News Posts matching "QuickPath Interconnect"

Return to Keyword Browsing

Haswell-ULT Processors Could Use 24 MHz BClk, New C-States, and MCM to Cut Power Draw

Going into 2013, Intel's tough balancing act between keeping a low power/thermal envelope, and advancing performance, all while staying on the 22 nm silicon fab process, will be care of its Core "Haswell-ULT" processor. The chip will feature some radical changes to traditional Intel processor design, which will help it achieve its design goals. According to a deck of leaked slides scored by Expreview, Intel plans to use additional C-states that drop the processor's base clock, and redesign the processor package to accommodate the PCH silicon, reducing the board footprint.

To begin with, Haswell-ULT will be designed to support 24 MHz base clock speed, which running in "deep" energy-saving idle states. Modern processors with FSB replacement interconnect technologies such as QuickPath Interconnect and HyperTransport need a base clock to time other components on the processor, and for low-level communications, while a bulk of the data is transported by the primary interconnect. Intel found a way to turn off the 100 MHz base clock signal (which is also used to time the PCI-Express root complex and integrated graphics core), and replace it with a 24 MHz clock, when the processor is idling. As the processor returns to lower (more active) C-states, the 100 MHz base clock is reapplied. The 24 MHz base clock is activated by three new power states, C8, C9, and C10, introduced by Haswell-ULT. The third slide below details what happens to the various components in the new C-states.

Intel Xeon E5-2690 Sandy Bridge-EP Performance Leaked

We sourced a presentation, allegedly by Intel, detailing its Sandy Bridge-EP platform, and giving out early performance figures. After successful and trouble-free launches of its Sandy Bridge architecture across all PC form-factors, including the recently-launched Core i7 "Sandy Bridge-E" HEDT (high-end desktop), Intel is taking the architecture to its ultimate market, enterprise, where processors derived from it will make up new lines of Intel Xeon processor families. Intel has two branches of enterprise variations the architecture, Sandy Bridge-EN, designed for high-density, low-power servers, and Sandy Bridge-EP, designed for high-performance servers and workstations. Sandy Bridge-EP is multi-socket capable.

Sandy Bridge-EP uses essentially the same piece of silicon as Sandy Bridge-E, but enabled with several of its features otherwise off limits to the Core processor family. These include twoQuickPath Interconnect (QPI) links, which facilitate high-bandwidth inter-socket communication in multi-socket systems, up to eight cores, sixteen threads enabled by HyperThreading, and up to 20 MB of L3 cache memory. Like its Core family cousins, Xeon Sandy Bridge-EP packs a quad-channel DDR3 integrated memory controller, and PC3-12800 (DDR3-1600 MHz) is its optimal memory standard, but unlike it, supporting up to 768 GB of memory (by two sockets, eight DDR3 channels in all, LRDIMMs). Other key features are listed in the first slide below.
Performance figures follow.

Intel to Bid Farewell to LGA1366 with Core i7-995X Extreme Edition

Intel is preparing its next high-end/enthusiast desktop platform for release in Q4 2011. That platform will be driven by a new socket, the LGA2011, and the new Intel X79 Express chipset. Before that, Intel will give its 2-year old current enthusiast platform, the LGA1366 and X58 Express, a fitting farewell with a new high-end processor model, the Core i7-995X Extreme Edition. The i7-995X is a six-core processor based on the 32 nm "Gulftown" silicon. It features a default clock speed of 3.60 GHz (27 x 133 MHz), 3.86 GHz max Turbo, though the BClk multiplier is unlocked to help with overclocking.

The six x86-64 cores are aided by Intel HyperThreading technology to give the OS a total of 12 logical CPUs (threads) to deal with. The processor features a triple-channel (192-bit wide) DDR3 memory controller, and connects to the X58 chipset over a 6.4 GT/s QuickPath Interconnect link. Intel will launch its new chip some time in Q3 2011, i7-995X is expected to be priced at $999 (in 1000-unit tray quantities), displacing the current SKU at this price, the Core i7-990X Extreme Edition. For now, a lucky few (read: "industry partners") have access to engineering samples.

