Nintendo Teaming Up With Nvidia For NX?

Source

This would be huge news if true. Finally a console maker is leaving those scrubs at AMD and working with the winning team! Maybe we will finally get a real console???

You do realize that this isn't the first console with an Nvidia GPU, don't you?

Also, it's using Tegra, a mobile chipset. It'll be considerably weaker than XB1, but will consume under 30W.

@GoldenElementXL: Yeah, maybe we'll get another real console like the Wii U.

I'm so sick of hearing about this stupid thing, nobody knows shit, just stop...

That website has zero credibility. Also, we know Nintendo is not gonna reveal the console at E3, yet this website expects a full reveal there lol.

Lol we'll have another Nvidia Shield. Total failure from the monopoly.

Would be cool for Nvidia to hook Nintendo up so they can continue to dominate the market.

@GoldenElementXL said:

Source

This would be huge news if true. Finally a console maker is leaving those scrubs at AMD and working with the winning team! Maybe we will finally get a real console???

http://semiaccurate.com/2010/07/14/sonys-psp2-powered-nvidias-tegra-line/

Charlie Demerjian asserted Sony's PSP2 has NVIDIA Tegra and it was proven wrong.

Have you realized NVIDIA Pascal is a near AMD GCN clone?

Both AMD GCN and NVIDIA Pascal's SM or CU building blocks has 64 32bit stream processors and 256 Kilobytes of register file.

To deliver XBO level performance, it would need about Pascal 14 SM units and 850 Mhz.

PC's Geforce GTX 1080 has 40 SM units with 1600 Mhz base clock with 1700 Mhz boost. This is like Radeon HD R9-390's 40 CU and overclocked to 1700 Mhz

For NX, Emily Rogers has asserted against Polaris, 2.5 TFLOPS, 2X over PS4 i.e. 3.6 TFLOPS.

-------------------

AMD's Mobile Polaris 10 with 32 CU and 1350 Mhz base clock yields 5.5 TFLOPS with 110 to 130 watts, Vega's 64 CU scaled version yields 11 TFLOPS.

Mobile Polaris 11 has 14 CU clocked at 1400 Mhz base clock with ~50 watts.

AMD Polaris also has high clock speed.

I thought Nvidia didn't do console GPU's because there isn't enough mark up for them to be bothered.

That's just NVIDIA PR. Watch what they do NOT what they say.

Well, AMD has several semi-custom SoC design wins for 2016 and 2017.

You do realize that Nvidia's new Tegra is more powerful than what the XB1 can produce right? The current one is already slightly stronger than a ps3.

The next Tegra should only be about 40% faster than the current one, so it's weaker than Xbone. It's still technically a cutting-edge, industry-leading 2017 console, but not in the way people want it to be.

While true, SemiAccurate has also said that NX is Tegra-based (though they called it a handheld), and Emily Rogers has apparently said that NX will use Nvidia chips. It all adds up. NX was using a mobile chipset all along. It will not compete with current-gen or try to get third-party support, but has instead been designed from the ground up as a cheap Nintendo box. Wii-like power consumption, attempt at a Wii-like gimmick, Wii-like price point, and increased first-party output. They've given up on competing and third-parties completely and are just doing their own thing. This is the future of Nintendo. They will never receive a western third-party game (other than mobile games) again, and that's okay.

Ideally, this would hit a $200 price point, but it'll probably be $250. All handheld games will work on it too. Will that be enough to keep them in hardware, or is this the end of the line?

TSMC fab'ed 20 nm Tegra X1 has about 500 GFLOPS (32 bit FP) and 1 TFLOPS (16 bit FP).

Terga X1 set-top-box is weaker Microsoft's Surface 4 Pro tablet!! Handheld version would be slower than Terga Shield TV.

XBO beats Surface 4 Pro

3DMark (2013) - Ice Storm Unlimited Graphics Score

R9-M370X(10 CU at 800 Mhz, 32bit FP 1.024TFLOPS) has 188,948 points http://www.notebookcheck.net/AMD-Radeon-R9-M370X.142763.0.html

If you double Shield TV's X1's 61735 points into 123,470 points, it would still inferior to XBO (12 CU, 853Mhz, 32bit FP 1.31 TFLOPS) and R9-M370X. (Apple MacBook Pro 2015).

