the drama of our articles.

[Tech Dav]

http://www.canardpc.com/img/dossier/divers/idf2008/ticktock.jpg

http://www.electronista.com/articles/09/08/12/sandy.bridge.in.q4.2010/


Intel always say Q4/Q1 :shrug:

[terrace215]

But i have heard it is partly AMD's fault but its not that AMD did not offer any competition but something much more devious than that. Intel has peaty good 32nm process the i3/i9 will be made on 32nm i am sure that its capable of putting out a SB in 32nm...


Devious? You mean Intel is taking SB directly to 22nm in early 2011, as that process is maturing faster than expected? :D

That would lap AMD, launching 22nm products as AMD launches 32nm. Hmmmm

[ajaidev]

Devious? You mean Intel is taking SB directly to 22nm in early 2011, as that process is maturing faster than expected? :D

That would lap AMD, launching 22nm products as AMD launches 32nm. Hmmmm

Naa lets say its a make or break thing, if it works SB will be way more comparative as compare to BD if it does not pan out they will release SB and the IB will be the real BD killer. :yepp:


Intel always say Q4/Q1 :shrug:


It does not work like that we had WM samples here around Q3 2009 and shipping has been started but for SB ppl are saying there will be no samples till 2011 now do u get it, the samples were suppose to coe around q3-q4 2010..

[informal]

Devious? You mean Intel is taking SB directly to 22nm in early 2011, as that process is maturing faster than expected? :D

That would lap AMD, launching 22nm products as AMD launches 32nm. Hmmmm

Yeah SB directly on 22nm, Dreamland here we come!:up:

[Solus Corvus]

Devious? You mean Intel is taking SB directly to 22nm in early 2011, as that process is maturing faster than expected? :D

That would lap AMD, launching 22nm products as AMD launches 32nm. Hmmmm
So you are suggesting that not only is Tick Tock not slowing down but they are in fact going to abandon it and skip ahead? On what evidence do you base this idea?

[terrace215]

Naa lets say its a make or break thing, if it works SB will be way more comparative as compare to BD if it does not pan out they will release SB and the IB will be the real BD killer. :yepp:


Well... SB is already going to splat BD. And IB is more or less a shrink, so that would only in theory indicate more cores, more clock, and/or lower power.

Explain your theory more.

[Nedjo]

Well... SB is already going to splat BD.

terrace you should seriously consider putting this statement in your sig :p:

[ajaidev]

Well... SB is already going to splat BD. And IB is more or less a shrink, so that would only in theory indicate more cores, more clock, and/or lower power.

Explain your theory more.

Its not my theory i got it from someone, someone i get intel ES "see my i9/i5 scores" off so i take it he knows more than well the avg joe.

You are not correct about SB crushing BD at least in multi threads i can tell you this with 80% surety "1 SB core<1 BD module". The magic dust intel plans is suppose to help with this and the result maybe something that is good enough or will not be so it can go anyway.

Extract Intel can magically ship the SB as soon as Q4 2010 if the thing does not work or 2011 if the thing does work simple.

[zalbard]

Maybe they are doing some further tweaking in order to compete with BD better, who knows.
But this does not come as a big surprise to be honest.
On the other hand, this is BSN... so not necessarily true.

[terrace215]

"1 SB core<1 BD module"

Well, it would be a DISASTER for AMD if that were not the case, as 1 BD module is nearly 2 cores, with shared FP & fetch/decode. The perf/W comparisons are going to be devastating.

[Nedjo]

The perf/W comparisons are going to be devastating.

and you can backup that "theory" by some relevant data, or you just continuing pulling the arguments from... err... air...
:rolleyes:

[informal]

It would be funny to see one BD module taking up similar or less die area than one SB SMT capable core,while outperforming it by a healthy margin in throughput and having much better power management capabilities(shutting off mini cores/fp;"turboing" each of the mini cores etc. ;SB on the other hand will need a reboot and BIOS modification to shut off SMT) and execution capabilities(+fastpath decoding resulting in possible mix of 8 instr. decoded in parallel;fp/sse execution along side the int workloads->in theory 2 x 4 way int cores + one 4 way fp unit,all three trying to catch peeks from the decoding stage;SMT like AVX execution in 2 FMAC capable FP/SIMD units that are shared between two mini int cores ).

