Ah. I seriously doubt that the measured power consumption of the GTX 1070 really stays below 150 watts running at top clock speed. Oddly enough, I see the reviewers just accepting nVidia's claims without reporting actual measurements, in contrast to the current tempest in a teapot over the RX 480.
Not sure why you say that.
The numbers. RX 480 is 5.7b transistors at 14nm, GTX 1070 is 7.2b at 16nm. Both engineering teams are top in their field, I doubt anybody seriously dropped the ball on architecture. And so far there are no reports of glaring deficiencies of the 14nm process. The nVidia part is clocked 35% higher. So: more transistors, higher clock speed, larger process, but the same or less power consumption? Doesn't add up, not even close. This leaves measurement error or as the likely explanation. BTW, I'm a big fan of both engineering organizations, but I like the AMD product more because I benefit from the open source driver. I hope this healthy competition runs far into the future and continues the trend to more powerful, cooler running and inexpensive graphics.
I don't remember a GTX 1070 rail test specifically (but at least HardOCP had a full-system test, which was as expected), but e.g. Tom's GTX 1080 power test was as rigorous as the RX 480, testing all rails separately etc, and it was staying under its specified 180 Watt on average. If there was a weird measurement, people would be all over nVidia as well...
I guess there's going to be a flurry of retesting on all fronts, should be fun.
For the reading impaired, he said "point and shoot photography", not "point and shoot camera". Point and shoot photography is a thing whether you do it with a cell phone or a more capable camera.
The question was about the already released GTX 1070, a larger, faster chip that despite that draws less power and clocks 50% higher than the RX 480.
Ah. I seriously doubt that the measured power consumption of the GTX 1070 really stays below 150 watts running at top clock speed. Oddly enough, I see the reviewers just accepting nVidia's claims without reporting actual measurements, in contrast to the current tempest in a teapot over the RX 480.
If they are on an equivalent process, it would mean an engineering issue. Or is GloFo's 14nm node not as good as the TSMC 16nm?
It's GloFo+Samsung, by the way. Speculation about the relative merits of the processes is just speculation until we see a lot better, trustworthy real life measurements. What we know for sure is that the two products are nowhere near the same price.
He said "point and shoot", which works far better on a DSLR than any cell phone, there are just so many reasons. Try it, your definition of "good enough" will immediately change.
Anybody has any idea whether GloFo's 14nm FinFET has some sort of disadvantage vs TSMC's 16nm? Otherwise it looks quite bad for the AMD engineers when they have to use more power than a much faster GTX 1070...
AMD's product is released and independently tested while nVidia's is only announced, so take those claims with a grain of salt. I believe the technical term for the situation is "FUD". Even if you accept nVidia's claims at face value, the 480 still comes out as great value and is shipping now. I guess the market agrees because the initial production run seems to be mostly sold out.
That's basically still how it is. The Linux driver performance is substantially worse than the Windows driver.
True mainly when the Linux port is actually a Direct3D wrapper. For a decent native port there are already cases where Linux soundly beats Windows, for example, here where Dota 2 on Linux with OpenGL 3.3 beats Windows DX11 by a wide margin. Not just that, but if you follow the Dota 2 scene, you know that Windows network lag and game crashes are painful and regular. About time to enjoy some of that buttery smooth Linux network experience and rock solid stability, don't you think? Especially when serious money is involved.
To answer your question though, I've heard through the vine that Nvidia drivers are easier to deal with on Linux, though I have not experienced it firsthand.
The AMD proprietary drivers (fglrx/catalyst) are roughly equivalent to nVidia in terms of stability and upgrade hassle, while the AMD open source driver (Radeon/AMDGPU) is the least hassle experience. Maybe Valve knows this for sure, but my impression is that most Linux AMD users stick with the default open source driver these days because the performance gap has closed up to the point that convenience outweighs it. Personally, I have had zero issues with the AMD open source drivers for several years on a variety of systems and haven't felt any motivation to switch in the closed source drivers, though some people might care if they need the latest OpenGL version right now, or more throughput for some corner case. Basically, reasons for using the closed AMD driver are fading away fast and likely will be gone completely in the current development cycle where we expect Vulkan and OpenGL 4.5 to land in the open source stack.
In "over two years"? How is 2 years a meaningful period of time for such a headline?
Because it covers the period when it was widely prophesied that the sky would fall on the smartphone boom. Turns out, the sky is only falling on Apple.
Smartphone cameras have been good enough for point and shoot photography for quite a while now.
Good enough to get a great shot of the backside of a baby running away or the place the kitten was one second ago, or great motion blur effects, sure. Good enough for somebody who has never used a DSLR and mainly collects headshots or tourist scenery. Nothing wrong with that, but please don't dismiss actual expertise in photography as snobbery. The optics just aren't there in a cell phone, sensors are far inferior and there is no such thing as an iris, among other issues. And DSLR technology does not stand still, far from it.
Oh yeah, Google is all full of smart people, don't you know. Saved the $10K of intern salary it would have taken to recode the algorithm in C++ in order to burn $100K of datacenter power.
