Haswell model specs leaked

[Fjodor2001]

oh of course they could sell it for a lot less, the Xeon E5-2620 goes for ~$400

But it's only 6C @ 2 GHz.

 

[Homeles]

But it's only 6C @ 2 GHz.

So?

 

[ShintaiDK]

The funny thing is that they could probably make an 8 core Haswell without iGPU, and it wouldn't occupy much more die area than a 3570K (perhaps 30% more). So if CPUs were priced based on die area alone, it could probably be sold for $300-400 or so.

But then there's the marketing aspect of this unfortunately that stops this from happening... :\

If you look away from everything else and only use diesize as a measure, then yes.

 

[cbn]

Also, it's interesting to note that a 96 EU chip with 4 CPU cores would probably allocate something like 70% of the die area to the iGPU, 20% to the CPU cores (including caches), and 10% to the rest like memory controller.

Then you're no longer buying a CPU with iGPU. You're buying a GPU with iCPU...

Yeah, that sounds like a pretty serious chip. I'll bet at least one company puts that on a desktop board complete with large cooler and overclocking options. (Either that or GT3 if substantially less expensive).

Yes, that could be a possibility. But going from 20->96 EUs without any solution to the memory bandwidth problem does not seem like a balanced design. So perhaps there is some truth to the eDRAM rumor after all, although some people here does not believe in that.

I think I read somewhere that tessellation was supposed to save memory bandwidth. So maybe in the future that can reduce the need for expensive embedded memory?

P.S. I noticed fast DDR3 memory is pretty cheap now. There are various 8 GB DDR3 2400 kits at newegg starting at $55.

 

[IntelUser2000]

Does anyone know the theoretical flops of Ivy's HD4k?

16EUs x 4 FMA per EU x 2 issue @ ~1.1GHz = 282GFlops.

That's peak FLOPs. The practical FLOPs probably end up at ~220GFlops.

 

[bunnyfubbles]

But it's only 6C @ 2 GHz.

That doesn't matter when it still consumes the same amount of materials to produce it as its an 8 core chip with 2 cores disabled and then its clockrates locked at an otherwise arbitrarily low number as I seriously doubt their yields are bad enough to where all those chips are only able to achieve 2GHz.

The point is that, physically speaking, there isn't much difference between the $400 chip and the $2000 ones. If there was a way to unlock those 2 cores like with a pencil mod of old, and then had the ability to overclock the CPU, we could essentially have a $2000 chip for $400.

But alas, neither are remotely possible. If we want to overclock we have to spend $600-1000 for one of the i7s, and if we want 8 cores and reasonably high clockrates we need to spend $2000.

This is why I said earlier that an 8 core i7 would be almost too good to be true because of how expensive the faster 8 core Xeons are. An i7 3980X, if priced at the traditional $1000 mark, would be half the price of a Xeon and then of course it would be able to hit 33-50+% higher clockrates if you can cool it, of which even a pretty high end water setup on top of that $1000 would still land you far short of the locked down 8 core Xeon.

 

[IntelUser2000]

Bunnyfubbles is right. They would have very hard time squeezing a 3.5GHz chip into 150W TDP, even if they could.

Then there's a issue of whether they will do that. Ivy Bridge E is still a year away*, so maybe its not as far fetched, but what's the point anyway? If you consider enthusiasts buying 2600Ks and 3770Ks as minority, the ones buying E chips would be minority out of those enthusiasts.

*Latest rumor puts Xeon E5 v2 at mid-2013 and the enthusiast derivative later towards Q3 and even Q4.

 

[bunnyfubbles]

the point is that the original rumor of the 3980X was that it was supposed to come out Q3 or Q4'12, but, assuming this current rumor is accurate, the 3980X was delayed and Intel gave us the 3970X in the mean time, and that with any luck we'll have the 3980X in Q1'13, plenty of time to get use out of it until Ivy-E.

And, yes, while the niche is very small, I'm sure there are enough eager enthusiasts wanting to get their hands on a beefier chip (myself included). I'm also sure there are original 980X users out there (of which the 980X is nearly 3 years old!) who have held fast to their 6 core chip wanting something more a bit more potent than the 3930K or 3960/70X.

 

[gmaster456]

Intel should come up with a new name.

Core
Core 2
Core i
Core i 2nd generation
Core i 3rd generation

Are they really going to call Haswell Core i 4th generation?

 

[bunnyfubbles]

Intel should come up with a new name.

Core
Core 2
Core i
Core i 2nd generation
Core i 3rd generation

Are they really going to call Haswell Core i 4th generation?

Core is brand name, like Pentium

Instead of "Pentium 3" and "Pentium 4", we have model numbers the designate generation.

I don't see whats wrong with it, and I think it actually works very well, and is very similar to how GPUs have been named for a while now.

 

[ShintaiDK]

Intel should come up with a new name.

Core
Core 2
Core i
Core i 2nd generation
Core i 3rd generation

Are they really going to call Haswell Core i 4th generation?

Yes they are, and I dont see the problem with it.

 

[Fjodor2001]

Yes they are, and I dont see the problem with it.

It's like with the iPad. You have the iPad, the new iPad, and the newer than the new iPad, which currently is the newest. Soon we'll have the newer than the previously newest iPad.

Totally clear...