[Chumbucket843]

shared decode is actually quite a good idea considering how much die space x86 and its >1000 complex instructions use. it is also the biggest source of power consumption in a modern cpu even with techniques like trace cache, lsd, macro-fusion, etc. we dont even know how much die space a BD module uses yet. its obvious that amd has taken focus off of single threaded performance. i am pretty sure adding another integer core would not add that much die space as one might assume.

[pausert20]

I'm not sure how true the topic article is. Intel showed off a SB at the Fall IDF. We don't even know what stepping that was. If it was an A0 then it might be tight to come out at the end of 2010 but if it was a A1 they have already spun it once to fix stuff. That would put them in a very good way to make EOY 2010 launch.

[terrace215]

It would be funny to see one BD module taking up similar or less die area than one SB SMT capable core

It will be even funnier when one BD module is 50% larger than one SB SMT core, not that die area is the factor to focus on.

[informal]

It will be even funnier when one BD module is 50% larger than one SB SMT core, not that die area is the factor to focus on.

Let's wait and see on that one.

[Helloworld_98]

It will be even funnier when one BD module is 50% larger than one SB SMT core, not that die area is the factor to focus on.

except one module is two cores, so even if one BD core is clock-for-clock the same as SB without SMT, a BD module will be miles ahead of sandy bridge in performance compared to a SB core with SMT.

so if the module is 50% bigger than a SB core, then that would mean BD could have 50% more cores in the same area meaning higher performance even if SB did have a slight IPC advantage.

[R101]

Maybe they are doing some further tweaking in order to compete with BD better, who knows.


+1
I think this might just be the case. They have the lead now and can focus solely on the SB vs BD battle.

I think it's gonna be epic.

[Particle]

One thing is certain. Customers are going to get to pick and choose between great and great.

[terrace215]

so even if one BD core is clock-for-clock the same as SB without SMT

No chance. Not even close.

[spursindonesia]

Before things are really unveiled, anything is just a guessing game, so let's just open our mind and be kind to others opinion. :)

[geo]

high time we get competition in the CPU area

[flippin_waffles]

Yup, and now that intel isn't bribing the entire tech industry to run AMD out of business anymore (or so they say), if the damage isn't done already, that's exactly what we should get. It's just a bloody shame we never got a chance to see where the industry could have been today if it was allowed to progress unrestricted.

[H2O]

Here's an interesting read on SB:
http://news.softpedia.com/news/Intel-039-s-32nm-Sandy-Bridge-Gets-Unveiled-115948.shtml

The IGP sounds promising, however I would not be surprised to see 4, 6, and 8 core variants without a IGP, aimed at people like us.

[ajaidev]

Here's an interesting read on SB:
http://news.softpedia.com/news/Intel-039-s-32nm-Sandy-Bridge-Gets-Unveiled-115948.shtml

The IGP sounds promising, however I would not be surprised to see 4, 6, and 8 core variants without a IGP, aimed at people like us.

Your assumption is very very very close..... :yepp:

I have heard something about SB that falls quite on the point of what you assumed :D

[Vozer]

Intel showed off a SB at the Fall IDF. We don't even know what stepping that was. If it was an A0 then it might be tight to come out at the end of 2010 but if it was a A1 they have already spun it once to fix stuff. That would put them in a very good way to make EOY 2010 launch.
I think it was an A1.

http://i279.photobucket.com/albums/kk141/overclock_photobucket/misc/sandy_02.png

[Helloworld_98]

No chance. Not even close.

Don't come to conclusions so quickly,

you're forgetting BD is a huge step up from K8, the architecture is completely different so could be far better than SB clock-for-clock