Broadwell @ 14 nm already taped out and working

[Nemesis 1]

Broadwell, unlike Haswell, will have a new GPU architecture. Haswell just has minor tweaks, more EUs and lower clocks to save power. We don't know how it will perform yet, but hopefully it will be more proportionally than Haswells disappointing jump.

I call troll . So tell us Why is Haswell Performance increase disapointing ? Were did you see benchmarks to make such a statement . Your calling 2x improvment disappointing. Trinity was what 20% faster than llano . Thats a disappointing performance increase . Haswell will be 30% faster than trinity yet your calling it a disappointing increase . If what INTEL says is true . Your trolling and telling a differant story than Intel . 1 of you Is LIEING Intel or YOU . I have zero reason not to believe INtel . You on the other hand you who has seen NO benchmarks its dissapointing . Everthing I seen presented at IDF was impressive . YOU trolled and lied and your getting away with it . AS usual. IVY bridge made a bigger jump than trinity . All trails covered.

 

[cytg111]

If you're willing to spend $400 go Sandy Bridge-E with the Intel Core i7-3820. If you're willing you can wait until Ivy Bridge-E, both use the same LGA 2011 socket. Either of these will last you comfortably until Broadwell.

I'm considering Ivy Bridge-E myself since I already need a better motherboard.

Early reports are showing that Haswell will not use either the LGA 2011 or the LGA 1155 desktop socket.

I dont see the "Bridge-E" angle here, unless you need those cores now.
My assumption is this : There is select few programs outthere that will actually use those 2 extra cores and i theorize that those select few programs are also candidates to be early adopters of AVX2, wich, if what we hear is true, will result in massive performance++... much more that 50% moar coar's.

Just saying.

 

[NTMBK]

I call troll . So tell us Why is Haswell Performance increase disapointing ? Were did you see benchmarks to make such a statement . Your calling 2x improvment disappointing. Trinity was what 20% faster than llano . Thats a disappointing performance increase . Haswell will be 30% faster than trinity yet your calling it a disappointing increase . If what INTEL says is true . Your trolling and telling a differant story than Intel . 1 of you Is LIEING Intel or YOU . I have zero reason not to believe INtel . You on the other hand you who has seen NO benchmarks its dissapointing . Everthing I seen presented at IDF was impressive . YOU trolled and lied and your getting away with it . AS usual. IVY bridge made a bigger jump than trinity . All trails covered.

2x improvement is a disappointment merely because Intel's graphics performance was already so far behind. Take a look at these results:

If HD5000 (or whatever top Haswell graphics is called) doubles that, it will pull ahead of Trinity- but then Kaveri will be getting GCN graphics, probably putting it ahead of Haswell again. (Improvements from VLIW4->GCN were far greater than VLIW5->VLIW4, so I expect a bigger improvement than from Llano to Trinity.)

However, Inetl's rate of progress is still incredibly impressive. If they keep up this pace, they will be beating AMD on graphics in just a couple of generations.

 

[tweakboy]

Dude if haswell has a hexcore im all over it.

IF not... i guess my gulfie wont be replaced anytime soon.

THe 2 top models of Haswell are Hexacore 12 threads. But expensive 600 and 1000 dollars.

 

[NTMBK]

THe 2 top models of Haswell are Hexacore 12 threads. But expensive 600 and 1000 dollars.

Is that Haswell, or Haswell-E? If that's in the standard desktop socket, then that's pretty neat.

 

[ShintaiDK]

We heard big talk about Trinity GFX too. That only planned out to be a very modest increase. Kaveri will be even smaller if they dont solve the bandwidth issue. And nothing shows that its gonna happen before DDR4 goes mainstream. And even then, the bar is just raised, not gone.

 

[Fjodor2001]

Has there been any indication of what Intel intends to do with the 14 nm chip processing technology advancements that Broadwell brings?

They will be able to fit twice the amount of transistors on the same die area. So some options are:

1. Moar CPU cores.
2. Moar GPU cores (EUs), resulting in better GFX performance.
3. Moar GPU cores (EUs) but clocked lower (same as on Haswell GT3), resulting in modest GFX performance increase and lower power consumption.
4. Keep same transistor count (basically means keep CPU and GPU core count about the same), which will require a smaller die area. That would effectively result in lower TDP and lower manufacturing cost per chip (=> lower price for customers, and/or more profit for Intel ).
5. Anything else?
6. A combination of some of the options above.

Has anything been indicated? Or what seems most reasonable?

 

[VirtualLarry]

Has there been any indication of what Intel intends to do with the 14 nm chip processing technology advancements that Broadwell brings?

They will be able to fit twice the amount of transistors on the same die area. So some options are:

1. Moar CPU cores.

Yes!

 

[Ferzerp]

Has there been any indication of what Intel intends to do with the 14 nm chip processing technology advancements that Broadwell brings?

