WITH THE LAUNCH of Intel's latest round of Xeon processors, further questions must be asked as to where exactly does the rise of Xeon leave Itanium.
The less than enthusiastic takeup of Intel's multi-billion dollar Itanium chips has garnered significant column inches over the years but the steady advance of its Xeon processors has generated a somewhat more muted reception. Perhaps this is because the industry expects Intel to be the dominant player in every market it is involved in. However it seems that now Intel has inadvertently put pressure on its Itanium architecture with the considerable success of Xeon.
Browsing through Intel's press materials for the Xeon E7 processor launch, there are several signs that Intel is pushing its x86 Xeon processor as the Swiss-army knife of server chips. Under the heading "Performance + reliability + security = Intel Xeon processor formula for mission-critical computing", it's obvious that Intel is trying to muscle in on missioin-critical enterprise server territory, and is claiming that Xeon, not Itanium, chips can provide exactly what customers want from large, fault-tolerant servers.
You don't have to go far to notice Intel's second bit of product positioning. One of the highlights, Intel claims, is the Xeon E7's ability to "democratise mission-critical computing by accelerating the migration away from proprietary computing environments". By proprietary environments Intel is of course referring to big-tin servers that have traditionally been based on RISC processors.
Looking at the impressive roster of server vendors supporting Intel's latest Xeon chips, including datacentre vendors such as Bull, Cray, Fujitsu, Hitachi, IBM, NEC, Oracle and Unisys raises the question once again, does Intel really need to bother with Itanium?
Given the recent comments made by Oracle, citing Intel's focus on x86 as the reason for its decision to drop support for Intel's Itanium servers, Intel seemed to add more credence to Oracle's decision with the launch of the Xeon E7. It said, "The days of IT organisations being forced to deploy expensive, closed RISC [reduced instruction set computer] architectures for mission-critical applications are nearing an end".
Although Itanium isn't a RISC chip, Intel spent billions designing it to take on traditional RISC servers, and given that Intel has the majority of enterprise server vendors already on board, it seems even the vendors are looking at a world beyond huge monolithic servers.
As a side note, the reference to RISC chips being used in mission-critical servers is a gross simplification. The RISC versus CISC debate died out decades ago, with many chip engineers saying that these days chips blur the lines of what is RISC or CISC. That said, Intel is using the term RISC to conjure up images of old servers that come with their own branded racks, while its Xeon chips end up in a multitude of servers that are available from both mainstream vendors such as Dell, HP and Supermicro and also the old guard corporate datacentre vendors such as Cray, IBM and Unisys.
Mainframes still have their place in the large scale server market, but with the growth of cloud computing and virtualisation, smaller servers in greater numbers have become the order of the day for many customers. Hugh Jenkins, product business manager at Dell, said that his customers want x86 servers, adding, "10 years ago it was a different story". Dell built its server business off x86 processors, whether they were from Intel or AMD, and Jenkins said that Dell "does not offer monolithic server designs for a good reason".
Jenkins referred to Dell's latest line of mission-critical Poweredge R810, R910 and M910 servers, which all run Intel's latest Xeon E7 chips, saying that the firm expects to have the machines available to customers within 120 days. Given that sort of turnaround, it's not surprising that enterprise customers are looking to machines that can be swapped out in weeks rather than decommissioning a server that probably took months to install.
Alongside an impressive array of hardware vendors, Intel also announced software vendors such as IBM, Microsoft, Oracle, Red Hat, SAP and VMware for its Xeon chips. Three out of those six have publicly dropped support for Intel's Itanium architecture, so Intel's move to position the Xeon as the heir apparent to Intel's Itanium is a shrewd one.
Intel's enterprise channel director Richard George said that cloud computing puts greater demands on the interconnects between servers, an area where Intel is strong. So while Itanium might slowly be put out to pasture, Intel is far from becoming a failure in the wider server market.
The sad thing for Intel is that it will take many years for it to wind down Itanium, regardless of how it markets Xeon processors. However with the cloud paradigm bringing power in numbers, it seems that traditional big tin servers really have seen their day. µ
Itanium was doomed years ago. I am surprised Intel kept it alive as long they as it did.
Give Intel their due...ever since their Core and Core 2 series appeared, they seem to be back in their groove.
The new Sandy Bridge-based Xeons are great server CPUs, from all early accounts.
Itanium will not be lamented.
And on the RISC side, I wouldn't pronouce that architecture dead just yet. Especially ARM. A bit off topic, but according too the article linked below, the ARM "niche" is rather large.
"There are more than 20 billion ARM-based processors in the world and a huge number are inside mobile phones."
The current price/performance crown fits squarely on the ARM head.
Read more: http://www.theinquirer.net/inquirer/news/1929158/arm-celebrates#ixzz1IxsfzzZU
The Inquirer - Computer hardware news and downloads. Visit the download store today.
its simple, the 1st answer that caused EPIC to fail was when memory speeds slowed wayyy down compared to cpu speeds, EPIC needs x2 the bandwidth of x86 for the same instructions, so already its crippled, whereby Intel throws 24MB cache to try and help but the issue is still the same, EPIC was designed when CPU's were slow and memory was comparably fast, when this flipped over, they had a big problem, so at this point its screwed for performance in comparison to CPU's that don't require as much bandwidth.
2nd thing that blew it out of the water was AMD64, without this it might still have made it into the market, this was a disruptive tech that intel didnt see coming and destroyed the plans for itanic.
How does these new 10-Core Xeon Chips compare with IBM Power 7 series?
I understand the installed based of Power 6/Unix, but if you were to buy new systems do you go with IBM Powerpc or Intel?
Intel chips have served the mission-critical computing market since 2001, when Stratus Technologies teamed up with Intel engineering to make x86 work in a fault-tolerant server architecture.
ftServer systems are not only compact at 4U, but cost-effective for critical application support ... for Windows, Linux, VMware and Hyper-V. Love Intel as we do, it takes more than hardware alone to serve this market properly, specifically specialized internal system diagnostics and after-sale, pro-active monitoring and manageement that prevents downtime from occuring.
If Intel has such a huge downer on "proprietary computing environments", then why did it create one?
The 8086 was a 16-bit architecture that built on the firm foundation of the 8-bit 8080; the 386 was a 32-bit architecture that built on the firm foundation of the 8086; so when AMD announced the Opteron, which was a 64-bit architecture that built on the firm foundation of the 386, its not like it came straight out of left field, is it?
Similarly, nearly every other RISC architecture failed as the performance of x86 improved over time and its cost-effectiveness became compelling. MIPS, ARM, SPARC, Alpha, PA - these were all once viable "workstation" chips. Where are they now? Niches (albeit in the case of ARM, fairly wide niches). So when not if Itanic failed, its not like it was a huge surprise to anyone, is it?
In fact, it seems that the only people who *didnt* think Itanic was doomed from day one is Intel itself. So why did they bother? What compelling business case was made for Itanic? Was it, like NetBurst, a triumph of marketing over common sense?
Dont make me wait for Craig Barretts memoirs to find out!