The fact that the nx was delayed till next year let's me believe the specs are changing and all these rumors are even less credible

I don't trust Nvidia. They are a shady company that try's to screw everyone over to make maximum profit, including their loyal customers. Nintendo better watch there backs.

Nvidia has proven problematic partner to work with in consoles, I don't think any of the big three want to join forces with them again. Plus Nvidia already went out and bad mouthed the big three when they went with AMD this time around.

If Nvidia is gonna enter console market again it'll be when room is made and they enter themselves with their own hardware.

Anyhow, I don't think Nintendo is going for beefy hardware. That would be a huge gamble trying to lure western third parties when Nintendo still proves problematic courting them. Console-handheld hybrid or bust. At least then they'll grab the Japanese third party enthusiasm.

@ronvalencia said:

@b_rich84 said:

@techhog89 said:

You do realize that this isn't the first console with an Nvidia GPU, don't you?

Also, it's using Tegra, a mobile chipset. It'll be considerably weaker than XB1, but will consume under 30W.

You do realize that Nvidia's new Tegra is more powerful than what the XB1 can produce right? The current one is already slightly stronger than a ps3.

TSMC fab'ed 20 nm Tegra X1 has about 500 GFLOPS (32 bit FP) and 1 TFLOPS (16 bit FP).

Terga X1 set-top-box is weaker Microsoft's Surface 4 Pro tablet!!

XBO beats Surface 4 Pro

3DMark (2013) - Ice Storm Unlimited Graphics Score

R9-M370X(10 CU at 800 Mhz, 1.024TFLOPS) has 188,948 points http://www.notebookcheck.net/AMD-Radeon-R9-M370X.142763.0.html

And NX can only be about 40% faster than this unless they increase power consumption compared to X1 significantly. What a joke. Nintendone.

@NFJSupreme: it wasnt delayed. a release date was never announced.

also when someone is thinking of a rumor they need to remember to keep it plausible. this fell apart when Pascal was mentioned.

the NX, in terms of development, is 99% done. the spec is done, the testing is almost done....the system is locked down and has been for a while. at this stage is about proving the manufacturing side is working and making sure it can be assembled as efficiently as possible. they don't just think "hey lets make a console", pick some random hardware up and call it done after a few months. the NX has been in the oven since the wiiu launched.

if the rumor said maxwell then maybe, maybe, there could be some sembelance of truth to it. but its not going to use polaris or pascal or zen or vega. at best it may take some elements from those as nvidia/AMD may have parts of those GPUs that are ready to be deployed in a custom SOC...maybe. we saw it with the GPU in the 360 for example. but even that is unlikely.

Here is a sample of plausible rumors (note: not true....just its possible):

nintendo using ARM for its CPU..thats a plausible rumor.

nintendo partnering with intel to use a custom Atom processor...thats possible.

nintendo have managed to source HBM and will be using it as a unified memory for the CPU and GPU.....not impossible.

the NX console will actually be based on a very beefed up snapdragon SOC...yeah maybe.

@osan0: why is AMD not plausible? Also "last minute " changes can happen since the ps4 got a ram boost after initial dev kits where out. So assuming there was truth to a 2016 release the fact it was pushed back ( again assuming, we are all working off rumors here) make me think the specs could be changing (slightly not dramatically) which makes me doubt how accurate these rumors are.

@NFJSupreme: sorry im missig something....why do you ask "why is AMD not plausible?"? the NX is an AMD based console more than likely.

a ram boost is a minor tweak (was it actually boosted or was it just more of the 8GB being made available to games due to OS optimisation? either way minor tweak). a switch from an AMD to an nvidia GPU is a more fundimental rethink. that affects tools, engines, code optimisation...everything. its way too late to change it....a change like that could bring a very big delay. it would be like changing from an x86CPU to a powerPC CPU..its a fundimental change in how the console works.

the CPU architecture, GPU architecture, type of ram, memory setup...these are locked down for the NX. they can play around with things like clock speeds, amount of ram, number of shaders, number of CPU cores (though at this stage those better be going up, not down. revisions down would be really bad now for developers). they can tweak and test that up until mass production begins really. but a change to the architecture is a far bigger change.

