Sign in with
Sign up | Sign in
Your question

Two New Sockets For Nehalem in 2008

Last response: in CPUs
Share
October 19, 2006 8:42:42 PM

http://www.vr-zone.com/?i=4182
Nehalem adds IMC+CSI coming most likely Q308/Q408 with two new sockets.

Socket B will be LGA1366.
Socket H will be LGA715.

More pins on Socket B because of the IMC. Not sure about anything else.

More about : sockets nehalem 2008

October 19, 2006 8:44:13 PM

Quote:
http://www.vr-zone.com/?i=4182
Nehalem adds IMC+CSI coming most likely Q308/Q408 with two new sockets.

Socket B will be LGA1366. This one has IMC.
Socket H will be LGA715. Maybe for desktop?


LGA1366 8O

LGA715...... with CSI?
Anonymous
a b à CPUs
October 19, 2006 10:17:56 PM

Would this mean that the CSI interconect will require less pins then the FSB? I guess it does make sens since some of the goal are higher bandwith per pin and so one.

I dont know if one will be aimed at desktop and the other at server, if not, it could me two tottaly different platform, one for performance and one for "celeron" like.

This would actually be in line with Intel trying to become a platform compagny and not a CPU cie.

Centrione, vPro, ViiV, etc
Related resources
a b à CPUs
October 19, 2006 10:23:28 PM

Quote:
Would this mean that the CSI interconect will require less pins then the FSB? I guess it does make sens since some of the goal are higher bandwith per pin and so one.

I dont know if one will be aimed at desktop and the other at server, if not, it could me two tottaly different platform, one for performance and one for "celeron" like.

This would actually be in line with Intel trying to become a platform compagny and not a CPU cie.

Centrione, vPro, ViiV, etc


CSI = Serial
FSB = Parallel

Yes... less pins..lol.
Anonymous
a b à CPUs
October 19, 2006 10:54:59 PM

Heh, that's what I figured but information on CSI is scarce and Im lazy, so I couldn't really verify that, mainly because of the lazyness and I know someone like you will enlighten me =).

:lol:  Is that why Parralel port are huge and Usb small/Noob :lol: 
a b à CPUs
October 19, 2006 11:08:53 PM

Quote:
Heh, that's what I figured but information on CSI is scarce and Im lazy, so I couldn't really verify that, mainly because of the lazyness and I know someone like you will enlighten me =).

:lol:  Is that why Parralel port are huge and Usb small/Noob :lol: 


Most companies have been switching the channels between the computer and disk drives from parallel to serial because of synchronization difficulties. Although it would seem that a parallel cable with multiple channels for data would always have a faster data transfer rate than a single data channel, keeping the bits aligned in a parallel channel requires more complex electronics. Signals on multiple data channels can arrive at the receiving side at different times and must be aligned in order to turn the bits into meaningful bytes. It also costs less to design a faster serial channel than to create and build the necessary circuitry to keep faster parallel channels properly aligned.
October 20, 2006 12:21:43 AM

How true. That's why SATA is better than ATA. And why for a while Hypertransport was better than FSB. Somehow, Intel found a way around that, but it seems they'll be going serial after all. When you think about it, it's much more logical. You can put several send/receive channels to increase the bandwidth. Bwahaha.
October 22, 2006 9:03:52 PM

Could it be that Socket H is not designed for Nehalem at all, but rather the notebook segment?

As I recall those are still working on Socket M (479-pin). That would be a good upgrade, not to mention that it makes a lot more sense.
October 22, 2006 11:36:50 PM

Santa Rosa in Q2 2007 will use Socket P. Still 479-pins just rearranged to prevent 800MHz FSB Meroms from working on 945 mobile chipsets. Incidentally, it also prevents companies from skimping and trying to use 667MHz FSB Meroms in Santa Rosa so companies will have to clear inventory before then.
October 23, 2006 12:59:34 AM

Quote:
Could it be that Socket H is not designed for Nehalem at all, but rather the notebook segment?

As I recall those are still working on Socket M (479-pin). That would be a good upgrade, not to mention that it makes a lot more sense.


