Intel will offer five GPU SKUs in its upcoming Haswell based processors and a further two for its server chips.
Intel has been placing greater emphasis on its graphics performance in the last few generations of its chips, with Ivy Bridge finally bringing full profile OpenCL capability. Now the firm has released some more details on the GPUs that will be available with its Haswell processors, with five options for desktop and laptops that the firm claims are up to twice as fast on its favoured 3DMark benchmarks.
While Intel showed off performance figures using various versions of its favoured synthetic 3Dmark benchmark, the firm also said that its next generation GPU will support Microsoft’s DirectX 11.1 and OpenGL 4.0 along with OpenCL 1.2. The firm also said it will put DRAM on the same package as the GPU.
Intel also revealed what GPUs certain processors will have. The Core i7-4650U, a 15W TDP chip for ultrabooks, will have a GT3 GPU, while the Core i7-4558U, a 28W TDP mobile chip, will have a higher performance GPU aimed at larger laptops. The firm will drop its internal GTx codenames in favour of more marketable monikers.
Before Intel plasters its HD Graphics and various other stickers on the GPUs, the firm said GT1 GPUs will be destined for processors carrying the Pentium and Celeron branding, with GT2 being the most common GPU found in processors branded as Core. Intel’s GT3 and GT3e GPUs will come in three flavours for particular TDPs, with the firm using the desktop Core i7 4770K and Core i7 4770R as examples of chips with different TDPs, 84W and 65W respectively, with different GPUs and benchmark performance.
While Intel’s GPUs in Haswell processors are still unlikely to be enough for gamers, it is clear that Haswell chips will provide another significant jump over Intel’s previous generation GPUs, which is particularly important for laptop users who are starting to see higher resolution displays but are forced to run with integrated graphics.
As we draw closer to the launch of Intel’s 4th generation Core CPUs, or Haswell, it is no wonder that we are starting to see more leaks and one showing Intel’s Core i7 4770K overclocked to 5GHz at 0.9V certainly drew a lot of attention.
An impressive overclocking achievement was spotted by Ocaholic.ch and shows a CPU-Z validation of Core i7 4770K overclocked to exactly 5005.83MHz at just 0.904V. As far as we can tell, Hyper-threading was disabled and it is not clear if the CPU is actually stable enough to run anything, but in any case, it is still an impressive result, especially at such low voltage.
The rest of the specs include 4GB of DDR3 memory and ASRock’s upcoming Z87 Extreme4 motherboard.
Intel is rather slow when it comes to the adoption of new wireless standards. Most, if not all, notebooks based on Intel platforms today feature 802.11n capable wireless and with the help of a few antennas it can get you between 150 and 450Mbits.
In reality 802.11ac is usually much slower than 150 to 450Mbits but since the middle of last year 802.11ac routers started to show up all around the world. This new standard can get you to 866Mbits and even higher, but Intel has been rather slow to adopt it.
Intel has promised that both Shark Bay notebook and desktop platforms for 2013 will get support for 802.11ac. The card is based on a 2×2 dual band configuration and will support speeds up to 867Mbits per second, in addition, it will support wireless 1080p display, Intel smart connect, Intel Vpro (only with Y and U processors for notebooks) and Bluetooth.
This is the first product based on 802.11ac but we believe that with time Intel will add more choices to its wireless portfolio as 3×3 802.11ac configuration should potentially run even faster. It will be interesting to test this new card in the real world and see if 802.11ac wireless can get you any faster than 802.11n in real life applications.
A few years ago it would have been impossible for Intel to acquire AMD, simply due to regulatory constraints put in place by the FTC and the European Union. Intel had more than a 60 percent of the PC and notebook market, so picking up AMD, a company that has some 20 percent of the market, would make Intel a real monopoly.
