NVIDIA Fires Shots at AMD’s 7nm Tech - Claims "Can Create Most Energy-efficient GPU in the World Anytime"

[MeatFeastMan]

1 hour ago, thorhammerz said:

Good luck finding Radeon VII's in stock anywhere near the price of an RTX 2070.

 

2070's in stock for ~$500 (or slightly cheaper). Radeon VII's at.... $700? That's 2080 pricing (and therefore, what consumers will compare it to).

You won't be laughing in 2 years time when the 2080 is dropping frames off a cliff due to having 8gb of vram, meanwhile the Radeon VII is perfectly smooth. We're seeing glimpses of what is to come already, some games use over 10gb at 4k now. I will never regret buying a Radeon VII over the 2080. The 2080 vs Radeon VII is just another case of what happened with the 290x and it's Nvidia rivals.

 

Look at the 290x, it's STILL a viable card. Where are it's rival cards from Nvidia. In the bin, dead by now. Why? Because they lacked vram. The 780ti has gone down the drain because it only had 3gb.

 

Now look at Turing. The 2060 has 6gb, the 2080 has 8gb. It's criminal as far as I'm concerned. It's clear Nvidia are setting traps for buyers. They want you to buy their current generation, and then buy next generation by forcing you into a bad experience because they gave you no vram to play with. More money for them. I will never compromise on longevity and other people should do the same.

 

Also, when was the Radeon VII a 2070 rival? It was always meant to compete against the 2080. Yes, it falls short for the most part. But it is definitely in the same ballpark. And in a lot of heavy compute workloads it wipes the floor with even the 2080ti.

 

 

[FloRolf]

Too bad they still can't make the most money efficient GPU in the world ¯\_(ヅ)_/¯

[pas008]

15 minutes ago, MeatFeastMan said:

You won't be laughing in 2 years time when the 2080 is dropping frames off a cliff due to having 8gb of vram, meanwhile the Radeon VII is perfectly smooth. We're seeing glimpses of what is to come already, some games use over 10gb at 4k now. I will never regret buying a Radeon VII over the 2080. The 2080 vs Radeon VII is just another case of what happened with the 290x and it's Nvidia rivals.

 

Look at the 290x, it's STILL a viable card. Where are it's rival cards from Nvidia. In the bin, dead by now. Why? Because they lacked vram. The 780ti has gone down the drain because it only had 3gb.

 

Now look at Turing. The 2060 has 6gb, the 2080 has 8gb. It's criminal as far as I'm concerned. It's clear Nvidia are setting traps for buyers. They want you to buy their current generation, and then buy next generation by forcing you into a bad experience because they gave you no vram to play with. More money for them. I will never compromise on longevity and other people should do the same.

 

Also, when was the Radeon VII a 2070 rival? It was always meant to compete against the 2080. Yes, it falls short for the most part. But it is definitely in the same ballpark. And in a lot of heavy compute workloads it wipes the floor with even the 2080ti.

 

 

while you are looking into  your crystal ball about futureproofing  can you tell me if valve is planning to release a 3rd installment of one of their series soon? lol ok then

which is a ridiculous argument itself in tech, 10gb at 4k already really? but they bought 2070 for 4k lol?

moot point

you should buy the best for what you need now, budget, or the best thats out otherwise everyone should would just keep waiting and nothing will be bought which leads to no options

 

[cj09beira]

On 4/1/2019 at 6:21 AM, vorticalbox said:

Well riding a bicycle is more efficient than driving a car by a massive amount but you would want to ride one across a state.

its though, how efficient are we, if you take into account how much energy is wasted to make a single kilo of food, i bet we would be less efficient than cars

5 hours ago, ryao said:

 

Nvidia’s GPUs are significantly more energy efficient than AMD’s:

 

https://www.anandtech.com/show/13923/the-amd-radeon-vii-review/19

 

Their drivers are also better. Unlike AMD’s drivers, Nvidia’s drivers support GPU resets if the GPU hangs yet hangs seem to occur far less on Nvidia hardware than on AMD graphics hardware. I had nothing but ATI graphics hardware from 1998 to 2006 when I got fed up and switched to Nvidia. I have enjoyed far fewer graphics issues since then. I am told that little has changed. The DXVK author puts up with the lack of GPU reset support in AMD’s drivers, but uses Polaris instead of Vega because the driver issues with Pascal are more severe.