I think Socket H is for server chips.
Multiple CSI and IMC is useful in server segment.
a b à CPUs
October 23, 2006 2:10:39 AM

heh and to think this new CSI might give out conroe just as greater boost as AMD got from K7 -> K8 (yes fanboys K7 was slower then P4, K8 was faster then P4, P4 WAS good, IN ITS DAY - lets not forget the Northwood, especially the Northwood C), even if AMD gets out a better design to beat conroe, this will put Intel in the lead again perhaps?
October 23, 2006 5:23:27 AM

Santa Rosa will be a platform upgrade for the existing Core 2 architecture, along with Socket P in H1 2007.

Socket H (715-pin) as well as Socket B (1366-pin) will be released along with Intel's new micro-architecture (to replace Core 2) in H2 2008, as stated in the original post (as well as some Intel's roadmaps).
October 23, 2006 5:28:32 AM

If, as you said, CSI and IMC are more useful in the server segment, I think Socket B with its larger pin count would be a better fit.
October 23, 2006 8:53:12 AM

Quote:
heh and to think this new CSI might give out conroe just as greater boost as AMD got from K7 -> K8 (yes fanboys K7 was slower then P4, K8 was faster then P4, P4 WAS good, IN ITS DAY - lets not forget the Northwood, especially the Northwood C), even if AMD gets out a better design to beat conroe, this will put Intel in the lead again perhaps?


CSI is only useful for multi-processor environment, especially for 4-way and up.
a b à CPUs
October 23, 2006 9:53:28 PM

Quote:
heh and to think this new CSI might give out conroe just as greater boost as AMD got from K7 -> K8 (yes fanboys K7 was slower then P4, K8 was faster then P4, P4 WAS good, IN ITS DAY - lets not forget the Northwood, especially the Northwood C), even if AMD gets out a better design to beat conroe, this will put Intel in the lead again perhaps?


CSI is only useful for multi-processor environment, especially for 4-way and up.

Hence why K8/Opteron was first (before coming to desktops)?
October 23, 2006 10:05:04 PM

Quote:
heh and to think this new CSI might give out conroe just as greater boost as AMD got from K7 -> K8 (yes fanboys K7 was slower then P4, K8 was faster then P4, P4 WAS good, IN ITS DAY - lets not forget the Northwood, especially the Northwood C), even if AMD gets out a better design to beat conroe, this will put Intel in the lead again perhaps?


CSI is only useful for multi-processor environment, especially for 4-way and up.

Hence why K8/Opteron was first (before coming to desktops)?

Maybe the sequency of release only :wink:

The only advantage of HTT over FSB for single processor is the easier layout of motherboard. The bandwidth provided by HTT / FSB is more than sufficient for K8.
October 23, 2006 10:29:37 PM

I'm waiting for LGA1337. :tongue:
October 24, 2006 12:57:30 AM

Have fun with that, the waiting game.
October 24, 2006 1:24:47 AM

Wasn't that the pin count people speculated would be released with AM2, or to be released with K8L (not LGA of course)?
October 24, 2006 1:55:12 AM

Ok I think there is enough CSI already.

CSI: Las Vegas, CSI: Miami, CSI: New York... :twisted:
October 24, 2006 1:57:07 AM

You actually watch that? Sorry but I can't watch it. I prefer NUMB3RS.
January 7, 2007 10:33:16 PM

Quote:
You actually watch that? Sorry but I can't watch it. I prefer NUMB3RS.


I cant help but comment this. Numbers is the most ridiculous and lamest (in its true meaning) series i have EVER seen, Sorry but its true. Well... me and my buddies laughed some when we watched one and only episode,

Watch and have fun but just in case - dont think anything they do has anything to do with reality,
January 7, 2007 10:41:50 PM

And whats up with the E in numbers. Since when do they try to advertise l33t script in public television?

Imagine that : N3x7 on your f4vori73 ch4nn31: "Who w4n7s 7o b3 a mi11ion4ir3"

You know the funnt thing is its actually what it is called: numb-three-rs.
Just google numb3rs. Paranoia.
January 7, 2007 10:43:03 PM

Lol, yeah... I don't really watch NUMB3RS... I just kind of watch anime... it takes my mind off of the depressing thing that is our world *faints*
January 7, 2007 11:20:39 PM

lawl. :tongue:
January 7, 2007 11:51:17 PM

Quote:
CSI is only useful for multi-processor environment, especially for 4-way and up.

