TweakTown NewsRefine News by Category:
According to the latest rumors, NVIDIA is preparing a new mobile GPU in the form of the GeForce GTX 990M. With the GTX 980M already being quite powerful, it still isn't as fast as a desktop GTX 980, but the new GTX 990M will fix that.
NVIDIA's purported GeForce GTX 990M would reportedly be as fast as current GTX 980M SLI configurations, with the GTX 980M being based off of the GM204 die with 1536 CUDA cores. A laptop powered by GTX 980M SLI beats a desktop GTX 980 by around 10%, so for a GTX 990M to beat both of those setups, we're talking about some serious mobile performance.
One of the big problems that NVIDIA will have is power consumption, with the GTX 980M SLI using around 180W TDP. Either NVIDIA is going to have to work some magic on the purported GTX 990M, or we're looking at around 150-180W TDP on the mobile GPU. After using the ASUS ROG G751 gaming laptop and its included GeForce GTX 980M for a few months now, it doesn't get too hot or loud at all. All NVIDIA would need is some half decent cooling, which all gaming laptops feature these days, and we could be looking at mobile 4K 60FPS gaming on the GTX 990M.
It looks like MSI is preparing what could be the best GeForce GTX 980 Ti on the market with its 980 Ti Lightning card, a beast that the company was teasing at Gamescom in Cologne, Germany over the last few days.
MSI confirmed they were making their GTX 980 Ti Lighting, with no prototype card on display. We should expect the MSI GeForce GTX 980 Ti Lightning card to sport the usual custom PCB, beautiful design and kick-ass cooling. Rumor has it MSI is going to have a stock GPU clock of 1217MHz resulting in a Boost clock of 1342MHz, but I would dare say we should expect more than that considering the ZOTAC GeForce GTX 980 Ti AMP! Extreme Edition is clocked higher than that, without a custom PCB.
We should hopefully have the MSI GeForce GTX 980 Ti Lightning in for review as soon as its released, so we'll keep you updated on what could be the most exciting GM200-based card yet.
An interesting turn of events for AMD's latest video cards has been discovered by 'TX12' from the OCN forums, where he has a universal tool that lets you unlock any Hawaii, Tonga or Fiji-based GPU to its full potential.
Some users have already managed to unlock their air-cooled Radeon R9 Fury cards to a full-blown Fury X, which requires water cooling. In the case of the Fury X, AMD disabled a number of CUs on the GPU, but this universal tool lets you unlock them and enjoy the full 4096 stream processor count from the Fury X, on your air-cooled Fury.
There are dangers of doing this, as you can completely brick your very expensive card, so use caution. The thread on the OCN forums has a full rundown of how to do it, with benchmarks and screenshots of a fully unlocked Fury to Fury X, increasing its stream processor count from 3840 to 4096.
When AMD flew us out to their event in Sydney, Australia, one of the more exciting video cards that we got our hands on was the super-small Radeon R9 Nano.
The Radeon R9 Nano is based on the same Fiji architecture that both the R9 Fury X and R9 Fury are based on, it also contains High Bandwidth Memory (HBM) but it is super, super small. Well, now Flikr user 'Elmy' has the card, and according to Guru3D 'Elmy' is really Anthony Lackey. Lackey has done extensive promotional work for AMD, so we're sure to be looking at a genuine card here folks.
As you can see, the new Radeon R9 Nano is plugged into MSI's new X99A GODLIKE GAMING motherboard, and barely spills over the edge of the x16 PCIe connector. Insanely small, and it's said to pack the power of the Radeon R9 290X with similar power numbers, too. We should have a sample on its way shortly, so expect a few from us in the coming weeks.
Vince 'K|NGP|N' Lucido has broken some single video card 3DMark world records with the help of his special edition KINGPIN NVIDIA GeForce GTX 980 Ti card supported by an EVGA X99 motherboard, an EVGA high power supply, an Intel Core i7 5690X and 16GB of quad-channel DDR4 memory.
With a few records being broken in one go, KINGPIN posted world-best scores across 3DMark Fire Strike, 3DMark Fire Strike Extreme and 3DMark Fire Strike Ultra, earning 25,233, 13,091 and 6,988 respectively.
If you're interested in picking up one of these beastly video cards for yourself or are looking for more information, head here.
Said to compete with AMD's Radeon R7 370 and provide customers with a relevant upgrade from their NVIDIA GTX 750 Ti's, specifications for the GTX 950 have come to light recently, or at least rumors have.
Based on GM206 silicon (as seen in the GTX 960), we are told to expect 768 CUDA cores, a TMU count of 48 a ROP count of 32, a memory bus of 128-bit, wrapped up in 2GB of GDDR5 memory. Paired with this is a rumored clock speed of 1150 to 1250 MHz, a 1350 - 1450 MHz boost and 6.60 - 6.75 GHz memory.
Powered by a singular 8-pin connection direct from your power supply, this card may draw only 90W. Further said to support 2-way SLI for those looking at a little extra grunt, display options may include dual-link DVI, HDMI 2.0 and DisplayPort 1.2.
We knew that NVIDIA's Pascal architecture was going to deliver a massive update over the current Maxwell-based offerings in the GeForce GTX 980 Ti and Titan X, but the transistor count is going to be insane. Titan X features 8 billion transistors while Pascal will reportedly contain an insane 17 billion transistors.
NVIDIA will be tapping TSMC's 16nm process for its Pascal architecture, as well as using HBM2, which should see a massive increase in horsepower. But even with 17 billion transistors, the Pascal-based GPU will be "significantly smaller" than the 28nm-based Maxwell GPUs, reports Fudzilla. NVIDIA will be making use of HBM2 on the next-gen video cards, offering up to 32GB of the next-gen VRAM technology on its highest end card.
Expect around 50% or more performance over the already fast GTX 980 Ti, which will see NVIDIA easily dominate AMD's Radeon R9 Fury X. But where does this leave AMD? Right now, AMD is in dire need of a huge architectural change, as Fiji didn't really bring anything new to the table. All AMD has done is used HBM1, but it's benefits weren't really shown on Fury X, apart from the card being smaller than usual. NVIDIA is really going to leapfrog AMD next year with its triple-punch in 16nm + Pascal + HBM2.
It looks like NVIDIA is all systems go to launch its Maxwell-based GeForce GTX 950 next month according to HWBattle, where we now have some specs and a better idea on price to share.
NVIDIA will be pricing its GeForce GTX 950 at under $150, $50 less than the GTX 960 which kicks off at around $199. The new video card will be based on NVIDIA's successful Maxwell architecture, with a GM206 GPU and 786 CUDA cores, 2GB of GDDR5 spread out over a 128-bit memory interface. We should expect memory bandwidth to be around 107GB/sec, while it uses a single 8-pin PCIe power connector using up to 90W of power.
NVIDIA will position this at MOBA and casual gamers, but that sub $150 pricing is going to be very enticing for many gamers.
While the Radeon R9 Fury X is already one of the smallest flagship video cards ever released, the upcoming R9 Nano is going to shrink that down to a whole new level. The R9 Fury X measures in at around 19cm, but the R9 Nano is just 15nm.
How much power does the R9 Nano have behind it? Well, according to some leaked benchmarks, the R9 Nano would feature similar performance to the massive R9 290X, but with half the power consumption. Considering the size of the card, this is quite the achievement, with most of the thanks going to High Bandwidth Memory (HBM). Another benefit over the Fury X is that the Nano features an air-cooler, and not an all-in-one liquid cooler like the Fury X.
Back to the benchmarks, where DGLee from IYD.KR reports that the Unigine Heaven benchmark running at 4K with the R9 Nano pushing out 26FPS. These numbers are from AMD themselves, showing off the bandwidth per stream processor. The R9 290X has 320GB/sec of memory bandwidth, with 2816 stream processors, leaving the R9 290X with 0.1136GB/sec. The R9 Nano with its 4096 stream processors and 512GB/sec bandwidth courtesy of HBM, delivers 0.1250GB/sec. When it comes to the power efficiency, the R9 Nano has 0.152 FPS/watt, while the R9 290X features 0.076 FPS/watt... a large increase in the R9 Nano's favor.
The release of the R9 Nano happens next month, where you can be sure we'll be running it through its paces here at TweakTown.
With our exclusive report that AMD would be making just 30,000 units of its HBM-powered Radeon R9 Fury X, the news that it is receiving help from United Microelectronics Corporation (UMC) in the form of the company entering volume production of TSVs (Through Silicon Vias) which is used in the production of HBM (High Bandwidth Memory).
AMD has already acknowledged the supply issues of the R9 Fury X, but with the release of the air-cooled Fury and the imminent release of the R9 Nano, these HBM supply issues are only going to get worse. UMC had kept AMD at bay until July 20, where it entered volume production of TSVs, which means the bottleneck that AMD is experiencing, should be gone.
UMC said: "AMD has a successful history of delivering cutting-edge GPU products to market. This volume production milestone is the culmination of UMC's close TSV collaboration with AMD, and we are happy to bring the performance benefits of this technology to help power their new generation of GPU products. We look forward to continuing this fruitful partnership with AMD for years to come".