Source: DonanimHaber

Eurocom Announces the First Server-Class Notebook with 6-core Intel Xeon Processors

Eurocom Corporation, the world's leading developer of highly personalized, high-performance notebook PCs and energy efficient All-in-One PCs, is now supporting the groundbreaking Intel Xeon X5690 Processor in its ine of Mobile Server and Mobile Workstation notebooks. The EUROCOM Mobile Server Technology significantly reduces the time and cost of deploying short-term LAN based solutions and of extending usage of existing business applications to users wherever is required.

Eurocom’s Panther line of high performance Mobile Workstations and Mobile Servers will be bolstered by the introduction of the Intel XEON X5690 Processor. Eurocom Mobile Workstations and Mobile Servers are ideal for the professional who frequently travels yet needs access to high performance computing. All Eurocom systems are designed to be fully configured with many optional components to choose from to allow each customer the opportunity to customize a system to fit their needs exactly.

Intel Readies New Xeon X5690 Processor, Fastest Six-Core

Intel is readying its fastest six-core processors ever. While the company is working to introduce the Core i7 990X Extreme Edition for launch sometime in Q4 2010, it's already begun sampling the Xeon X5690. The X5690 is a dual-socket ready LGA1366 processor based on the 32 nm Westmere-EP silicon. It carries a nominal clock-speed of 3.46 GHz (26.0 x 133 MHz), with a Turbo Boost speed of 3.60 GHz (27.0 x 133 MHz). It features a QuickPath Interconnect speed of 6.4 GT/s, and supports triple-channel registered DDR3 memory with its integrated memory controller. Cache levels include 256 KB L2 cache per core, and 12 MB shared L3 cache.

The processor is compatible with most server boards based on the Intel 5500 and 5520 chipsets with BIOS updates. A quick overclocking feat by Fitseries3 on the EVGA Classified SR-2 motherboard yielded a speed of 5 GHz making use of the high BClk multiplier. The chip is likely to carry a rated TDP of 130W. Intel may release this processor in the next quarter, either displacing the X5680 from its US $1,730 price point, or occupying a higher one. The CPU-Z validation of the mentioned overclocking feat can be found here.

EVGA Unleashes Classified SR-2 Motherboard and Power Supply

EVGA just released two of its most indigenous designs that marks a complete evolution from being a reference-design products vendor to an enthusiast-grade hardware giant: the much talked about EVGA Classified SR-2 dual-LGA1366 enthusiast-workstation motherboard, and power supply unit (PSU) tailor-made to suit systems based on the SR-2 motherboard. To begin with, the Classified SR-2 motherboard is a dual-LGA1366 motherboard supporting two dual-socket capable Intel Xeon E5500 series, and X5600 series quad-core and six-core processors. It is based on Intel's enterprise-grade 5520 chipset with ICH10R southbridge.

The Intel 5520 supports two QuickPath Interconnect paths to each of the sockets, which in turn can communicate with each other. Since each socket (processor) has its own triple-channel DDR3 memory controller, it has its own six DDR3 memory slots. The 5520 gives out two PCI-Express 2.0 x16 lanes which connect to two NVIDIA nForce 200 chips that give out two PCI-Express 2.0 x16 links each. With external x8 switching thrown into the mix, the motherboard manages a staggering seven PCI-Express x16 slots.

Core i7 Dual-Socket Operation Possible?

Intel's Core i7 processor in the LGA1366 package is currently leading in the performance front, some of the fastest processors for the desktop platform fall in this league. Despite lacking a directly competitive processor, the platform had its share of competition thanks to motherboard manufacturers competing within themselves. EVGA for one, is set to unveil a bleeding-edge motherboard later this year, which supports two LGA1366 processors. So far, enthusiasts working for EVGA have only disclosed that the motherboard would work with Intel's DP Xeon 5000 series LGA1366 processors, which are officially capable of working on dual-socket server and workstation motherboards.