 

[Homeles]

It's like with the iPad. You have the iPad, the new iPad, and the newer than the new iPad, which currently is the newest. Soon we'll have the newer than the previously newest iPad.

Totally clear...

Not even a remotely valid comparison...

A progressive series of numerals is not equivalent to Apple's inconsistent naming scheme.

 

[Torn Mind]

The Core i-series usually has it's model number after it in advertisements. The generation number corresponds to the far left number in the model number. So, it's easy to see a Core i3 3xxx and know that it's a later generation than a Core i3 2xxx while the ix part of the name tells you which tier of performance it is in, although i7s don't distinguish between LGA 1155 and LGA 2011. The naming scheme is similar to the Radeon naming scheme.

It's a helluva lot easier than the Core 2 Duos nomenclature, where a T6xxx is a Penryn while a T7xxx is a Merom.

Apple used to have a numerical scheme with the Ipad 2, but they dumped that with "new iPad". Intel is not doing "newer newer i-series".

 

[Fjodor2001]

Not even a remotely valid comparison...

A progressive series of numerals is not equivalent to Apple's inconsistent naming scheme.

Fair enough, the Intel Core naming is far better. Especially since they have a digit in the model number indicating the generation.

But I don't get why they change from e.g.:

2500K->3570K (instead of 3500K)
3570K->4670K (instead of 4570K)

I've not been able to find the logic behind that. Has anyone else?

 

[Homeles]

The 4670K has a 100MHz bump in clock speed over the 3570K. As far as the "70" designation goes, I have no idea.

 

[Fjodor2001]

The 4670K has a 100MHz bump in clock speed over the 3570K.

Does it really? I thought both were 3.4 GHz / 3.8 GHz Turbo. See this and this.

Also, it's not like the second digit in the Core model number in general has any direct indication of frequency as far as I can tell?

 

[Torn Mind]

Fair enough, the Intel Core naming is far better. Especially since they have a digit in the model number indicating the generation.

But it I don't get why they change from e.g.:

2500K->3570K (instead of 3500K)
3570K->4670K (instead of 4570K)

I've not been able to find the logic behind that. Has anyone else?

Only thing I can think of is to troll the rumor spreaders.

 

[ShintaiDK]

Fair enough, the Intel Core naming is far better. Especially since they have a digit in the model number indicating the generation.

But it I don't get why they change from e.g.:

2500K->3570K (instead of 3500K)
3570K->4670K (instead of 4570K)

I've not been able to find the logic behind that. Has anyone else?

We need to see the dualcore lineup and mobile lineup too. Plus anything related to OEM models only. Then we can see if there is logic or if its just PR.

No reason for AMD either to skip the 4000 series on APUs. But again, marketing is marketing. And that applies for all products in all segments. Why is is a Radeon HD7970 not just called a Radeon 797. Why aint a Geforce GTX680 not just called Geforce 68. Why are memory and SSDs usually named all kinds of crap?

Names simply sell. Maybe 4570K didnt sound good in chinese and 4670K was better. Who knows.

 

[Homeles]

Does it really? I thought both were 3.4 GHz / 3.8 GHz Turbo. See this and this.

Also, it's not like the second digit in the Core model number in general has any direct indication of frequency as far as I can tell?

Whoops, mixed it up with Sandy.

 

[skipsneeky2]

The name scheme is fine for these chips,simple for most people to know which generation it is if we follow the 2000,3000,4000 series as we do with our video cards as well.

 

[Haserath]

It's like with the iPad. You have the iPad, the new iPad, and the newer than the new iPad, which currently is the newest. Soon we'll have the newer than the previously newest iPad.

Totally clear...

Core 2 to Core I, maybe a little bit.

But having generations makes perfect sense.

It would be more like iPad 1-4 if Apple didn't screw with the naming scheme.

 

[cbn]

NUC, Mac Mini etc?

My vote for a bigger cooler and mobo overclocking options for both Haswell CPU and GT3 iGPU to further differentiate it from being stock ULV laptop guts minus screen, battery and keyboard.

See below to get an idea of the NUC 4" x 4" board size vs. stock desktop (? TDP) LGA push pin cooler:

Of course, a somewhat taller NUC form factor enclosure would have to be designed as well depending on large of a cooler is desired.

 

[Idontcare]

Fair enough, the Intel Core naming is far better. Especially since they have a digit in the model number indicating the generation.

But I don't get why they change from e.g.:

2500K->3570K (instead of 3500K)
3570K->4670K (instead of 4570K)

I've not been able to find the logic behind that. Has anyone else?

The 4670K has a 100MHz bump in clock speed over the 3570K. As far as the "70" designation goes, I have no idea.

Maybe it is to designate the "P1270" Process Name that stands for 22nm?

So 4670K really should become 4672K? (P1272 is 14nm)

Just making guesses, yes I get that it doesn't stand to reason then why the 2500K wasn't called the 2568K in such a schema. (P1268 = 32nm)

 

[Charles Kozierok]

Intel's model naming has blown goat barf ever since they discovered the word "Core". It is so ridiculously overloaded at this point that nobody knows what anyone else is talking about half the time. It would be like if Ford suddenly decided to call every vehicle that rolls off the assembly line a variation of "Car" with a number on it.

Hey guys: there are lots of words in the English language. Pick a few of them and use them for model names. If you don't like any of them, you can even make up new ones!