We've been wondering what cost the custom GeForce GTX 1080 cards will be, with NVIDIA stating the lowest will cost $599, while their GTX 1080 Founders Edition card costing $699. Well, according to a recent Colorful press release, Colorful's "custom-designed GTX 1080 video cards" will "retail below the [price of] the GTX 1080 Founders Edition".
With the cost of the GeForce GTX 1080 Founders Edition from NVIDIA set at $699, this means we should see a varying cost between $599-$699 for the overclocked GeForce GTX 1080 cards. That might sound like news, but think about the cost. The GTX 1080 costs $599, but the Founders Edition which is capable of over 2GHz on its new 16nm-based GP104 GPU, so what will custom cards be clocked up to if something with a single fan and 1x 8-pin PCIe power is already the best card on the market?
The custom GeForce GTX 1080 cards should, in theory, be faster than the GTX 1080 Founders Edition. But, it looks like they'll be priced cheaper than the Founders Edition, which is - well, great? NVIDIA loses out because most gamers and enthusiasts will buy a custom GeForce GTX 1080, while system builders and more professional customers will find comfort in the new NVIDIA reference design GTX 1080 in the form of the Founders Edition.
NVIDIA has just unveiled their latest Tesla M10 video card, which crams a huge 4 x Maxwell-based GPUs onto the single PCB. The new Tegra M10 card isn't for normal, or even enthusiast gamers, but it's aimed at virtual PCs and virtual application workloads.
NVIDIA's new Tegra M10 has 4 x GM107 GPUs, with each core featuring 640 CUDA cores and 8GB of GDDR5 for a total of 2560 CUDA cores and 32GB of GDDR5. The four GPUs would normally suck up a considerable amount of power, but the Tesla M10 has a TDP of 225W, which is pretty amazing considering how much technology is crammed onto this card.
The card can support up to 64 x virtual GPU instances, and 28 x H.264 1080p 30FPS streams at once. Each user will have 512MB of VRAM, with the entire GPU filled with vGPU instances. If you've got 2 x Tesla M10 cards in your machine, then you can support 124 separate vGPU instances.
GALAX has started its big push for its upcoming GeForce GTX 1070 and GeForce GTX 1080 cards, where it has used an interesting style of four strange figures as its marketing push.
The red guy in the hat is the new face of the GALAX GTX 1070 line, while the blue clown looking guy is going to be the face of the GTX 1080 series. We haven't heard what the other two guys will be used for, but we're sure it's going to be the GTX 1060. Earlier today, we reported on Inno3D's upcoming GeForce GTX 1080 and GTX 1070 cards, while we also reported on the mid-range GTX 1060 being a price/performance champion when it comes to VR gaming.
The official GeForce GTX 1080 Founders Edition is a beast of a video card, but the AIB partners are cooking up some delicious alternatives, like the iChill GTX 1080 from Inno3D. This is on top of the news that the mobile GTX 1080M will be a beast, beating out the Titan X, and the mid-range GTX 1060 being a mainstream VR champion.
Inno3D's upcoming products based on the GeForce GTX 1070 and GTX 1080 will feature a triple-fan cooling solution, with two LED stripes on either side of the fans. We could even see the two LED strips acting as GPU load or temperature indicators, and not just there 'just cos'.
Inno3D's iChill GTX 1080 will sport a custom board with a 6+8-pin PCIe setup, which will fly past that 120% TDP limit and really begin seeing the legs on the GP104 stretch. We should hopefully see Inno3D unveil its new video cards at Computex 2016.
We've just reported on the exciting promise of the GeForce GTX 1080M, offering Titan X level performance in a notebook, but what about the mid-range market? That's what the GTX 1060 will be for, as well as doubling as a price/performance VR monster.
Well, the GeForce GTX 1060 is expected to be unveiled in the near future, with the GP106-based card set to feature a 256-bit memory bus. The 256-bit memory bus will find itself with either 4GB or 8GB of GDDR5 RAM, which will handle most games thrown at it, including VR headsets. We could also expect NVIDIA to fill the bigger gap between the GTX 1060 and GTX 1070 with the GTX 1060 Ti, which would allow the company to possibly offer the GTX 1060 with 4GB of RAM, while the GTX 1060 Ti could have its framebuffer bumped to 8GB.
We should expect around 1280 CUDA cores on the GTX 1060, knocked down from the 1920 CUDA cores on the GTX 1070 and 2560 CUDA cores on the GTX 1080. The GTX 1060 will fight off the slew of mid-range Polaris cards that AMD is preparing for launch, so expect a huge battle in the mid-range graphics market from both NVIDIA and AMD in the coming months.
NVIDIA has been ramping up towards a big Computex 2016, but the news just got much more exciting, with rumors that the company will unleash its next-gen GeForce GTX 1080M graphics solution at the show. We already know that the GeForce GTX 1080 is the new king, but it looks like the GTX 1080M will be the new mobile king.
