Processors

Intel's new processor an upgrade over Itanium predecessor

Last week, Intel began shipping a new processor called the Itanium 9300 based on its 65 nm manufacturing process. This is a significant upgrade over its Itanium predecessor.

Last week, Intel began shipping a new processor based on its 65 nm manufacturing process. Though it will not start showing up in the market for up to 90 days, it is a significant upgrade over its Itanium predecessor; it sports four cores and eight threads per core and operates best with DDR3 memory. Formerly dubbed Tukwila, the Itanium 9300 processor uses a socket design that will remain unchanged for upcoming Itanium versions, including one code-named Poulson that will be manufactured using Intel's new 32 nm manufacturing process.

Intel's new offering was announced on the same day that IBM announced the release of its new Power7 chip, which is destined for a higher-end market. Intel has been experiencing growth in its server processor sales lately, and some of this growth appears to be due to the slow but steady increase in cloud computing. Companies with large data centers like those common to cloud vendors continually upgrade servers to improve their levels of service and much of the benefit of these upgrades is going to Intel.

Intel is also rebranding one of its high-end desktop chips for the server market. The Gulftown Core i7 980X will be known simply as Westmere EP and will feature up to six cores with 12 threads each at 3.33 GHz. In addition, Intel is planning some big things for its 32 nm manufacturing process including a chip code named "Sandy Bridge," which will have an all-in-one core that includes the CPU, graphics, and memory controller. The 32 nm process will also reduce the Thermal Design Power (TDP) to as low as 65W for a chip with graphics on the processor. Production of the 32 nm chips isn't expected until the first quarter of 2011, but it could be a boon to customers who want a smaller thermal and power consumption footprint.

Intel has had a spate of bad luck recently. Problems with older versions of its Itanium chip caused blue screens in Windows operating systems and have led to the company losing market share to rival AMD. Manufacturing problems delayed shipment of Tukwila, with stories of the delays going back more than a year. Though the microprocessor giant probably isn't losing much server market share to AMD, delays in its new chip lines might lead customers to consider some of the new offerings from IBM and Sun.

I think that some cracks are starting to show in Moore's Law, which states that the number of transistors that can be economically put on a chip will double about every two years. Intel's difficulty in ironing out the problems with its Itanium line may be a sign that the pace of progress may slow some. But writers have been talking about the end of Moore's Law for more than a decade, so maybe I am just the latest one to make that observation.

TechRepublic's Servers and Storage newsletter, delivered on Monday and Wednesday, offers tips that will help you manage and optimize your data center. Automatically sign up today!

16 comments
ian.obrien
ian.obrien

I thought Itanium was a VLIW (Very Long Instruction Word) processor designed to over come the RISC limitations

Andy J. Moon
Andy J. Moon

I love that technology keeps changing and evolving, it fits my personality and even the description of my astrological sign (Gemini) says that I deal with change well. However, it just seems to me that at some point, we won't be able to keep putting more and more transistors onto a wafer of silicon. This isn't my specific area of expertise, so maybe I don't know what I'm talking about, but at some point the physical limitations of the silicon seem like they will render this particular "Law" obsolete. The fact that it took Intel so long to iron out the problems with this particular processor makes me think that we may be approaching the limits of this particular area of computing. What do you think, am I making too many assumptions or is Moore's Law finally in danger of slowing down?

9naomi9
9naomi9

Itanium is EPIC. A hybrid between VLIW and RISC. More dynamic than VLIW but less than RISC. EPIC is the most static you can be without having to recompile everytime you change the CPU design.

marv13
marv13

As "nano technology" evolves, there will come a point when the laws of classical Newtonian Physics fail, and Quantum Mechanics rears its head. What to do when your little particle decides to behave like a wave? hahaha However, I believe that the transistor as a high speed switching device will become obsolete before Moore's Law fails.

Dan.Finley
Dan.Finley

I'm glad your all excited about the Power 7 release but I think you going a little far. What are the problems you speak of regarding the 9300 series?

Andy J. Moon
Andy J. Moon

Sorry, I should have dug a little deeper to ascertain the chip architecture. I will try to get a correction.

seanferd
seanferd

Already long a requirement in creating chips. Not sure what Newtonian mechanics are all that useful in designing anything to do with microprocessors.

Andy J. Moon
Andy J. Moon

...and haven't had kool-aid in years. ;) The problems I referred to are the ones that kept the new processor from shipping in 2008, as was originally expected (per the second to last link in my OP, which talked about this chip being delayed and was published in February of 2009). I don't know why you think I am so giddy about Power7, it got a mention in a single sentence in my article while the rest talked about different Intel offerings.

rmerchberger
rmerchberger

... and still might for imbedded purposes; but don't quote me on that part. http://www.intel.com/design/i960/i960_linecards.htm As I understand it, people who've worked directly with the i960 line of processors tend to think they were the best to come from the minds of Intel... but I'm a Moto guy myself, having started designing with the 6809. I mean, can you really go wrong with a processor that has a SEX instruction? ;-) Laterz, "Merch"

seanferd
seanferd

Since Intel is always bashing RISC, and I have seen no mention of it in reference to this processor. They have made at least one in the past, though.

Dan.Finley
Dan.Finley

The 9300 is not even in general release and you say users have problems with them, can you explain that? Also the linked article states that Intel moved from FBDIMM to DDR3. How is that to be considered a manufacturing problem?

seanferd
seanferd

Or do a search, whatever. Intel had problems with the processor. So did users.

Dan.Finley
Dan.Finley

Is the "problem" you speak of the switch from FBDIMM to DDR3 memory architecture?