Read more.Featuring the Gigabyte OC.
Read more.Featuring the Gigabyte OC.
Performance increase
GTX 560Ti vs GTX 660 : 2011 > 2012 - 120% Faster (1 year)
GTX 660 vs. GTX 760 : 2012 > 2013 - 120% Faster (less than 1 Year)
GTX 760 vs. GTX 960 : 2013 > 2015 - 110% Faster (2 Years)
GTX 960 vs. GTX 1060 : 2015 > 2016 - 182% Faster (1 year)
GTX 1060 vs. GTX 1660 : 2016 > 2019 - 116% Faster (3 year)
What the F happened in 2016?
I didn't follow the GPU market back then, was some competing AMD card around?
Rx480 meant they had to do something.
As MonkFish said, the AMD Polaris parts (RX470/480) offered a very compelling price/performance ratio when they launched in 2016 (before prices got silly). However, nvidia had essentially let this happen with a series of pitiful releases as you show in your table. The 960 in particular was appalling (essentially they tried to capitalise on the good press the 970 was generating).
I've only ever (that I can remember) returned one graphics card because it was pathetic, and the 960 holds that honour. Mine was both pathetic, and misbehaved on displayport with the monitor it was bought to drive. Thankfully AMD had released the 470 by then, and it made an attractive alternative for about 20ukp more.
I think you both missing the point here.
The RX480 (Jun 29th, 2016) was released in the 10xx series period, just after the 1070 (Jun 10th, 2016).
The 1060 was already planned like all the other x60 cards. To prompt that huge jump in performance from 900 to 1000 series something must have going on way before, you don't design a GPU chipset in a month.
Was it the Radeon R9 3xx series that made the pressure for that big jump in performance to nVidia cards?
Was nVidia holding back doing little 20% steps because they didn't have any competition?
Now they are back at little steps because AMD didn't came out with some remarkable competition?
Too many questions
Any reason why the Founders edition of the 10series was used against 2 AIB (one of which, the subject of this article, is overclocked)? Why not use the best of all 3 series? I'm pretty sure there's 1060s out there which are more powerful & quieter & less hot than the FE edition..
It's fairly simple: Pretty much everyone but Apple skipped the 20nm node as it was supposedly only good for mobile devices. Which is odd, as I thought mobile and graphics were a similar requirement of density and low power, *shrug*
Anyway, the 960 was the last 28nm card. It was on par with the AMD R9-380, so it was as good as it should be and Nvidia weren't holding anything back in what was an honest race.
Is it fair in your table to compare the 560ti to the 660? If you compare the plain 40nm non-ti 560 to the 28nm 660 you would get a bigger jump which is more like what I would expect. But then all those cards from 660 to 960 inclusive are on the 28nm process, so you only get a lot more transistors if the die size goes up.
Then finally we go from 28nm to 14/16nm cards. The R9 380 gets trounced by the RX 480, the GTX 960 gets to look silly compared to the GTX 1060.
Sadly the 12nm label is pretty much just that, it is a finesse on the last year's tech with no improvement in density and so no significant additional transistors. Note also that if you go back to cards like the Geforce 3 on what now sounds like a silicon feature size you could drive a bus down, every 18 months we got a new process node and about a doubling in performance. With the 1060 we had a two process node jump and still that only got us a 180% speed, physics is no longer kind to us. The jump to 7nm is another two process jump missing out 10nm, after that companies might have to get creative.
I am watching AMDs chiplets layout with interest, if they can put multiple CPU chiplets on a carrier then why not multiple GPU chiplets?
Some 3GB 1060 and 4GB 960 results would have been really interesting here.
The 2GB 960 is clearly hitting a memory wall in some of those tests. Someone with a 3GB 1060 might be hitting the same issues, so an upgrade from that to a 6GB 1160 might be worthwhile.
You are right, somehow when I was looking for those data, I didn't saw a 560 (non ti) thinking they didn't make one, which would be weird.
Correct line would be:
GTX 560 vs GTX 660 : 2011 > 2012 - 139% Faster (1 year)
Also a maybe more adequate generational comparison would have been:
GTX 1060 vs. GTX 2060 : 2016 > 2019 - 150% Faster (3 year) But we are far off if you consider the price point!
Of course over the years the improvements slowed down a little, but I just concluded that they are simply playing with our pockets.
ginolatino ..all your figures are rubbish ... go back and recalc . 100% +20% =20% Faster , kapiche? LOl
Mr_Jon (21-03-2019)
That's harsh, they got a word wrong, that doesn't make the figures rubbish just badly presented.
It is usual in benchmarking to have a 100% baseline to compare against, so a 20% faster card will be reported as 120% of the baseline card, so really I'm sure we all knew what was meant. Just mentally substitute "Faster" with "of previous card" and the world is nice.
Compared to the choice of cards in the article this thread is supposed to be for (which this table helps highlight) it seems a tiny mistake. And yes I agree it is a mistake, but what should have been an "Even Better If" sounds like a lynch mob.
It's a valid point. It did confuse me initially how they got those stats which didn't seem to add up. While this site is tech orientated, those less in the know may have gotten the wrong idea quite easily. Saying something is a set amount faster has a particular meaning which shouldn't be confused. Albeit it was originally brought up a little rudely...
There are currently 1 users browsing this thread. (0 members and 1 guests)