Print 23 comment(s) - last by mlau.. on Jun 12 at 2:16 AM

2.6GHz Opterons with land-grid arrays

The Computex 2006 trade show is over (at least for DailyTech) but we still have a few tidbits of information here and there.  Over the past few days we've gotten a good look at some of the AMD Socket F motherboards and chipsets, and today we have a few shots of the CPUs at the DailyTech labs. 

The OSA2218GAA6CQ is a dual-core 2.6GHz AMD Opteron Revision F processor, code named Santa Rosa.  The CPU features DDR2-667, Pacifica virtualization support, Presido security extensions and 1207 land-grid array pins. Like current Intel processors, AMD's Revision F Opterons have the entire pin interconnects on the motherboard, and the processors simply use connected pads to interface with the motherboard.  The processor also features enhanced memory RAS and registered DIMMs.  An AM2 version of the processor dubbed Santa Ana is expected later this year as well.

The general consensus from manufacturers is that the July 11 embargo date that was originally set by AMD public relations may be overly optimistic.  The motherboards and CPUs appear ready, so the root of this speculation may revolve around the actual channel launch date and not the OEM availability.

AMD is expected to get at least one more revision out of the LGA-1207 Socket F; the Deerhound quad-core Revision G processor is also slated to use the LGA-1207 socket.  After Deerhound, AMD roadmaps show plans for quad-core K8L processors using HyperTransport 3.0, but these same roadmaps have not confirmed which socket interface the processors will use.

Comments     Threshold

This article is over a month old, voting and posting comments is disabled

By PrinceGaz on 6/10/2006 11:43:08 AM , Rating: 2
What a lot of pins! Will the ever increasing number of pins ever end? At this rate we'll be approaching two thousand pins by 2010!

RE: Whew...
By TheDoc9 on 6/10/2006 2:20:11 PM , Rating: 2
One day the main CPU will almost literally be the size of the motherboard. Containing dozens of general and specialized cores, running at 30 GHZ+, and slide in behind the motherboard. They'll work with 'sheets' of processors like a high end server of today, only dozens of times faster.

can't wait

RE: Whew...
By biohzrd on 6/10/2006 7:06:25 PM , Rating: 2
I hardly think so. The amount of pins may increase, but the size of the cores are exponentially decreasing. We have two cores running incredibly faster on a processor that is much smaller than we did 10 years ago. My prediction is that there will have to be a new kind of interface developed, or a central processor that feeds instructions to the rest of the cores, which would require less pins. Although, there would inherently be some lag associated with that idea, so who knows.

/can only hope CPU's dont become the size of a motherboard.

RE: Whew...
By livinloud on 6/10/2006 10:16:00 PM , Rating: 2
Intel is actually planning to build a CPU that as it's RAM direcly over the DIE of the CPU. Just Like a L3 cache but with 1 or 2 go or more maybe. So closer is the CPU to the better it is, so putting it just on top is a real good idea. For people who goes Intel Channel Conference you will what i'm talking about. It's like the new FB-DIMM ... something new on the way for Intel Server Platform.

RE: Whew...
By tygrus on 6/11/2006 10:55:41 PM , Rating: 2
Eventually the CPU will become less dominant and will be integrated with memory chips or modules and not the other way around. Computing is becoming more reliant on data storage and transmission.
It's been discussed before for putting logic for data manipulation (array/matrix) or search into RAM modules to boost system performance.
It's like having a IBM (et. al.) Cell SPE processor in each RAM chip and the system processor doing the supervising.

RE: Whew...
By shadowzz on 6/11/2006 3:40:11 AM , Rating: 2
or a central processor that feeds instructions to the rest of the cores,

Don't give Sony any ideas! Oh wait, that's right, I forgot about Cell.

RE: Whew...
By Goolic on 6/11/2006 12:12:02 PM , Rating: 2
Cell is IBM project not sony, sony don't have as nearly as sufficient expertise in processor thecnics that are involved in Cell - like systems...

And in my mind Cell is a completely new architeture that can modify our computing as long as programs are adapated to run optimally on them !

Sun's Niagra (don't remenber the name of the lunched product) has some of the principles of Cell, For servers it's simply astouding ! Incledible performance (equivalento to 2 opterons) in one package with lower power usager than a single one opteron.

Forguive my english, it sucks !

RE: Whew...
By tygrus on 6/11/2006 11:00:26 PM , Rating: 2
Sun's Niagra is still not as good as it could be. Some of the benchmarks still favour the latest 2S AMD Opterons and the Intel Woodcrest Xeon's but the Sun chip is still up there. The next iteration of Niagra will be considerably better and by then the others will improve as well, the competition will continue.

RE: Whew...
By mlau on 6/12/2006 2:16:02 AM , Rating: 2
Niagara only really shines if you give it _lots_ of concurrent work to do, e.g. 32 threads of Apache serving webpages. Then it outperforms a higher clocked opteron by a large margin.
It is not designed for raw integer performance like the x86 line.

