Video Cards & GPUs News - Page 348

Something we talked about a few weeks ago now looks to be true: AMD will release two versions of its Radeon R9 390X. One of them will rock the next-gen HBM, while another will use the standard GDDR5 VRAM. WCCFTech is reporting that they noticed some juicy news on the ASUS forums, with the following units:

What we do think this means, is that AMD will release a Radeon R9 390X with 4GB of HBM, while the 8GB version will rock GDDR5. We've heard through our industry sources that HBM is experiencing seriously low yields, which will stop AMD from slapping 8GB of HBM onto the cards. This move will allow AMD to sell more R9 390X cards as they'll only be using 4GB of HBM, versus 8GB of High Bandwidth Memory.

The bigger question is: will the Radeon R9 390X be enough to compete against the GM200-powered NVIDIA GeForce GTX 980 Ti? What time of performance leap are we to expect from the HBM-powered R9 390X, over the nearly two-year-old R9 290X?

As we get closer to the official announcement and launch of the Radeon R9 390X from AMD, all we have to enjoy for now are leaked benchmarks that show the "Fiji XT" card beating out the GeForce GTX Titan X, barely.

With its super-fast HBM, the AMD Radeon R9 390X beats out the NVIDIA GeForce GTX Titan X, the unreleased and expected GTX 980 Ti, and every other single GPU solution from NVIDIA in the average scores from 19 benchmarks according to a leaked look at the next-gen card from AMD.

When it comes to power consumption, it looks like AMD's next-gen Fiji architecture and High Bandwidth Memory aren't enough to save it from the perils of high power consumption. The leaked benchmarks show that the R9 390X uses 289W of power, which is just 3W away from the R9 290X which is quite the consumer of power. Comparing this to the Titan X which uses 256W, and the GTX 980 Ti which uses 235W, AMD is once again consuming a large amount of power in order to beat NVIDIA.

With AMD on the verge of unveiling its new Fiji XT-based Radeon R9 390X powered by HBM, NVIDIA isn't just waiting around sitting on its hands. NVIDIA is reporting preparing to roll out its new GM200-based GeForce GTX 980 Ti, with 6GB of GDDR5.

NVIDIA's rumored GeForce GTX 980 Ti will be made from feature the GM200 GPU, the same one that was found in the Titan X, except that the 980 Ti will feature 6GB of framebuffer, versus the 12GB found on Titan X. As for pricing, WCCFTech's source had it listed at around $954 USD, but we should expect NVIDIA to release it much cheaper than that, especially to compete against AMD.

After the price, the second big question is: when will the GeForce GTX 980 Ti arrive? Rumors have pegged NVIDIA at releasing a new product before Computex, but then there's some that say during Computex, and even into July. NVIDIA has the power of waiting right now, especially with NVIDIA owning 76% of the discrete GPU market.

According to the latest rumors, we should expect the Radeon R9 390X to launch with 4GB of HBM, while a dual-GPU version of the Fiji XT-based card will arrive with 8GB of HBM. For those who have been keeping up, this is a very, very interesting move, if the rumors are true. If you want to catch up on how revolutionary HBM will be, we wrote a detailed piece on High Bandwidth Memory yesterday.

Fudzilla is reporting from "insider sources" that AMD will launch the Radeon R9 390X with 4GB of HBM for an MSRP of around $849, while the dual-GPU version of the card, which should arrive as the Radeon R9 395X2, will include 8GB of HBM. The Radeon R9 395X2 (that's what we're calling it for now, this could change at any moment) should arrive sometime later in the year, or 2016 - depending on HBM yields, I'd say.

The sources stated that AMD had plans to launch the Radeon R9 390X with a price of $799, but this is no longer the case. The sources also added that the R9 390X will be competing directly against the NVIDIA GeForce GTX Titan X, which launched at $999 and is still around the $999 mark on Amazon still. According to Fudzilla's sources, the HBM-powered Radeon R9 390X will win in some benchmarks, and lose in others against the GDDR5-based Titan X.

According to VideoCardz.com, we should expect rebranded AMD Radeon R9 200 series cards based on the Hawaii architecture to arrive next month with a disguise, as the Radeon R9 300 series.

While the HBM-based Radeon R9 390X will arrive in two flavors: 4GB and 8GB (and maybe one model with GDDR5 and another with HBM), there will be other Radeon R9 300 series cards based on the Hawaii architecture. These should arrive as the Radeon R9 385, and R9 380 - but those specifics could change. But these new cards will feature slightly higher Core Clocks, and a nice jump on Memory speeds.

The Radeon R9 290X has a Core Clock of 1GHz, but the new R9 300 series rebrand will have 2816 stream processors, while its Core set at 1050MHz, a 50MHz jump. The Memory Clock on the other hand, jumps from the 1250MHz found on the R9 290X, to 1500MHz on the new cards, according to VideoCardz.com. This will give that particular card based on the Hawaii XT GPU around 384GB/sec of memory bandwidth, up from the 320GB on the R9 290X.

TSMC has come out swinging lately, teasing that the shift into 16nm FinFET is going to be quite big for GPUs. The Taiwanese manufacturer said that the move from 28nm to 16nm, and in particular, the 16nm FinFET+ process, will deliver around 40% more performance.

This 40% improvement in performance will not consume any additional power, which should have both NVIDIA and AMD smiling from ear to ear. This means if they were to spin up an NVIDIA GeForce GTX Titan X on 16nm FinFET+ and have the same performance, it would consume 50% less power. Alternatively, for the same power, they would be able to cram in a huge 40% performance gain. Impressive stuff, shrinking down to 16nm.

TSMC will begin volume production of its 16nm FinFET in Q3 2015, which means we could expect the first GPUs based on the smaller node towards the end of the year, or early 2016. We are predicting that flagship GPUs released from this new 16nm process will be, at an absolute minimum 30-40% faster, all while using the same power draw of around 200-250W. Along with HBM, we could see some serious improvements of 80-100% over the flagship cards we see today. HBM2 (something we saw at NVIDIA's GTC 2015) is due next year, with 1.2TB/sec of memory bandwidth, up from the 640GB/sec that we should see on the AMD Radeon R9 390X, and a big gain from the $999 Titan X and its 336GB/sec.

Can you believe that we're just over a month from the unveiling of AMD's next-generation GPU architecture? We have already had our industry sources tell us that AMD will be launching its Radeon R9 390X at E3 2015, but according to WCCFTech, AMD will be launching its next-gen video card on June 24.

There seems to be quite a lot of conflicting information, but right now we know that AMD will not be launching the Radeon R9 390X at Computex, as we thought back in early March. Instead, our sources have said Fiji XT will be launched at E3 2015 during the first PC conference, but AMD does need to have its own event for a launch this important.

AMD's Radeon R9 390X is shaping up to be something special as it'll be the first video card to feature High Bandwidth Memory, or HBM. This next generation memory is packaged alongside the GPU die on a single piece of interposer, rather than on the circuit board surrounding the video card as we've had up until this point. This is going to usher in a different video card than we're used to, where AMD can make their flagship R9 390X much shorter than usual thanks to saved PCB space.

Back in February, we wrote a story about NVIDIA reiterating that TSMC was its 'very important' foundry partner as it heads into the world of 16nm manufacturing.

Well, NVIDIA is back to reiterate once more that it still loves TSMC, and that it's working with the Taiwanese manufacturer on not just 16nm technology, but 10nm. Jen-Hsun Huang, NVIDIA's CEO and co-founder said: "We are constantly evaluating foundry suppliers, ... [but] we largely purchase from TSMC, the vast majority of our wafers we buy from TSMC. We are in 20nm, we are expecting to ramp 16nm. We are deeply engaged with TSMC for many, many nodes to come, including 10nm".

This could be happening as Samsung is kicking ass with its 14nm FinFET production, but with the current legal battles between NVIDIA and Samsung, could NVIDIA be deflating the rumor balloon that it will ditch, or at least shift some production over from TSMC to Samsung in the near future? NVIDIA has added that it's not "too obsessed about the process technology", which is an interesting statement from a company that pushes the boundaries of GPU technology with each successive release.

We might be enjoying a time of total NVIDIA dominance in the VGA market with the GeForce GTX 960, GTX 970, GTX 980 and Titan X giving gamers everything they wanted, and more, but the future of GPUs is nearly upon us.

AMD is ready to launch its new Radeon R9 390X which will arrive with HBM1 technology, but NVIDIA is beginning to talk about their Pascal architecture, which should arrive sometime in 2016 as the PK100 and PK104 GPUs. Jen-Hsun Huang, NVIDIA's CEO and co-founder, said during their latest financial report that the Pascal-based GeForce products will be arriving with HBM2, which should increase the available memory bandwidth from the ~300GB/sec or so right up to a huge 1.2TB/sec.

We have reported that AMD's next, next-gen Radeon R9 490X would use HBM2 with over 1TB/sec memory bandwidth, too. NVIDIA should be tapping some of that sweet 14nm FinFET technology by then, so the future GeForce cards will be radically smaller, cooler and insanely fast. NVIDIA is only starting to tease them now, with Huang saying: "I cannot wait to tell you about the products that we have in the pipeline. There are more engineers at NVIDIA building the future of GPUs than just about anywhere else in the world. We are singularly focused on visual computing, as you guys know".

KFA2 has announced its new white NVIDIA GeForce GTX 960 EXOC video card to the public, boasting this new color scheme covering the shroud, fans and back plate, but not on the PCB.

Featuring a factory overlocked core at 1279 MHz, 1367 MHz GPU Boost and 7 GHz of memory, this provides a little extra boost when compared to reference speeds of 1127 and 1187 MHz respectively.

As for cooling, an aluminum fin-stack heat sink combined with 8mm thick nickel-plated copper heat pipes ensures your card stays cool. The display options include one DisplayPort 1.2 port, two DVI and a single HDMI 2.0 option.