• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Hynix HBM Presentation - many slides!

About time. GPU performance has been very stagnant for the past few years

Well I wouldn't say stagnant..

AMD HD 7970 Launched 2011 (Paper) Available Jan 2012.
AMD R9 290X Launched 2013. 290X VS 7970 Comparison


Fermi' GTX 480 Launched 2010.
Kepler' GTX 680 Launched 2012. 680 VS 480 Comparison
Maxwell' GTX 980 Launched 2014. 980 VS 680 Comparison


Big jumps between each gen, the 980 almost doubling the performance of the 680, and at the same power use level is very impressive, and still way better VS CPU progress 5% per year, let's face it in CPU's X99 is the only remotely exciting release. AMD aren't even planning a proper new CPU architecture release until 2016.. At least with GPU's we have seen significant jumps every two years.
 
Last edited:
You forgot the Titan Boom. That was a big jump. :D

Although if you are still on 1080p, I could see how you might get the idea it's stagnant. The push has been toward 4K. For example, the 290X really shone there compared to the increases elsewhere. GPUs are being tuned for 4K perf: more ROPs, VRAM etc. We saw bigger gains there because before it had been just a dream. Also bottlenecks (which Mantle tried to address, at least in part) kept 1080p FPS along a more shallow curve. Things are happening now across the board that will see computing change drastically in the next few years.
 
Last edited:
You forgot the Titan Boom. That was a big jump. :D

Although if you are still on 1080p, I could see how you might get the idea it's stagnant. The push has been toward 4K. For example, the 290X really shone there compared to the increases elsewhere. GPUs are being tuned for 4K perf: more ROPs, VRAM etc. We saw bigger gains there because before it had been just a dream. Also bottlenecks (which Mantle tried to address, at least in part) kept 1080p FPS along a more shallow curve. Things are happening now across the board that will see computing change drastically in the next few years.

Forget the Titan that was a 'compute' card according to Nvidia :p

If you look at the architecture jumps, 'Fermi' 2010, 'Kepler' 2012, Maxwell '2014'. We have seen big leaps in performance and power consumption improvements. Much better than what we get in CPU.

GPU market def is not stagnant. The 980 almost doubles the performance of the 680, two years later. Imagine if a CPU had double the performance of say the 2700K, that would be a big deal, with GPU's we got that and people are like 'Meh' :D

Imagine two years from now, GPU's will be even more amazing, likely double the performance again, people be like 'Meh'. CPU performance increase 5% / 10% in the same period 'Wow' :p
 
I was going back over the foundry news from earlier in the year and found this:

http://www.digitimes.com/news/a20140701PD201.html

It says that Apple will move to 14nm @ Global early 2015, coincidentally the same time AMD are supposedly launching 390X. This would free up 20nm wafers for use on GPUs.

Also there's a great thread from June here:

http://forums.anandtech.com/showthread.php?t=2389030&page=4

It's fairly unlikely they'll stop production at TSMC, it's more likely that TSMC's production ramp is just under way, they start off low then when they've got yields up and have a little experience they start installing more equipment in place of old in other fabs. So you might only start with 20-50k wafers a month but after 6 months might open up another fab at 20nm and jump from 50k to 100k, or 250k a month.

I'd guess that basically Apple have been making chips for around a year on 20nm, going through the risk production and low yield stage but getting some chips then producing their 20nm A8's. Whenever 14nm starts risk production at GloFo and Samsung it seems Apple will be using that but they'll likely just start making A9's, in the same way, low yield, go through the production ramp aiming for a release of new devices the same time next year. i-**** 7 or whatever it is next September. They'll continue making A8's at TSMC until they stop making devices with A8's in them.

What it does mean in the longer term future is that the early lower volume production at TSMC likely won't be taken up by Apple again, so the next big process move will probably be back to the more usual setup of AMD/Nvidia taking a lot of the early capacity.

Whatever the actual situation is chips take what will be pushing 2 months or maybe even more to make at 20nm, Apple can't just switch in Feb and have a seamless transition and in the same way if AMD do release a 20nm chip in Feb, it will have taped out 6 months ago and start production most likely some time in the next month or so.
 
Back
Top Bottom