TweakTown NewsRefine News by Category:
It's about time we got some competition out of China, where they're close to unveiling details on a new 8-core processor designed and developed entirely within the country of China. The new CPU is a custom design and ready to take on competitors Intel, AMD and ARM in the PC, server and supercomputer categories.
Development on the chip began over a decade ago in 2001, by Loongson Technology, a group that receives funding from the Chinese Academy of Science. The first processor they developed was the 32-bit Godson-1, which was followed by the 64-bit Godson a few years later and since then, the chips have been used in low-power notebooks since 2008. Late last year, the Shenwei supercomputer was turned on, and powered by the same design.
The latest chip out of the company is the Godson-3B1500, which sports a 1.35GHz clock speed, and is capable of performing 172.8 gigaflops of performance while chewing just 40W of power. The chip is manufactured under a 32nm process, sporting 1.14 billion transistors. Loongson CPUs aren't compatible with Microsoft's WIndows OS, which will hinder its adoption - big time. There are still plenty of places this chip is destined to reach - mobile devices being an absolutely huge market, for example.
For as long as I remember, I've loved having the ability of upgrading the CPU in my computer - back in the days of the original Pentium (and even before that), and the glory days of the Celeron 300A - the option was always there, but it could have its days numbered.
According to a report, starting with Intel's Broadwell generation of processors, Intel will only offer mainstream desktop CPUs in BGA packaging - killing upgrade options, and hurting PC makers. Starting with Broadwell, Intel's CPUs will depart the current land grid array (LGA) and micro pin grid array (µPGA) packages, and only offering chips in ball grid array (BGA) form factors, just like their Atom processors.
Japanese site, PC Watch, has said that the Haswell processors might be the last desktop chips arriving in LGA packaging. This would mean you won't be upgrading your CPU anymore, with no way of changing the CPU on the motherboard when this happens. It all begins with Broadwell in 2013, where all mainstream desktop processors being made available in BGA.
This might not sound like a threatening move by Intel, but by doing this - you're going to receive a CPU physically soldered onto your motherboard. This is no easy feat, and will be done in very sophisticated manufacturing facilities. Maybe we're going to see a throwback to the old school days of physically modding your motherboard or CPU to achieve something better - or an upgrade in this case.
Some leaked roadmaps of Intel's next-gen Atom tablet chips have been made available, teasing the new chips that will be baked into tablets in early 2014. The new chips sport quad-core CPUs, and DirectX 11-capable GPUs. All of this is built on a 22nm process, named Valley View-T.
Bay Trail-T should arrive before hand, and will be given out to OEMs to bake into their tablets. We should see these new Bay Trail-T-based devices teased at CES 2014 (we're already talking about CES 2014!!) and the main competitors should be working with ARM Cortex-A15 architecture SoCs, or NVIDIA Tegra 3 SoCs. Intel's new Bay Trail-T, when compared to its predecessor, Clover Trail, is quite impressive. Built on a 22nm process, it pushes new lower power consumption numbers, as well as increased performance. Clover Trail finds itself on the 32nm process, and its competition, the ARM Cortex-A15 being built on the 28nm process.
Moving on to Valley View-T, where we'll see Intel drop the hyper-threading capabilities, but increase the core count from two to four. Moving from a dual-core design, to a proper quad-core design should see performance improvements of around 50-60%. Backed up by a new GPU based on the HD 4000 series, as well as new LP DDR3-1067 RAM, we should expect performance to increase, big time.
If you're in the market of building a new server, folding farm, or other system in which you wish to stuff many cores, the new Opteron 6300 series of processors could be the ones you're looking for. AMD has officially launched the new series of Opteron server processors, which feature the updated Piledriver cores.
AMD is saying that performance of these new chips is up to 24 percent better than that of the Opteron 6200 processors, which are based upon the older Bulldozer architecture. Piledriver cores are an updated version of Bulldozer that look to lower power consumption, while increasing performance.
The new chips feature varying core counts, from lowly quad-cores all the way up to 16-core monsters. Clock speeds range from 1.8Ghz all the way to 3.5GHz. With Turbo Core, the max frequency of some parts will boost all the way to 3.8GHz. The majority of the parts are in the 115W TDP category with the 6386 S being a 140W part.
We all know Samsung doesn't just twiddle their thumbs after their release a flagship smartphone, where earlier this year we saw the release of the phenomenally popular Galaxy S III, and now we're seeing the rollout of the 5.5-inch Galaxy Note II. These two smartphones represent just a sliver of what Samsung actually sell, and now we're hearing more talk on the Galaxy S III's successor, the Galaxy S IV.