They will be able to fit twice the amount of transistors on the same die area. So some options are:

1. Moar CPU cores.
2. Moar GPU cores (EUs), resulting in better GFX performance.
3. Moar GPU cores (EUs) but clocked lower (same as on Haswell GT3), resulting in modest GFX performance increase and lower power consumption.
4. Keep same transistor count (basically means keep CPU and GPU core count about the same), which will require a smaller die area. That would effectively result in lower TDP and lower manufacturing cost per chip (=> lower price for customers, and/or more profit for Intel ).
5. Anything else?
6. A combination of some of the options above.

Has anything been indicated? Or what seems most reasonable?

Remember that the node change has to come with a resulting decrease in die area per processor *or* a massive increase in cpu price. Each fab upgrade costs much more than the last, and the way that they are funded is by making the resulting processors cheaper to manufacture. The way this is done is by squeezing more procs into a wafer (since the wafer cost doesn't go up in the same way that the cost of the infrastructure needed goes up).

Sure, we end up with more transistors, but it's not 2x per shrink (because of the resulting die area reduction).

There is nothing special about 14nm in this regard... Each process shrink exhibits this behavior. Over time, the effects really add up.

edit: to give it some perspective, to manufacture a 3770k on the same process that the 486 was made on (ignoring the reasons this wouldn't work), you would need a die area of around 331,000 mm^2. If square (IB isn't square, but I am lazy and it makes the math easy), this would be around 575mm per side. This is a processor die that would be almost 2 ft to a side. Wow.

 

[jpiniero]

Has there been any indication of what Intel intends to do with the 14 nm chip processing technology advancements that Broadwell brings?

The only real rumored feature at this point is moving the chipset on package for all mainstream processors and not just the ultrabook one like in Haswell.

 

[ShintaiDK]

Has there been any indication of what Intel intends to do with the 14 nm chip processing technology advancements that Broadwell brings?

They will be able to fit twice the amount of transistors on the same die area. So some options are:

1. Moar CPU cores.
2. Moar GPU cores (EUs), resulting in better GFX performance.
3. Moar GPU cores (EUs) but clocked lower (same as on Haswell GT3), resulting in modest GFX performance increase and lower power consumption.
4. Keep same transistor count (basically means keep CPU and GPU core count about the same), which will require a smaller die area. That would effectively result in lower TDP and lower manufacturing cost per chip (=> lower price for customers, and/or more profit for Intel ).
5. Anything else?
6. A combination of some of the options above.

Has anything been indicated? Or what seems most reasonable?

The question is rather what Skylake will bring on 14nm. I think the PCH will dissapear. But for Broadwell it will be identical to Haswell on that area.

Number 1 wont happen with Broadwell, that I can pretty much guarantee you.
Number 4 wont happen either. Intels prices are also already in the perfect ratio, meaning we get inflation corrected price increases every year. Lower TDP tho, yes.

GPU wise Broadwell will have gen8 GPU uarch, Haswell will have 7+. I assume you think they will have the same. Haswell GPU is based an improvement/expand of Ivy.

 

[Nemesis 1]

2x improvement is a disappointment merely because Intel's graphics performance was already so far behind. Take a look at these results:

If HD5000 (or whatever top Haswell graphics is called) doubles that, it will pull ahead of Trinity- but then Kaveri will be getting GCN graphics, probably putting it ahead of Haswell again. (Improvements from VLIW4->GCN were far greater than VLIW5->VLIW4, so I expect a bigger improvement than from Llano to Trinity.)

However, Inetl's rate of progress is still incredibly impressive. If they keep up this pace, they will be beating AMD on graphics in just a couple of generations.

An honest person would not cherry pick as you did above.. The other point . I won't buy Haswell for its desktop graphics . Maybe at 10NM I think about it.. We seem to be forgetting the desktop is a shrinking market that AMD already has fair amout of share in . But its a shrinking market . Trinity doesn't do so good in the notebook area. Comparred to IB. and once you get down to Ultra books IB is just as strong as Trinity . Just as I said 8 months ago it would be .

 

[Ferzerp]

We seem to be forgetting the desktop is a shrinking market that AMD already has fair amout of share in .

This gets rather exaggerated. We've been told its shrinking for years, but I think that mid 2012 was the first time that there has been any decrease in desktop processor shipments year over year. I don't have the stats in front of me, but if I recall correctly, it wasn't a significant decline, so whether it is indicative of a shrinking market, or just statistical noise wasn't clear.

This is one of those "facts" that people have heard over time (tech journalists), and repeated because it sounded correct without ever researching it.

I wish I had access to all the IDC stats, but I don't. If anyone can find the real numbers, please let me know.

edit: potentially, I am confusing this stat with the entire x86 market. If anyone can find the stats, please let me know.

edit2: I was. http://www.pcworld.idg.com.au/article/430324/tablets_smartphones_dent_pc_shipments_second_quarter/ references the info I was talking about. 0.1% total reduction. I was lumping desktop and laptop together in my head. Still, nice info for all the "PC is dying" FUD.

