Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

2016 nMP

Status
Not open for further replies.
G

goMac

macrumors 604
Apr 15, 2004
7,663
1,694
koyoot said:
If you ask about CUDA, I genuinely suggest doing some research and educating yourself about what is happening in CUDA world. Let me put some things this way: Otoy is doing CUDA cross platform, cross vendor , to serve some purpose. There is huge background story happening, and happened when AMD launched one thing: GPUOpen. Nvidia is afraid of the impact it can have on their sales, and mindshare. What is more, there is huge hint in that story about who will dominate HPC in upcoming months in terms of raw performance(Pascal, or Polaris/Vega).
Click to expand...

Some big industry names are already preparing for this. Carmack the other day outright said AMD GPU hardware is faster, but Nvidia's software is what preserves their dominance.

Take that advantage away and even the current AMD GPUs are starting to show their real performance advantages. Chickens are coming home to roost for the Nvidia camp. Going to be regrets for whoever bought into the 700 series and higher.
 
koyoot

koyoot

macrumors 603
Jun 5, 2012
5,939
1,853
goMac said:
Some big industry names are already preparing for this. Carmack the other day outright said AMD GPU hardware is faster, but Nvidia's software is what preserves their dominance.

Take that advantage away and even the current AMD GPUs are starting to show their real performance advantages. Chickens are coming home to roost for the Nvidia camp. Going to be regrets for whoever bought into the 700 series and higher.
Click to expand...
The absurdity of situation with CUDA is that people are claiming that 4 TFLOPs FP32 GPU from Nvidia is more powerful than 4 TFLOPs FP32 GPU from AMD. Tide has changed for AMD dramatically in recent months. I am not sure however how it will affect the mindshare.
 
lowendlinux

lowendlinux

macrumors 603
Sep 24, 2014
5,460
6,788
Germany
koyoot said:
The absurdity of situation with CUDA is that people are claiming that 4 TFLOPs FP32 GPU from Nvidia is more powerful than 4 TFLOPs FP32 GPU from AMD. Tide has changed for AMD dramatically in recent months. I am not sure however how it will affect the mindshare.
Click to expand...

You have to remember AMD lost mindshare not because they produced bad GPU's so much is their SW support and drivers have been abysmal for the last half decade.

As soon as I can get a decent optimized driver out of the box I'll pay the extra electicity required to be on red team but that has to happen first and it needs to be cross platform. The new AMDGPU FOSS drivers seem to working well still not on par with Catalyst in most benches but it's still beta, the newer Windows drivers don't seem to be doing much.

For now CUDA is here and it works and people need to do stuff and make money now.
 
fuchsdh

fuchsdh

macrumors 68020
Jun 19, 2014
2,028
1,831
lowendlinux said:
You have to remember AMD lost mindshare not because they produced bad GPU's so much is their SW support and drivers have been abysmal for the last half decade.

As soon as I can get a decent optimized driver out of the box I'll pay the extra electicity required to be on red team but that has to happen first and it needs to be cross platform. The new AMDGPU FOSS drivers seem to working well still not on par with Catalyst in most benches but it's still beta, the newer Windows drivers don't seem to be doing much.

For now CUDA is here and it works and people need to do stuff and make money now.
Click to expand...

Eh I think the rebadging of their 7XXX series and the fact that the Fury were very situational GPU buys hurt them too. But the biggest issue has been the same that Microsoft faces trying to unseat Apple's products—it's not good enough to produce a solid alternative, AMD has to make something demonstrably much better than Nvidia to really make up ground quickly.
 
koyoot

koyoot

macrumors 603
Jun 5, 2012
5,939
1,853
fuschd, you are right. What worries me here is that people believe that Polaris 11 will be better than Fury X, which is not true. Expectations are very high for Polaris, AMD is also very confident about their upcoming lineup of products. The problem is that true replacement for Fury X is Vega 10 which is for Q3, due to HBM2 availability.