DDR2 667 Support ?
By KHysiek on 6/10/2006 5:48:14 AM , Rating: 2
Really impressive.


This is some joke, introducing definitely slower (than current gen.) server processor whwne Intel starts selling somethin definitely faster (than S940 Opterons).

RE: DDR2 667 Support ?
By Xeeros on 6/10/2006 6:40:58 AM , Rating: 2
Yeah Intel went from laughable to actually making something good although the new naming scheme is still not my favorite.

I think AMD is just hoping to hold onto the crown just long enough to gain a few more points in their marketshare. Intel's new Core 2 series is looking good and AMD better have some big plans for K8L. Although if the Core 2 Duo series does hold up its benchmarks that ive seen so far then maybe Dell will have a great year ahead of them

RE: DDR2 667 Support ?
By MartinT on 6/10/2006 6:41:44 AM , Rating: 2
Exactly my thought, DDR2-667 will hurt them at the worst time possible.

RE: DDR2 667 Support ?
By psychobriggsy on 6/10/2006 7:42:58 AM , Rating: 2
Right, so can you find registered ECC DDR2-800 DIMMs for sale?

RE: DDR2 667 Support ?
By SexyK on 6/10/2006 10:42:13 AM , Rating: 1
Right, so can you find registered ECC DDR2-800 DIMMs for sale?

RE: DDR2 667 Support ?
By PrinceGaz on 6/10/2006 11:41:57 AM , Rating: 2
They are unbuffered, not registered modules.

RE: DDR2 667 Support ?
By Warren21 on 6/10/2006 2:40:03 PM , Rating: 2
Psychobriggsy is right. As of yet the only FB-DIMM memory you can find is DDR2 533 (PC2-4200) and 667 (PC2-5400), though even THAT (667) is hard to come by, so until the memory makers can get the act together, this isn't too much of a problem (as of yet)

I'm confused
By Poximex on 6/9/2006 8:48:21 PM , Rating: 2
I'm confused...

So if the LGA-1207 Socket F is a new socket for AMD, why would they only be able to get one more revision, G, out of it. It seems like a new socket would have a much longer lifetime than that.

RE: I'm confused
By KristopherKubicki on 6/9/2006 8:49:38 PM , Rating: 2
That is just the only one they have confirmed. But it would seem improbable that engineers could change the HyperTransport protocol and keep the same socket for K8L.

RE: I'm confused
By tonjohn on 6/9/2006 10:16:11 PM , Rating: 2
AMD won't confirm such info at this time.

But one of the big reasons behind Socket F is that AMD was planning for the future. There is no doubt in my mind that K8L will be fully supported by Socket F.

However, things on the desktop market don't look so bright as it looks like AM2 probably won't support K8L, atleast not quadcore K8Ls.

RE: I'm confused
By breethon on 6/9/2006 10:57:57 PM , Rating: 2
It seems like a new socket would have a much longer lifetime than that.

Yeah....just look how long Socket 939 lasted.....I should have saved my money or spent more time reading....*SIGH*

socket life is over-valued
By proamerica on 6/10/2006 12:00:36 AM , Rating: 4
Socket life is over-valued. Typically if you wait so long that you are going to actually benefit in a substantial way by spending hundreds of dollars on a new CPU then you probably want to get a new motherboard anyway. Why, oh why some of you ask. Well the reason is because a new motherboard is going to maximize the value of your CPU. Throwing a state of the art CPU into an old motherboard is a waste.

RE: socket life is over-valued
By quickk on 6/10/2006 12:11:54 PM , Rating: 2
I don't think that socket life is overvalued. Way back in 1998 I bought an AMD BX6 motherboard. At that time the fastest processor was the pentium 2 450 MHz, which sold for over $1000 (cdn). I got the celeron 300A and overclocked it to 450+ and saved lots of moolah. A few years later, I upgraded the cpu to a celeron 2 566 with the use of a slotket adapter. After a nice 50% overclock, my processor was once again amongst the fastest available.

A few more years went by and I eventually replaced the aging celeron2 with a nice Tualatin celeron 1200MHz. I overclocked it to 1455MHz, and it still runs just fine. My in-laws now use this system as their primary computer with the motherboard now being about 9 years old!

My point here is that I went through 3 cpu generations while still using the same motherboard. I would be nice if we could still do the same thing today, even if it's through the use of some adapter.

RE: socket life is over-valued
By quickk on 6/10/2006 12:15:41 PM , Rating: 2
Throwing a state of the art CPU into an old motherboard is a waste.

While you might not get the same increase in performance you would get by also buying a new motherboard, you can still get a very significant performance increase. Sometimes you just don't have the money to upgrade everything.

"It's okay. The scenarios aren't that clear. But it's good looking. [Steve Jobs] does good design, and [the iPad] is absolutely a good example of that." -- Bill Gates on the Apple iPad

Copyright 2016 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki