Intel Haswell, Shark Bay platform, Quad-Core, 22nm, LGA1150 - Chiphell

[Olivon]

Chiphell

Thanks to Hardware.fr

[K404]

Another socket??

[naokaji]

Another socket??

Does it really matter though?

a) look at S775 that was around forever but you could not run every S775 Cpu in every S775 Mainboard, so whats the point of having the same physical socket if there are other differences that lead to incompatibilities?
b) all current Intel Chipsets are seriously lacking if it comes to SATA 3 and USB 3, so we will need new Chipsets and with it new Mainboards anyway, so if we are already going to replace the Mainboards for other reasons, does it then still matter that the socket changes?

[K404]

Does it really matter though?

a) look at S775 that was around forever but you could not run every S775 Cpu in every S775 Mainboard, so whats the point of having the same physical socket if there are other differences that lead to incompatibilities?
b) all current Intel Chipsets are seriously lacking if it comes to SATA 3 and USB 3, so we will need new Chipsets and with it new Mainboards anyway, so if we are already going to replace the Mainboards for other reasons, does it then still matter that the socket changes?

/A I partially concede on, but theres still a much larger compatability pool.... a 2005 CPU will work on a mid 2008 mobo. X58 has the same level of support and it's a GOOD thing.

/B..... yes... Intel should have had S-ATA3 and USB3 dealt with by now.... again, making us need new boards for the latest features. At least 3rd party controllers mean we can still have the features and aren't dependent on Intel.




Maybe my recent opinion is tainted by s1156 and s1155. One pin difference? Everything about the socket says "ha ha, f***ed you over, buy more hardware" Now we get 1150 as well. 2015 relase: s1147.

[UltraNEO*]

Maybe my recent opinion is tainted by s1156 and s1155. One pin difference? Everything about the socket says "ha ha, f***ed you over, buy more hardware" Now we get 1150 as well. 2015 relase: s1147.

Yeah and Intel recently learnt a new strategy, instead of making sockets with completely random numbers that no-body remembers!
They'll follow the year, for 2011 they'll release LGA 2011... what happens if there's a delay?

[Particle]

Yeah and Intel recently learnt a new strategy, instead of making sockets with completely random numbers that no-body remembers!
They'll follow the year, for 2011 they'll release LGA 2011... what happens if there's a delay?

Sockets are named two ways. There's usually a letter designation and a number designation. The number designations have been more popularly used for Intel products around the time Socket 7 was replaced by Socket 370 and later 478. I'm not sure if either of them had letter designations, but sockets after that did. The ubiquitous LGA 775 is also called Socket T, LGA 1366 is Socket B, and LGA 1156 is Socket H for example. The numbers aren't random or named for years (in the case of LGA 2011). They're named for the number of contacts in the socket (plus or minus a few that are sometimes used for keying on PGA products).

It also occurs to me that you might have just been joking, but it's hard to tell with inflectionless text on the Internet.

[bamtan2]

Does it really matter though?

a) look at S775 that was around forever but you could not run every S775 Cpu in every S775 Mainboard, so whats the point of having the same physical socket if there are other differences that lead to incompatibilities?
b) all current Intel Chipsets are seriously lacking if it comes to SATA 3 and USB 3, so we will need new Chipsets and with it new Mainboards anyway, so if we are already going to replace the Mainboards for other reasons, does it then still matter that the socket changes?

they could reuse sockets longer. please don't try to re-rationalize something that is obvious. intel has always been the leader and they can do whatever they want. there is no law against being an and obsoleting your products. normally competition allows you to switch, voting with your money, but intel has no good competition.

[freeloader]

they could reuse sockets longer. please don't try to re-rationalize something that is obvious. intel has always been the leader and they can do whatever they want. there is no law against being an and obsoleting your products. normally competition allows you to switch, voting with your money, but intel has no good competition.

If anyone is unhappy about the socket switch, then don't buy their products. Totally agree with you.

[M.Beier]

Another socket??

Passing SNB, IVB on laptop, and going straight for this =)

[HKPolice]

This better hit 6Ghz stable on air with 22nm + intergrated VRs.

[zanzabar]

This better hit 6Ghz stable on air with 22nm + intergrated VRs.

how would an onboard pwm be good for enthusiasts, intel cannot put a good one on its server boards so i dont see how an integrated one will be any good for overclocking. im sure it will help laptops and OEMs but not the enthusiast.

[sin0822]

how would an onboard pwm be good for enthusiasts, intel cannot put a good one on its server boards so i dont see how an integrated one will be any good for overclocking. im sure it will help laptops and OEMs but not the enthusiast.

Its not integrated. The PWM chip is still on the board, so is the VRM. Currently on boards there are 3 major VRMs that power the CPU, each one provides a different output voltage, in the case of current CPUs, its VCCIO(Vcore)(This is usually 8,16,12,24 phases). VCCIO(or QPI'/VTT voltage, usually 1-2 phases), and then VCCSA(System Agent voltage, 1-2 phases).

Now with this instead of having 16+2+2 youll just have something like 16, and the CPU will reduce that vcore into the other voltages needed.

[Chrono Detector]

Another socket again Intel? You guys just started using 1155 and now they are going to make another one.

[Tao~]

Every 2 years a new socket for mainstream and another for enthusiast.

BTW, I dont see any expected performance improvements listed - all thats interesting to it are the new instruction sets, AVX 2.0 and FMA. They are obviously focusing on mobile form factors but the significance of these power saving technologies will become relevant only after IVB numbers are out - which IMO should be significantly better than SNB in the power department.

But does anyone else feel Intel is again going back to its relaxed state akin to the pre K8 era ?

[Hornet331]

BTW, I dont see any expected performance improvements listed - all thats interesting to it are the new instruction sets, AVX 2.0 and FMA. They are obviously focusing on mobile form factors but the significance of these power saving technologies will become relevant only after IVB numbers are out - which IMO should be significantly better than SNB in the power department.

Second slide.

Also by now I would think ppl would have seen that with every new tick there is a new socket.

[STaRGaZeR]

Second slide.

Also by now I would think ppl would have seen that with every new tick there is a new socket.

That would be a tock.

BTW, everything that uses SSE2 now will get a MAJOR boost with AVX2.

[3NZ0]

Another socket again Intel? You guys just started using 1155 and now they are going to make another one.

If they had stuck with the same socket but required a new motherboard with a corresponding chipset/vrm then you would be complaining about that too.

Heck, even llano is only getting the socket it currently uses. The next revision will be fm2.


I love the progress of tech and this looks just as interesting as SB when that came out but the very heavy focus on mobile/low power is making me think about the inevitable move away from high performance desktops for the mass market. The m18x was a rather rude awakening for myself in that respect, it packs an enormous punch for something very portable compared to a desktop.

[PatRaceTin]

new socket as i expected


amd need to read this haha

[haylui]

AMD needs to have something perform like this

[Lanek]

I will play the devil advocate, but at the moment you can buy an IvyBridge, use it on your actual S1155 board, and then if you need or want it, move to S1150 boards..... I better like this, of just need buy an high end processor + high end motherboard at the same time. here i can buy 1 and then wait for the second. ( without saying, all brand board are not allways there at start ).

At the same time, i remember have change my motherboard each time a new chipset was released with better overclocking capabilities on the Socket S775 times ...

[informal]

So what is Haswell exactly? Is it a brand new x86 core with redesigned pipeline or an IB "evolution" (or to be more precise "Core" evolution)? We know we will get AVX2 with 256bit integer instruction support and FMA3. But what about the core itself?

[tool_462]

So what is Haswell exactly? Is it a brand new x86 core with redesigned pipeline or an IB "evolution" (or to be more precise "Core" evolution)? We know we will get AVX2 with 256bit integer instruction support and FMA3. But what about the core itself?

I'm nowhere near an expert, but from what I've read it is an "evolved" 14 stage pipeline.

[savantu]

So what is Haswell exactly? Is it a brand new x86 core with redesigned pipeline or an IB "evolution" (or to be more precise "Core" evolution)? We know we will get AVX2 with 256bit integer instruction support and FMA3. But what about the core itself?

IB is as related to Core as FX is related to K7...

I'd expect first of all to have true 256bit data paths in the core and not using the int paths to do 256bit/cycle ( while nice and area friendly, it's not going to allow large speedups with avx and you have a warming up latency of around 70 cycles IIRC ).

As for the uarch itself, it's done by the Nehalem team. It should bring some revolutionary stuff.

[TEDY]

geeze wtf is wrong with intel?

[RedBull78]

Wait a min, I think someone should call up Intel and inform them that motherboards are not phones...