 

I dislike Nvidia’s use of a binary blob for a driver, but upon looking at AMD’s hardware which on Linux has an open source driver stack, I concluded that even if I expended enormous time on fixing the drivers, energy efficiency would never match Nvidia.

amd's drivers and hardware are nowhere near what they were more than 10 years ago, i have been using an rx480 for close to 2 years now, and had no gpu related crashes that did not involve gpu overclocking, and never have my pc been restarted because of a gpu crash, on linux which i have been using now for around 6 months i have not had any problem what so ever 

[thorhammerz]

56 minutes ago, Stefan Payne said:

https://www.mindfactory.de/Highlights/MindStar

689 is reasonable for a card with 16GiB VRAM

2070 has 8GiB VRAM

R7 has 16GiB VRAM

And yet still paying 37% extra for... 10-15% more performance?

1 hour ago, Stefan Payne said:

And look at Level1 Techs Youtube Channel. The Video with EposVox, where he used a Radeon 7 and it worked, where it didn't with nVidia.

Link + timestamp, si vous plait?

 

Quoting an instance of X card working in a situation where Y card doesn't isn't particularly new, but I suppose cherry picking data points is inevitable when there's so few to begin with (kinda like the number of cards produced, or the lack thereof... amirite?)

52 minutes ago, MeatFeastMan said:

-snip-

Fear not! A comparison of performance 2 years into the future will be conducted in about.... 2 years into the future !

[Stefan Payne]

6 minutes ago, thorhammerz said:

And yet still paying 37% extra for... 10-15% more performance?

Yeah, that's normal in the nVidia Infrastructure, isn't it?!
Like 1070->1070ti->1080 for example.

Everytime you see a slight increase with a big increase in cost...

 

But its OK when we do it, right??

6 minutes ago, thorhammerz said:

Link + timestamp, si vous plait?

The whole Video is about Radeon 7 in Adobe Premiere.

 

 

6 minutes ago, thorhammerz said:

Quoting an instance of X card working in a situation where Y card doesn't isn't particularly new, but I suppose cherry picking data points is inevitable when there's so few to begin with (kinda like the number of cards produced, or the lack thereof... amirite?) 

Because its an instance that proves that the 16GiB VRAM are an advantage and 8GiB VRAM just not enough.

Also HBCC works and is also an advantage.

 

We have 8GiB in the Consumer Market since 390 Series - 4 years ago!!

And now, 4 Years later, you still claim that 8GiB VRAM are enough?!
 

RLY?!

[Blademaster91]

1 hour ago, MeatFeastMan said:

You won't be laughing in 2 years time when the 2080 is dropping frames off a cliff due to having 8gb of vram, meanwhile the Radeon VII is perfectly smooth. We're seeing glimpses of what is to come already, some games use over 10gb at 4k now. I will never regret buying a Radeon VII over the 2080. The 2080 vs Radeon VII is just another case of what happened with the 290x and it's Nvidia rivals.

 

Look at the 290x, it's STILL a viable card. Where are it's rival cards from Nvidia. In the bin, dead by now. Why? Because they lacked vram. The 780ti has gone down the drain because it only had 3gb.

 

Now look at Turing. The 2060 has 6gb, the 2080 has 8gb. It's criminal as far as I'm concerned. It's clear Nvidia are setting traps for buyers. They want you to buy their current generation, and then buy next generation by forcing you into a bad experience because they gave you no vram to play with. More money for them. I will never compromise on longevity and other people should do the same.

 

Also, when was the Radeon VII a 2070 rival? It was always meant to compete against the 2080. Yes, it falls short for the most part. But it is definitely in the same ballpark. And in a lot of heavy compute workloads it wipes the floor with even the 2080ti.