Disappointment can be very huge. Not because Polaris will be bad. But not as good as people hyped themselves for it.
 
fuchsdh

fuchsdh

macrumors 68020
Jun 19, 2014
2,028
1,831
koyoot said:
fuschd, you are right. What worries me here is that people believe that Polaris 11 will be better than Fury X, which is not true. Expectations are very high for Polaris, AMD is also very confident about their upcoming lineup of products. The problem is that true replacement for Fury X is Vega 10 which is for Q3, due to HBM2 availability.

Disappointment can be very huge. Not because Polaris will be bad. But not as good as people hyped themselves for it.
Click to expand...

Yeah, it's certainly possible. At least this go around AMD doesn't seem to be smoking their own press releases which might tamp down some of that at least.

Still, in a world of stagnant processors at least GPUs are still going through serious changes to keep nerds interested.
 
koyoot

koyoot

macrumors 603
Jun 5, 2012
5,939
1,853
AidenShaw said:
Yes, again the big "if" that starts every story about AMD processors.
Click to expand...
We do not know anything about it, so thats why I put "If" there. But on everyones side(optimists/skeptics) I would not underestimate AMDs capabilities of delivering. Zen comes from the guy who brought Macroscalar architecture to iOS with A6 CPU. At the very least the CPUs will be competitive.
 
M

Mago

macrumors 68030
Aug 16, 2011
2,789
912
Beyond the Thunderdome
koyoot said:
We do not know anything about it, so thats why I put "If" there. But on everyones side(optimists/skeptics) I would not underestimate AMDs capabilities of delivering. Zen comes from the guy who brought Macroscalar architecture to iOS with A6 CPU. At the very least the CPUs will be competitive.
Click to expand...
AMD has some hits (amd64, the first APU) But the problem is they somewhat don't keep hungry and maybe Zen cpu will be as good (i think it will beat any Intel cpu (Xeon too) at launch and for a while at least on raw performance), but how long AMD would keep the hype enough to be again in the enthusiasts minds?

In mid term is known Intel will focus on TCO than performance as market strategy, so say goodbye to the Intel performance era, now arrives the Intel economy era, does AMD has any chance to survive in a world hungry for $ or the enthusiasts will give them enough market share?
 
koyoot

koyoot

macrumors 603
Jun 5, 2012
5,939
1,853
Mago said:
AMD has some hits (amd64, the first APU) But the problem is they somewhat don't keep hungry and maybe Zen cpu will be as good (i think it will beat any Intel cpu (Xeon too) at launch and for a while at least on raw performance), but how long AMD would keep the hype enough to be again in the enthusiasts minds?

In mid term is known Intel will focus on TCO than performance as market strategy, so say goodbye to the Intel performance era, now arrives the Intel economy era, does AMD has any chance to survive in a world hungry for $ or the enthusiasts will give them enough market share?
Click to expand...
If you will look at WHOLE of the market, not only enthusiasts, but whole market and consider changes that are happening: Low-Level APIs, VR, HPC, GPU farms/GPU Clouds, Virtualization on GPU, servers, APUs, consumer electronics you will see that AMD has not only enthusiast segment to feed with, but a lot more.

Think about APU with 8 cores, powerful Polaris GPU and HBM2 memory. Or even 4 core APU with Polaris GPU and HBM2. Each of the cores with Broadwell level of performance. Which segment it would fit?

Not only enthusiast market will benefit, because for the first time in years, we can see price drops, because of good competition. IF Zen will deliver - that is the only requirement here.
 
M

Mago

macrumors 68030
Aug 16, 2011
2,789
912
Beyond the Thunderdome
koyoot said:
Think about APU with 8 cores, powerful Polaris GPU and HBM2 memory. Or even 4 core APU with Polaris GPU and HBM2. Each of the cores with Broadwell level of performance. Which segment it would fit?
Click to expand...

It's tricky market, but it probably is aimed at HPC and VR optimized PC/Consoles, Intel also has 10 core i7 cpu on its road map for 2016, I've read there will be also Zen cpu with 20+ cores.

