Close
Report Comment to a Moderator Our Moderators review all comments for abusive and offensive language, and ensure comments are from Verified Users only.
Please report a comment only if you feel it requires our urgent attention.
I understand, report it. Cancel

Digital Foundry

Tech Focus: Where Next for Intel?

Tech Focus: Where Next for Intel?

Wed 25 Apr 2012 7:00am GMT / 3:00am EDT / 12:00am PDT
Digital Foundry

Digital Foundry on the new Ivy Bridge launch and future plans for mobile and graphics

Intel has done it again. This week's release of its new Ivy Bridge processor products reveals - yet again - best-in-class processing power and efficiency, further cementing the firm's position as the number one provider of PC CPUs.

There's no secret about Intel's strategy for dominance in this market: the company operates on a "tick tock" strategy that sees the release of a new architecture on an established fabrication node, following it up the following year with a hardware revision based on the same principles but die-shrunk to run on smaller, more power-efficient chips. So, the new Ivy Bridge is seemingly a "tick" then - the smaller, more efficient version of last year's "tock" - Sandy Bridge.

"Ivy Bridge is a stepping stone on the way to addressing Intel's major weaknesses and failures - its lacklustre graphics support and inability to address the exploding mobile market."

This time though, something's different. Intel is calling Ivy Bridge a "tick+" - suggesting that it is more than just a more power efficient version of the existing product. What constitutes that extra plus point is the real story behind Ivy Bridge and sets the scene for Intel's typically ambitious plans for the future. First up there's the nature of the process reduction, the foundation on which the company plans to dominate processing, graphics and mobile chips: the current 32nm Sandy Bridge has given way to a new 22nm process based on three dimensional "Tri-Gate" transistors.

Moore's Law suggests that the amount of transistors that can occupy an area of silicon should double every 18 months to two years, effectively doubling computational power. The problem Intel faced in sustaining Moore's Law was that power leakage was increasing the smaller it made its transistors, meaning a new approach was required. By moving to 3D transistors, surface area for electrons to travel triples and the power leakage problem is comprehensively addressed. The move to 3D transistors also sets the stage for further die-shrinks - Intel's roadmap suggests we'll see 14nm chips next year and 10nm in 2015, making Moore's Law safe for a good while yet.

However, the real story is that Ivy Bridge sees a significant improvement in Intel's integrated graphics technology - an area that arch-rival AMD has dominated with its CPU/GPU combo chips, or APUs as it likes to call them. The idea here is that the traditional set-up of a main CPU backed up by a dedicated graphics core elsewhere within the system is replaced by a single design that sees both main processing elements sitting on one chip. While high-end gaming graphics performance is well out of reach of these integrated processors, the fact is that the majority of the PC market doesn't actually require enthusiast-level GPU performance. Indeed, it's swiftly becoming the case that GPUs are becoming increasingly more useful for tasks mostly unrelated to graphics and it's a trend that's set to continue: the rise of DirectX 11 and OpenCL are more about taking traditional CPU tasks and porting them across to the massively parallel graphics architecture which is more suited to certain tasks.

GPU acceleration has some phenomenal potential. We've already seen how it is being used in some browsers to speed up webpage rendering, but there's a wealth of opportunity here, especially in content creation applications - editing footage and then preparing an encode for YouTube for example, not to mention video decoding. This is why Intel has dedicated engineers working on QuickSync - a hardware accelerated encoding and playback tool.

But why the need to integrate graphics hardware into the CPU? Isn't it a waste when so many PCs already come with dedicated hardware sitting in the PCI Express slot?

The answer to that is very straightforward. Desktop PC market share is losing traction at pretty much the same rate that laptop take-up is increasing, and an integrated CPU/GPU makes a lot of sense here: the cost of these machines drops significantly if no dedicated graphics chip is included, and battery life increases too. Low power graphics acceleration on laptops results in some massive advantages for the end user too in terms of functionality: HD media decoding is a feature that will be used day-in, day-out and the idea of being able to watch movies and TV shows without an enormous power drain is essential.

Having absolutely dominated AMD in terms of CPU processing power ever since it moved over to its Core-based products, Intel is now looking to improve on its hitherto patchy graphics technology. It's safe to say that the company's previous form here has been poor, with many recent games operating at single-digit frame-rates - or not running at all. Driver revisions haven't had anything like the same level of care and attention put into them that we see from vendors like AMD and NVIDIA.