 

 

Except you buy a GPU for your needs and for the games and apps you run now, and you won't know what a game will need for system requirements later. What games are using 10GB of VRAM or more? There are plenty of games that will take the VRAM but not actually utilize it. And personally I'd rather play games at 1440P than spend a lot more money to game at 4K high FPS. Arguing only about the VRAM is kind of pointless, you have to look at how the cards actually compare in games and the radeon VII is slower than a 2080.

[cj09beira]

11 minutes ago, thorhammerz said:

And yet still paying 37% extra for... 10-15% more performance?

Link + timestamp, si vous plait?

 

Quoting an instance of X card working in a situation where Y card doesn't isn't particularly new, but I suppose cherry picking data points is inevitable when there's so few to begin with (kinda like the number of cards produced, or the lack thereof... amirite?)

Fear not! A comparison of performance 2 years into the future will be conducted in about.... 2 years into the future !

i saw it before, basically its one of those edge cases where having the extra vram is the difference between the card finishing the project or not

[MeatFeastMan]

15 minutes ago, Blademaster91 said:

Except you buy a GPU for your needs and for the games and apps you run now, and you won't know what a game will need for system requirements later. What games are using 10GB of VRAM or more? There are plenty of games that will take the VRAM but not actually utilize it. And personally I'd rather play games at 1440P than spend a lot more money to game at 4K high FPS. Arguing only about the VRAM is kind of pointless, you have to look at how the cards actually compare in games and the radeon VII is slower than a 2080.

Right, but if there was a graph that showed average fps of the two cards as time went on (Over maybe a period of 4-5 years), you'd find the Radeon VII would likely come out on top. Because although the maximum fps is likely higher on the 2080, the overall fps will be more choppy and the experience would be worse.

 

Games like Black Ops 4 (It's horribly optimized but still an example), Far Cry 5 with the textures on in some more intensive areas of the map, The Division 2 at 4k again in some areas of the map. They all eat up VRAM for fun. Resident Evil 2 is on the brink of passing the 2080. And there's a few more that are on the edge of the 2080's vram limit. Once those numbers are consistently over the 9gb mark, it'll drop frames and the experience overall as well as the minimums will plummet. Within 2 years I expect us to be on the 10gb number at 4k, and around at least 7gb at 1440p consistently.

 

I'll try to find the article AMD published a month or two ago showing the benefits over the extra vram and HBCC on Radeon VII and Vega. It's really useful. Yes, it's from AMD but it doesn't seem to me like a biased article much if at all.

 

Edit: Here we go, take a look: https://community.amd.com/community/gaming/blog/2019/02/07/the-radeon-vii-s-16gb-of-memory-tackles-the-most-demanding-games-and-content-creation-workloads

 

 

 

[thorhammerz]

18 minutes ago, Stefan Payne said:

Yeah, that's normal in the nVidia Infrastructure, isn't it?!
Like 1070->1070ti->1080 for example.

Everytime you see a slight increase with a big increase in cost...

Still doesn't make it anymore "worth" buying for the performance you get . And do note, this applies to comparing between Nvidia cards as well - price to performance is still price to performance regardless of who makes the damn thing.

Quote

But its OK when we do it, right??

The whole Video is about Radeon 7 in Adobe Premiere.

 

Because its an instance that proves that the 16GiB VRAM are an advantage and 8GiB VRAM just not enough.

Also HBCC works and is also an advantage.

Oh right, the Puget Systems testing article they reference in that video? The one where the R7 barely trade blows with a GTX 1660 Ti (much less an RTX 2060, both cards which can be had for hundreds less than a Radeon VII) in Adobe Premeire, in the vast majority of use-cases? 

Quote

We have 8GiB in the Consumer Market since 390 Series - 4 years ago!!

And now, 4 Years later, you still claim that 8GiB VRAM are enough?!
 

RLY?!

Yes?  Would it "nice to have" 16/32 GB of VRAM? Sure it'd be, if you can afford it.

 

 Do the vast majority of users need it (much less pay for it)?

