TweakTown NewsRefine News by Category:
As always, this is just a rumor, and it comes from SemiAccurate, which by name alone isn't accurate all the time. But, Charlie has some interesting news, and according to his sources, "GT3 variants" of Haswell will sport an insane 64MB of memory sitting on the CPU package itself, right next to the processor die.
This memory will be enough to accommodate the frame buffer, which should help performance bottlenecks which are tied to the main memory interface. Earlier reports were talking about desktop variants of Intel's Haswell sporting a dual-channel DDR3 memory controller, and we all know that there's only so much bandwidth you can suck out of a dual-channel memory controller.
In an earlier report by Charlie, he said that Haswell's integrated GPU will have five times the shader performance of Sandy Bridge's IGP. If we take into consideration the 64MB of onboard memory, we should be in for some pretty big performance improvements, if this rumor turns out to be true, that is.
Details are leaking for NVIDIA's next-generation Tegra, which for now is called Tegra 4. Tegra 4 is built on a 28nm process, and according to VR-Zone's tipsters, should bring with it GPGPU support to use its graphics core for boosting CUDA, OpenCL and other general-purpose tasks.
The addition would push it a few notches up the road, as it would sport a graphics core that is multiple generations newer than what is in the Tegra 3, which dates back to the 2006-era PlayStation 3. What's being thrown around as a rumor right now is that NVIDIA could opt to use the current-generation Kepler core in Tegra 4.
NVIDIA CEO Jen-Hsun Huang has already teased that Kepler would arrive in smartphones, but he didn't state whether this would be sooner rather than later. If NVIDIA's trend continued, we won't see Kepler-based cores in Tegra for years, as the current Tegra 3 only sports NV40-class cores. There are more speculative rumors, though.
AMD have integrated a real-time operating system (RTOS) into their G-Series of processors. What does this entail, you ask? Well, AMD is now loading Green Hill's Integrity OS into the platform. The embedded system is capable of a bunch of tasks where space and power are at a premium.
This would include devices such as TV set-top boxes, other electronic home devices, industrial control systems, networking, the military, and medicine. You know, just a few of the largest markets in the world.
AMD's G-Series of processors combine a low-power CPU with a discrete graphics core into a single form-factor. The RTOS then provides intelligence upon which application specific code can be written. The Green Hill Integrity OS has integrated middleware functionality that includes support for a bunch of necessities like USB, Wi-Fi, Bluetooth, and IP routing and switching.
The tech is used across the market in devices like printers, network routers, and cars, which often have to handle actions immediately and with a minimum of overhead.
After a couple of delays, it looks like Intel is on track to deliver one side of the Ivy Bridge coin on April 29 according to two different sources, Sweclockers and Donanimhaber. Both desktop and mobile IBs will be released at the same time if these reports are correct.
Just over a week from now on April 8, Intel will announce Z77, Z75, H77 and B75 chipsets. On the same day, we should see a slew of reviews for Z75- and Z77-based boards, sporting Sandy Bridge-based CPUs. Ivy Bridge processors are expected to take the stage in the fourth week of April, somewhere between April 22 and April 28.
The announcement will only include quad-core models, Core i7 and Core i7 Extreme families, desktop Core i7, Core i5-3570K, i5-3570T, i5-3550, i5-3550S, i5-3450 and i5-3450S. The chips that get announced should be made available for sales on April 29, with benchmarks and reviews hitting the same day. Specifications and pricing on said processors is below:
On June 3, we should get the second side of the Ivy Bridge coin with the following chips: Core i5-3570, i5-3570S, i5-3475S, i5-3470, i5-3470S and i5-3470T desktop CPUs, and Core i5 dual-core mobile microprocessors. Below is a full list of third-generation Core-branded processors:
What do we have here? An official document on Samsung's upcoming Exynos 5 dual-core System-on-a-Chip (SoC). From the document, we can see it will sport 12.8GB/sec of bandwidth that runs through two 32-bit 800MHz memory ports, capable of providing support for both LPDDR3 and standard DDR3.
The chip will also support USB 3.0 and SATA 3.0, which is a great step forward for Samsung and future devices that will feature this Exynos chip. It will also feature support for HSIC and C2C interfaces, which will give the chip the capability of LTE. High resolution image sensors are also getting baked in, which should provide support for handling 8-megapixel at 30fps, very nice.
