Tag: 22nm

Intel announces Bay Trail tablet CPU: Part One

Wednesday’s IDF Keynote started by asking the audience to stand for a moment of silence in remembrance of lives lost on 9-11 in 2001. From there, it was business as usual with product hype and promises of future success.

Intel seems to be spotlighting health. It opened with a feel-good video of Jack Andraka, child prodigy and biology whiz. Andraka is a high school sophomore who won the youth achievement Smithsonian American Ingenuity Award in December 2012 for inventing a new method to detect a lethal form of pancreatic cancer.

From there, Intel moved into its theme of “The Internet of Things.” One thing that aroused curiosity was a dull white plastic wristband on every seat. It became an attention-getter later in the programme. In the meantime, everyone got a shot at the podium to talk about their pet project.

Doug Fisher, VP General Manager Software and Services Group, gave a few brief remarks, then introduced Dr. Herman Eul, VP General Manager Mobile and Communications Group. He started off with a video about MTV and Intel getting together to improve the audience’s experience because they do not really understand how wireless works, and what are its limitations.

 
Eul said the goal is to make the mobile platform smarter, the CPU more powerful, and the imaging performance better. He did a brief introduction of “Bay Trail,” the next-generation Atom Z3000 ,  focusing on it being used as a gaming platform. He showed that it is capable of running Windows – which is called heavy legacy software – or running Android OS, Apple OS, Chrome OS, or Linux OS. Bay Trail is a 64-bit processor, built using Intel’s Silvermont 22nm micro-architecture. There will be six variants of the chip available – with dual and quad-core configurations. Clock speeds will range from 1.8GHz to 2.4GHz.

Bay Trail’s Hardware and Software supports:  

  • Windows (32/64-bit) and/or Android and/or Chrome
  • Displays resolutions up to 2500 x 1600 (Retina display)
  • Dual independent displays
  • Intel Wireless Display (WiDi) technology
  • Up to 4GB of LPDDR3 RAM
  • USB 3, HDMI, Displayport, SD card, NFC, 4G, Wi-Fi, GPS
  • X 11, Open GL 3.0 graphics
  • Up to 13MP camera on the rear with Zero shutter lag, burst mode, digital video stabilization, 1080p recording at 60FPS and up to 2MP on the front.

Eul then brought Victoria Molina on stage, a fashion industry consultant and former executive for Ralph Lauren, Levi’s, and the Gap, who explained her virtual shopping experience application. They developed it using the Intel Android SDK in about a week  – but gave no information on the experience level of their programmers.

Molina said the most important part of this application is the fit map, an important factor in making the apparel attractive on the wearer, to attain a “cool” outcome. The application uses an avatar based around the person’s measurements, height and weight, and a facial photograph. The shopper goes out to the web site where they want to shop and chooses the clothing to virtually try on before purchasing. Next, the website pulls up sample clothing from their product lines.

After you build your ensemble of clothing, then you can adjust the clothing so the fit is tight, medium, or loose. After deciding on your look, you go through the “Cat Walk” show-n-tell process. That means the avatar is dressed with each one of the outfits in the size and drape you want and it looks like you are a model on a fashion show runway. Molina said, “This will revolutionise the online shopping experience. Because of the huge “cool factor”.

Next, Intel focused on a Bay Trail small-form-factor tablet running and editing videos. Eul invited Jerry Shen, chief executive of Asus, to introduce its T100, a 2-in-1 Bay Trail notebook with over ten hours of battery life. “We are very excited about the Bay Trail quad-core promise,” Shen said.

Asus is more optimistic than Intel regarding battery longevity. Intel claims Bay Trail tablets could weigh as little 14.1 ounces and offer more than eight hours of battery life when the users are watching high-definition video.

Neil Hand, Dell’s VP of Tablets, showed its  Venue 8-inch, Windows 8.1, Bay Trail tablet that is going to be shipping soon. He said it has 4G LTE.
 
