TR Forums

Realisticaly, what time frame are we looking at here? June? December?

In the past, have early cards on a new arcitecture / smaller die process had problems/bugs?

I'm so close to buying a 760 but if Maxwell will actually be shipping by June I can hold off.... if it's gonna be late into 2014 well I might as well buy now.

Are there any details known about what we might be able to expect? I'm guessing the equivilent 860 (or whatever they call it) would most likely be the same 256bit interface, with more cores and higher clocks?

Three weeks, according to the Internet rumor machine. But they're not starting with the high end this time, rather they're starting in the middle with the 750 Ti. Definitely worth waiting to see what happens over the next few months.

Since GPUs fabricated on TSMC's new 20nm process won't be in full production until the second half of the year, whatever they're bringing out now will be fabricated on the same 28nm process as everything since the GeForce GTX680.

JustAnEngineer wrote:

Since GPUs fabricated on TSMC's new 20nm process won't be in full production until the second half of the year, whatever they're bringing out now will be fabricated on the same 28nm process as everything since the GeForce GTX680.

It was interesting to read about this as I thought "volume production" was full production, but apparently I'm wrong. Safe bet that they won't make a third flagship part on 28nm again though.

I was really looking forward to an earlier launch too. More time to earn EVGA bucks I guess

Ahhhh, too bad. I guess I'll go ahead an order a 760 now.

The high-end Maxwell chips look especially interesting if they do in fact come with an integrated ARM cpu ... still not exactly clear on what role they will be playing, something to offload instructions from the main CPU?

melk wrote:

The high-end Maxwell chips look especially interesting if they do in fact come with an integrated ARM cpu ... still not exactly clear on what role they will be playing, something to offload instructions from the main CPU?

Completely random guess, but I'm going to say it has to do with memory management and anything else the drivers need to do overhead work on but want to sidestep the CPU. Maxwell will be the first GPU with fully unified virtual memory, so to avoid calling the CPU for memory management overhead it could turn to the ARM core on the GPU for that? Probably lots of other compute-related overhead it can be used for as well.