Mysterious Graphics Chip from Nvidia, the ‘JM601’ GPU, Spotted in Shipping Manifest on Zauba

[KoM]

What ever it is, you know it will be amazing. Everything Nvidia make is cutting edge and better then the rest.

 

[Perfect_Chaos]

Looking at average performance on techpowerup at different frame rates, the 780Ti to 980Ti is pretty much 30% faster. Same from the 970 to 980Ti. I think Its usually roughly around that from one high end card to the next as well.

1070 (or whatever it will be called) will probably = the 980Ti then i suppose. I will not budge until at least the 1080Ti otherwise it's not worth it, that would be my limit. No way in hell will i get a Titan either

 

[Dicehunter]

Looking at average performance on techpowerup at different frame rates, the 780Ti to 980Ti is pretty much 30% faster. Same from the 970 to 980Ti. I think Its usually roughly around that from one high end card to the next as well.

1070 (or whatever it will be called) will probably = the 980Ti then i suppose. I will not budge until at least the 1080Ti otherwise it's not worth it, that would be my limit. No way in hell will i get a Titan either

I had a Titan X and in single card configurations to be honest they aren't any better than a 980 Ti.

I could only get the Titan X to 1400 on the core but with the 980 Ti I can run it stable at 1520 and it does out perform the Titan X in the tests I did.

I'll also be waiting for the 1080 Ti Classified

 

[andybird123]

Looking at average performance on techpowerup at different frame rates, the 780Ti to 980Ti is pretty much 30% faster. Same from the 970 to 980Ti. I think Its usually roughly around that from one high end card to the next as well.

1070 (or whatever it will be called) will probably = the 980Ti then i suppose. I will not budge until at least the 1080Ti otherwise it's not worth it, that would be my limit. No way in hell will i get a Titan either

erm, if you look at pretty much any of the aftermarket 980ti's it shows the 780ti at ~60% and the 980ti at 100%... the reciprocal of 60-100% is 166%, so the 980ti is 66% faster than a 780ti, before manually overclocking both, and iirc the 980ti tends to overclock better than the 780ti

And that is with both cores on 28nm

 

[Seanspeed]

It delivered massively, much lower power consumption, enabled small form factors to have a big punch. Huge memory bandwidth and more efficient than GDDR5, i.e less amount of HBM can do the same task than more GDDR5.

HBM 2.0 looks even better from rumored spec, twice the speed and option for higher capacity.

Right. Shouldn't confuse Fury's generally disappointing performance for the price with that being down to HBM. And you can see where the improvements are at higher resolutions, too. Very worthwhile.

Though of course it'll be interesting to see how GDDR5X compares.

Well the last 55nm card was the 285, two nodes latter (40nm -28nm) the first full fat card was the Titan, which was about THREE times faster than the 285. So the first 16nm card (two nodes down from the currant 28nm cards) could in theory be 3 times faster than what we have now.

Now of course I do not expect the first Pascal cards to be 3 times faster than a TitanX, but I really do think we will be seeing a much bigger leap than the suggested 50%.

14/16 isn't really a two node jump, though. More like 1.5 or something.

 

[bru]

14/16 isn't really a two node jump, though. More like 1.5 or something.

I'm not sure which ones are half nodes and which are full nodes, all I do know is that we have been on 28nm for far too long and they missed out 20nm, so that makes it a two node jump in my book.

Similar to the NVidia 285 to original Titan ( being the best 55nm card to the first full fat 28nm card) so we could have 'UPTO' 3 times the performance of a TitanX if we follow the same pattern.
Now obviously I don't really expect 3x performance, but I will be severely disappointed if it only comes out at 50% more

 

[Seanspeed]

I'm not sure which ones are half nodes and which are full nodes, all I do know is that we have been on 28nm for far too long and they missed out 20nm, so that makes it a two node jump in my book.

Not really important what your book says, no offense. It's not a full two node jump. These 14/16nm FF chips are basically just 20nm with some manufacturing improvements as straight 20nm wasn't working very well. Intel is the only one who has a *true* 14nm process in place. That would be a proper two node jump.

I dont disagree with you that we'll get a better-than-50% improvement, but I'm telling you this to temper expectations somewhat. It's basically a one node jump with some improvements.

 

[maonayze]

erm, if you look at pretty much any of the aftermarket 980ti's it shows the 780ti at ~60% and the 980ti at 100%... the reciprocal of 60-100% is 166%, so the 980ti is 66% faster than a 780ti, before manually overclocking both, and iirc the 980ti tends to overclock better than the 780ti

And that is with both cores on 28nm

I personally think it is a good thing that we have been on 28nm for quite a while as it shows how good both Nvidia and AMD were at squeezing more power out of that ageing tech.

Hopefully we will see great cards from both sides in the next few years and the solidification of DX12 and Vulkan low level API's.

The futures bright, the futures Red & Green

 

[Dalandius]

I personally think it is a good thing that we have been on 28nm for quite a while as it shows how good both Nvidia and AMD were at squeezing more power out of that ageing tech.

Hopefully we will see great cards from both sides in the next few years and the solidification of DX12 and Vulkan low level API's.

The futures bright, the futures Red & Green

Aye they both push themselves more when there is competition. I want them both to have great cards.
Ideally a third maker to push them more would be great. Intel made the right kind of sounds a bit back but that all seems to be just sound and fury with no substance