NVIDIA Launches GeForce GTX 1060 To Take On AMD's Radeon RX 480 (hothardware.com)

Reader MojoKid writes: NVIDIA just launched their answer to AMD's Radeon RX 480 mainstream card today, dubbed the GeForce GTX 1060. The GP106 GPU at the heart of the GeForce GTX 1060 has roughly half of the resources of NVIDIA's current flagship GeForce GTX 1080. NVIDIA claims the GTX 1060 performs on par with a previous generation high-end GeForce GTX 980 and indeed this 120W mainstream offers an interesting mix of low-power and high-performance. The new GeForce GTX 1060 features a new Pascal derivative GPU that's somewhat smaller, called the GP106. The GP106 features 10 streaming multiprocessors (SM) with a total of 1280, single-precision CUDA cores and eight texture units. The GeForce GTX 1060 also features six 32-bit memory controllers, for 192-bits in total. GeForce GTX 1060 cards with either 6GB or 3GB of GDDR5 memory will be available and offered performance that just misses the mark set by the pricier AMD Radeon R9 Nano but often outran the 8GB Radeon RX 480. The GeForce GTX 1060 held onto its largest leads over the Radeon RX 480 in the DirectX 11 tests, though the Radeon had a clear edge in OpenCL and managed to pull ahead in Thief and in some DirectX 12 tests (like Hitman). The GeForce GTX 1060, however, consumes significantly less power than the Radeon RX 480 and is quieter too.You may also want to read PCPerspective's take on this.

Comments

Deja vu!

[fustakrakich]

All over again.

What is this? An Alzheimers test?

Re:Deja vu!

[msmash]

No, today's post sheds more insight into the GPUs, and how they fare against each other -- with benchmark numbers etc.

When will VideoCards peak?

[jellomizer]

We had a good run From 1995-1998 with the SVGA cards that did 1024x768 with 32bit color. Then that 3D acceleration came out and buying a good video card became much more difficult.
With Displays going up to 4k we should be getting to a point where increase of resolution will not matter, And 3d performance on those displays should be quick enough.
While Mores law is in effect our bodies are not adapting as fast as the technology, so there should be a point where the Video from a computer will meet a threshold where playing such upgrade games isn't going to be important.

Much like how we don't talk much about Sound cards.

Re:When will VideoCards peak?

[drinkypoo]

With Displays going up to 4k we should be getting to a point where increase of resolution will not matter, And 3d performance on those displays should be quick enough.

Quick enough for what? When we reach photorealism at dual 4k, then we can maybe talk about peaking. We're a long, long way off from that.

Re:When will VideoCards peak?

[110010001000]

Moores law has been dead for quite a while now. Really digital computing is reaching a dead end in itself. If you have noticed, the single thread performance of Intel CPUs is only around 20% of what they were 5 years ago. The days of exponential growth are well over. That is why they have just been adding more cores and cache and trying to improve memory technology. All the low hanging fruit has been picked.

Re: When will VideoCards peak?

[Type44Q]

When we reach photorealism at dual 4k, then we can maybe talk about peaking.

When a single mobile GPU can drive a pair of small 8Ã--8k 120hz stereoscopic displays, then we can maybe talk about peaking. :)

Re:When will VideoCards peak?

[greenfruitsalad]

exactly. even if we increased the processing power of graphics cards 20 fold right now, we still wouldn't even have real-time raytracing of inanimate scenes. let alone trees with moving leaves, human hair or people wearing realistically looking fabric (of fur).

Re:When will VideoCards peak?

[by (1706743)]

...44.1kHz 16 bit audio is relatively trivial...

So, is is there an analogous specification for video cards? The 44.1kHz @ 16bit is pretty easily justified (Nyquist–Shannon + reasonable dynamic range). Can a visual equivalent be easily justified? That is to say, at sort of "eye limited" (retina, in Apple lingo) resolution and field of view, how many polygons can be said to make up the human perception of reality, and what sort of graphics processing muscle would be required to drive this?

I of course have no idea, just wondering out loud. Just trying to approach OP's question in a pseudo-scientific fashion.

Re:When will VideoCards peak?

[ShooterNeo]

As it so happens, the demands of VR headsets mean that video cards available now are nowhere NEAR adequate. As a poster south of me says, you need at LEAST dual 4k - one for each eye - and fovea tracking - and at LEAST 90 FPS. All the time. With minimal latency.

Believe it or not, but not even the most expensive GPU money can buy - heck, not even unreleased GPUs that Nvidia has in Tesla cards (they are "released" but you can't use em as a graphics card) - is anywhere close to being able to push this kind of resolution and framerate at a low latency.

It literally cannot be done with current chip construction. Maybe with a 4 or 8 GPU solution.

Re:When will VideoCards peak?

[Lonewolf666]

Well, lets do a bit of math then to figure it out.

According to https://www.nde-ed.org/EducationResources/CommunityCollege/PenetrantTest/Introduction/visualacuity.htm, 20/20 vision is

the ability to resolve a spatial pattern separated by a visual angle of one minute of arc

Lets take that as a given for the sake of the argument, and assume that we want just enough dpi on our screen that one pixel shows up at a visual angle of one minute of arc. So the screen can just match the resolution of the eye.

Lets also assume that the largest screen we might ever want is as wide as the viewing distance from our eyes to the center of the screen. Think a 32'' screen on your desk, at arms length away, or a 100'' screen 2-3 meters away in your living room.

Then the viewing angle at the left/right edge of the screen is arctan (0.5) = 26.6 degrees and the total viewing angle from left to right is 53.2Â.

The requirement of one pixel = one minute of arc translates to 60 pixel per degree and to a horizontal screen resolution of 3192 pixel. A bit less than the 4K resolution that is already on the market. I haven't watched 4K material yet, but I found that WQHD (2560x1440 pixels) on a good monitor is already pushing the limit of my eyesight.

So I guess 3840x2160 will end up being the "44.1kHz 16 bit of video". Most people won't really see the difference when the resolution is pushed higher.
 

But what about the DPC latency?

[mrraptor98]

Initial benchmarks look lower than the RX 480 and the price higher [with actual retail availability no better]. When you add the DPC latency issues, I wouldn't touch it with a 10 foot pole. I actually have a GTX 1070 I'm sending back now just because i don't want to deal with possible Nvidia DPC hell.

The elephant i

[edxwelch]

Well, it doesn't look too good for AMD. Their "super efficient" RX 480 uses much more power than the 1060 and is slower.
On the bright side is the price of the 480 is only $200 (well, eventually it will be ;) ) and also AMD's version of aync compute works far better than Pascal (see: http://wccftech.com/nvidia-gef... and http://www.eurogamer.net/artic... )

Re:The elephant i

[Freedom Bug]

The 1060 is faster the 480 in old games. However newer games will use technologies like asynchronous DX12 and Vulkan. Ashes of the Singularity and Hitman are good examples of the former, and the Vulkan build of Doom is a good example of the latter.

The 480 is faster than the 1060 in those 3 games. Doom/Vulkan is a *lot* faster on the 480.