Eul talked briefly about upcoming Merryfield, a 22nm SoC which is build on the Silvermont architecture specifically for smartphones. We were told that Airmont, a 14nm process engineering SoC with all the features of Bay Trail for tablets, is on schedule for Q3 2014 release.

Finally, Eul satisfied our curiosity by showing his audio DJ idea which activated those dull white plastic bracelets that were sitting on each chair. A video was projected onto the giant screens in the auditorium showing the Keynote audience and the wristbands lighting up in synch with Eul’s music.

The presentation took another turn with Kirk Skaugen, Senior VP General Manager PC Client Group at Intel which will be covered in part two.

Intel's usual troops missing from IDF stage

For Tuesday morning’s keynote presentation at IDF-SF 2013, there were none of the usual Intel standard bearers. New CEO, Brian Krzanich, did a major part of the presentation along with Renée James, Intel’s President.

Before the presentation started, one of the old guard, Mooly Eden was spotted standing in the aisle way wearing his signature cap.

Also sitting in the audience’s VIP seats was former Intel CTO Justin Rattner. Rattner retired in June of this year and we missed his imitations of TV’s Mister Wizard.

CEO Krzanich gave his overview of the “new and improved” Intel. Krzanich laid out Intel’s vision and described how Intel is refocusing – away from its traditional CPU centric design philosophy to a system centric solution based around SoCs (system-on-a-chip) and broader integration.

Intel’s foundry capabilities were touted as reducing the die size to 20nm which is now shipping, with 14nm in the works. This will allow wearable computers. The obvious ones are smart watches – Intel’s engineering sample is many generations behind the competition in looks. The not-so-obvious areas they’ll address will be in the healthcare industry.

Krzanich said: “Innovation and industry transformation are happening more rapidly than ever before, which play to Intel’s strengths. We have the manufacturing technology leadership and architectural tools in place to push further into lower power regimes. We plan to shape and lead in all areas of computing.”

He continued: “Intel plans to lead in every segment of technology from the traditional to the emerging. Intel will continue with its data center revolution/evolution by increasing the computing power and lowering the kilowatts used in the rack space.” Krzanich stated that “the traditional PC is in the process of reinventing itself” with most notably tablets and 2-in-1 PC platforms.

The CEO said that Intel is introducing this week “Bay Trail,” Intel’s first 22nm SoC for mobile devices. “Bay Trail” is based on the company’s new low-power, high-performance Silvermont microarchitecture, which will power a range of Android and Windows designs.

[Remember Intel’s commitment to Wimax?-Ed]

Krzanich showed the first Intel phone with the 22nm SoC with Intel data LTE and voice 3G. He claimed that “by next year you will see LTE data and LTE voice in the same phone”. Then, he showed a demonstration of LTE Advanced. LTE advanced will have carrier activation switching from 30Mbps (Megabits per second) to 70 Mbps. He said the San Diego group is working on this. Could this be Qualcomm?

Krzanich announced the Intel Quark processor family. The new lower-power products will extend Intel’s reach to growing segments from the industrial Internet-of-Things to wearable computing. It is designed for applications where lower power and size take priority over higher performance.

The tablet marketplace is a key ingredient for the atom processor family. “The Hallway tablet systems price point will go below $100 by Q4 2013,” Krzanich said. 

However, the ARM and MIPS based 7-inch tablets have been there for over a year with good quality graphics, wi-fi, and reasonable gaming performance. Intel has some hurdles to jump over to gain a bigger chunk of that marketplace.

Renée James, Intel’s President, talked about the company’s involvement in the healthcare world and wearables.

Referring to health care as it relates to technology, she gave an example: “one person’s complete genomic data is approximately 1 PB, or 25 filing cabinets of information”.

“Genomic data cost for one person was in the hundred thousand dollar range less than four years ago,” James said. “Soon it will be in the $1000 range, which makes it plausible for use as a cancer fighting tool.”