AMD also is researching on server/hpc with ARM and just released its first ARM cortex a53 Opteron and announced an much improved version for 2017 on cortex a72 or maybe a custom cortex as Qualcomm with as many as 64+ cores aimed at compute severs too (Opteron A1100 it's aimed at low end Web servers and alike).
 
Last edited:
koyoot

koyoot

macrumors 603
Jun 5, 2012
5,939
1,853
Not exactly. On the bottom there is a chart which shows conversion from 28nm GLoFo to 14nm LPP, and conversion of the area.

14 nm LPP from GloFo is 0.29 of 28 nm GloFo, which is denser a bit than 28 nm TSMC. By the look of that chart Fury X on this process would be under 200mm2.

http://imgur.com/68y9WUg
 
ActionableMango

ActionableMango

macrumors G3
Sep 21, 2010
9,613
6,909
Competition is good for us end-users. Even if you are an Nvidia supporter who hates every single thing about AMD, you should hope for a strong AMD product. A strong AMD product will inspire Nvidia to stay strong too.

They drive each other forward, and we all benefit.
 
  • Like
Reactions: Flint Ironstag and Bubba Satori
Bubba Satori

Bubba Satori

Suspended
Feb 15, 2008
4,726
3,756
B'ham
ActionableMango said:
Competition is good for us end-users. Even if you are an Nvidia supporter who hates every single thing about AMD, you should hope for a strong AMD product. A strong AMD product will inspire Nvidia to stay strong too.

They drive each other forward, and we all benefit.
Click to expand...

Agreed.
I've owned 2 AMD cpus and 2 ATI gpus, 2 Intel cpus and 2 Nvidia gpus.
Not a partisan, just want whats best at the time.
 
koyoot

koyoot

macrumors 603
Jun 5, 2012
5,939
1,853
First presentation of VR with Ubisoft Eagle Flight, and I completely fell in absolute love with VR.

Absolutely tremendous.

Polaris 10, small, laptop GPU. DirectX12 demo in Hotman locked at 60 FPS in 1440p.

http://pic.yupoo.com/ztwss/FomEtouo/7UI3V.jpg
I do not know what to say. It is less than 75W GPU.
 
Last edited:
koyoot

koyoot

macrumors 603
Jun 5, 2012
5,939
1,853
flat five said:
is that legal? (i mean, i assume it is or otoy probably wouldn't be boasting about their accomplishment)

..but, what gives? why are they able to do that?
Click to expand...
There is a lot of background story, but it looks like I can tell more right now. Nvidia is threatened by OpenGPU, thats why they told Otoy, gave them money and engineers to bring CUDA to other platforms. It is not open as you might think, still.

What is more, it looks like Nvidia also considers that they will not be able to compete with Polaris in HPC market, that is what makes them more likely to "open" CUDA. They may close it again with Volta.
koyoot said:
First presentation of VR with Ubisoft Eagle Flight, and I completely fell in absolute love with VR.

Absolutely tremendous.

Polaris 10, small, laptop GPU. DirectX12 demo in Hotman locked at 60 FPS in 1440p.

http://pic.yupoo.com/ztwss/FomEtouo/7UI3V.jpg
I do not know what to say. It is less than 75W GPU.
Click to expand...
My bad. Was under impression that Polaris 10 is the smaller one GPU, which is not true. This was P10 and this is bigger one. The GPU shown at GDC was Polaris 11.
 
Last edited:
  • Like
Reactions: flat five
flat five

flat five

macrumors 603
Feb 6, 2007
5,580
2,657
newyorkcity
koyoot said:
There is a lot of background story, but it looks like I can tell more right now. Nvidia is threatened by OpenGPU, thats why they told Otoy, gave them money and engineers to bring CUDA to other platforms. It is not open as you might think, still.
Click to expand...
seems likely.

also, i don't really imagine nvidia fully opening cuda.. they'd probably be dumb to since they can still make money off it.. currently, i think cuda may serve two roles from a financial point of view.. selling nvidia gpus and licensing the software.. if they quit using cuda as a sales pitch for the hardware, they can still probably profit off the software.
 
Status
Not open for further replies.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom