Core i7-4770K

good answer but their have been changes in the architecture of just about every CPU, yet pin count tended to increase over time as CPUs became more capable, and even withing a single company higher end products always (that I am aware of) come with more pins. Yet Intel has seemingly stagnated pin counts have gone down the last 3 sockets. (since they are so close we can assume they are actually in need of no different pins just want to change them for naming sake, or even just to sell us new motherboards. )
 
that has been the strategy of intel from the very 1st place unlike amd.
and this may well be too,but from nehalem to sandy there was a no of changes,so i guess socket was changed.sandy to ivy not a big change.hence same socket.haswell is rumoured to have quite a large no of improvements over ivy.
 
I think pin counts may be more related to power requirements (efficiency) than ability, and Haswell has been designed to be as low-power and efficient as possible. I did a quick comparison between pin counts and max TDP going back a ways for the Desktop Core i7.

Ivy Bridge --> 1155 pins --> 77W
Sandy Bridge --> 1155 pins --> 95W
Sandy Bridge-E --> 2011 pins --> 130W
Gulftown --> 1366 pins --> 130W
Bloomfield --> 1366 pins --> 130W
Lynnfield --> 1156 pins --> 95W

A pin here or there is probably a rebalance of power needs versus data bandwidth if I was guessing.

Via http://en.wikipedia.org/wiki/Intel_Core#Core_i7 (Nahelem core) and http://en.wikipedia.org/wiki/Intel_Core#Core_i7_2 (Sandy Bridge core)
 
You might be correct/May be onboard vrm stuff has anything to do with it.
 
yea.. the pin count being related to power.. and greater efficiency.. is most likely...and better usage of pins..
 
There are really three reasons for pin (or contact or ball or whatever) count increases

1: higher current requirements
2: requirement for more/faster interfaces
3: higher integration moving interfaces that were previously in the chipset into the CPU

Afaict for mainstream desktop stuff the jump in pins from 478 to 775 pins was caused by increasing power requirements (and probablly the fear that they would have to increase power requirements even move in future) and then jump from 775 pins to 1156 pins was caused by moving from a FSB/northbridge architecture to integrated PCIe and memory controller.

Since then current consumption has stayed pretty static, interfaces have stayed pretty much the same (16 lanes of PCIe and dual channel memory) and the integration level has stayed pretty much the same (memory controller, graphics and fast PCIe integrated, the rest on the chipset). So afaict there just isn't any need for more pins.

With high end deskop/midrange 1-2 socket server parts the pin counts jumped to 1366 when intel replaced the front side bus with two QPI links and a triple channel memory controller and jumped again to 2011 when intel added an extra memory channel and a shedload of onchip PCIe (40 lanes of PCIe 3.0, and four lanes that can be either DMI or PCIe 2.0). They also produced a cut down version with only 24 PCIe lanes, three ram channels and one QPI link that has 1356 pins but afaict it didn't really take off.
 
Socket 1150 is basically because this would be the first SOC chip and as mentioned built-in VRM and some more changes, but mostly because broadwell (14nm) would later on use the same socket, so it's almost like future-proofing it from now only.
 
Grss429 said:
sucks, I hate everytime a new CPU comes out you have to buy new mobo
Click to expand...

You dont have to buy a new CPU everyime a new one comes out. Called future proofing. It will be a loooooooooooooooooooooooooooooooooooooooooong time until something can truly best my 3930K.
 
tangoseal said:
You dont have to buy a new CPU everyime a new one comes out. Called future proofing. It will be a loooooooooooooooooooooooooooooooooooooooooong time until something can truly best my 3930K.
Click to expand...

That's what everyone thought about the i7 920s.
 
tangoseal said:
You dont have to buy a new CPU everyime a new one comes out. Called future proofing. It will be a loooooooooooooooooooooooooooooooooooooooooong time until something can truly best my 3930K.
Click to expand...

IB-E will be a little faster then in 18 months or so when they release the socket 2011 version of Haswell it will will really get beat. Same story as the 980X and the 3960X.
 
Illyria89 said:
Socket 1150 is basically because this would be the first SOC chip and as mentioned built-in VRM and some more changes, but mostly because broadwell (14nm) would later on use the same socket, so it's almost like future-proofing it from now only.
Click to expand...

Haswell will not be a true SOC, there still is a separate southbridge.

As for the number of pins... a 5 pin difference is not a significant change. The pin change is probably to help avoid confusion between what works with what, although Haswell-E is supposed to use 2011 pins.
 
Mike_Lewis89 said:
That's what everyone thought about the i7 920s.
Click to expand...

Depends on what is being done with said processor. For the 99% of home-users out there doing nothing but Facebook, email, and web-surfing for reviews of as-seen-on-tv crap products, the i7-920 is still vastly overkill.
 
Well, I have the i7 860, and it chews through everything I throw at it in regard to games and such. (which is what I use my rig for mostly, apart from web surfing and other tidbits). I might consider Haswell though. But it's not like I -really- need an upgrade atm. Then again, I've sort of departed from the enthustiast path and my requirements aren't as big as they used to be.
 
I have a 920 and 930, sure they are nice but the newer cpus OC much better and surpass them, skipping this socket is a good idea but I think jumping in on haswell is the right move. It gets you in on the socket early and will give you one solid upgrade.

I have always tried to buy in early on any socket, not only does it let you have an upgrade it means if you upgrade any other computer it is likely to fall on the same socket and that is good because it gives you more options for swapping parts around.
 
Nitrogen said:
I think pin counts may be more related to power requirements (efficiency) than ability, and Haswell has been designed to be as low-power and efficient as possible. I did a quick comparison between pin counts and max TDP going back a ways for the Desktop Core i7.

Ivy Bridge --> 1155 pins --> 77W
Sandy Bridge --> 1155 pins --> 95W
Sandy Bridge-E --> 2011 pins --> 130W
Gulftown --> 1366 pins --> 130W
Bloomfield --> 1366 pins --> 130W
Lynnfield --> 1156 pins --> 95W

A pin here or there is probably a rebalance of power needs versus data bandwidth if I was guessing.

Via http://en.wikipedia.org/wiki/Intel_Core#Core_i7 (Nahelem core) and http://en.wikipedia.org/wiki/Intel_Core#Core_i7_2 (Sandy Bridge core)
Click to expand...

sandy bridge-e does 150 watts with the 2687W

http://ark.intel.com/products/series/61422
 
Tsumi said:
Haswell will not be a true SOC, there still is a separate southbridge.

As for the number of pins... a 5 pin difference is not a significant change. The pin change is probably to help avoid confusion between what works with what, although Haswell-E is supposed to use 2011 pins.
Click to expand...

I know that and broadwell would be a 'true' SOC, but Haswell, in other terms, would still be it's first SOC.

As far as the pin count goes, as far as I read, they said that Ivy Bridge-E would be coming out next year on the same socket of 2011, really not sure about Haswell.
 
Broadwell "true" SOC are still just rumors, and if that does happen, it will break all compatibility with Haswell boards. I doubt Intel will make their boards obsolete after only one generation, but it might happen. And if it does happen, we'll probably see Broadwell as BGA only CPUs as well.

IB-E will be on the same SB-E 2011 socket. Haswell-E will also use a 2011 pin socket, but the pinout will be completely different from SB-E pinout, rendering it completely incompatible. The integrated VRMs alone ensures that it will be a different socket.
 
Pin counts generally change because power or signal requirements change. Intel went from 1156 to 1155 because Sandy Bridge required signals that weren't provided on the 1156 platform. They could have used some of the reserved pins (if there are any, I haven't looked), but then they'd run into the problems they had on 775 when moving from P4/D to Core 2, or again when they moved from the Conroe to Wolfdale and their associated quads. There were many boards that could not work with the newer processors despite being the same socket. If they're going to have pissed off customers no matter what they do, I'm sure they'd much rather them be pissed off about knowing they'll have to buy new equipment up front over them being pissed off and thinking they've been deceieved.


Besides, AMD has been doing the same thing since 2003. Sockets 940, 939, AM2, AM2+, AM3, and AM3+ around 940 pins, but are gerneally incompatible with each other. Even parts that are backwards compatible lose newer features when mated with older parts.
 
I have a computer in my garage a pentium 3 with 512mb ram and a nvidia geforce 3 with a 40gb hd and windows 7 on it. Can browse the interent just fine and even run most modern games at 1152x868 res :)
 
noid370 said:
Wow! I give up. I cant keep up with these processors no more.
Click to expand...

For higher-end enthusiast/gaming, there's only a few you need to be concerned with, imo:

Intel: i5/i7 K series
AMD: FX series
 
You must log in or register to reply here.
Back
Top