However, a plausible rumor doing rounds has it that EVGA may attempt to make its motherboard work with Core i7 series processors - which are incapable to operate in pairs, at least from what we're led to believe. EVGA is attempting to do this by modifying the reference Intel design for the platform.

Intel Introduces Xeon W3565 Workstation/Server Processor

Intel updated its socket LGA-1366 Xeon processor portfolio with the W3565. Based on the Nehalem architecture, this quad-core processor is made for single-socket servers and workstations. It has near-identical specifications to the consumer-grade Core i7 960, in having a clock speed of 3.20 GHz (24 x 133 MHz), 8 MB of L3 cache, and a QuickPath Interconnect (QPI) speed of 4.8 GT/s. The QPI speed is perhaps the only thing that differentiates it from the W3570, which has the same clock speed, albeit a QPI speed of 6.4 GT/s. This one specification makes for a large price difference. While the new W3565 is priced at $562, the W3570 is priced at $999. With the introduction of the new processor, Intel is set to retire the 2.93 GHz W3540, and 3.06 GHz W3550.Source: TechConnect Magazine

Intel Introduces Core i7 960 Processor

Intel silently made its newest addition to the Core i7 desktop processor family official: the Core i7 960. This socket LGA-1366 processor is designed to work on motherboards with the Intel X58 Express chipset. After Intel's gradual replacement of the Core i7 965 Extreme Edition with the Core i7 975 Extreme Edition, there remained a scope for a non-XE Core i7 model with the same clock-speed as the i7 965 XE, albeit with an upwards locked bus frequency multiplier.

The Core i7 960 comes clocked at 3.20 GHz (24.0 x 133 MHz), with a maximum Turbo Boost speed of 3.43 GHz. It has four processing cores, and with the HyperThreading Technology, provides the operating system with eight logical CPUs. It comes with a QuickPath Interconnect speed of 4.8 GT/s, and supports up to 24 GB of memory with its triple-channel (192-bit wide) DDR3 memory interface. Each processing core has a dedicated L2 cache of 256 KB, while an 8 MB L3 cache is shared between all four cores. It is built on the 45 nm HKMG process, with a TDP of 130W.

The Core i7 960 is touted to displace the Core i7 950 from its existing price-point of around US $562 (in 1000 unit tray quantities). As evidence of that, popular American retailer has listed the OEM part (chip-only) at $589.99, a mere $10 higher than the retail box of the Core i7 950 listed on the same store.

nForce Lives: NVIDIA Plans LGA-1156 Chipset

NVIDIA enjoys its share of the pie with the existing socket LGA-775 platform, with a broad range of performance nForce, and value-oriented GeForce MGPU chipsets. With Intel's implementation of the FSB-replacement QuickPath Interconnect system interface with LGA-1366, a conflict ensued with regards to Intel licensing NVIDIA making chipset. Intel's contention stood that its older licenses did not cover the latest processors that implement integrated memory controllers and QPI. As a result, NVIDIA was forced to license its SLI technology to motherboard manufacturers for products based on Intel X58 chipset, and now Intel P55 chipset for the LGA-1156 series processor, and the future of NVIDIA Intel-series chipsets looked bleak at worst.

Fresh reports suggest that NVIDIA indeed has a chipset product lineup planned for Q1 2010, that supports socket LGA-1156 processors, codenamed MCP99. In addition to this, NVIDIA will also serve up two new LGA-775 chipsets, codenamed MCP85 and MCP89. What's common to all these chipsets is that they embed an integrated graphics processor (IGP). Some of these chipsets feature a 64-bit wide memory channel dedicated to the IGP's memory needs, apart from the 128-bit wide (dual-channel) system memory controller, on the LGA-775 platform. The dedicated memory channel allows the IGP to match entry-level graphics cards in terms of performance, without eating into the system memory. We predict this will be implemented in the same way AMD 780G, 785G and 790GX does, with dedicated memory chips present onboard. The LGA-775 compatible ones will further feature dual-channel DDR3-1333 standard supportive memory controllers for system memory. The MGPU will be named inside the GeForce 200M series.

