TweakTown NewsRefine News by Category:
AMD are continuing on with their APUs, having named a successor to their not-even-released 'Kaveri' APU expected in 2015. Kaveri is their Steamroller-based x86 CPU with a GCN-based GPU and HUMA memory architecture, and won't see the light of day until the second half of 2014.
But, AMD have named its successor already, which will be known as Carizo. Carizo will arrive in 2015, and will feature even more improvements on the unreleased Kaveri APU. Carizo will probably feature AMD's upcoming Excavator CPU core and hopefully an even better rack of GCN GPU cores. We should expect the memory architecture to receive an injection of speed improvements, too.
Earlier today on Twitter, Samsung announced that next week it will be launching a new Exynos Evolved 5 octa-core processor. The company's current flagship processor, the Exynos 5 Octa, was revealed earlier this year at CES and is just now began being integrated into handsets but as all things go in the electronics world, it appears that a new and improved model has already been developed.
Samsung says that the new processor will be more powerful and enhanced and while no other information was given, we're kind of hoping to see the Exynos built on a 22nm scale which could bring us closer to seeing 2 GHz to 2.5 GHz clock speeds. The current Exynos is built on a 28nm process and its flagship chip is clocked at 1.8GHz.
It was just two years ago that a 1 GHz processor in your smartphone was considered a big deal, but by today's standards 1 GHz is pretty slow. Today, ARM announced that it will be ramping up clock speeds in its mobile processors to the 3 GHz range in 2014.
A statement from the company reveals that it has plans to move away from the 28nm process and begin manufacturing chips on a 20nm process which will set the stage for the 3GHz core speed. TSMC, the fabrication house which has signed a deal to manufacture Apple's new line of chipsets, has said that the new process is expected to deliver a 30 percent increase in clock speed while at the same time a 25 percent reduction in heat. These gains can be attributed to efficiency gains from the next-gen manufacturing process.
ARM says that most of the new transistors added will be dedicated to graphics processing performance. The 25 percent reduction in power function will translate into a 25 percent boost in battery life. Since battery life is still the top complaint when it comes to smartphones ARM says that consumers should look forward to the new 20nm process.
When it was first announced that Intel was using thermal grease instead of solder to attach the CPU die to the heat spreader on Ivy Bridge chips, the overclocking world became very disappointed. The thermal interface material that was used really left a lot to be desired, and as such, Ivy Bridge chips were not as overclocking-friendly as previous generations.
This morning a forum post by user "Toppc" on the Taiwanese website Coolaler has unveiled what appears to be a next-generation Ivy-E ship, which has been delidded to unveil that Intel has once again switched back to solder as the connecting material between the heat spreader and CPU die.
The chip in question is said to be an Ivy Bridge-E Core i7 4960x, which should retail in the $1000 range on release. Fortunately for us, Toppc cared less about his expensive CPU and more by showing the world what was inside. With Intel switching back to solder, we can most likely expect very respectable overclocking numbers coming from Ivy-E this September.
This morning a slide surfaced that appears to be part of an Intel presentation and details information about the upcoming enthusiast platform from Intel, Ivy Bridge-E. The slide appears to be legit, but faking a PowerPoint slide is not very hard, so keep that in mind while reading.
The leak comes from website HardwareLuxx and gives us a new release date as early as September 4th, which is two months earlier than previously speculated. The slide also alludes to a very strict NDA around the new processors. The slide also discloses that the new Ivy-E processors are built on a 22nm process based on Intel's 3D Tri-Gate Transistors.
Continuing the leak, CPU World has posted information on processor SKUs starting with the i7-4960X, featuring 6 cores (12 threads) at 3.6GHz with turbo up to 4GHz and includes a total of 15MB L3 cache. This is the top-end SKU, which should retail for $999-$1099.
Engadget have had some hands-on experience with Qualcomm's new MDP devices, which are powered by the company's Snapdragon 800 SoC (MSM8974). The tablet the site used is an 11.6-inch device featuring a 1920x1080-pixel display.
Also inside, we find 2GB of LPDDR3 RAM, 32GB of built-in flash storage with microSD expansion capabilities, a 12-megapixel auto-focus rear-facing camera and a 2-megapixel camera up front. The device itself is quite thin, measuring in at just 0.46 inches thick (11.7mm) and has a 3400mAh Li-ion battery crammed inside. Connectivity wise, you're covered with LTE, Wi-Fi ac, Bluetooth 4 LE, GPS and NFC. Pressure and humidity sensors also make an appearance.