In the last two years the iPad, smartphones and ARM based tablets have changed the landscape, eating up Intel’s revenue and market share. It is true that most people, especially professionals and the business crowd, use x86 processors, but this is rapidly changing as home users are happy with emailing, browsing and playing some games on their iPad or other tablets. This puts Intel in a world trouble, as the PC market nosedived by 14 percent last quarter, due to a lack of interest for new devices and upgrade.
Tablets are becoming couch browsing devices, people use their smartphones to read news on the go and sometimes at home. More and more users don’t even touch their notebooks or desktops at home. With ARM staying the dominant instruction set in the phone and tablet space, Intel is facing a serious issue as Apple, Samsung, Qualcomm and Nvidia are all making money on ARM chips.
With this in mind, this would be the main reason for Intel to pick up AMD. AMD would not cost them that much, as Intel still has billions in bank, but with AMD, Intel would gain great graphics, something that the company has been struggling to crack for many years. It would make Intel slightly more competitive, but it would not solve all of its problems.
ARM manufacturers also face challenges, they need to produce more powerful chips and deliver a better user experience in order to win more notebooks and detachable devices, but this is going well with non-Apple based tablets. Apple uses ARM, so in the tablet world ARM is winning this fight, but Qualcomm and Nvidia as two independent chip manufactures could do a much better job at getting popular design wins. The Snapdragon S800 and Tegra 4 will get these two companies a step closer, while Apple will continue making good chips for iPads and iPhones. Let’s not forget about Samsung, as it makes many chips for its phones and tablets.
AMD gained 14 percent on May 1st, and an additional 5.9 percent yesterday, getting its stocks up to $3.41. Back on April 30th, AMD stock was trading at $2.68. In last three days of trading AMD gained 27.24 percent or $0.73 per share, which is a huge leap for a company with a 52-week low of just $1.81
Intel has announced that it will launch its next generation Haswell processors at Computex.
Intel showed running Haswell silicon to journalists last month at the Game Developers Conference (GDC) in a bid to talk up the upcoming chip’s GPU. Last Friday the firm announced what some already knew and many had already guessed, that it will launch Haswell at Computex in June.
Intel published a blog post on 26 April saying that the fourth generation Core processor known as Haswell would arrive in 3,337,200,000,000,000 nanoseconds, which worked out to just under 39 days. The countdown figure matched perfectly with the start of Computex on 4 June, and confirmed what an Intel insider said that the chip would be launched at Computex.
The fact that Intel is using Computex to launch its next generation chip is not surprising, given that there are few big IT shows during the summer and launching the chip later will not give the firm’s system builder and OEM partners enough time to gear up marketing for the lucrative back to school and holiday buying seasons.
While Intel’s Haswell launch is a big event for the firm, it isn’t the most important. Rather, the firm is expected to launch updated low-power Atom chips that it hopes will help it compete in the tablet market, a market that is growing, as opposed to the PC market that Haswell addresses.
Intel’s decision to launch at Computex means that the late spring computer industry show should be awash with updated notebook and desktop PCs, as well as the firm’s preferred ultrabook branded laptops.
We have already mentioned Intel’s one-chip Haswell platform on several occasions, but we have managed to get a few extra details about this chip. As we have stated many times before 1 chip Haswell has BGA packaging packed with SoC that integrates a Haswell CPU as well as Lynx Point LP PCH chipset inside.
The SoC packaging leads to lower production costs, power footprint and lower TDP, everything that you need in order to drive the SoC price down. We remembered Dave Orton, the former CEO of a company acquired by AMD that went by the name of ATI, explaining the importance of APUs and SoCs. The explanation is rather simple, the more you integrate the cheaper the chip ends up, the fewer pins you have and theoretically you can make more money. This conversation happened in the summer of 2007, roughly a year after AMD acquired ATI and announced its plans to produce Fusion APUs.
Since the top ARM chips such as Qualcomm Snapdragon 800 / 600 and Tegra 4 have multiple cores, chipset elements and graphics all on the same package, it was only natural for Intel to take the same approach with Haswell. Qualcomm, Nvidia and Intel are after the same market, tablets, notebooks, convertibles with a slight advantage that Intel has X86 and the other two don’t.