 

[ryao]

5 hours ago, CarlBar said:

 

Thats not how skewing in favour of somthing works.

 

I'd also be amazed if it's anywhere near that high. That said the ratio of AMD to NVIDIA is certainly more biased towards AMD than NVIDIA. Thats inevitable though. That average consumer tends to rely strongly on a combination of brand familiarity and third party OEM supplied parts, both of which favour NVIDIA heavilly. Actual whats the best choice for them analytics tend to be very low on the factors that influence NVIDIA vs AMD GPU sales.

 

Conversely people here are quite understandably more tech savvy, we do a lot more focusing on the best choice analytics, and price per performance tends to be a big deal for us. For people buying mid or low end graphics cards AMD has had a stranglehold for a while, (the 1660Ti and 1660 have broken that for now, but how long that will remain true is subject to change without notice), so they tend to get a lot of attention from the majority. (Most of us would like high end GPU's in our systems, but only a few of us can actually afford them).

 

Incidentally i'm typing this from a system powered by a 2080Ti in the graphics department.

 

 

What are you smoking and can i have some of it please. The R7 beats the 2070 be 10-15%. The 2080ti beats the 2070 by far more...

 

It is 74.75% vs 14.90% in steam’s survey:

 

https://store.steampowered.com/hwsurvey/Steam-Hardware-Software-Survey-Welcome-to-Steam

 

That makes it 5:1 nvidia to AMD. If the ratio is more favorable to AMD here than that, then these forums are skewed toward them relative to everyone else.

[BuckGup]

Right but can they actually do what they are claiming, no.

[ryao]

2 minutes ago, BuckGup said:

Right but can they actually do what they are claiming, no.

Who? AMD or Nvidia? It seems that the discussion claimed that AMD claimed that they could do real-time raytracing without special hardware.

[BuckGup]

58 minutes ago, ryao said:

Who? AMD or Nvidia? It seems that the discussion claimed that AMD claimed that they could do real-time raytracing without special hardware.

Nvidia and the title of the post. Also ray tracing has been shown off years before Nvidia announced it on ASICs made for mobile use

[ryao]

2 hours ago, BuckGup said:

Nvidia and the title of the post. Also ray tracing has been shown off years before Nvidia announced it on ASICs made for mobile use

Was it done in realtime in production with anything nontrivial?

[BuckGup]

9 minutes ago, ryao said:

Was it done in realtime in production with anything nontrivial?

Yup it was. Apple bought the company a few years after they released it and they dissolved them

[mr moose]

6 minutes ago, ryao said:

Was it done in realtime in production with anything nontrivial?

It was done in real time on mobile processors as a concept demonstration, the difference was that was not done on a dynamic scene.    What Nvidia are pushing is not only real time processing of dynamic scenes but the ability to bring such processing to power to desktop computers where prior rendering required render farms.

 

 

[Stefan Payne]

13 minutes ago, ryao said:

Was it done in realtime in production with anything nontrivial?

Yes, there are dedicated Raytracing Cards...

 

For example:

https://www.extremetech.com/extreme/161074-the-future-of-ray-tracing-reviewed-caustics-r2500-accelerator-finally-moves-us-towards-real-time-ray-tracing

https://www.pcper.com/reviews/Graphics-Cards/Caustic-Graphics-Ray-Tracing-Acceleration-Technology-Review/CausticOne-and-Ca

 

But hey, nVidia were the first!111

Even if the first Raytracing Accelerator is 33 years old. But hey, nVidia didn't steal anything, they wouln't do that, right?

 

Or the good old SGI Workstations...

 

Or this:

https://www.slideserve.com/lulu/rpu-a-programmable-ray-processing-unit-for-realtime-ray-tracing

 

Yeah, Raytracing is totally new and an Invention from nVidia...

[mr moose]

7 minutes ago, Stefan Payne said:

Yes, there are dedicated Raytracing Cards...