The phone side of things features a 4.3-inch 720p display with a 1500mAh battery. Engadget chucked the Snapdragon 800 SoC into the benchmarking ring, and found some truly ridiculous results. In just a few tests, it truly blows the competition away - completely. Things like AnTuTu 3.x score 13,836 on the current Snapdragon S4 Pro, compared to a whopping 33,828 with the Snapdragon 800 doing the work.
Earlier today, we covered the announcement of AMD's ARM-based server architecture codenamed Seattle. Following that announcement, AMD laid out the next generation of its x86 server compatible chips.
First up is Berlin which will replace the Opteron 3300 series at the end of this year. From looking at the specifications, Berlin appears to be a server-oriented version of the upcoming Kaveri APU which includes the Steamroller CPU microarchitecture. Berlin will include accommodations for servers including ECC support any internal caches or memory, as well as uniform memory access which is promised by AMD to be the next big thing.
Up next is Warsaw which appears to be a very small update to the current eight-core Opteron chips. It will only be offered in dual-chip package configurations with 12 to 16 cores. Warsaw will be built on a 32nm process and will feature the current Piledriver cores instead of the next-generation Steamroller architecture. It will be socket compatible with the Opterton 6300 series and AMD is promising power efficiency improvements even though it utilizes the older technology.
Last week during E3, AMD launched its new FX-9590 CPU, the world's first factory clocked 5GHz CPU. While the timing was a little odd for such a big deal product, it still managed to make some enthusiasts turn their heads and once again look AMD's way. Unfortunately in another weird move, AMD announced that the new CPU would only be available to system integrators for the time being.
AMD neglected to release any pricing information for the CPU, but lucky for us, a Tier 1 system integrator, PC SuperStore, posted pricing information on the AMD FX-9590 as well as the FX-9370, which is the 4.7GHz little brother. AMD seems to think that the 5GHz FX-9590 is worth a premium of $920, while the FX-9370 will only run consumers $346 for a mere 300MHz drop in clock speed.
As a long time AMD user, the idea of a $920 CPU simply turns me off entirely. I was actually considering sticking with AMD for my new system when the FX-9590 is released to consumers, but at that pricing point, I will just switch over to an Intel Core i7 Haswell 4770K, which may only have four cores compared to the eight of the AMD CPU, but the Haswell will use half the power, and with HyperThreading and Turbo Boost, it will outperform the CPU by leaps and bounds. Just for full transparency, if AMD would have priced the FX-9590 at around $450-$500, I most likely would have stuck with them.
The first half of 2014 will see AMD release its first ARM-based CPU, if the Wall Street Journal is correct. AMD, the company best known for it's APU line of processors is said to be building their first-ever ARM CPU as we speak. The chip is codenamed Seattle and is based on a similar ARM architecture to that used in smartphones and tablets currently.
Rumors have it that the first chip released will be an eight-core version with a sixteen-core model being released later down the line. Both processors will have a clock speed of 2GHz, but no word yet on cache sizes or any other CPU specifics. We do know that Seattle will be aimed at the server industry, which is a fast growing market in desperate needs of power saving CPUs.
This is not the first time that someone has aimed an ARM processor at the server world; Applied Micro Circuits did something similar with their X-Gene "server-on-a-chip," and it worked out fairly well for them. The major question is if AMD can manage to follow in AMC's footsteps and make this a profitable venture. I guess we will have to revisit this in a year to find out.
Intel have just launched their Haswell processors, or fourth-generation Core processor, but they will never stop. The 2014 platform known as Haswell-E will launch as an 8-core processor, oh baby, yeah.
Intel's Haswell-E platform will launch in 2014 and will be their enthusiast platform, we're going to get into some specs now so you'd better put a book on your lap or something. Haswell-E is set to debut in the second half of next year and will be Intel's first 8-core desktop CPU, so expect some serious CPU performance. We should see an offer of between 6- and 8-core CPUs with up to an amazing 20MB of L3 cache.
But, wait, there's more! 8 physical CPU cores means we're still greeted with Hyper-Threading, meaning 16 logical cores. Maximum TDP for Haswell-E should sit at around 130-140W with the processor itself sitting on the 22nm second-generation Hi-k process. Performance-wise, with the additional two physical CPU cores and the usual improvements in the architecture, we should expect a 33-50% performance increase over the 2014 platform.