Let’s not forget that Haswell 1- chip is much bigger than any ARM based top performers, but at the same time Haswell brings a lot more performance. Despite billions of transistors and 22nm SoC design tablets and Ultrabooks based on 1-chip Haswell or Haswell-ULT how some call it, you can expect 8- to 10-hour battery from products based on this Y processor line chips. This is a respectable score for PC like performance and having a scenario design power (SDP) typical expected TDP at 7.5W these products come close to the top ARM performers that have 5+W TDP.
Intel stresses that these chips won’t simply land in tablets and Ultrabooks. It plans to use them in detachable, foldable and similar designs usually represented as the result of an unholy coupling between a notebook and a tablet.
The bad thing is that Y line of tablet, Ultrabook, de-attachable, switchable SoC Haswell chips only comes in Q4 2013 so we are in for a pretty long wait.
Haswell will save your battery, Haswell has connected stand by, it promises higher performance per clock and for some it is important that. It also gets significantly better graphics.
Since Intel makes huge dies, it won’t be a problem to squeeze some L4 (fourth level cache) to boost memory bandwidth and lower the latency in some of its Haswell SKUs. The Haswell variant that is internally known as Crystal Well offers much larger L4 cache.
The size of the cache is not clear, but we heard that there can be up to 64MB cache dedicated for graphics. This does sound a bit too much,
From the mouth of engineers in the Far East, it could be that the L4 cache remains dedicated only for the GPU, but the other independent sources claim that L1, L2, L3 and L4 memory will be shared between CPU and GPU. We will have to look into which of these two theories is right.
Crystal Well is reserved for GT3 based highest end processors from Intel, and we have heard that it remains an exclusive technology for Core i7 processors. You will have to pay up to enjoy it.
L4 cache is nothing new to the GPU world and consoles have been using such cache to make the texture and antialiasing faster on them. Dedicated cache on GPUs has been considered by Nvidia and ATI (even before the 2006 AMD acquisition) for years. The main obstacle was always that the transistor count for GPU cache memory was very high and it would result in a huge chip, something that semiconductor manufacturers tend to avoid.
It will be interesting to see Haswell Crystal Well in action when it launches later this year, but we are certain that we can see a huge performance leap from Intel Ivy Bridge 4000 series graphics.
Intel and Cisco have struck a manufacturing partnership, to make Cisco’s networking chips on a contractual basis.
The scoop came from the Korea Times though neither company has made an official announcement. It quotes Lee Hee-sung, country manager at Intel Korea as its source. The deal with Cisco looks significant. If Intel successfully produces chips with designs offered by Cisco, then it will get additional momentum to effectively grow its foundry business.
Intel has been expanding its foundry businesss on a relatively small scale, with customers like Netronome, Achronix, Tabula, and most recently Altera.
The deals so far revolve mostly around FPGAs or field-programmable gate arrays. However this Intel is looking for new ways to make cash until its main PC business picks up. It also has a lot of half used Fabs needing something to do.
However the Koreans have stolen a march on the rest of the world as the Intel in the US has declined to comment and Cisco has not said anything yet.
Intel has released an OpenCL 1.2 software development kit (SDK) for its upcoming Haswell processors.
Intel launched an updated graphics driver yesterday, touting OpenCL 1.2 support for its present generation Ivy Bridge processors, and last week during a press conference it claimed that was actually designed for Haswell. The firm followed that driver release with a far more substantial OpenCL SDK release for its Ivy Bridge and upcoming Haswell processors.
Intel’s decision to support full profile OpenCL in its HD Graphics 4000 core marked the first time the firm acknowledged the power of GPGPUs. Now it seems that Intel is perfectly happy to help developers make use of the GPU for computing purposes in addition to its CPU cores.