 

For example:

https://www.extremetech.com/extreme/161074-the-future-of-ray-tracing-reviewed-caustics-r2500-accelerator-finally-moves-us-towards-real-time-ray-tracing

https://www.pcper.com/reviews/Graphics-Cards/Caustic-Graphics-Ray-Tracing-Acceleration-Technology-Review/CausticOne-and-Ca

 

But hey, nVidia were the first!111

Even if the first Raytracing Accelerator is 33 years old. But hey, nVidia didn't steal anything, they wouln't do that, right?

 

Or the good old SGI Workstations...

 

Or this:

https://www.slideserve.com/lulu/rpu-a-programmable-ray-processing-unit-for-realtime-ray-tracing

 

Yeah, Raytracing is totally new and an Invention from nVidia...

And yet here we are some many years later and the closest thing anyone has to real time RT in their desktop gaming is from NVIDIA.  You can drop the childish rhetoric about NVIDIA claiming to have invented it, I have yet to see anyone on these forums make such a claim.

[ryao]

On 4/3/2019 at 9:56 PM, Stefan Payne said:

Yes, there are dedicated Raytracing Cards...

 

For example:

https://www.extremetech.com/extreme/161074-the-future-of-ray-tracing-reviewed-caustics-r2500-accelerator-finally-moves-us-towards-real-time-ray-tracing

https://www.pcper.com/reviews/Graphics-Cards/Caustic-Graphics-Ray-Tracing-Acceleration-Technology-Review/CausticOne-and-Ca

 

But hey, nVidia were the first!111

Even if the first Raytracing Accelerator is 33 years old. But hey, nVidia didn't steal anything, they wouln't do that, right?

 

Or the good old SGI Workstations...

 

Or this:

https://www.slideserve.com/lulu/rpu-a-programmable-ray-processing-unit-for-realtime-ray-tracing

 

Yeah, Raytracing is totally new and an Invention from nVidia...

Comparing raytracing on SGI workstations to what Nvidia has done is laughable. That occurred on the CPU and was only done as part of render farms for 3D movies.Doing it as part of a video game for regular people just did not happen because regular people did not have SGI hardware and the hardware could not do real-time raytracing at any decent resolution.

 

The ray tracing card that you linked could not touch Nvidia’s Ray tracing capability, which is something like 1000 times higher. It was neither useful nor marketed for real-time ray tracing. Furthermore, it was something no end users would buy and no game developers would touch. It was even less mass market than this:

 

https://www.red.com/red-rocket-x

 

Coincidentally, the RTX cards make the Red Rocket X obsolete too:

 

https://fstoppers.com/news/nvidia-and-red-unveil-gpu-solution-8k-real-time-editing-320356

 

I do not understand the desire to minimize Nvidia’s achievement. Doing 10 billion rays per second to render a 4K scene in real-time is amazing. Nobody had done that with commodity hardware before Nvidia did it.

[mr moose]

20 minutes ago, ryao said:

Comparing raytracing on SGI workstations to what Nvidia has done is laughable. That occurred on the CPU and was only done as part of render farms for 3D movies.Doing it as part of a video game for regular people just did not happen because regular people did not have SGI hardware and the hardware could not do real-time raytracing at any decent resolution.

 

The ray tracing card that you linked could not touch Nvidia’s Ray tracing capability, which is something like 1000 times higher. It therefore was not useful for real-time ray tracing and consequently, was not marketed for it. Furthermore, it was something no end users would buy and no game developers would touch. It was even less mass market than this:

 

https://www.red.com/red-rocket-x

 

Coincidentally, the RTX cards make the Red Rocket X obsolete too:

 

https://fstoppers.com/news/nvidia-and-red-unveil-gpu-solution-8k-real-time-editing-320356

 

I do not understand the desire to minimize Nvidia’s achievement. Doing 10 billion rays per second to render a 4K scene in real-time is amazing. Nobody had done that with commodity hardware before Nvidia did it.

The card he linked to could do 4M rays per second, the RTX can do in the billions per second.   It doesn't matter which way we split the hairs, trying to dismiss nvidias contribution is just childish ignorance at best.  