James introduced Eric, an Intel employee who for over 20 years has been fighting cancer.

Eric came up and told his story about having his genomic data sequenced and taking that data to his doctors. About a month after they had the data they had a meeting with all his doctors including the East Coast doctors on Skype.

Eric said by having his genomic data, the doctors figured out that over the 20 year period of time, 90 percent of those drugs they had given for his cancer treatment could not work for him.

The doctors created a new set of drugs specifically typed for his genome, and in less than 90 days, he was completely cancer free and has remained cancer free. Understandably, Eric received resounding round of applause from the audience.

When one can see directly how technology impacts one person’s life in the extreme, we are all glad to be in this industry. 

Intel attempts to re-invent itself

For yesterday’s “keynote” presentation at IDF-SF 2013, there were none of the usual Intel standard bearers. Intel’s newly hatched CEO, Brian Krzanich, did a major part of the presentation along with Renée James, Intel’s President.

Before the presentation started, one of the old guard, Mooly Eden  was spotted standing in the aisle way wearing his signature cap.

Also sitting in the audience’s VIP seats was former Intel CTO Justin Rattner. Rattner retired in June of this year and we missed his imitations of TV’s Mister Wizard.

CEO Krzanich gave his overview of the “new and improved” Intel. Krzanich laid out Intel’s vision and described how Intel is refocusing – away from their traditional CPU centric design philosophy to a system centric solution based around SoCs (system-on-a-chip) and broader integration.

Intel’s foundry capabilities were touted as reducing the die size to 20 nm which is now shipping with 14 nm is in the works. This will allow wearable computers. The obvious ones are smart watches – Intel’s engineering sample is many generations behind the competition in looks. The not-so-obvious areas they’ll address will be in the healthcare industry.

Krzanich said: “Innovation and industry transformation are happening more rapidly than ever before, which play to Intel’s strengths. We have the manufacturing technology leadership and architectural tools in place to push further into lower power regimes. We plan to shape and lead in all areas of computing.”  There you go.

He said, “Intel plans to lead in every segment of technology from the traditional to the emerging. Intel will continue with its data centre revolution/evolution by increasing the computing power and lowering the kilowatts used in the rack space.” Krzanich stated that “the traditional PC is in the process of reinventing itself” with most notably tablets and 2-in-1 PC platforms. See?

The CEO said that Intel is introducing this week “Bay Trail,” Intel’s first 22nm SoC for mobile devices. “Bay Trail” is based on the company’s low-power, high-performance Silvermont microarchitecture, which will power a range of Android and Windows designs.

Krzanich showed the first Intel phone with the 22 nm SoC with Intel data LTE and voice 3G. He claimed that “by next year you will see LTE data and LTE voice in the same phone”. Then, he showed a demonstration of LTE Advanced. LTE advanced will have carrier activation switching from 30Mbps (Megabits per second) to 70 Mbps. He said the San Diego group is working on this. Could this be QUALCOMM?

Krzanich announced the Intel Quark processor family. The lower-power products will extend Intel’s reach to growing segments from the industrial Internet-of-Fangs to wearable computing. It is designed for applications where lower power and size take priority over higher performance.

The tablet marketplace is a key ingredient for the atom processor family. Krzanich said, “the Hallway tablet systems price point will go below $100 by Q4 2013.” However, the ARM and MIPS based 7-inch tablets have been there for over a year with good quality graphics, WiFi, and reasonable gaming performance. Intel has some hurdles to jump over to gain a bigger chunk of that marketplace.

Renée James, Intel’s President, talked about its involvement in the healthcare world and wearables. Referring to health care as it relates to technology, she gave an example “one person’s complete genomic data is approximately 1 PB, or 25 filing cabinets of information”. She said, “genomic data cost for one person was in the hundred thousand dollar range less than four years ago. Soon it will be in the $1,000 range, which makes it plausible for use as a cancer fighting tool.”