According to Intel OpenCL is particularly well suited to content creation applications, workloads the firm has been promoting. Last week at the Game Developers Conference Intel talked up its collaboration with video encoding project Handbrake, claiming that its Quicksync video codec could dramatically improve encoding times.
Given the prevalence of Intel’s GPUs in the market, the firm’s support of OpenCL is vital to widespread adoption. The problem for Intel is that should developers make good use of its GPU and its full computing capabilities, the firm’s investment in CPU engineering could be undermined.
Intel said that the GPU in Haswell chips will not only support OpenCL 1.2 but will be able to access main system memory, which opens up interesting computing possibilities if memory access is truly unbound. The firm didn’t release any technical details about its Instantaccess technology last week, but did say that it is a feature that has the potential to be widely used.
Intel’s OpenCL 1.2 SDK is available for Microsoft’s Windows 7 and Windows 8 operating systems.
Intel is working with the US Army’s Orlando simulation-research lab to create computerized war games capable of handling hundreds of participants at once.
The goal of the project is to create a computing network powerful enough to deliver interactive training simulations to large groups of players around the world. Of course those who have seen the flick Wargames can be re-assured that the computer will not be plugged into the US weapons grid.
Using “cloud computing,” the new system would eclipse not only the military’s current remote-training systems but also commercial “massive multi-player” websites. Intel’s Mic Bowman said that Second Life is the closest thing to what we’re doing, but even that limits the number of players to 60 or 80 per region. Apparently that’s not nearly enough for the kind of higher-level engagement training the Army needs to do, but Intel’s system will be able to support at least five times that many.
The research is part of Intel’s experiments in to Cloud computing. It wants heavy duty network software out of the project.
In the universe where Haswell comes as two-chip platform, formally referred to as Shark Bay 2-chip platform by Intel, there will be three mobile chipsets. The HM86 is targeting at mainstream consumers, HM87 targets Premium consumers and SMBs, while the top one QM87 targets the hard working corporate market.
Things get a bit simpler with the Shark Bay 1-chip platform. The value consumer chipset is simply called Baseline, while the next one is simply called Premium. It can’t get any simpler than that. The Shark Bay 1-chip Platform has same I/O, or what we call a chipset, integrated on either Haswell U or Haswell Y processors line.
The Premium chipset supports Windows 8 connected stand by, Intel Active Management Technology 9.0, Intel Small Business Advantage, ACHI and Raid Rapid storage technology, Intel Insider, Intel Anti-Theft Technology, wireless display and three independent displays.
The list goes on with 8 USB ports where two to four can be USB 3.0 ports, up to 6 devices with PCI express 2.0 5GT/s, four SATA ports capable of 6Gbps. The chipset doesn’t have VGA or LVDS as the CPU has the graphics on it, but the Premium chipset has two sensors interface with I2C and UART, 1.5 to 5MB firmware support, Anchor Cover, Platform trust technology and Platform Flash Armoring technology.
Getting away from numbers and naming the chipset simply Premium means that U and Y line of CPUs are meant for tablet and Ultrabook markets, where manufacturers want you to love the product as a whole, not specifics. Premium and Baseline chpisets for Y and U line Haswell processors capable of TDPs as low as 13W are coming in Q3 2013.
We found out peculiar fact that some members of Intel’s M Processor line, 35W dual-core products, won’t get the Haswell upgrade until Q4 2013. Haswell starts in very late Q2 2013 and is still scheduled to launch in late May or early June, but most of the parts are aimed at the very expensive quad core MX line and MQ line edition parts.
When it comes to dual-core 35W Core i7 and Core i5 Haswell parts they won’t come at least until Q4 2013, at least this is the current part. Core i7 3540 launched in this quarter, Q1 2013, and it is a 3GHz dual-core with four threads and a top turbo frequency set at 3.7 GHz. It is a 35W, 22nm Ivy Bridge part with 4MB of cache memory. A replacement part might be on the way with a slightly higher clock in Q3 2013, but the Haswell based replacement is set for Q4 2013. With an official price of $348 it is not really the cheapest kid on the block.