 

Most of us don't even care who made the first, the best or the biggest GPU, or even who makes the next best, so long as it is better and doesn't require I sell a kidney I'll buy it.

[RejZoR]

No one denied RTX isn't impressive as a feature. What we're criticizing is the fact it's a commercial flop. They promised all these games and what we got? 2 bloody games after months of waiting and 1 extra (Metro Exodus) half a year after RTX launch. And that's it. That DLSS or whatever the fuck it's even named is an useless tech. It gives better performance. And makes everything blurry. If it was reverse DSR and worked with any game I'd even be fine with it, but this nonse with "deep learning" bollocks and requiring games to have special support for it just makes it the most useless thing in the world.

 

NVIDIA should really thought this through better and make a rock solid deal with developers to deliver at least 10 titles with full RTX support on day 1 of card release or within 1 month from RTX cards release. But 3 games after months and months of waiting is just pathetic and useless. Bloody Tomb Raider still isn't released with RTX. People finished and forgotten about it already, no one's gonna even try it with RTX if it ever comes out with support they promised.

 

It's gonna be cool when ray tracing becomes a norm, but it won't be that soon no matter how hard NVIDIA is pushing it (which is not that hard really).

[mr moose]

10 hours ago, RejZoR said:

No one denied RTX isn't impressive as a feature. What we're criticizing is the fact it's a commercial flop. They promised all these games and what we got? 2 bloody games after months of waiting and 1 extra (Metro Exodus) half a year after RTX launch. And that's it. That DLSS or whatever the fuck it's even named is an useless tech. It gives better performance. And makes everything blurry. If it was reverse DSR and worked with any game I'd even be fine with it, but this nonse with "deep learning" bollocks and requiring games to have special support for it just makes it the most useless thing in the world.

 

NVIDIA should really thought this through better and make a rock solid deal with developers to deliver at least 10 titles with full RTX support on day 1 of card release or within 1 month from RTX cards release. But 3 games after months and months of waiting is just pathetic and useless. Bloody Tomb Raider still isn't released with RTX. People finished and forgotten about it already, no one's gonna even try it with RTX if it ever comes out with support they promised.

 

It's gonna be cool when ray tracing becomes a norm, but it won't be that soon no matter how hard NVIDIA is pushing it (which is not that hard really).

Sounds like Nvidia should hire you to run their business, because clearly they don't know what you do and haven't really been successful because of it.

[RejZoR]

I don't know, why do you think Intel is hiring so many people that are just very experienced users/testers and not people with PhD's in economics and shit? Coz they are able to view things from a different perspective that may not align with usual corporate mumbo jumbo. That's why.

[pas008]

11 hours ago, RejZoR said:

No one denied RTX isn't impressive as a feature. What we're criticizing is the fact it's a commercial flop. They promised all these games and what we got? 2 bloody games after months of waiting and 1 extra (Metro Exodus) half a year after RTX launch. And that's it. That DLSS or whatever the fuck it's even named is an useless tech. It gives better performance. And makes everything blurry. If it was reverse DSR and worked with any game I'd even be fine with it, but this nonse with "deep learning" bollocks and requiring games to have special support for it just makes it the most useless thing in the world.

 

NVIDIA should really thought this through better and make a rock solid deal with developers to deliver at least 10 titles with full RTX support on day 1 of card release or within 1 month from RTX cards release. But 3 games after months and months of waiting is just pathetic and useless. Bloody Tomb Raider still isn't released with RTX. People finished and forgotten about it already, no one's gonna even try it with RTX if it ever comes out with support they promised.

 

It's gonna be cool when ray tracing becomes a norm, but it won't be that soon no matter how hard NVIDIA is pushing it (which is not that hard really).

What kind of logic is this?

Dx9 10 11 and 12 games didn't just pop up right away they needed hardware support first plus you don't want to alienate a huge audience by focusing on using newer tech as a developer you dabble and offer a little or some support until it becomes norm or mainstream

Dx12 been out for how long? Vulkan?

How many games are using them?

Ok then shit takes time sadly

 

But you may have your commerical flop opinion but then again you just showed how your logic works