James introduced Eric, an Intel employee who for over 20 years has been fighting cancer. Eric came up and told his story about having his genomic data sequenced and taking that data to his doctors. About a month after they had the data they had a meeting with all his doctors including the East Coast doctors on Skype. Eric said by having his genomic data, the doctors figured out that over the 20 year period of time, 90 percent of those drugs they had given for his cancer treatment could not work for him. The doctors created a new set of drugs specifically typed for his genome, and in less than 90 days, he was completely cancer free and has remained cancer free. Understandably, Eric received resounding round of applause from the audience.

Intel’s Silvermont SoC ready for ARM wrestling

Intel is finally starting to take the mobile market seriously, three years too late for anyone to care. The chipmaker has finally revealed its next generation Silvermont microarchitecture, and although it is late to the party, it looks like an impressive piece of tech.

For years Atoms were built using ancient architectures and off the shelf chipsets, but Silvermont is a different beast. It is a 22nm system-on-a-chip and it is the first Atom to use out-of-order execution. It also features 3D tri-gate transistor technology and a very scalable design, which means Intel could theoretically come up with eight-core parts. Some Silvermont parts will use graphics derived from Intel’s HD 4000 core, used in Ivy Bridge chips, which means they should end up quite fast. 

Basically Intel crammed Silvermont with the latest tech it has to offer, and that’s what makes it significant. Intel is finally taking ARM seriously.

In terms of performance, the new microarchitecture is three times as powerful as the cores used in current Atom SoCs, which are already capable of holding their own against many ARM chips. Silvermont chips can wipe the floor with the current crop of ARM SoCs with relative ease. 

The added performance doesn’t come at a price. In fact, Intel says the new chips can cut power consumption five times compared to existing Atoms running at the same performance level. Performance per watt is crucial in smartphones and tablets. It was Intel’s undoing for years, but it seems to have nailed it at last. 

Silvermont will appear in several flavours. Merrifield chips will cater to smartphones, while beefier Bay Trail chips are reserved for tablets. Avoton will take care of microsevers. Merrifield and Bay Trail should basically deliver the performance of three to four year old PC chips to tablets and phones, which sounds very impressive indeed. It has the potential to transform Microsoft’s fledgling Windows 8 into a proper tablet operating system, which means Silvermont is yet another nail in the Windows RT coffin.

The bad news? We’ll have to wait a bit longer to see what Intel has cooked up for the ARM gang. Silvermont phones will show up sometime next year, which means ARM will continue to dominate the market for the time being. Bay Trail tablets are expected later this year, running Windows 8.1 and Android

Quarter of wafer capacity dedicated to sub 40nm processes

As chipmakers struggle to go beyond 22nm and 28nm, it seems older processes are dying faster than Gangnam Style. A recent report by IC Insights shows that more than a quarter of installed wafer capacity is dedicated to sub-40nm process geometries. 

At the end of 2012 sub-40nm plants were capable of churning out more than 3.9 million wafers a month, accounting for 27.3 percent of total worldwide production. At the same time 40nm to 60nm processes account for 18.8 percent, while 80nm to 60nm has a 7.6 percent share. 

Mature processes, used for high-voltage ICs in the 80nm to 400nm range, still make up about 35 percent of total wafer production. 

Intel is already at 22nm, while other foundries are still at 28nm, although TSMC and Globalfoundries are rather upbeat when it comes to their 20nm transition. Although the market for PC CPUs and big GPUs is shrinking, strong demand for mobiles is boosting demand for high performance SoCs and high density DRAM.

Big Blue spills beans on 22nm

IBM suits have adjusted their ties and released a chip which they think proves they can get funky with all these young low powered whipper snappers.

Details of the 22-nanometer chip were shown off at the the International Electron Devices Meeting (IEDM) in San Francisco.

IBM insists that the new chip, which could be aimed at mobiles, has serious business use in servers so the chip might not be as with-it as it appears.