Sources are claiming that Samsung recently produced a test version of their 28nm High-K Metal Gate (HKMG) processor which goes by the codename "Adonis", that will eventually be baked into their next-gen flagship smartphone. Adonis is a quad-core design, which will most likely end up with the name Exynos 5400, and uses ARM's Cortex A15 architecture. The new chip should ship with better GPU abilities, that reportedly use very little power. Samsung currently use the Cortex A15 architecture in their Exynos 5250, but that is built on the older, and less efficient 32nm process.
ARM have announced two new Cortex-A50 series chips that will bring forth 64-bit compatibilities to ARM SoCs. ARM have a bunch of licensees of the new chips including AMD, Broadcom, Calxeda, HiSilicon, Samsung and STMicroelectronics.
ARM's press release points to two specific chips, the Cortex-A57, which is a high-performance solution that would most likely see the light of day in servers, and the second is the Cortex-A53 which has the same capabilities of the A57, but is much more power-efficient. Better yet, the two chips can be combined into the single package if needed with the A53 cores taking care of low-impact workloads and stay on while the system sits in the idle, while the A57 cranks up only when the workload requires it.
This type of processor layout is called "big.LITTLE", according to ARM, who already offer licensees the ability to pair up a Cortex-A15 CPU with a Cortex-A17 chip to receive similar results. ARM says that the new processors should scale well enough so that they get baked into smartphones, tablets, laptops and servers. Being 64-bit capable is another huge feature, but they'll also fully support 32-bit programs and operating systems, too.
AMD licensed the ARM architecture last year and not much was said about it. Move to today and you'll find a massive new use of the ARM technology. AMD is planning on building server chips (Opteron) using a 64-bit ARM architecture. Though there are no product announcements today, some see this as an important first step for the company.
Interestingly, AMD has a processor license, not an instruction set one, so AMD will be integrating an ARM-designed core into their Opterons, rather than designing their own. This has both its advantages and disadvantages. For one, AMD doesn't have to spend R&D costs designing a new ARM core, however, they aren't spending R&D costs to design a core that could function better in the server market.
ARM-based Opterons would be great in the microserver market. However, that is currently a small market, though its a market that Intel doesn't currently compete in. If AMD is successful with this, they could end up having some very exclusive access to the market, at least until Intel throws lots of money at building a chip.
The new AMD Vishera desktop CPUs are proving to be quite the overclockers. Vishera is the codename for the new desktop chips that are taking advantage of the updated Piledriver cores. They're said to offer 15 percent more performance per clock. It appears they have also retained their incredible overclocking ability.
As you can see in the image above, the new record saw the FX-8350 overclocked to 8.176GHz, which blows away the previous record of 7.443GHz. The feat was achieved by NAMEGT. NAMEGT utilized a Crosshair V Formula-Z motherboard and a dual-channel 8GB memory kit. Of course, liquid nitrogen was used for cooling.
The core voltage was a super-high 1.932V. The chip features a multiplier of 29x and an external frequency of 281.94. This resulted in an ultimate clock speed of 8176.5MHz, an impressive feat. It will be great to see what these chips can really do when more overclockers get them in hand after their launch. New clock speed record? Seems like a possibility.
Intel's next-gen Atom currently slides along with the name of "Avoton", and is the chipmakers upcoming system-on-a-chip (SoC). Avoton will be based on 22nm Silvermont architecture and will reportedly sport 8 cores sharing 4MB of L2 cache (1MB per core).
Avoton's eight-core processors will feature clock speeds of between 1.6GHz and 2.4GHz, while including OOE (out-of-order execution) as well as Turbo Boost. Current Centerton-based Atom processors should be left behind performance wise by these new Edisonville platform-based Atom processors.
Turbo Boost will see the clock speeds reach around 2.7GHz which is quite good considering these new SoC will only come with a TDP of 6 to 8.5W. The Avoton SoC will have a TDP of somewhere between 6 and 20W. Avoton will improve upon Centerton's formula by throwing more features on-die, which will include gigabit ethernet, USB 2.0 and USB 3.0.
It always starts the same way with Intel: they release their new, high-end architecture to the top of the product line at the start. The remaining older silicon and design continues to be sold as the low-end chips to deplete supply. The failed high-end chips and excess supply then start trickling down into the lower-end chips.
Ivy Bridge was launched all the way back in April, a fair time ago considering the speed at which PC innovations continue to move at. It was only last month that Ivy Bridge finally made it into the Pentium line of Intel CPUs, and by Q1 2013, it will be nearing an entire year since the original launch of the architecture.
The new Celeron chips will reportedly be named G1610, G1610T and G1620. They will still feature two cores and 2 MB of L3 cache. They will also continue to lack Hyper-threading and Turbo Boost. The new Ivy Bridge models will support faster memory as well as a slight speed bump on the bottom model. The top two processors will feature a lower TDP.