TweakTown NewsRefine News by Category:
NVIDIA Editor's Day 2014 - Another interesting technology unveiled by NVIDIA in Monterey Bay was MFAA, a new anti-aliasing technique that isn't as harsh on your hardware.
MFAA stands for Multi-Frame Sampled AA, where NVIDIA finds more information inside of every pixel on your screen, anti-aliasing the scene much better. It has the performance hit of 2x MSAA, but the look of 4x MSAA, which will provide gamers with a better looking image, without the performance hit they're used to with traditional AA methods, like MSAA.
NVIDIA Editor's Day 2014 - One of the more interesting things shown off by NVIDIA at its Editors Day was DSR, or Dynamic Super Resolution. DSR is capable of rendering a 4K image, and then downsampling it to 1080p, which has multiple use cases.
First, you can render at 4K, and then display this higher resolution image on your not-so-high resolution display, such as a 1080p panel. This provides a sharper image, without you needing to upgrade your display.
Above, we have the normal image on the left, with DSR on the right.
NVIDIA Editor's Day 2014 - One area of NVIDIA's new direction with Maxwell is virtual reality, with a totally new focus on VR with something NVIDIA calls VR Direct. VR Direct provides multiple technologies from the Maxwell architecture, and previous technologies from NVIDIA, to make the VR experience much better.
First, we have low-latency, which will reduce the amount of time it takes for the scene to be rendered on your PC, and displayed to the VR display sitting just an inch or so from your eyes. Second, we have VR SLI: something that will help your GeForce GTX SLI setup render the two scenes (one to each eye) much better, resulting in improved performance and better yet, scaling between your GPUs.
Latency is something NVIDIA is very concerned about, with the company wanting to remove as much of it as possible, in order to get a better VR gaming experience. As you can see, right now we have around 50ms of delay between the image being rendered and outputted to something like the Oculus Rift, but NVIDIA have some new technologies to remove much of this delay.
MFAA also makes an appearance, which will provide yet another layer of improved visuals. Auto Asynchronous Wrap begins doing work on the scene before the next frame, as you can see in the image above. The dotted line is where the next frame begins, but the GPU can now shift some of that ms delay (to before the frame), instead of after - resulting in delay to your eyes.
VR DSR is going to be the big thing here, with Dynamic Super Resolution able to render the scene at 3840x2160, or 4K, and then down to 1080p (or 1440p on the CV1 if the rumors are correct). This will provide a better-looking image, without needing the display on the Rift needing to be 4K, just yet.
For the past week or so, we have been sitting on the information of NVIDIA's new GeForce 900 series GPUs, with the two latest cards now official: the GeForce GTX 980 (our review is right here), and GeForce GTX 970. NVIDIA is pricing the GTX 900 series very competitively, with the GeForce GTX 980 priced at $549, and the GTX 970 at only $329. These are some incredible price points considering the feature set, improved performance, additional features and reduced TDP. NVIDIA will be discontinuing the GTX 780 and GTX 770, shifting the price of the GTX 760 to just $219.
For starters, we have the new Maxwell architecture, which is where NVIDIA is pulling this rabbit from a hat from. NVIDIA has some serious magic from Maxwell, with the star of the show, the GTX 980, really pushing the boundaries of what is possible from a GPU, without requiring a nuclear reactor to power it. For starters, the "GM204" has 5.2 billion transistors, 2048 CUDA cores, 128 Texture Units, 64 ROPS, and a 256-bit wide memory bus with 4GB of GDDR5.
NVIDIA has achieved this huge jump over the GK104 thanks to twice the performance of the GK104 with the GM204, and two times the performance-per-watt over GK104, too. There's an improved schedular, new datapath organization, and over 40% delivered performance per CUDA core on GM204. The memory architecture behind Maxwell has also received an injection of improvement, with enhanced compression algorithms, and enhancing caching effectiveness, and when compared against the GTX 680, we have some big improvements.
It looks like we are just days away from meeting the new GeForce GTX 900 series from NVIDIA, with a very detailed look at MSI's offering of the GTX 970 GAMING card, which features the company's TwinFrozr V cooler.
MSI's GTX 970 GAMING features TwinFrozr V, which includes two 100mm propeller-blade fans, with a beautiful black-and-red design. The GeForce GTX 970 itself features 1664 CUDA cores, with MSI's factory-overclocked 1140MHz and Boost clock of 1279MHz. 4GB of RAM is on the card with a 256-bit memory bus, with a TDP of just 148W. MSI requires one 8-pin and one 6-pin, which is up from the two 6-pin PCIe power connector requirements of the reference GeForce GTX 970.
One of the weird things that MSI has done to its GTX 970 GAMING card, is remove some of the DisplayPort outputs that the reference GTX 970 sports. The reference GTX 970 and GTX 980 GPUs include three DisplayPort outputs, but the MSI GTX 970 GAMING has just one DisplayPort, one HDMI and two DVI ports.