Biggish Blue researcher S. Narasimha said IBM’s 22-nm process uses partially depleted silicon-on-insulator. IBM has apparently prototyped “a number of server processors” in the node that achieve latency below 1.5 ns and 750 MHz random clock cycles.

Narasimha refused to say what IBM will do with the 22-nm node, he said the goal was to provide 25 to 35 percent boosts to what has gone before.

This meant that server processors could run up to 5.5 GHz and others with up to 80 Mbytes embedded DRAM, EE Times reports.

TSMC expects Q3 results boost

TSMC announced today it had increased sales by 32 percent, while reports claim that the foundry is set to up its capex spend amid talk of new orders from Apple.

In financial results released today, the Taiwanese foundry said that unconsolidated sales were up to $1.64 billion during August, increasing two percent over the previous quarter and 32 percent from the same point in 2011.  Consolidated sales increased 31.5 percent over August 2011.

Commenting on the results, TSMC’s chief financial officer, Lora Ho, said that third quarter revenues are now expected to be slightly higher than previous estimates.

According to CENS, TSMC is also set to increase its capital expenditure budget by 25 percent, up from $8 billion to $10 billion during 2013.  

Although TSMC is not expected to make any official announcements until later this year, the increased spending would enable expansion of production at its current leading edge 28nm process, as well as beginning volume production of 20nm chips.  Cash would also be spent on advanced lithography tools, and beginning development of 16 nm chips.

The increased capex budget will also dovetail with the growing rumours that Apple is to ditch its mobile patent sparring partner, Samsung, with regards to chip production for its smartphones.  
The battle between the two has become increasingly savage and Apple appears to be looking for alternative component suppliers.

Citing sources close to Chinese-language Economic Daily News, CENS claimed that plans for TSMC to begin pilot production for Apple in the first half of 2013, with volume production to kick off in the second half of the year.

TSMC to spend $2.4 billion on capacity expansion

TSMC has been given the go ahead to spend $2.4 billion on expanding its advanced chip process production, as well as constructing a new fab.

Following a board of directors meeting, the Taiwanese mega-foundry was granted permission by the board to “expand and upgrade” production of its most advanced process 28nm chips.

The decision followed a meeting regarding TSMC’s July sales results which showed an increase of 37.3 percent from the same point the previous year, hitting $1.6 billion for the month.

Supply of 28nm chips has been tight for many customers as TSMC continues to ramp up production. In the second quarter, 28nm chips accounted for seven percent of total wafer revenues. So with sales up and strong demand for its advanced process chips, it seems that the board at TSMC is happy to expand production.

The board also gave the nod for $378 million to be spent on building a new fab as well as fitting it with the appropriate facilities.

The details are scant so far, and it was not clear in the statement when the fab construction would begin, though it is likely that the any building would support production of chips at more advanced process. The chip maker is set to begin 20nm production next year. 

TSMC recently announced its intentions to begin development at the 20nm process with ARM on FinFET server chip designs.  ARM later announced it would be extending its partnership with GlobalFoundries.

Here comes our Intel Core i7 3770K review

Here is our review of the Core i7-3770K processor, Intel’s highest-end Ivy Bridge-based processor. There’s a lot to be discussed about it, but we’ll start from the top.

 

Sandy Bridge exits the scene

Sandy Bridge was a tock in Intel’s design and manufacturing tick-tock strategy. A ‘tock’ is usually a new architecture on a mature process and, as such, normally results in better yields and much better revenues for the manufacturer. Ivy Bridge, its successor, is a tick. It’s a new process on a slightly tweaked architecture so carries with it a risk of lower yields.

Sandy Bridge was a successful move for Intel, in particular in the processor graphics department. The 32nm-built processor fully integrated the graphics core and improved GPU performance over its stepdad, Clarkdale. It introduced an extended instruction set named AVX, video hardware encoding features and an optimized branch prediction amongst other improvements. The now-famous K-series offered unlocked multipliers and some serious overclocking headroom, which proved to be an enthusiast’s delight. It has proved to be a very successful design and was bound to be hard for Intel to do better.

