Nvidia's Pascal is 10X faster than Maxwell. Then graphics will no longer be the bottleneck. I hope AMD keeps up the pace, I own too many shares. If AMD has something close to Pascal the large array of CPU cores will come in handy.
I can assume Intel will release a larger 16 CPU core chip and prosper when this happens.
AMD's advantage is integration of the CPU & GPU on the same die. They have good match of technology, they only need to keep up the fast development pace. I always thought this task is hard for AMD with limited resources. AMD should partner with a Sony, or Apple to help with development.
ACE, the news you are trying to keep secret is already out. The Teen cheaply mounted a gun onto a cheap drone, and with a little automation software this drone could shoot anything that moved.
There is a lot of development for the automated sight and shoot 50 caliber turret. There is a group of teenagers who created a website for developers to work together to make this killing machine. (Search auto turret) You could mount this on the auto driven cars that Google and Apple and others are making and you now have a mobile kill machine. You could launch 500 of the cars working together (communication links can easily synchronize their work).
There is a new oven with a built in camera that can identify the food that is placed in the oven. The computer auto recognizes the food, auto adjusts the temp and time, and auto does everything. Perfect cooking every time.
This can be done today, and we are just getting started. In 10 years the processing power of computers will be 10X faster.
You have not seen anything yet. Wait a few more years (10) and you will see the havoc of the robot revolution.
1. Military wars will be won by those with the best robot warriors. I did not say the country with the best robot warriors. A single company could easily own this earth.
2. Jobs, most productive jobs will not need humans. Farming, Mining, Construction, Shipping are first to go. Then Doctors, Nurses, Teachers, Bankers. (I don't think Engineers and Entertainers need to worry, but there are robots designed to replace the oldest profession)
Virtual Reality will be big. I can predict how this technology will cripple society. High School kids with low self esteem ( there are millions ) will enter the virtual reality and become a super hero. They will never want to leave this dream world.
"Billions of people have cellphones. How fast did that happen? A decade. For virtual reality I'm predicting 4 years. I think we're going to see mass acceptance for VR," Crytek's David Bowman told Blasting News after a demo at E3. The company reportedly put 50 people working on VR and has been showing demos of 'Robinson' with one of the latest Oculus Rift prototypes.
There are probably a few technical and cost reasons AMD did not design the large Interposer. The large Interposer does not improve performance and the only benefit is cutting the cost of memory in half.
The large multi-GPU Interposer is a worthy goal even if it was difficult to develop. AMD took years to develop the current small Interposer, so a larger one might have been a bridge too far for AMD.
Nvidia's Pascal claims to be 10X faster than Maxwell, and I have inside knowledge that Nvidia has good performance tools that can test the performance of new architectures. The 10X is a real value.
The only way I can see a 10X improvement would be if Pascal was a multi-chip solution. I believe Nvidia is using the very large Interposer.
AMD demo system is using Intel's CPU. This has to hurt the pride of AMD's to management. AMD's Zen must be getting full attention.
The dual Fury board is not what I wanted. Nvidia has a long history of selling more boards than AMD and AMD needs every ounce of performance they can get. In this battle AMD fumbled the ball in a game they have a history of losing.
I just now saw a screen shot of the dual Fury board. Sad.
This board does not share the memory. Sad.
AMD had an opportunity to knock it out of the park, but failed.
AMD should have created a single large Interposer that both GPU dies and the shared HBM memory.
Thanks for the information gem I was looking for.
I know how to design chips, and the Interposer has 0 transistors. The base die that sits under the HBM memory stack contains those extra transistors. You can bank on this fact.
There are 4 base dies, 4 HBM stacks. Each die contains 400 Million transistors.
What are these 400 Million transistors doing?
I/O buffers + Transceivers = less than 1 million transistors.
Memory controller = at most 4 million transistors.
The remaining 395 Million transistors ?
Processing near memory CPU ?
With the memory controllers placed on the base die under the HBM stack the dual Fury design is something Nvidia cannot compete with. The memory controller supports a dual port design. This allows both Fury dies to access the same memory stack with not bottlenecks. The dual Fury is going to be a virtual reality beast. (and small size)
I agree, AMD will not produce many Quantum systems. The Quantum should be used as Nintendo's next game console. Nintendo should enter from the high end with virtual reality enabled.
Nintendo could use the dual FURY design as a development platform. The production game console should use the 14nm chips will be available in 2016. The 14nm chips have enough capacity to integrate all of the transistors from the 2 Fury chips + Zen.
Does this make AMD a lot of $$$ ?
I usually discuss new technology more than I discuss $$$. AMD will earn some money and I still own a lot of AMD. But, Pascal is forcing me to buy Nvidia.
I gave you a thumbs down. This message board is inverted.
Carrizo supports H.265 and NOT Intel. Fixed the number and capitalized the I in Intel.
This one ? Did not look inside.
May 31, 2015
Nvidia finally outs the 980Ti
Computex 2015: Exactly what we told you about last year UPDATED
It looks like Nvidia is finally launching the 980Ti we told you about last August.
The article must be old, I checked going back 1 week. I should have guessed that the Nvidia lovers are using old news.
This isn’t particularly surprising for us as we had already told you four weeks ago that Samsung and GF had beaten TSMC to FinFETs. The timing is also important because Globalfoundries had already announced that it plans to begin volume production of 14nm FinFET products in the first quarter of 2015. Luckily the considerable yield “fix” that Samsung recently figured out means that GF can begin mass production with a solid baseline rather than having to start slowly like Samsung.
If you remember a month ago we reported that Nvidia is listed under TSMC’s partners list for 16FF while AMD was suspiciously absent. On the GPU front this development means that AMD’s decision to make FinFET GPUs at Globalfoundries will soon pay dividends. As the delays faced by TSMC will not affect AMD’s schedule with Globalfoundries. We’ll continue to follow this closely as it’s interesting to see how this plays out and if it leads to any competitive advantage or disadvantage between AMD and Nvidia.
Intel might have emptied their bank, but they now have new talent. I work for Altera. Soon an major AMD stockholder working for Intel, Irony.