![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Yes 30% is roughly average but that did not include a massive price hike. For the same price the percent increase in performance was virtually zero between the 2080 and 2080 Ti in other words.
Not necessarily going up, just not going down proportionally with the node.
Thats not cost per transistor.
That was the point, cost per transistor is flat, not decreasing as previous and not increasing as Factum said.
Nvidia did not go to 7nm process, 12nm -> with your logic, as usual, I guess 7nm Amper 3080 Ti will be over $2000
Nvidia did not go to 7nm process, 12nm -> with your logic, as usual, I guess 7nm Amper 3080 Ti will be over $2000
Looks like that will be the case if you are a future Nvidia user. $1200 for a 12nm (modified 16nm node) to Nvidia 7nm -> $2000 and justification will be what? It is a luxury lol. Ahmmm AMD Radeon VII, even when the 7nm was relatively new, lower yields, using expensive HBM, 16GB etc. was what? Less than $800. It is more Nvidia wanting you to pay them more for their pockets than the increase in making the card. I really don't think prices will go higher than Nvidia current pricing but lower due to competition. AMD 5xxx series is getting Nvidia to become a little bit more reasonable and Nvidia immediately released the Super line -> Better performance and cheaper pricing when AMD had competitive performance at a much lower cost.
lol, you have no clue what I can and cannot buy
. I buy AMD/Nvidia at my pleasure, not theirs.
lol, you have no clue what I can and cannot buy
Not whining, laughing, big difference. When I go to a restaurant, I don't go to the highest priced one with crap food because it cost more but the one that has the best service and food. Pay the same amount for the same performance as my 1080 Ti - that would be stupid. spend almost double the amount for 30%+ - lol more stupid. I expect Nvidia to learn from their mistakes and provide something better is all or I will just eat some place else. Now the tech finally shows it has potential which I also expect to be realized more with Ampere - if so, I may buy several like I bought 4 pascal cards in the past and have two now.
Cool! Epyc and Next Gen Nvidia Tesla. AMD better support Nvidia and visa versa in the long run. Probably some very good reasons keeping Nvidia for AI, software, experience with, performance etc. AMD so far has taken more a wait and see for specific AI hardware as in like Tensor cores and relying more on their Platform capability to push AI. Now will this Supercomputer use NVLink? or just pcie4?
Maxwell was also a very keen moment from Kepler - those two big jumps pushed Nvidia way ahead particularly in the mobile market where ATI/AMD had a good standing. Yes big jump -> no need to jump to Turing with little to no jump for the same price. I don't think that is too hard to comprehend and many took it the same way. The many promises of Turing has been very slow to come, while I like Nvidia taking chances, I do not like how Turing was promoted and the overall performance did not warrant an upgrade. People have many reasons why not going ahead so everyone can be different. Ampere makes more sense on my end to look at when available which also means AMD as well. In my case it will be a high end, high resolution, high refresh monitor/GPU combination. Zero reason to upgrade with current monitors and the monitor I really want is more like 2nd half of this year. We will see.
I didn't think that something the size of V100 was possible. Wasn't it like 815mm2?
Yields for TSMC 7nm are said to be very good, even the 5nm yields are above the curve per time frame compared to 7nm. 7nm TSMC fabs open up much more to everyone when Apple goes to 5nm so Nvidia should have plenty access for big and smaller chips. Will Nvidia wait to push their larger size Amper chips down for gaming? Usually with Nvidia on a new node, the bigger size GPU's come later but if AMD has something very competitive, I hope Nvidia pushes up the time table (Nvidia usually refuses to take 2nd place, which is one quality I do like, don't like some of the marketing BS that goes along with it). March will be interesting or could be.
980 Ti was only 30% faster than 780 Ti. What was impressive about Maxwell was the performance they were able to squeeze out of 28nm when everyone thought it was already being pushed to its limit.
30% but then Maxwell OC so well it pushed it up past 40% easily - slaughtering AMD Fury cards to no end - It was an embarrassing time for AMD.
True, but this was also the last generation from NVIDIA where one could easily up the power limit. Even without breaking the power limit, though, I believe most easily overclocked to 1.5 GHz from 1.05 GHz.
Maxwell killed AMD in the mobile market followed up by Pascal - literally slaughtered. Now with Turing and AMD Navi, AMD is making a come back. How long and successful that comeback will depend upon Ampere which should be better efficiency wise than Turing. Then again AMD will have refreshes as well, plus RNDA 2 will come out, so AMD may continue to make a recovery in the mobile market. Long ways to go there.