Too many boring rumours. This thing will either be more powerful than Galactus or weaker than Aston Villa.

@DocSanchez: or both.

wait...what if all the rumors are true....from a certain point of view....*head explodes* :P

I don't believe Emily Rodgers said anything about Nvidia. Just that NX won't be X86.

Nintendo has no long-term future if their consoles continue to be 1st party only boxes. Only 12 million people will buy a console for only Nintendo games. Again, the only way Nintendo changes that is if they come out with a new IP on the level of Minecraft, which is easier said than done.

On NeoGAF someone who spoke with her privately said that she'd heard about Nvidia being involved in NX, which is why the Polaris rumors were "wacky."

I added more to my previous post.

So, you're telling me that you'd never buy Nintendo console for their games at any price? What it's it's just $200? $150? $99?

Also, are you really trying to say that lack of third-party support was Wii U's only problem? Seriously?

The Wii U's main problem was it not having a killer app that appealed to people outside of Nintendo fans.

Nintendo needs another big hit on the level of Wii Sports if they want to sell more than 12 million consoles.

The NX isn't going to be successful just by having new installments of aging IP's and a couple of new IP's on the level of Splatoon released here and there. They need really big hitters. They need something that's going to get PC, XB1 and PS4 owners on board.

I'd pay up to $249.99 for a Nintendo console if it had games I wanted to play.

So, you think that the price, the value for the the money, the controller, and the marketing all played essentially no part in Wii U's failure? It was all just because it was a Nintendo box and launching at a cheaper price with better marketing wouldn't have helped it even a little bit?

Yuck. Nvidia gpu based consoles have been crap.

NVIDIA Shield TV cost $199.99 with 16 GB storage or $299.99 with 500 GB storage. It doesn't include Blu-Ray drive.

FinFET enables NVIDIA to increase clock speed minimal power consumption increase.

NVIDIA's X1 SoC is already using TSMC's 20 nm process node.

Expect around 40 percent performance increase from TSMC's 20 nm to 16 nm FinFET.

Marketing wouldn't have made much of a difference as the console would have still been held back by the fact that its just a Nintendo box. Majority of people who want to play Nintendo games would rather save money and purchase a cheaper Nintendo handheld. People who don't like Nintendo games wouldn't have bought a Wii U even if it had the best marketing behind it. This is the age of the internet. People our age and younger don't watch TV as much and don't need to wait for commercials to know when a product is coming. People are more in the know now than ever before. I'm not saying marketing doesn't help, but its no longer as important as it once was.

@GoldenElementXL said:

Source

This would be huge news if true. Finally a console maker is leaving those scrubs at AMD and working with the winning team! Maybe we will finally get a real console???

Maybe you forgot that Nvidia is more expensive than AMD and chances are they get something weak because of it,PS3 style.

That would also mean probably that it would not be backward compatible since emulating AMD GPU will surely land nintendo in court with AMD unless they agree on it before hand,i remember how MS had to pay Nvidia royalties because of the 360 emulation of xbox games.

@ronvalencia said:

@techhog89 said:

@emgesp said:

@techhog89 said:

@emgesp said:

That website has zero credibility. Also, we know Nintendo is not gonna reveal the console at E3, yet this website expects a full reveal there lol.

While true, SemiAccurate has also said that NX is Tegra-based (though they called it a handheld), and Emily Rogers has apparently said that NX will use Nvidia chips. It all adds up. NX was using a mobile chipset all along. It will not compete with current-gen or try to get third-party support, but has instead been designed from the ground up as a cheap Nintendo box. Wii-like power consumption, attempt at a Wii-like gimmick, Wii-like price point, and increased first-party output. They've given up on competing and third-parties completely and are just doing their own thing. This is the future of Nintendo. They will never receive a western third-party game (other than mobile games) again, and that's okay.

Ideally, this would hit a $200 price point, but it'll probably be $250. All handheld games will work on it too. Will that be enough to keep them in hardware, or is this the end of the line?

I don't believe Emily Rodgers said anything about Nvidia. Just that NX won't be X86.