Really? Would it not also be good in a multi-core environment? I had thought that was at least one reason that Intel is now looking to retire the FSB and move to IMC/CSI. To improve the performance of inter-core communication. Not true?? :?

Intel is thinking about 8 or more cores sharing a single socket around the end of 2008 into 2009 timeframe, no?

-john
January 7, 2007 11:54:55 PM

Yeah... anyways, how many of you guys are going to buy into this whole CSI thing from Intel? Sounds like... that show... that I don't watch.
January 8, 2007 12:11:30 AM

Quote:
CSI is only useful for multi-processor environment, especially for 4-way and up.

Really? Would it not also be good in a multi-core environment? I had thought that was at least one reason that Intel is now looking to retire the FSB and move to IMC/CSI. To improve the performance of inter-core communication. Not true?? :?

Intel is thinking about 8 or more cores sharing a single socket around the end of 2008 into 2009 timeframe, no?

-john

Ya know... it doesn't seem like a bad idea I mean just look at what HTT did for AMD, might just do the same thing for Intel. Then again, we'll just have to wait and see!
a c 102 à CPUs
January 8, 2007 2:41:11 AM

The LGA 1366 chip makes perfect sense if you think about it. AMD has no less than 939 pins on a non-mobile CPU with a dual-channel integrated memory controller and high-speed serial interconnect link. AMD put 1207 pins on the Socket F Opterons, and I assume that they are all there for a reason (can't find a pinout to see for NC's.) I'd expect Intel to be roughly around those numbers, perhaps higher if they intend to keep doing high-watt MCMs like the Kenstfield/Clovertown. They'd need more power pins than the 1207 Opterons would in that case.

I am guessing the LGA 715 needed a rewire of the socket that 771 or 775 could not handle, like AMD did with AM2 from 939. I am guessing that this socket won't have all that long of a life as the IMC version of the CPU ought to have a sizable advantage over the FSB version and people won't want to buy the slower version unless there is some reason that they'll have to (the 1366 chips being only Xeons and taking $$$ ECC RAM, for one.)

And to apache_lives, the K7 was not slower than the P4 in all cases. The P4 Willies plain sucked and got beaten by not only the K7s but PIII Tualatins as well. Yes, the Northwoods were the top-dog, especially the P4C, but they were the only P4 to beat the AMD CPUs. The Williamettes didn't, and the K8 killed the Prescotts and Cedar Mills. And it was not HTT that gave AMD its edge on the desktop with the K8, in fact HTT on the desktop is basically just to replace the functionality that AMD lost by going with an IMC rather than an FSB. Part of it was the IMC and part of it was the widening of the excellent K7 core for the addition of the 64-bit capabilities. However, HTT really played a big deal on the server front and this should help the Xeons regain speed in the MP segment.
a b à CPUs
January 8, 2007 6:52:01 AM

Quote:
heh and to think this new CSI might give out conroe just as greater boost as AMD got from K7 -> K8 (yes fanboys K7 was slower then P4, K8 was faster then P4, P4 WAS good, IN ITS DAY - lets not forget the Northwood, especially the Northwood C), even if AMD gets out a better design to beat conroe, this will put Intel in the lead again perhaps?


CSI is only useful for multi-processor environment, especially for 4-way and up.

We found all this FSB talk was a load of BS when a quad core core 2 series cpu smashed AMD's "scalability" :lol: 
a b à CPUs
January 8, 2007 6:56:48 AM

Quote:
heh and to think this new CSI might give out conroe just as greater boost as AMD got from K7 -> K8 (yes fanboys K7 was slower then P4, K8 was faster then P4, P4 WAS good, IN ITS DAY - lets not forget the Northwood, especially the Northwood C), even if AMD gets out a better design to beat conroe, this will put Intel in the lead again perhaps?


CSI is only useful for multi-processor environment, especially for 4-way and up.