As Sandy Bridge bows out of the market, you’ll see boxes and boxes at heavily discounted prices right now. The brutal slashing began a week before the launch, emptying shelves and making room for the shiny new toy to come. 

 

Ivy Bridge arrives, not late, not early

Lo and behold, the Ivy Bridge, Intel’s 3rd generation Intel Core processor with processor graphics (as the chipmaker calls it). Not really late to the party, nor early, just on time considering it is Intel that’s pushing the market forward. Despite rumours of delays and a bit of crossed lines between some Intel execs, the CPUs officially launched this Monday.

As of now, Intel is introducing 14 new Ivy Bridge-based SKUs. These include one mobile extreme edition, four standard mobile versions, five desktop and four low-power ones. In order, these are:

Core i7-3920XM, Core i7-3820QM, Core i7-3720QM, Core i7-3612QM, Core i7-3610QM, Core i7-3770K, Core i7-3570K, Core i5-3570K, Core i5-3550, Core i5-3450, Core i7-3770T, Core i7-3770S, Core i5-3550S and the Core i5-3450S.

The 3 prefix in the numbering is the generation, ie: 3rd generation Core processors, while the rest of the number represents the model itself. The letter suffixes represent variants. K for multiplier unlocked, S models are low power and T models are ultra-low power. You can see below the full spec sheets.

 

Intel Ivy Bridge desktop CPUs

Low Power Ivy Bridge CPUs

Mobile Ivy Bridge CPUs

Some facts about Ivy Bridge

Ivy Bridge is the successor to Intel’s Sandy Bridge microarchitecture. It isn’t a completely new design, but a spin on its predecessor, built on a smaller process and introducing a few new tweaks to the original recipe… some of them more than just pure performance tweaks. Still, we need to state some facts about Ivy Bridge, even before we start the testing. There are two parts to the Ivy Bridge architecture that need focusing on.

First of all, Intel proudly parades Ivy Bridge as the first 22nm “3D” (ie: tri-gate) transistor-based processor. Yes, 3D is all the rage even on CPUs. Simply put this means it’s stacking the gates on its transistors keeping current leakage down (allowing Intel to scale its CPUs to 22nm and beyond) as well as providing some valuable space savings. Transistors built on this 22nm process also require less power, which has amounted to some substantial power savings on the CPUs.

Ivy Bridge integrates a more advanced graphics core onto the die, the HD 4000, a DirectX 11 (ie: hardware tessellation), DirectCompute capable part, which now shares the CPUs own L3 cache. The Intel HD 4000 processor graphics features 16 Execution Units (let’s call them shaders), Clear Video Technology (to offload video decode) and Quick Sync Video, which is hardware based encoding and decoding, which, we’ll see, works quite well. Intel claims up to twice the performance of the graphics in its Sandy Bridge predecessor.

Ivy Bridge is a 1.4 billion transistor processor with a die size of just 160mm2, by comparison, Sandy Bridge had 1.16 billion transistors and a die size of 216mm2. Despite a higher transistor count, the more efficient power design of the 22nm “3D transistors” still rack up the power savings from 95W on the 2700K to 77W on the 3770K. You can see the “processor graphics” die area has become considerably larger than its predecessor

Sandy Bridge Die, labeled

Ivy Bridge die, labeled

Sandy Bridge (on top) and Ivy Bridge (below), you can see that the processor graphics element has swollen up considerably in the latter.

The new architecture comes hand in hand with a new chipset family, the 7 series, codenamed Panther Point. This chipset is compatible with both Sandy Bridge and Ivy Bridge, but not first generation Core products.

