Next Generation NVIDIA Ampere reportedly to offer 50% more perf at half the power

10 hour per day x 365days x0,3KW x 0,25cents = 273€ per year

 

Your point is?

 

Give me POWAH and give me COMPETITION!!!! I want this to be true and as well have AMD drop their flagship for half the price of THEFTVIDIA. Sure it'll still be around fifteen percent maybe even twenty percent slower, but I am hoping something will give...

AMD is already on a "hella" roll. Now they need to smash the GPU market with something truly breathtaking!..!

 

I don't care about cost to run these things.... Like I even looked at the bill for quad TITAN's....? I want power and that is simply it..... I'll always find a way to cool it.

Shoot AMD wise I think I have the fastest 5700xt on air at that according to firestrike Ultra stats. Search it YOUDIEMOFO rx 5700 xt and that is my powercolor red devil shining on air. All the while the waterblock sits lonely in its box....

 

Nvidia already has a working MCM gpu design.

It appears you have a defect in your application of the English language.

You used a - to bridge a foolish statement with an ignorant statement.

I addressed the Foolish one, Walk off Karen.

 

Please refer to my previous posts. Anyway, it's been nice not knowing you as it is for probably a healthy number of people.

 

Unless you are saying Ampere is already MCM, it doesn't really matter what they have in their laboratories. As long as those designs are stuck in the labs, they change nothing. I should have written a working MCM design ready for the market. That's what I meant.

 

Doesn't matter how you spin it, MCM is never going to be viable in real time 3d gaming graphics, it will serve for ai and enterprise compute but not gaming.

 

I'm sure that's what Intel thought about CPUs as well, before AMD kicked their ass. At the end of the day, if the fundamental GPU technology doesn't change radically, MCM is the only way forward if significantly more performance is desired. Giant monoliths aren't the way to go, unless PC gaming is only meant for the rich, which isn't really sustainable. I'm no semiconductor engineer, but I don't see why the brilliant minds working for AMD and Nvidia couldn't figure it out. It's what they do for living, after all. Seems like a small step compared to suddenly switching to optical chips or something.

 

I think in 10-20 years there will be only one computing unit that will do both like it was in the beginning. I know its an idiotic statement but I'm not asking anyone to wait 20 years

 

Somebody said to me "how can you justify £700 on a graphics card (1080Ti)", i said "I can in the same way you justify £350 a month on cigarettes".

 

10 hours per day ? Is your job streaming or something ? anyway but 25 cents per kwatt hour ? where is that ? In my country we have variable rate on Electricity and i pay from 17to19 euro cents per kwatt hours and that's expensive i know in the states it can get as low as 11 usd cents per kwatt hour so 273 euros for you so the same usage for someone can be 120.45 usd a year that's why people often will not care if a card consumes 200 watts and gives 10% more performance than a card that consumes 150 watts in your case it will be 45 .625 euros spread across few bills and on the 11 usd cents will be 12.045 cents a year , but i repeat you assume people game for 10 hours everyday very very few can do such thing a lot of people in here will be lucky if they get 4 hours to game on the weekends .

edit : this moment the exchange rate is 1euro = 1.12 usd ...so for euro cents just add 12% to the price for simplicity shake just +10% except if 306 usd a year instead of 300 usd a year will change your perception

 

And?

 

Unless my memory has failed me, both NVidia and AMD have already done MCM designs before and none of them were successful. Basically ultra high end niche products. The GTX 295, GTX 590, GTX 690, HD 6990, Asus Ares etcetera. You'd think that something as parallel as 3D graphics would easily make use of MCM designs, but it had the same inherent problems as SLI and Crossfore. The 3D engine, game and drivers all had to be optimized for the technology to work properly, which created a lot of additional work for developers. Also, there was a lot of CPU overhead with running these types of GPUs, but with Vulkan and DX12 available now, that should be ameliorated.

At any rate, I'm sure that MCM technology has come a very long way since then. If Ampere is MCM capable, it would be very interesting to see how it differs compared to previous MCM designs. If they could make it so that the game or application only sees one GPU, then that would take a massive burden off of developers.

 

Looking at game benchmarks between the 2080ti and 1080ti there was a 25% performance improvement for the 2080ti. For the 2080 against the 1080ti it's in the range of 0-10% (depending on the game). Of course this might have stretched a bit after a while with the performance improvements made in later driver revisions.

 

And those are the people crying about how they can't get out to do anything let alone afford to.....

 

Nope it won't. Wouldn't care if it were $800 a year. It's my cup of tea..... Get it. Some people smoke cigarettes others play high fidelity bad ass games cranked up at Uber resolutions.

 

Just leave the man be ... he has his hobby and admire his dedication and the time and money invested. Do you think that if someones buys 4xTitans really cares about electricity ? Just think about it for a second.

 

Around 300$ per month is the average fast food or delivery cost for a full-time working adult. Imagine telling them to use their kitchen to make some healthy food. They would shout about electricity!

 

No, those were just cards that had two individual, separate GPUs installed on them. Those GPUs were originally designed to work alone, and thus they worked like two regular cards in Crossfire/SLI, more or less. If you create an MCM GPU, the chiplets inside will be designed from the ground up to work as a single unit, just like they do in the Ryzen CPUs. Increased latency and other issues are the things the engineers need to figure out.