We found all this FSB talk was a load of BS when a quad core core 2 series cpu smashed AMD's "scalability" :lol: 
January 8, 2007 12:38:20 PM

Quote:
heh and to think this new CSI might give out conroe just as greater boost as AMD got from K7 -> K8 (yes fanboys K7 was slower then P4, K8 was faster then P4, P4 WAS good, IN ITS DAY - lets not forget the Northwood, especially the Northwood C), even if AMD gets out a better design to beat conroe, this will put Intel in the lead again perhaps?


CSI is only useful for multi-processor environment, especially for 4-way and up.

We found all this FSB talk was a load of BS when a quad core core 2 series cpu smashed AMD's "scalability" :lol: 

Just scale to 8 cores now :wink:
January 8, 2007 4:24:50 PM

Quote:

Yes Northwood was an awesome chip. Then Intel crapped out the prescott.
Who knows if AMD will ever take the lead again. I dont think thats even a goal of theirs. They are moving into different segments of the market besides cpu's. They got where they are today by being number two. They are content with being number 2. You dont have to be number one to be succesful.


Maybe. But being in the #2 spot affords you far fewer chances to set the course for the industry.
a b à CPUs
January 8, 2007 10:42:32 PM

Quote:
heh and to think this new CSI might give out conroe just as greater boost as AMD got from K7 -> K8 (yes fanboys K7 was slower then P4, K8 was faster then P4, P4 WAS good, IN ITS DAY - lets not forget the Northwood, especially the Northwood C), even if AMD gets out a better design to beat conroe, this will put Intel in the lead again perhaps?


Yes Northwood was an awesome chip. Then Intel crapped out the Prescott.
Who knows if AMD will ever take the lead again. I don't think thats even a goal of theirs. They are moving into different segments of the market besides cpu's. They got where they are today by being number two. They are content with being number 2. You don't have to be number one to be successful.

To me it looks like there getting desperate to get back to number one, and as for trying other new markets to get ontop - take a look at 4x4 and AMD live - both to take on Intel, and there fighting to get Dell.

The only "new" market is there video card market, buying up a company to see it drop in value, and rebadging things. Intel atm has a weak integrated video market and AMD is taking that on, perhaps this will be the only successful idea and result from the buy out of ATi, but atm it seems like the biggest mistake.
a c 102 à CPUs
January 9, 2007 12:18:11 PM

It depends. You are right in the fact that you have a better chance of doing better after trying to ram overpriced and underperforming junk down the throats of consumers (e.g. RAMBUS, FB-DIMMs, the P4 Williamette and Prescott, MS Windows, HD-DVD/Blu-Ray and DRM) if you're the biggest in the market. But you don't have to be the biggest to get *good* tech adopted in the market as the market will want to get it on its own. That's why AMD set the ISA for desktop and server chips for the next 20 years or so and managed to capture about half of the 4-8P server market in 4 years from first entering that market.
January 10, 2007 5:11:02 AM

Quote:
It depends. You are right in the fact that you have a better chance of doing better after trying to ram overpriced and underperforming junk down the throats of consumers (e.g. RAMBUS, FB-DIMMs, the P4 Williamette and Prescott, MS Windows, HD-DVD/Blu-Ray and DRM)


At least FB-DIMMs have an advantage in the market designed for, and yes I am talking about performance. Rambus RAM didn't have any real advantage. FB-DIMM+5000X chipset may be slower than DDR2+975X in PC apps, but take a look at below.

http://www.anandtech.com/storage/showdoc.aspx?i=1799&p=...

WD360 Raptor kicks U320 SCSI's ass thoroughly until the database benchmark, the ones the SCSI drives are meant for.

5000X chipset market segment prefers sustained memory bandwidth over burst that is more useful for PC markets, similar for FB-DIMM. 5000X chipset also has a snoop filter, which is irrevalent for PC users, but the guys getting 5000X systems will.

Put IBM's monster Power 5 to a PC(with a supporting OS, etc), and you'll get a crap system.
a c 102 à CPUs
January 10, 2007 1:01:25 PM

The FB-DIMMs' advantage is not in memory performance at all. It was designed to allow massive *amounts* of RAM on a board. However, I guess if your app needs a ton of RAM, more slower RAM is much better than less faster RAM for performance.
!