Core i5 3380 remains the fastest dual-core 35W Pentium part until Q4 2013 Haswell reinforcement. The 2.7GHz / 3.4GHz turbo clocked dual-core will remain the fastest in this league at least until Q3 2013, when it might get slightly faster version of the Ivy Bridge based core, but it won’t be replaced by Haswell 35W dual-cores before Q4 2013.
Intel definitely wants to prioritize the quad-core 55W i7-3940XM $1096.00 replacement called Core i7 4930MX and Core i7-3840QM replacement in $568 market segment, branded as Core i7 4800 MQ, as it can simply make more money on these pricey these parts. These quad-core Haswell parts start selling in Q2 2013, followed by 17W Ultra low voltage dual-cores in Q3 2013 and only after these two lines rolls out, Intel will introduce the rest of the Haswell line-up.
According to a report over at Hardware.info that managed to get their hands on an internal Intel document, it appears that Intel’s Haswell platform might have a problem with its USB 3.0 host controller.
Although it is not as serious as the Cougar Point SATA 3Gbps bug, the USB 3.0 controller on Haswell platform will have issues with the S3 sleep mode and devices that are connected via USB 3.0 port. Apparently, when waking from S3 sleep, applications that are accessing the data from, for example, USB 3.0 storage device might freeze and force the user to reopen them manually.
Thankfully, the bug will be more of a nuisance rather than a problem as any loss of data is excluded. Intel does not plan to delay the launch and it is still scheduled for mid-2013, according to an Intel representative comment for Hardware.info. Intel is apparently still researching what other consequences this issue could possibly have and plans to resolve the problem in a future CPU stepping.
Intel claims its parts can outperform ARM chips in benchmarks and its manufacturing process lead should help it deliver faster and smaller chips. However, in spite of Intel’s claims, few vendors seem interested in its mobile chips.
Speaking to CNN, Intel mobile chief Mike Bell stressed that Intel has the software and systems competence to be the most successful player on the market. He pointed out that Intel can develop software to get the most out of its hardware and that Intel single core chips outperform multicore ARM designs.
“It’s a question of whether you’d rather have a jet engine or two propellers,” said Bell.
Granted, Bell has to tout the company line, but his engine comparison works both ways. Crop dusters and ultralight planes don’t need jet engines, or two piston engines for that matter. That is what really matters and Intel knows it. Not everyone needs a turbojet or turbofan, and not everyone needs an Intel core, especially not in mid- to low-end devices.
Intel believes its next generation 22nm mobile parts, with integrated LTE, will allow it to score some tablet and smartphone partners in late 2013 or 2014. However, Intel will have nothing to take on new A15 class ARM chips this year.
So far we are aware of a few different Haswell graphics cores. The main distinction is that the G3 is more powerful and faster, while and GT2 is smaller, less powerful and obviously more efficient.
The fastest of all of them for notebook computing is GT3 with cache memory on board that ends up with Intel HD Graphics 5200 branding. This part goes into high performance notebooks and we can imagine that this core will find its place in quad core Haswell processors.
The GT3 for Ultrabooks will end up branded as Intel HD Graphics 5200 for the faster one and Intel HD Graphics 5100 for a somewhat slower version. We don’t know the clocks or the actual specification for the 5000 series, just brands. GT3 for Ultrabooks doesn’t come with on package memory, probably due to a limited space on this dual core.
Intel GT2 HD graphics target high end notebook computers at high clocks and they end up with Intel HD Graphics 4600 branding. Intel GT2 HD graphics for Ultrabooks will end up branded as Intel HD Graphics 4400 for the faster one and Intel HD Graphics 4200 for the slowest of the GT2s for the mobile.
Intel also claims that same graphics numbering system applies across client and server products.