By using this site, you agree to our Privacy Policy and our Terms of Use. Close

Forums - PC Discussion - AMD New Polaris GPU Revisions w/50% Better Perf/Watt (NX delay reason?)

I'd still prefer two devices: A home console and a handheld that share the same OS and play the same games at different settings (cartridges instead of Blu-Rays, Cross-Buy, Cross-Play, one account for both devices). It would be way more comfortable for the user to just pick the product he/she likes instead of forcing people to buy a tablet.

The tablet sounds like a huge compromise to me and would be expensive for a handheld console. If Nintendo makes two devices instead, a home console and a handheld, they could have a home console as powerful as a PS4 that's really small (no Blu-Ray drive and no physical hard drive!) and sell that for $250-300. The handheld could be $199 at launch. Get a free game if you buy both machines. But I guess that's just wishful thinking at this point.



Around the Network

A source tell me, rx 460/470 customised... I can't confirm anything, but he told me 2 différents résolutions for home console games, i suppose 1080 on TV and 720 for streaming to handheld devices



linkhouse said:
NX wasn't delayed! March 2017 was NX first official release date!

I doubt Nintendo intentionally missed the holiday, so by delay we mean why Nintendo couldn't reach the ideal date of holiday 2016. Nintendo even suggested that the only reason that its launching in march is so all the software will be ready so a internal delay seems like it definietly happened. 



Oh dear...AMD just can't keep their shit together. 480's wattage was really ridiculous compared to nVidia, glad that they've finally managed to sort it out, but FFS, that consumer trust is going the way of dodos for them.



se7en7thre3 said:

 

Just speculation but if Nintendo were to use this, they can essentially create gcn 2.0 engineering w/cartridges.  PS4pro will be an ugly hulking beast while NX , if using these new polaris revisions, can come in sleeker than OG PS4 with more power.  

 

There is no one single info/rumour saying that NX is using AMD, all point that NX is Nvidia Tegra based.



Around the Network
se7en7thre3 said:

Just speculation but if Nintendo were to use this, they can essentially create gcn 2.0 engineering w/cartridges.  PS4pro will be an ugly hulking beast while NX , if using these new polaris revisions, can come in sleeker than OG PS4 with more power.  

That's blowing it out of proportion I think. :P

fleischr said:
That makes sense. Nintendo likes to keep wattage low on handhelds for battery life reasons.
They're pretty keen on minimizing power draw for their home consoles too.

These chips aren't meant for handheld devices.

JRPGfan said:
WTF? 50% improvement alone from better manufactureing?

Thats nuts. Makes you wonder if the factory that manufactured those chips at start wherent properly calibrated or something.

It's not as uncommon as you think.
Whenever a new node is used, there are tweaks to the process as time goes on.
28nm for example went from 250w on the 7970 to the FuryX at 275w... Transister counts doubled. - And there was a performance difference of 40% to 300% depending on benchmark.

Global Foundries is using Samsung technolgy anyway, Samsung managed to reduce power consumption by 15% and increase switching speed by 15% when it moved from it's LPE (Low-Power-Early) to LPP (Low-Power-Plus) process. Aka. 2nd Generation Samsung 14nm.

This is likely based on the 3rd Generation Samsung 14nm process known as LPC (Low-Power Compact).

numberwang said:
If Zen hits the store, AMD may have something competitive again.

I still have an Athlon 64 X2 running as a legacy device for office use every day... ye olde lives on.

It will be competitive in the mainstream markets most likely, not the high-end, just like Polaris.

shikamaru317 said:

Yeah, I think this is what Nintendo was waiting for when they delayed NX. This improvement affects embedded chips which is exactly what Nintendo would be using for NX. There was a rumor recently of NX using a 3.6 tflop AMD GPU and that matches perfectly with an underclocked RX 470.

Nintendo wouldn't be using an embedded chip. They will likely opt for a Semi-Custom design.

eva01beserk said:

For a hybrid if it reaches 2.5tf on console mode and like 1.25tf on mobile mode, then I would be happy as hell. I mean thats close to x1 performance for a handheald and even better than a wii u. No more than $300 guesses are around $250 then it wwwill still kick ass.