Nintendo has no long-term future if their consoles continue to be 1st party only boxes. Only 12 million people will buy a console for only Nintendo games. Again, the only way Nintendo changes that is if they come out with a new IP on the level of Minecraft, which is easier said than done.

So, you're telling me that you'd never buy Nintendo console for their games at any price? What it's it's just $200? $150? $99?

Also, are you really trying to say that lack of third-party support was Wii U's only problem? Seriously?

NVIDIA Shield TV cost $199.99 with 16 GB storage or $299.99 with 500 GB storage. It doesn't include Blu-Ray drive.

FinFET enables NVIDIA to increase clock speed minimal power consumption increase.

NVIDIA's X1 SoC is already using TSMC's 20 nm process node.

Expect around 40 percent performance increase from TSMC's 20 nm to 16 nm FinFET.

Rumor is that NX won't have an optical drive, and knowing Nintendo it won't have an HDD either. So, $200 is possible. Also, don't forget that Nvidia like fat margins. Shield TV was probably making a good $70-80 per unit at launch.

Wii U version 2.0.

Option 1

NX handheld device's weaker mobile Tegra SoC could be boosted by another Shield TV like GPU via non-standard link e.g. NVlink.

Shield Tablet = $199, modified with smaller screen and built-in game controller

Shield TV = $199, with 16 GB and separate game controller. Supplemental Computing Device...

Total price around $400.

Total system performance just below R9-M370X and XBO.

Option 1 kills New 3DS.

----------------------------

Option 2

NX handheld device just a dumb device with a small screen and builtin game controller

Shield TV = $199, with 16 GB and separate game controller

Total price around $299.

Total system performance around Intel Surface 4 Pro Tablet.

So it'll be terrible price:performance when compared to AMD/PS4/XBO/Neo.

Got it.

Have people considered that perhaps Nintendo need what Nvidia are offering with the Tegra as a solution for this particular unknown device of theirs.

What is the AMD equivalent?

Shield TV uses Terga X1 which is fab'ed on lower power TSMC's 20 nm process.

NVIDIA's Shield Tablet

AMD aborted their TSMC 20 nm switch.

http://www.kitguru.net/components/cpu/anton-shilov/amd-officially-cancels-20nm-chips-takes-33-million-charge/

Regardless of AMD or NVIDIA solution, NX's power level is a cause for concern.

How would this be good?...

Not only would it be a nvidia GPU but it won't be a X86 platfrom meaning once again say bye bye to multiplatforms and third party games that look good(harder to code for).

@GoldenElementXL: First of all if someone is a scrub team thats NVIDIA , the reasons .. too many ! You can google if you dont know em already.

Now , NVIDIA or AMD , doesnt matter. For months now im calling out Nintendo for NX hardware. We once again hearing for fancy controllers and fancy ways to play games which all that are PRICEY and in order for NX to compete it needs a viable price versus the already established PS4/X1 and their new versions that most likely will come sooner than NX . How many times i need to say it ? Hardware >> if controller has fancy stuff on it or if console is a transformer or if it makes coffee when you playing MARIO

So whats the deal here ? Will NX cost more than PS4 in order to provide fancy stuff ALONG with competitive hardware or whats the deal ?

NX might be once again a low end hardware that after 2-3 years wont be able to deliver 3rd party games ( if not at all ) because of its hardware limitations ( along sales ) and another one of their systems will end up in life support or dead after 3 years like it happened more or less with Wii and Wii U.

Thats something you sheep need to consider .. NVIDIA or AMD , shouldnt be your main concern right now , if not at all.

Besides , to understand how hypocrites NVIDIA guys are if this is true , NVIDIA was dissing AMD for being a PC traitor when all consoles went AMD all the way , and that they are the real deal etc ... I wonder if they end up in NX what theyll say afterwards lol

FinfET yields about 1.7X perf/watt or 70 percent improvement e.g. 980 ti (~8 billion transistors) and 1080 (~7.2 billion transistors) has similar transistors count and the general clock speed gain was from 1000Mhz to 1700Mhz.

http://www.overclock3d.net/articles/gpu_displays/amd_has_two_polaris_gpus_coming_this_year/1

Staff from the Radeon Technology Group did admit that the bulk of the efficiency improvements that we will see with AMD's newest GPUs will come from the so-called "FinFET Advantage", with PCPER stating that is is "on the order of a 70/30 split".