NVIDIA is positioning itself to launch its new Maxwell-based GPUs within the next couple of days, with two new products to be launched: the GeForce GTX 980 and GTX 970, but now we're hearing more details on pricing.
We should expect NVIDIA to start the price of its GeForce GTX 980 at $599, while the GTX 970 is rumored to be priced at a damn competitive price of just $399. At $399, NVIDIA could start cutting into its own GTX 780 pricing, which could be seriously good for consumers and gamers. We are expecting a slew of non-reference GTX 980 and GTX 970 cards to be made available, with varying cooling setups.
When it comes to the specifications of the new GeForce GTX 980, it will be a Maxwell-based card on the 28nm process, with 2,048 CUDA cores, 128 TMUs, and 32 ROPs. The GTX 970 will feature 1,664 CUDA cores, and 104 TMUs. Both variants will feature a 256-bit memory bus with 4GB of GDDR5 RAM.
By now you should know that NVIDIA is ready to launch its GeForce GTX 980, but AMD wants to take some of that thunder away with the rumors that the company is set to release its Radeon R9 390X GPU.
The news is coming from VideoCardz, who is reporting that Asetek, the company who made the watercooler for the dual-GPU Radeon R9 295X2, is working on the cooler for the upcoming R9 390X. The new cooler will keep the VRM and memory under its fans, with AMD wanting to keep its reference design cards to sound much quieter than previous GPUs.
No specifications are known on the R9 390X, but the name "Fiji" is being thrown around for this upcoming family of GPUs.
Right on the back of our news on the sighting of ASUS and GIGABYTE branded GeForce GTX 980 and GTX 970 cards, WCCFTech is reporting that the GeForce GTX 980 will offer 10% more performance on the GTX 780 Ti, all with a 170W TDP - an incredible feat, if the rumors are true.
The rumor is coming from a post on the Chilhell forums, with some impressive 3DMark scores posted. The post mentions that the GTX 980 has 32 RPS and 128 TMUs, with its 256-bit bus and 4GB of RAM, with an model featuring 8GB of RAM to probably show up, too. The biggest thing here is that at stock clocks, the GTX 980 is 10% faster than the GTX 780 Ti, but with a TDP of just 170W compared to the TDP on the GTX 780 Ti which is still a modest 250W.
If this is true, under the same 28nm process NVIDIA have pulled off something quite incredible here, which might justify the skip over the GTX 800 series. With a TDP of 170W on the GTX 980, we have a 60% performance-per-watt increase on the same 28nm process over the GTX 780 Ti, which is something worth talking about. With a $499 price tag, this could be one heck of a GPU, something we should hear about very soon. With a reported press event this week, we could even hear about it in the coming days.
It looks like NVIDIA is all set to launch its next-gen GPU, which should be skipping the GeForce GTX 800 series and moving onto the GTX 900 series with two GPUs to be launched at first.
According to Synnex, an international distrbutor of PC hardware, ASUS is preparing two GeForce GTX 900 series cards. First, we have the reference GeForce GTX 980 with what should arrive as the reference cooler for the GTX 980, then a second model: the STRIX GTX 970. This model should feature ASUS' custom cooling setup, the DirectCU II. We will see a quieter, cooler card with factory overclocking applied.
Then we have two entries for the GIGABYTE GeForce GTX 900 series, also from Synnex. The first model is the GeForce GTX 980, another reference card, backed up by a custom GTX 970. GIGABYTE's GeForce GTX 970 G1 Gaming card should also feature a custom-designed cooling solution, and overclocked. Both the GTX 980 and GTX 970 cards are listed with 4GB of RAM across a 256-bit bus, so we're not going to see massive changes in the skipped-a-generation GPUs at the GPU-hungry 4K resolutions, and above.
Most will remember Matrox, a company that dealt in mainly professional video cards that were great for multi-monitor setups before the likes of AMD's Eyefinity and NVIDIA's Surround Vision technology hit the market. Well, Matrox is coming back, and is releasing GPUs that are powered with AMD technology.
Matrox's next-gen video cards will be powered by Radeon GPUs, with the announcement stating that "key features of the selected AMD GPU include 28nm technology with 1.5 billion transistors; DirectX 11.2, OpenGL 4.4 and OpenCL 1.2 compatibility; shader model 5.0; PCI Express 3.0 and 128-bit memory interface." We should expect Matrox to be using a lower performance GraphicsCore Next (GCN) part, something from the Cape Verde GCN family. We could also see Matrox lean toward AMD's FirePro W600 cards, with Matrox sprinkling its custom-developed software applications such as Matrox PowerDesk on top.
This is an interesting move, as it frees up R&D that Matrox would spend on hardware costs, leveraging AMD's technology and putting more of its time into the software side of things. It helps AMD, as AMD can sell countless GPUs at higher "professional" prices, increasing their share of the professional market with AMD-powered Matrox cards.