Intel supplied a DZ77GA-70K motherboard which is powered by the Z77 chipset and was launched a week prior to the Ivy Bridge release. The DZ77GA-70K, as most Intel motherboards, have all the shiny LEDs and the looks of a deadly killer, but is very tame when it comes to overclocking and basically stepping out of bounds, even though its EFI BIOS is one of the best we’ve seen to date. From system monitoring to dialing up the clock on the CPU, it’s all dead simple. Our overclocking experiments with the motherboard yielded a humble 1.4GHz overclock (3.5 to 4.9GHz), that we are sure was too easy to achieve, yet too hard to overtake on this particular motherboard – something Asus or Gigabyte will pick up and take to the next level. Still the EFI BIOS is gorgeous and simple to use.

The 7-series chipset includes Intel Rapid Storage Technology 11, USB 3.0, Thunderbolt support, SATA 3.0, PCIe 3.0 and up to 3 independent displays (depending on configuration). It’s what the 6-series could have been, in essence.

7 Series Chipset Overview

 

Benchmarks

Our Engineering Sample Core i7 3770K is the counterpart to Intel’s Core i7 2700K Sandy Bridge, both clocked at 3.5GHz and both sport four cores / eight threads. Both have the same Turbo Boost speed of 3.9GHz and both are in the lab for our Apples to Apples comparison. Intel promised something in the vicinity of 7/15 percent pure CPU performance increase, and almost twice as much in “media” processing, thanks to its new graphics core, so let’s see what we get.

We’ll begin with a few CPU benchmarks. We aren’t holding our breaths on this, to be fair, Ivy Bridge didn’t introduce any revolutionary new magic tricks.

Cinebench R11.5 score

In Cinebench R11.5, the HD 4000 GPU is clearly marking the difference. The 3770K pulls ahead of its predecessor by a comfy margin.

Passmark Int/FPU score

Passmark is a simple fire’n’forget benchmark that assesses PC performance on several levels. We’ve focused on FPU and Integer performance. The Ivy Bridge FPU is tremendously more efficient than its predecessor, beating it by a 67 percent margin. Overclocked, the 3770K scales very well.

PCMark7 Computation score

The PC Mark 7 benchmark suite tests all PC subsystems, but we’re actually interested in the Computation score, here. Ivy Bridge and Sandy Bridge are almost 1-for-1.

POV-Ray Biscuit

POV-Ray is a ray tracing benchmark that relies on CPU muscle to render its target image.

SuperPI score

Purely mathematical in nature, Super PI maxes out single-threaded performance to calculate PI, in this case to the 2 millionth place.

SANDRA 2010 AES 256 bandwidth

The 3770K features a new encryption engine that allows it to squeeze a lot more data down the pipe.

SANDRA 2010 Arithmetic score

WinRAR Compression 320MB time

WinRAR Compression shows the minor edge the 3770K offers over the Core i7 2700K. A bit meh, if you ask us.

Now onto some strictly graphics-oriented benchmarks.

The HD 4000 end of business warrants its own analysis. With its 16 Execution Units and CPU-shared LLC (Last Level Cache) the HD 4000 is now spelling out some doom and gloom for the low-end discrete graphics business.

3DMark Vantage score

The inevitable 3DMark Vantage benchmark shows off DirectX 10 performance for the HD 4000 graphics. Granted it’s nothing to write home about, but it seems Intel is finally getting somewhere with its graphics processors.

3DMark 11 Performance score

3DMark 11 performance is nothing to sneeze at, considering that DirectX 11 support is brand new to the Intel lineup. We did get some artifacts in some scenes, but we believe this to be a driver issue, more than the hardware getting uncomfortable with the benchmark.

Dirt 3

We put Dirt 3 at max settings and Intel’s processor graphics survived the ordeal. If you scale down AA, you can game quite well on Intel’s new toy.

Metro 2033

We threw Metro 2033 at it as a crash test. The Metro 2033 – Frontline benchmark, running in DX 11 mode with Very High details, was like a slide-fest at times, but, again, scale back the details and image quality just a little bit and you’ll find something playable.