AMD also stated their FinFET has 70 percent or 1.7X improvements.

20 nm is a half node process improvement i.e. for TSMC, it has reduce power consumption. Are you better than Apple?

Your "20 nm being failed node" assertion is debunked by Apple i.e. TSMC's 20 nm is better than TSMC's 28 nm..

.Note Shield TV X1's result vs Surface Pro 4.

http://www.notebookcheck.net/AMD-Radeon-R9-M370X.142763.0.html

Radeon R9-M370X (10 CU at 800mhz, 1.024 TFLOPS) scored 188,948.

-------------------------------

https://developer.nvidia.com/dx12-dos-and-donts

On DX11 the driver does farm off asynchronous tasks to driver worker threads where possible

With DX11, NVIDIA DX11 drivers already using key DX12 like speed up methods i.e.

1. asynchronous tasks

2. threads i.e. that's multiple threads.

Under DX12, AMD catches up to NVIDIA in relation to asynchronous and multiple threads.

980 Ti at 1380 Mhz yield about 7.7 TFLOPS

Fury Nitro yields about 7.5 TFLOPS

Without Gamework's tessellation overdraw BS and both GPU vendors has multithreading and async task, the TFLOPS between AMD and NVIDIA are similar.

Firstly, isn't this just for the handheld device? Secondly, this Tegra would use pascal wouldn't it?

Thirdly, how much does it matter with Nintendo games?

I think we should wait until we see what this device is about before we get upset about Nintendo's choice of hardware.

@zelus: Wrong. Nvidia X1 is on TSMC's 20 nm process node not on 28 nm process node and X1 is targeting lower power consumption. The reason why AMD avoided TSMC's 20 nm is due to "bang for the buck" issue relative to FinFET.

http://techfrag.com/2015/07/09/amd-scraps-20nm-process-tech-to-focus-on-finfet/

“In the past I talked about 20nm node that we did some designs on,” said Lisa Su, CEO AMD at the company’s financial analyst day. “We have started some initial designs, we have run some silicon, but those parts are probably not going to go into production because we think we can get much more bang for the buck out of FinFET technologies going forward.”

Another 70 percent or 1.7X improvement for 28nm to 16 nm FinFET claim ARM Ltd.

Your assertion for 20 nm equal zero improvements from 28 nm was debunked by ARM, Apple and Qualcomm.

@zelus:

Simply put Nvidia and AMD do not calculate gflops the same way. A 1 tflop gpu from Nvidia isn't the same as a 1 tflop gpu from AMD, because they aren't calculating that figure using the same equation. Much of the advantage comes down to drivers on the PC environment between Nvidia vs AMD, but in a closed box a 1 tflop gpu from Nvidia is always going to out perform a 1 tflop gpu from AMD. Its not exactly easy to compare the two as apple to apple equivalents.

That's unsupported assertion. Shield TV's X1 has about 500 GFLOPS 32bit FP and it's 3Xless than R9-M370X's 1 TFLOPS 32bit FP in Ice Storm Unlimited benchmark.

NVIDIA has applied very aggressive clock gating for their X1 SoC.

NVIDIA's Shield TV is not running Microsoft Windows.

Without tessellation overdraw BS and both camps has the same multi-threading speed up features, Hitman DX12 (lower level API) shows similar FLOPS between NVIDIA and AMD.

Emily Rogers already dismissed NX being on par with PS4.

@zelus said:

20nm has little thermal or performance benefits compared to 28nm half-node its partly why AMD or Nvidia doesn't think much of it and didn't end up using it in dedicated graphics. It doesn't have much of a market outside mobile where any thermal savings is worth having. 20nm's shortcomings are why development in dedicated graphics cards slowed down for a period of time. Nvidia and AMD not using the process for dedicated graphics and the numerous reasons why only takes a short google search to look into. Instead both went for FinFET alternatives after long delays.

