NVIDIA Maxwell Architecture

Here's a question, since they released the GTX 980, would they eventually release the TI version of the card?

Yeah its likely there will be a waxwell Titan or Ti in the future 

the 970 & 980 are out, and they run on the GM204 chip. If you follow Nvidia you know that flagship chips are x10 codenamed. that is to say we should keep our eyes peeled for GM110 chips. This will be the next flagship.

This will most likely be the GTX980ti featuring the GM210 core, Or we could see another off-shoot in naming like "Titan"

yupp.. maybe nvidia skipped the 8-series because they want the 20nm to be 1000x or some shit like that. 

the 800 got turned into mobile chips. That is to say GM1xx chips are mobile, GTX800 desktop was scrapped apparently and the 980&970 are GM2xx

but they're all the same 28nm process, but why they couldn't keep the 800 naming for maxwell pardon one or two changes I don't know.

well the big loss in transistor count for pretty much no loss of performance is good improvements in power, and temp. This is great leap for optimizations but if 900 series bios can be hacked in some way to bring to 780ti then we'll have 1.4x performance of 980 just thoughts... maybe not possible. Haven't researched the architecture yet.


Yeah, i susspect the 980 Ti will be the 20nm (or 14nm?) GPU. i can't wait to get my 970. $329 for performance that seems to be a little better than the 780, and 4GB RAM.... yes please ^_^


980 and 780ti are not the same chip. No amount of bios unlocking will do anything.

Just saying the 780 has 2304 CUDA cores. The ti had 2880 cores. The 980 has 2048 with a base clock about 200 MHz higher than the 780 cores and the 970 has 1664 CUDA cores with a base clock about 100 MHz higher than the 780 with similar performance as the 780.

Thats how big of a difference. Not to mention the chip itself is 4 5ths the size of the GK110. Even smaller if you ignore the disabled cores on the 970 unit. 

That is a chip, With a little more cores than the 770 with a much much lower TDP therefor less heat output (As seen on the 750 ti's) and a smaller footprint.

Like NJM said. Nvidias naming with their units are X04 then X10. We had the GK110 (titan) and the GK104. We did have other models like the GK106, GK107, GK108, GK208 but they were all for low end models. Made based off the GK110 chip and an unreleased GK210.

We have seen a 40% core count drop with a performance increase and power consumption degrease along with a price drop. Thats insane and with Gainward and palit already announcing cards with an overclock stock.

Im waiting for Benchmarks of EVGA's and MSI's high end cards (Successors of the Lightning, ACX and Classified) as with more efficient coolers and better VRM's, Because fan power consumption hits the cards overall TDP and limits its potential but K|ngp|n has already hit over 2000MHz on LN. highest ever seen. Stock card with voltage mod. And this is just the GM204. Nvidia is holding back on the GM210 sadly but they dont want to show their hand. This is going to be a stupidly fast generation and next generation is going to be even better.

Still going to hold onto my 780 but I am going to have to pick up a second one if I want to keep up with the 980.

The 980 is a GM204, 780 and 780 ti are GK110. Understand what your trying to say but the 980 would have to be a GM210 nerfed for that to work and even then the chips for many years have had the extra cored physically disabled through lazer cuts.

the GM210 will be a larger chip. Next generation of chips (Mentioned in the past) will be 16nm I believe?

well I don't think you understood; there are many bios tweaks at works in maxwell arch -> maybe there's way to hack 780ti to employ some of them.

Does not work. Architecture and chip is 100% different. The cores are more efficient at crunching numbers than the cores in the 780 ti. Its a physical different, not a software issue.

These chips are large though, core count may be down, but the die size is up over the cards they are replacing, 970 die size is almost identical to R9 290s die for instance which was much larger than the 770/680. So while core count is down, transistors and die size is up, although they are still performing better in terms of transistor count or die area than the predecessor.

The most amazing thing here, is how high they are getting on the overclocks, they are really pushing frequencies extremely high, which is why they are doing so well compared to say R9-290 with similar die size.

They arnt though? My 780's die size is about 560? and the new cards are down at about 430. Smaller than the cores they were replacing.

Asthetically, i don't like the blue accents on the new EVGA cooler, as my build will be red/black. Definitely loving the performance and price of the 970. it will be my first dedicated GPU. i was honestly thinking it would be $400, with 780+ performance. but $329? so crazy. 

They almost had to with the pricing because if AMD's specs on the cards are proven true (and if their pricing stayed as agressive as it has been), it could have been more lopsided than the 770 vs the 290 based on the price points. But of course since we haven't heard from AMD we have no idea whats in store...

Yeah and the 970 for instance, it really replaces the 680/770 and is much larger, that is how they got that extra performance to match the R9 290. Infact look back to the 600 generation, forget about all these 780 variant cards that came out and are not part of the traditional nvidia stack.

The 980 is significantly larger than the 680/770 and performs on parity bassically with the 880ti. for slightly more than the traditional release price of a card in that bracket. The 970 is much larger than the 670/760 or even the 770 and performs much better. The 960 will probably perform much better than the 760 and have a bigger die. These replace the Gm104 parts.