 

[Nemesis 1]

So when is Kaveri do? Befor or after broadwell'? IF amd wants to do mask every 6 months they can go for it . You seem to forget the more powerful AMD makes its weak Igpus the fewer discret card they sell. AMD has to dance to a tune . Intel has no such restrictions. Itel is going to beat AMD/ATI without spending 6 billion dollars on an industry both intel and AMD are destroying . LOL.

 

[Fjodor2001]

The only real rumored feature at this point is moving the chipset on package for all mainstream processors and not just the ultrabook one like in Haswell.

Yes, but how much die area does that occupy? Not that much, I think.

So there should be plenty of additional transistors left to do fun stuff with. For example they could both double the amount of CPU cores and GPU cores, while keeping the same die area!

 

[Nemesis 1]

This gets rather exaggerated. We've been told its shrinking for years, but I think that mid 2012 was the first time that there has been any decrease in desktop processor shipments year over year. I don't have the stats in front of me, but if I recall correctly, it wasn't a significant decline, so whether it is indicative of a shrinking market, or just statistical noise wasn't clear.

This is one of those "facts" that people have heard over time (tech journalists), and repeated because it sounded correct without ever researching it.

I wish I had access to all the IDC stats, but I don't. If anyone can find the real numbers, please let me know.

edit: potentially, I am confusing this stat with the entire x86 market. If anyone can find the stats, please let me know.

Everyone else in tech world is wrong and your correct . You sound like me in theology.

 

[Ferzerp]

Everyone else in tech world is wrong and your correct . You sound like me in theology.

See the correction I made to myself 1 minute before your post.

edit: It is, however, important to take any projections or analyses of tech journalists with a grain of salt. They tend to love to make projections, predictions, etc. They all too quickly forget that they are journalists, and not computing experts. Many of the so called "trends" are often just fabrications or things that a certain journalist would love to see, and so pretends it is the case.

Sadly, this is the way with all media. I just pay more attention to tech.

If any of you know journalists in real life, ask yourself if you really believe that they are an expert on everything they write about.

 

[AtenRa]

So when is Kaveri do? Befor or after broadwell'? IF amd wants to do mask every 6 months they can go for it . You seem to forget the more powerful AMD makes its weak Igpus the fewer discret card they sell. AMD has to dance to a tune . Intel has no such restrictions. Itel is going to beat AMD/ATI without spending 6 billion dollars on an industry both intel and AMD are destroying . LOL.

Kaveri will come next year (2013) to counter Haswell.

APUs will not eat Discrete GPU market share, Discrete GPUs will be updated as well every year.

 

[IntelEnthusiast]

Freaky.

Intel must be staffed by aliens.

Hey, my head is not that big.

 

[Idontcare]

An honest person would not cherry pick as you did above..

You're above that kind of pointless underhanded credibility attack, be true to yourself and don't do this. NTMBK is a respectable poster, if you find yourself confusing their posts as mindless cherry picking then your first suspicion ought to be that perhaps you have misinterpreted the post or the convo's direction itself. Not everyone is out to get you or fool you.

 

[Idontcare]

Freaky.

Intel must be staffed by aliens.

Hey, my head is not that big.

Tsavo didn't rule out body-snatching aliens

Makes those blue-man group commercials all the more understandable now...:whiste:

 

[Nemesis 1]

You're above that kind of pointless underhanded credibility attack, be true to yourself and don't do this. NTMBK is a respectable poster, if you find yourself confusing their posts as mindless cherry picking then your first suspicion ought to be that perhaps you have misinterpreted the post or the convo's direction itself. Not everyone is out to get you or fool you.

Reallly that wasn't a best case . Thats as close as it gets between IB Igpu and trinity . If thats the case . Clearly it wasn't cherry picking . But if thats a case were its Best case than indeed its cherry pickin . Intel 22nm Soc OoO dual and single cores next year 2013 out in the market . IDC he can be a respectful member . That doesn't mean He can make crap up no were does it say in any Chart that The next atom OoO will be out in 2014 . No were. Kaverti who says that is AMDs answer to Haswell. Is it at 28nm a half node or is it on the GF 20nm . If its on 28nm Kaverti isn't bring much more than trinity. The road maps you guys went and got . Medfield was released along time ago and yet its brother for tablets is not yet released. Does than mean Medifield isn't in phones right now today. You guys are looking at the complete total release . I am looking at just the smart phone 22nm SoC

 

[TuxDave]

Hey, my head is not that big.

Sometimes I can act a little big headed. =\

 

[ShintaiDK]

Tsavo didn't rule out body-snatching aliens

Makes those blue-man group commercials all the more understandable now...:whiste:

And new XCOM game out...think about that. I heard Intels Hillsboro facilities D1x is actually the secret Area 52.

 

[TuxDave]

And new XCOM game out...think about that. I heard Intels Hillsboro facilities D1x is actually the secret Area 52.

So when Obama was invited to talk at Hillsboro... /cue body snatchers...