How do you know it's close to the Xbox 1 in terms of performance or better than a Wii U? There is more to performance than flops.

globalisateur said:
I hope Nintendo will chose Nvidia for their tablet. AMD would be a terrible choice. And it's too late for using that tech for NX anyways.

Samsung had laid out their fabrication plans well in advance, Global Foundries licenses Samsung's technology in it's Fabs, Nintendo would have been aware of this for a long time.

I hope Nintendo doesn't go with nVidia and goes with high-end AMD hardware in a fixed-console, if the console is going to be mobile Centric, then I hope it is nVidia or an ARM chip using a fully-fledged Bifrost GPU.

shikamaru317 said:
malistix1985 said:
This is also good news for Scorpio

Yep, Scorpio should use less than 200 watts. PS4 Pro meanwhile uses 310 watts max according to Sony. 

Sony will likely release a revision to take advantage of this new fabrication process in time, all these chips are being made by the same company remember.

malistix1985 said:
shikamaru317 said:

Yep, Scorpio should use less than 200 watts. PS4 Pro meanwhile uses 310 watts max according to Sony. 

Its not unthinkable that sony releases a PS4 Pro slim that has a UHD-player and uses less energy around the Scorpio release

I would be all for a 4k-Bluray support.
The Playstation 4 had the potential to be the ultimate Multimedia hub for the home, but they gave that opportunity away to Microsoft with the lack of CD and 4k Blu-Ray support.
A PS4 Pro Slim could turn that around.

se7en7thre3 said:

Scorpio Im betting will be based off Vega, in the article it says over 7tflop for rx580 (130w) so scorpio will be a cut down version of that or get  bumped up performance.

Polaris can hit those flop numbers as well with aggressive clocks and binning.
Which brings home the point that Polaris and Vega with the same "Flops" will still perform very differently, Microsoft will go with whatever is the most  price/performance efficient at the time.

JEMC said:

Tweaktown isn't as bad as wccftech, but it's not the most credible site either.

I found TweakTown's article: http://www.tweaktown.com/news/54433/amd-improved-radeon-rx-480-50-better-perf-watt/index.html

Funnily enough, that article also lacks proper sources and says the opposite thing of the wccftech one. TweakTown claims that AMD will use these improvements on the desktop cards (the 480 more specifically) while Wccftech claims that AMD will use the improved efficiency on their embedded products first.

I don't really believe any of those rumors.

The Fabrication "rumors" are the real legitimate tidbit.

CGI-Quality said:
numberwang said:
If Zen hits the store, AMD may have something competitive again.

I still have an Athlon 64 X2 running as a legacy device for office use every day... ye olde lives on.

Let's hope. No more $1200 Titans/$1700 Intel-E Series CPUs!



I was legitimately looking at 10-Core Xeon's today. Because it would be cheaper than upgrading to LGA 2011-3~ *Grumbles*

JEMC said:
shikamaru317 said:

Well, we do know that AMD's naming./numbering scheme for Polaris does allow for hardware revisions. So maybe these will be sold as 465, 475, and 485.

I can see AMD launching xx5 revisions as they showed in that slide explaining the naming of the new cards. What I don't believe is the 50% improvement claimed by those sites.

Up to 15, maybe 20% by next year as the process gets more mature and refined could be, but a 50% improvement in only 4 months since the 480 launched? No way.

Depends on what they are comparing it to.
This is the 3rd Generation 14nm Finfet Process by Samsung.
The 2nd Generation brought a 15% improvement.
Thus if it's improvement is a 50% increase against the 1st generation 14nm process, then it's only a 35% increase from the 2nd generation 14nm process.

shikamaru317 said:

It does sound unlikely, but supposedly Polaris GPU's aren't supposed to be using as much power as they are, some kind of manufacturing defect caused the poor efficiency. Assuming that is the case, then a 50% improvement could in fact be possible. It would have to have been a pretty big manufacturing defect though, I can't imagine anybody making that big of a mistake. But we shall find out soon enough, supposedly these improvements are already showing up in embedded products and will show up in mobile and desktop variants eventually, if the rumors are true. 