The nitty-gritty is that the new Exynos chip sports an ARM Mali-T604 GPU and a dual-core ARM Cortex-A15 which should be clocked between 1.7 and 2.0GHz, where we should expect performance to reach 14,000 DMIPS, which is 0.2DMIPS/MHz per core faster than Qualcomm's new Snapdragon S4 CPU, and 1.0 DMIPS/MHz per core faster than NVIDIA's Tegra 3.
Everyone knows how GlobalFoundries' 32nm yields started out; its no secret that they were quite poor. AMD even went so far as to go onto record regarding the issue, blaming the poor yields for their lackluster financial performance. The occasional rumor was even heard that AMD may part ways with them at 28nm.
But, the firm is claiming that its 32nm process ramped up quicker than its 45nm process, allowing more than double the amount of wafers to be made during the first five months of production. GlobalFoundries CEO Ajit Manocha acknowledges the initial poor yields, but explains that some operational and organizational changes were made. He hints that these changes will carry over to 28nm.
"In just one quarter, we were able to see more than a doubling of yields on 32nm, allowing us to exit 2011 having exceeded our 32nm product shipment requirement," he says. "We are committed to moving ahead on 28nm with GLOBALFOUNDRIES." AMD appears to be happy with GlobalFoundries and they seem to be planning on using them for their 28nm production. Only time will tell if the changes have really helped with 28nm.
Sandy Bridge-E owners can breath a sigh of relief, their LGA2011-based CPUs will still be the king of the hill for a while longer now thanks to an X-bit Labs report. Ivy Bridge-E, the hypothetical successor of Sandy Bridge-E, which is an Ivy Bridge take on the LGA2011-HEDT platform, won't see the light of day until 2H of 2013.
It's being reported that we won't see Ivy Bridge-E for quite some time because of Ivy Bridge's delay, which has caused a massive shift in Intel's roadmaps. This will also delay the Haswell-based chips several months to March-June 2013.
This all doesn't really mean much, as Intel aren't really being pushed by their competition, AMD, at all. Could this be a sly way of just delaying their products, and polishing them in case AMD do come out with something unexpected? Or are Intel just widening the gap of CPU releases because, well, they can?
IBM's Holey Optochip is capable of 1 terabit per second transfer speeds, holey transfer speeds, Batman!
IBM have reached quite the milestone, with researchers at the company demonstrating a prototype optical chip that is capable of transmitting up to 1 terabit of data per second. Engineers managed to built the Holey Optochip using readily available components, meaning that the chip could actually be manufactured in high volumes for commercial implementation sooner, rather than later.
IBM created the Holey Optochip as part of their continued efforts to use light rather than electronics over wires to transmit huge amounts of data. Surprisingly, the Holey Optochip was created using a standard 90nm-based CMOS chip, which they then drilled 48 tiny holes through the back of it.
What this allows is for 24 receiver and 24 transmitter channels, with each capable of moving 20 gigabits of data per second. This results in a product that can move 960 gigabits (nearly 1 terabit) of data per second. For this next bit, I hope you're sitting down: this means you could transfer 30,000 HD movies in 60 seconds. 30,000 in 60 seconds.
Intel have just published publically accessible information all of their upcoming 22nm-built Ivy Bridge processors. This is as official as it gets, before we see the tasty CPUs themselves on the stores of retailers around the world.
Intel has shown something that they call a 'flipbook', which is something that would normally be a printed product guide that would normally be given out to its various partners as a hand-out to let their customers keep track of what all the CPU models are and their specs.
The PDF was posted on one of Intel's partner sites called the Retail Edge Program, and it's not accessible by the general public. But, this is the Internet, and we have tricks and stuff. The specific PDF can be found by simply searching for the various upcoming CPU models. Oh yeah.
Intel have confirmed that their next-generation, 22nm-fabbed Ivy Bridge chips will go on sale eight-to-ten weeks later than they had originally planned, according to Executive Vice-President of Intel, and chairman of Intel China, Sean Maloney.
Maloney talked about the delay in his first interview to discuss Intel's business in China, where he told the Financial Times that the start of sales that sport Ivy Bridge had been pushed back from April, where he says "I think maybe it's June now". First off you "think", Sean? How can you not know this?
Maloney said the adjustment was due to problems with the manufacturing process, and also stated that the global launch of the new Intel-powered smartphones would follow the launch of other global launches this year. An Intel-powered Lenovo-based phone would become available in other countries four to five months after it's initial launch in China. Maloney also denied reports that Ivy Bridge was pushed back due to the over-stocked inventories of Sandy Bridge-based notebooks, citing manufacturing process issues.