ComputeMark score

Considering Intel’s HD 4000 is now a OpenCL/DirectCompute capable part, we ran ComputeMark on it. The HD4000 part scored a quarter of the discrete competition.

TechARP x264 Benchmark

Finally, our media encoding test is where Intel’s HD graphics part stretches its legs. The HD 4000 graphics with its new media encoding engine chews away at frames almost as well as a discrete part.

 

Conclusion

Overall the Ivy Bridge core offers some meager performance gains over Sandy Bridge, good power savings and some great potential if you like to overclock your CPUs. The Core i7 3770K’s direct competition hails not from AMD (it hasn’t for a while now) but from its direct predecessor, the Core i7 2700K.

Over the next weeks you’ll also see that Ivy Bridge brought with it a bevy of new hardware releases, from motherboards to RAM to SSDs, as one way or another you do get quite unique advantages if you buy hardware that has been optimised for Intel hardware. The optimisations, however, revolve mostly around the motherboard and its chipset rather than the CPU, so if you see Z77 bundles with Core i7 2600K processors at a good price, you might want to consider the deal. As much as HD 4000 graphics are an improvement over their Sandy Bridge predecessor, many will keep on asking why bother with processor graphics in the mid-to-high end of things, considering most discrete GPUs will simply annihilate it. Ivy Bridge does bring DX 11 compute capabilities which we can only expect Intel will leverage down the line. Our media encoding results with the HD 4000 were close to the results we had with a discrete (GTX 460 1GB) GPU, which is nothing short of amazing. Gaming, while it might not be its forte, is definitely on the menu. Add to that the fact that you can combine the processor graphics with a discrete part, it’s up to Intel to bring to the fore some additional features.

Sandy Bridge was, admittedly, a hard act to follow, but Ivy Bridge is more than a speed bump with minor architectural improvements. It’s an important shift in design and manufacturing for Intel. In its own right, Ivy Bridge is a formidable opponent even for some higher-end Extreme Edition CPUs. It happens to also have a great deal of potential for forthcoming software and driver updates, like OpenCL/DirectCompute support. “Potential” is the operative word here, and it might not shake you to your core (no pun intended) and make you rush out to buy it.

If you can do without the power savings, overclocking tweaks and processor graphics, you might be better off picking up a Core i7 2600K/2700K on sale, but if you were about to buy one, this supersedes – dollar for dollar – what the 2700K had on offer, on just about every level. If you already own a Core i7 2600K or 2700K, you needn’t go digging in your pocket for the upgrade money just yet.

Intel will make FPGAs on contract

Intel has announced that it will make chips for another developer of field-programmable gate arrays (FPGA) using its 22nm fabrication process.

 

Tabula designs programmable logic chips for network infrastructure and has apparently signed up with the Intel Custom Foundry, a division of Intel’s technology and manufacturing group.

 

Intel said that it will make Tabula’s ABAX family of 3D programmable logic devices (3PLD) which will set new marks in density and performance for logic, memory, and signal processing.

 

ABAX devices have a mix of configurable I/Os, including 920 general-purpose I/Os and 48 6.5Gb/s SerDes.

 

In a statement, Tabula said that as FPGAs have become larger, and with the advent of 100Gb Ethernet requirements, the traditional FPGAs are reaching their performance limits.

 

Its Spacetime programmable fabric is more balanced and have shorter interconnects than traditional FPGAs. It means that it is possible to clock the entire fabric at the same frequency.

 

However, to do all that, it needed Intel’s 22nm process technology which uses tri-gate transistors.

 

It thinks that the combination of process and architecture will mean Tabula to produce high-performance programmable circuits that consume significantly less chip area than circuits implemented with traditional FPGA fabrics.

 

Sunit Rikhi, vice president of technology and manufacturing group at Intel, said he expects the tri-gate transistors will attract other customers.

 

Intel has worked closely with Tabula throughout the product design cycle to optimise Tabula’s 3PLD family with Intel’s 22nm manufacturing process and design kits, he said.