The GeForce GTX 1080M will reportedly be faster than the GTX Titan X, which sells for $999 and is a desktop video card, not a notebook solution. The new GTX 1080M has been spotted on the ASUS website, with the company teasing benchmarks for a gaming notebook that beats everything else on the market.
NVIDIA had a fantastic mobile solution with the desktop GTX 980 being unleashed into notebooks last year, but this new GTX 1080M is demolishing everything else in its path. It's faster than AMD's Radeon R9 Fury X, R9 295X2, and NVIDIA's own GeForce GTX 980 Ti and Titan X - oh, and I have mentioned already, it's a notebook part - not a desktop part, which makes it even more incredible. The GTX 1080M pushes a huge 20,811 in 3DMark 11, while the Titan X and R9 295X2 trail behind.
We should expect the full GTX 1080 to find its way into the GTX 1080M part, with 2560 CUDA cores, and GDDR5X memory. Thanks to the GTX 1080 not requiring much power at all, it makes sense that the mobile GTX 1080M part would be just as power efficient, maintaining its huge horsepower. NVIDIA has quite a lot going on during Computex 2016 this year, and you can be sure we'll be all over it when it takes place in just over a week from now!
AMD has just announced its press conference during Computex 2016, with AMD set to unveil their new Polaris GPUs and Zen APUs on June 1. There'll be plenty of top shelf AMD staff there, with AMD's CEO Lisa Su and Radeon Technologies Group boss Raja Koduri, with both of them being joined by AMD's CPU and APU boss, Jim Anderson.
The company said in its press release: "The event is slated to feature presentations from AMD executives including AMD President and CEO Lisa Su; Senior Vice President and General Manager, Computing and Graphics Business Group, Jim Anderson; and Senior Vice President and Chief Architect, Radeon Technologies Group, Raja Koduri. A real-time video webcast of the event will be accessible on AMD's Computex page (www.amd.com/computex) and on AMD's Investor Relations home page (ir.amd.com). A replay of the webcast can be accessed a few hours after the conclusion of the live event on both pages and will be available there for one year after the event".
While we are super pumped for the Polaris-based video card side of things, I think I'm equally as psyched about seeing just how good Zen turns out to be. AMD has quite a lot riding on the new Zen-based side of its CPU/APU business, as they're being shrinked down to thet 14/16nm node like the new Polaris-based GPUs. We're expecting a decent amount of improvement over the current Vishera- and Kaveri-based APUs, with Zen said to really ramp things up for AMD and have them competing against Intel on a much firmer ground.
NVIDIA has been dominating discrete GPU market share for a while now, but AMD has made a comeback in the last few months with its Radeon R9 300 series and the HBM1-powered R9 Nano and R9 Fury cards. In Q4 2015, AMD's discrete GPU sales increased by 6.69%, thanks to the release of the R9 380X.
On the other hand, NVIDIA's discrete GPU sales were down 7.56% after it released the GTX 950. This means that AMD could secure itself 7% of the discrete GPU market share in a single quarter, which is impressive. It won't be long until AMD fully unveils their next-gen Polaris 10 and Polaris 11-based video cards, but we'll have to wait until late this year according to the new rumors, or early 2017 for the HBM2-powered Vega GPU.
We've just wrapped our review on the NVIDIA GeForce GTX 1080, with it blowing everything else out of the water, but one of the things I can't wait for is the cheaper GTX 1070.
NVIDIA's GeForce GTX 1070 is going to be a video card that represents huge value for money, priced at $379. The specifications were hush-hush until now, but it looks like the GTX 1070 will feature 1920 cores, down from the 2560 cores found on the GTX 1080. The GP104-based GeForce GTX 1070 is expected to be faster than the current GM200-based GeForce GTX 980 Ti, and even the Titan X.
We can work out the rough GPU clock of the GTX 1070 if we look at the ~1600MHz Boost clock, comparing it to the 1733MHz Boost on the GTX 1080. The GTX 1080's default GPU clock speed is 1607MHz, so I'm expecting the GTX 1070 to fall somewhere in the 1400-1450MHz range. Coupled with 8GB of GDDR5 on a 256-bit memory bus, we're talking about 256GB/sec of memory bandwidth - which is a kick down from the 320GB/sec on the GTX 1080 with its 8GB of GDDR5X on the same 256-bit memory bus.
Last week's Radeon 16.5.2 drivers included some optimizations for Doom, although no specific numbers were included. Now AMD has released a beta driver in 126.96.36.199 with yet more performance improvements, and this time with numbers in hand.
The company claims up to 35% better performance on the R9 390 over the previous drivers. Anyone on the 290 series should absolutely download these as well given the similar GPU architecture; it doesn't hurt to try for older cards either.
In my personal testing with a 290X, Ultra settings were giving me anywhere from about 40 to 110 FPS with frequent stuttering and a general feeling of input lag with 16.5.2; updating the drivers sees Ultra run smooth as butter, rarely if ever dipping below 60 FPS and often staying over 100, with little or no stuttering and no feeling of input lag. I'll go so far as to say this is likely the biggest performance improvement I've ever seen from a driver update.