With Ivy Bridge, the integrated graphics processor seems to show some real, healthy signs of improvement. Based on benchmarks found in Anandtech's Ivy Bridge review, we see some impressive feats: Crysis Warhead, Starcraft 2, DiRT3, Battlefield 3 and Metro 2033 at 1366x768 running at over 30 frames per second. While resolution is low and quality settings usually at the medium or low levels, that's still console-beating performance - but it's still a good 30 per cent behind what AMD is achieving with its existing Llano CPU/GPU (or as the firm calls it "APU") architecture, set to be revamped imminently with its new Trinity offering.

In the meantime, Ivy Bridge is being touted by Intel as offering 20 per cent more computational power overall while being 20 per cent more power efficient than the existing Sandy Bridge technology. What reviewers have actually found is a 5-23 per cent performance advantage and little to no improvement on power efficiency at idle, with the savings only really kicking in when the CPUs are being pushed to the limit - something that doesn't really happen that often in general usage.

"Ivy Bridge is being touted by Intel as offering 20 per cent more computational power overall while being 20 per cent more power efficient than the existing Sandy Bridge technology."

The overall consensus seems to be that Ivy Bridge is a very good product, with a useful GPU performance boost but nothing like the generational increase in power that we saw with its predecessor. However, it gives us a sense of where Intel is heading with its future products and it lays the production groundwork for its next generation chips, which should address what many may consider a monumental misstep in Intel's strategy: its almost criminal negligence in the tablet and smartphone markets.

Put into perspective, Intel's failure in any mobile market outside of laptops is simply staggering bearing in mind the R&D and production resources at its disposal. And it's not as if it has no track record in providing good mobile technology - the original Intel Atom found in a multitude of netbooks is a remarkable piece of engineering and at max power, sips an impressive 2 watts of power. What it lacked was any similarly power efficient integrated graphics hardware - something Intel has remedied with this year's Saltwell Atom architecture, which adds a PowerVR SGX540 and sees the production process drop to 32nm (most of the ARM chips are at 40/45nm). The Orange Santa Clara, Lenovo K800 and XOLO X900 are three new smartphones based on this chip, ironically running the Android operating system.

While insiders reckon that the ARM chips beat Intel in terms of design, the company has the advantage of the most advanced production facilities in the world - its chips will be smaller and more power efficient for the foreseeable future, giving it a key advantage over its competitors in the smartphone market.

Curiously, Intel's plans with regards taking on ARM for the tablet market seem less clear. The obvious solution is to utilise the same Atom technology it's targeting for smartphones, making use of the larger form factor and bigger batteries to allow for higher clock speeds. However, it seems that the next "Tock" in Intel's relentless progression in CPU architecture suggests a stunning combination of both efficiency and performance. Remarkably, it seems as if Ivy Bridge's successor will be targeted at desktop, laptop and even mobile markets.

Codenamed "Haswell", not a huge amount is know about the architecture but well-informed rumours suggest a 2.5xGPU boost over Ivy Bridge, combined with power-saving technology that allows for Intel laptops to run on standard usage patterns for up to ten days between charges, while at the same time being able to process push notifications (update email etc) while in standby. This suggests very exciting times ahead for Intel's thin and light Ultrabook platform - especially when you factor in the mouthwatering tablet/laptop hybrid Intel showed off at the recent IDF 2012 conference in Beijing.

At the moment there is a delicate balance of power when it comes to desktop, mobile and graphics technology: Intel dominates desktop and laptop CPU marketshare, ARM virtually owns mobile while AMD's integrated GPUs are best-in-class (and its desktop parts are competitive too). Ivy Bridge looks for all the world like a consolidation of Intel's strengths, but the signs are there that it's more of a stepping stone to a sustained assault on it major competitors. In response, AMD will be betting heavily on its new Trinity APUs, and ARM's new A15 architecture promises to offer incredible performance - whether the new Atoms and Haswell have what it takes to see Intel win out will make for a very interesting 2013...

2 Comments

Barry Scott Software Design

19 2 0.1
Atom + PowerVR is a product that Intel cannot sell.
Driver support is between none existent or very poor quality depending on what OS you run.

The rumours are that it gets killed off with ValleyView (Atom + IvyBridge Graphics).

Posted:2 years ago

#1

Andrew Ihegbu Studying Bsc Commercial Music, University of Westminster

464 173 0.4
Atom + PowerVR wont be sold by intel, it will be sold by Samsung, or whatever company decides to put it in thier phone/tablet.

The driver support line is absolute rubbish. PowerVR graphics are everywhere in the mobile space, from Android to Blackberry OS, hell iPhone 1G used it. Driver support will exist, because the developers that make the product make drivers for it in every case, unless I've misinterpreted you, which I hope I have.

Edited 1 times. Last edit by Andrew Ihegbu on 28th April 2012 2:12am

Posted:2 years ago

#2

Login or register to post

Take part in the GamesIndustry community

Register now