Simply put Nvidia and AMD do not calculate gflops the same way. A 1 tflop gpu from Nvidia isn't the same as a 1 tflop gpu from AMD, because they aren't calculating that figure using the same equation. Much of the advantage comes down to drivers on the PC environment between Nvidia vs AMD, but in a closed box a 1 tflop gpu from Nvidia is always going to out perform a 1 tflop gpu from AMD. Its not exactly easy to compare the two as apple to apple equivalents. Depending of feature set relative performance between the two go up or down and AMD has made large improvements, but with any arch improvements between the current consoles and a newer more modern Nvidia gpu and what remains of the edge between products gflop vs gflop its still fairly safe to suggest those rough numbers that Nintendo doesn't need equivalent tflops to beat Xbox One or PS4 performance wise with new hardware. Same would be true with a newer AMD gpu, but fundamentally at least in that case the gflops are still calculated the same way. Its been part of why Nvidia has historically out performed AMD in benchmarks even when the Nvidia card has a lower amount of gflops besides the greater software support Nvidia supplies to the PC market.

@ronvalencia said:

@zelus said:

Assuming 16FF process much less if its an Pascal based design the wild conclusions majority of you are coming up with is far too pessimistic. Not that its a bad habit when it comes to Nintendo. 20nm is a failed node. The difference between a 16 FinFET design and the older 20nm design would likely be more than 40%. Then there is the simple reality Nvidia gflops aren't the same as AMD gflops. A 1 tflop to 1.1 tflop Nvidia gpu would likely preform similarly to Xbone while you likely could match PS4 with a modern Nvidia design somewhere around 1.3 to 1.4 tflops. It all depends on the TDP Nintendo is going for, but I don't think they need a max TDP of more than 50 watts to match PS4 even if the on paper gflops seems less than PS4 assuming they are going with Nvidia in reality and the chip is fairly modern.

FinfET yields about 1.7X perf/watt or 70 percent improvement e.g. 980 ti (~8 billion transistors) and 1080 (~7.2 billion transistors) has similar transistors count and the general clock speed gain was from 1000Mhz to 1700Mhz.

http://www.overclock3d.net/articles/gpu_displays/amd_has_two_polaris_gpus_coming_this_year/1

Staff from the Radeon Technology Group did admit that the bulk of the efficiency improvements that we will see with AMD's newest GPUs will come from the so-called "FinFET Advantage", with PCPER stating that is is "on the order of a 70/30 split".

AMD also stated their FinFET has 70 percent or 1.7X improvements.

20 nm is a half node process improvement i.e. for TSMC, it has reduce power consumption. Are you better than Apple?

Your "20 nm being failed node" assertion is debunked by Apple i.e. TSMC's 20 nm is better than TSMC's 28 nm..

.Note Shield TV X1's result vs Surface Pro 4.

http://www.notebookcheck.net/AMD-Radeon-R9-M370X.142763.0.html

Radeon R9-M370X (10 CU at 800mhz, 1.024 TFLOPS) scored 188,948.

-------------------------------

https://developer.nvidia.com/dx12-dos-and-donts

On DX11 the driver does farm off asynchronous tasks to driver worker threads where possible

With DX11, NVIDIA DX11 drivers already using key DX12 like speed up methods i.e.

1. asynchronous tasks

2. threads i.e. that's multiple threads.

Under DX12, AMD catches up to NVIDIA in relation to asynchronous and multiple threads.

980 Ti at 1380 Mhz yield about 7.7 TFLOPS

Fury Nitro yields about 7.5 TFLOPS

Without Gamework's tessellation overdraw BS and both GPU vendors has multithreading and async task, the TFLOPS between AMD and NVIDIA are similar.

If they're "Industry leaders", then yes.

What about 16 nm FinFET Pascal? NVIDIA's X1 SoC is already using TSMC's 20 nm half node improvement.

Nintendo's 1st party games wasn't able to change Wii U's result for being last.

Emily Rogers already dismissed 2X over PS4, on par with PS4 level and 2.5 TFLOPS level.

Regardless of AMD or NVIDIA selection, Nintendo seems to execute another Wii U version 2.0.