Core i9 ''Gulftown'' Comes to Life

Intel's posterboy processor for the 32 nm Westmere architecture, the six-core Gulftown is now living, breathing silicon. The company seems to have already dispatched samples of the chip. Gulftown is based on the LGA-1366 socket. Featuring 6 cores and 12 threads with HyperThreading enabled, it holds 12 MB of L3 cache to support the additional data load over the QuickPath Interconnect.

A noted enthusiast has two Gulftown processors running in a dual-socket setup. This 12 core, 24 thread monstrosity uses 24 GB of DDR3 memory using 4 GB modules (perhaps 2 x 3 modules). The processors are running at 2.40 GHz (18 x 133 MHz). The machine was put through WPrime multi-threaded benchmark. It crunched WPrime 32M in a little over 6 seconds, and 1024M in 145.6 seconds. Going by older information, Gulftown should be implemented in a commercial product in Q1 2010, when Intel plans a host of other important product launches. When released as Core i9, the processor will target the premium enthusiast market.

Source: XtremeSystems Forums

Intel Previews Intel Xeon 'Nehalem-EX' Processor

Intel Corporation today previewed a new Intel Xeon processor codenamed "Nehalem-EX." The processor will be at the heart of the next generation of intelligent and expandable high-end Intel server platforms, which will deliver a number of new technical advancements and boost enterprise computing performance.

In production later this year, the Nehalem-EX processor will feature up to eight cores inside a single chip supporting 16 threads and 24MB of cache. Its performance increase will be dramatic, posting the highest-ever jump from a previous generation processor.

''Real Men Use Real Cores'': AMD

AMD finally stepped out of its shell after Intel's launch of its newest line of Xeon processors based on the Nehalem architecture. In an interview with TechPulse 360, AMD's Pat Patla and John Fruehe took on Intel's recent marketing drive for Nehalem Xeon products. The conversation revolved mainly around the issues of platform costs, and the features the new Xeon processors introduce (or reintroduce) to the server/enterprise computing industry, namely the company's proprietary FSB-replacement, QuickPath Interconnect, and HyperThreading.

The two first took on Intel's marketing, particularly on its material that said that the slowest Nehalem Xeon chip was faster than the fastest Opteron chip, saying that Intel's statements weren't backed by real figures. The two also alleged that Intel's server platform was too expensive and delivered lesser value in an ailing state of the economy. Perhaps the most audacious statement from AMD since the somewhat famous "only real men have fabs" statement by Jerry Sanders III, came from this interview, where AMD responded to a question on HyperThreading saying that "real men use real cores". "We’ve got real cores across our products. HyperThreading is basically designed to act like a core except that it only gives 10 to 15 percent performance bump for real applications workload." they said. Is AMD making a real point, or fighting fire...erm marketing with marketing? Find out in this interview.

Core i5 to Feature Lower QPI Multiplier

Intel's newest line of CPUs, the Core i5 series is generating quite some buzz ahead of its release, with each piece of detail emerging about the processors adding to the attention it's gathering. Expreview has learned that the Core i5 series processors will feature a slightly different system interface in comparison to their Core i7 cousins.

Coming to the QPI interface between the CPU and NB dice, Intel will raise the QPI base frequency at the expense of lowering the multiplier. For example, the Core i7 uses a 20+ multiplier with a base frequency of 133 MHz. This frequency could be scaled up to 166 MHz at 24x multiplier due to the theoretical speed-limit of QPI at 8.00 GT/s. With the Core i5, Intel will set lower QPI multiplier values (around 16X), while upping the base frequency to around 250 MHz.

Source: Expreview
Return to Keyword Browsing