Polaris is actually energy efficient. And does well with what it is given.

The problem is people are comparing it against nVidia, nVidia has had a massive edge in efficiency since Maxwell, when nVidia introduced their tiled based approach and took on some things they learned from building Tegra.

Simply put, nVidia is just better. - Will this likely change the landscape in favor of AMD in regards to price/power/performance? Not likely. That is probably a few years away yet.

Game_God said:

Wii U has 352 GFLOPS, so half a TFLOP (500 GFLOPS) makes it roughly 45% more powerfull than Wii U... not very impressive unless it's a handheld, which is nice enought power in that case...

Comparing chips on flops alone is highly inaccurate.

bunchanumbers said:
Game_God said:

I don't understand what you meant here... :/

I don't know how reliable this is:

http://kyokojap.myweb.hinet.net/gpu_gflops/

But every site I read says 352 GFLOPS...

There was already this big argument about it, but Neogaf verified it at 175 or so. Who knows if its true though.

It's 176. The Wii U is more efficient which is why, when games are made it's-way, it can have a massive lead over the Xbox 360 and Playstation 3 graphically.

JEMC said:

Summary: using embbeded products to extrapolate the performance or power consumption of the desktop versions doesn't work.

We cannot forget either that Embedded products tend to not be aggressively binned. They need to be built for reliability first, so they usually run with more volts than they need.



--::{PC Gaming Master Race}::--

There is probably more then one raison for delayed...
A another one could be have better production if it is a succès, make anough console for the end of 2017... Not having the problem of Wii without stock...
A rumors speak about 10 millions consoles... for Wii, it was 6 millions...



bunchanumbers said:
Game_God said:

I don't understand what you meant here... :/

I don't know how reliable this is:

http://kyokojap.myweb.hinet.net/gpu_gflops/

But every site I read says 352 GFLOPS...

There was already this big argument about it, but Neogaf verified it at 175 or so. Who knows if its true though.

One of the huge issues that made it a certainty that it was 176 gflops is the wii u uses a older cheaper 40nm fabrication process which is more power hungry and yet its power consumption was far too low to be 352 gflops. It simply didn't consume enough power. The main reason people were believing it was 352 gflops was if anything a simple belief that Nintendo wouldn't release a console in 2012 that was of such low performance as 176 gflops but then the cpu performance was evidently much weaker from day one.

Still even at 176 gflops still more powerful than 360 and PS3 gpu's. It has later much more efficient architecture. In one generation Radeon gpu's become 30% more efficient and the wii u has 32MB of ultra fast memory for the gpu. It's probably still overall more powerful unlike the cpu performance which is much weaker than both 360 and PS3.

That's what worries me about the NX its hard to predict how low Nintendo will go in performance terms. Even if they go for a decent chipset in theory how much performance will they strip off it by lowering speeds, reducing features or using cheap slow memory chips?

I'm going to assume very low performance unless there is evidence to the contrary and I would suggest other's do the same rather than give Nintendo the benefit of the doubt when the first NX games show weak performance, missing graphic features etc.

 



bunchanumbers said:
Game_God said:

I don't understand what you meant here... :/

I don't know how reliable this is:

http://kyokojap.myweb.hinet.net/gpu_gflops/

But every site I read says 352 GFLOPS...

There was already this big argument about it, but Neogaf verified it at 175 or so. Who knows if its true though.

It's 176. The Wii U is more efficient which is why, when games are made it's-way, it can have a massive lead over the Xbox 360 and Playstation 3 graphically.

Do you have any source for the 176 part, the only place I read about it was from neogaf forums...



https://www.youtube.com/watch?v=n29CicBxZuw

01001011 01101001 01110011 01110011 00100000 01101101 01111001 00100000 01110011 01101000 01101001 01101110 01111001 00100000 01101101 01100101 01110100 01100001 01101100 00100000 01100001 01110011 01110011 00100001

Actually, it was confirmed that nVidia was the chip manufacturer for NX, not AMD as people previously speculated.



I describe myself as a little dose of toxic masculinity.