Slight correction: As I posted in my thread earlier, it seems that Emily's info was based not on gaming performance, but from a direct comparison of specs on paper. Look in the last couple of pages in the thread on NeoGAF to If Nintendo is using an Nvidia GPU, it could have specs that look like they're on-par with XB1 on paper (such as 1.3 TFLOPs or 768 cores), but actually perform at or above PS4 level. So, all we can really say for sure at this point is that it doesn't take on PS4k. Also, Emily said that the specs were close to XB1, so that rules out the console using a straight TX1 completely no matter how you look at it.

Also, again, the 2.5 TFLOPs post by 10k was SPECULATION SPECULATION NOT A RUMOR BUT INSTEAD SPECULATION!!!!!!!!!!!!!! It has been implicitly ruled out by the latest post, but that post that you keep clinging to was not a rumor by 10k. Why do I have to tell you this every time? You can't be this dense.

What about NeoGAF? Why don't they come here? Most Nvidia fanboys didn't realize NVIDIA DX11 drivers are already using DX12 like speed-up methods while zero on AMD DX11 drivers. Under DX12, both AMD and NVIDIA has async and multi-threading. DX12 wouldn't solve AMD's tessellation problem, but that's doesn't refer to shaders/stream processors' power i.e. FLOPS

Here is a box powered by NVIDIA GPU i.e. Shield TV.

X1 has 500 GFLOPS 32bit FP with 3X slower than R9-M370X (1 TFLOPS) in Ice Storm Unlimited benchmark.

NeoGAF's ozfunghi claimed

You can't compare Nvidia flops to AMD flops as 1:1. A 1TF Nvidia part will always outperform a 1TF AMD part.

http://www.neogaf.com/forum/showpost.php?p=203859900&postcount=1948

NeoGAF's LordOfChaos's counter argument against NeoGAF's ozfunghi

Pedantic sidenote, but no, not always, their architectures flip flop on ALU performances ratio to shader performance to fill rate and everything else. Sometimes a design will have more FLOPs for the other performance metrics, sometimes less.

GCN is currently an architecture where a certain performance level has a higher Gflops number than the equivalently performing Nvidia one. Wasn't always the case, won't always be the case.

It also depends on how much the GPU maker cares about compute. i.e, Intel cared about GPU compute to not be left at a disadvantage, so the lowly Iris Pro had a higher Gflop number than you'd expect for its performance level, and it showed, even besting the 750M at compute while losing to it in gaming performance. So higher performance per flop isn't necessarily a good thing either, just different design choices on who prioritizes compute more.

@techhog89 said:

Also, again, the 2.5 TFLOPs post by 10k was SPECULATION SPECULATION NOT A RUMOR BUT INSTEAD SPECULATION!!!!!!!!!!!!!! It has been implicitly ruled out by the latest post, but that post that you keep clinging to was not a rumor by 10k. Why do I have to tell you this every time? You can't be this dense.

Emily Roger has stated closer to X1 i.e. it could fall either weaker than XBO or slightly stronger XBO.

"2.5 TFLOPS" was fromReddit [verified] User shares NX info: x86 Architecture, Second screen support etc. : Page 61

Emply Rogers' 10K debunk was fromRUMOR: NX more powerful than PS4, Splatoon/Mario Maker ports in development : Page 139

1st post for "RUMOR: NX more powerful than PS4, Splatoon/Mario Maker ports in development" thread referred to parent thread "Reddit [verified] User shares NX info: x86 Architecture, Second screen support etc".

Emply Rogers' debunk killed 10K argument which includes parent thread Reddit [verified] User shares NX info: x86 Architecture, Second screen support etc..

From http://neogaf.com/forum/showthread.php?t=1207224

10K claimed

"The NX will be more powerful than the PS4. "By a noticeable amount"

The NX will use a custom Polaris-like GPU

"2.5 TFLOPS" is an example for 10K's "The NX will be more powerful than the PS4. "By a noticeable amount" and "The NX will use a custom Polaris-like GPU" view point.

2.5 TFLOPS number derived from Polaris 11.

Why do I have to tell you this every time? You can't be this dense.

Except this may not be the main Nvidia GPU. It could just be for the controller we don't know.

That's why I said we should wait to see what the deal is and how this could be used.