Given MS can tune specifically for Windows while Vulkan needs to remain somewhat generic then DX will always have a performance advantage even when working with basically the same concepts
Sorry, you lost me here. The entire concept of Vulkan (and DX12, both being derived from Mantle) is to run as close to the hardware as possible, so where does your imagined performance advantage come from?
from what has been seen of benchmarks vulkan gives you great performance cross platform but DX12+ will give better performance on windows
Given that Vulkan and DX12 are both derived from AMD's Mantle, you are blowing smoke. It's a safe bet which one is better designed, also which one will be more widely deployed (hint: handsets outnumber desktops by a wide and increasing margin).
I don't pretend that it's for people who want the highest performance from the latest games or anything approaching that.
But if you look at the Dota2 Vulkan and the Unigine results, it is approaching that. Vulkan on Linux within 1.5% of windows on AMD and 9% on nVidia. Not that OpenGL is any slouch - Unigine OpenGL are within 9% of windows DirectX, showing the huge difference between native support and translation layers. But developers already voted with their allegiance: few had the luxury of supporting two dissimilar rendering platforms and almost all picked the one with the biggest market and modest performance advantage.
That equation changes now. See this this or this if you have any doubts. The new equation is, if you want one platform that delivers top performance across Android and Windows you go with Vulkan, end of story. Desktop Linux is the lucky beneficiary of that.
Slashdot Mirror
Ah. I seriously doubt that the measured power consumption of the GTX 1070 really stays below 150 watts running at top clock speed. Oddly enough, I see the reviewers just accepting nVidia's claims without reporting actual measurements, in contrast to the current tempest in a teapot over the RX 480.
Not sure why you say that.
The numbers. RX 480 is 5.7b transistors at 14nm, GTX 1070 is 7.2b at 16nm. Both engineering teams are top in their field, I doubt anybody seriously dropped the ball on architecture. And so far there are no reports of glaring deficiencies of the 14nm process. The nVidia part is clocked 35% higher. So: more transistors, higher clock speed, larger process, but the same or less power consumption? Doesn't add up, not even close. This leaves measurement error or as the likely explanation. BTW, I'm a big fan of both engineering organizations, but I like the AMD product more because I benefit from the open source driver. I hope this healthy competition runs far into the future and continues the trend to more powerful, cooler running and inexpensive graphics.
I don't remember a GTX 1070 rail test specifically (but at least HardOCP had a full-system test, which was as expected), but e.g. Tom's GTX 1080 power test was as rigorous as the RX 480, testing all rails separately etc, and it was staying under its specified 180 Watt on average. If there was a weird measurement, people would be all over nVidia as well...
I guess there's going to be a flurry of retesting on all fronts, should be fun.
For the reading impaired, he said "point and shoot photography", not "point and shoot camera". Point and shoot photography is a thing whether you do it with a cell phone or a more capable camera.
The question was about the already released GTX 1070, a larger, faster chip that despite that draws less power and clocks 50% higher than the RX 480.
Ah. I seriously doubt that the measured power consumption of the GTX 1070 really stays below 150 watts running at top clock speed. Oddly enough, I see the reviewers just accepting nVidia's claims without reporting actual measurements, in contrast to the current tempest in a teapot over the RX 480.
If they are on an equivalent process, it would mean an engineering issue. Or is GloFo's 14nm node not as good as the TSMC 16nm?
It's GloFo+Samsung, by the way. Speculation about the relative merits of the processes is just speculation until we see a lot better, trustworthy real life measurements. What we know for sure is that the two products are nowhere near the same price.
He said "point and shoot", which works far better on a DSLR than any cell phone, there are just so many reasons. Try it, your definition of "good enough" will immediately change.
Anybody has any idea whether GloFo's 14nm FinFET has some sort of disadvantage vs TSMC's 16nm? Otherwise it looks quite bad for the AMD engineers when they have to use more power than a much faster GTX 1070...
AMD's product is released and independently tested while nVidia's is only announced, so take those claims with a grain of salt. I believe the technical term for the situation is "FUD". Even if you accept nVidia's claims at face value, the 480 still comes out as great value and is shipping now. I guess the market agrees because the initial production run seems to be mostly sold out.
Now I'm left to wonder if the infamous Radeon mouse cursor corruption bug also exists on Linux.
Fixed last year. Catalyst bug, I have been running the open source driver for the last few years so I never saw it.
That's basically still how it is. The Linux driver performance is substantially worse than the Windows driver.
True mainly when the Linux port is actually a Direct3D wrapper. For a decent native port there are already cases where Linux soundly beats Windows, for example, here where Dota 2 on Linux with OpenGL 3.3 beats Windows DX11 by a wide margin. Not just that, but if you follow the Dota 2 scene, you know that Windows network lag and game crashes are painful and regular. About time to enjoy some of that buttery smooth Linux network experience and rock solid stability, don't you think? Especially when serious money is involved.
To answer your question though, I've heard through the vine that Nvidia drivers are easier to deal with on Linux, though I have not experienced it firsthand.
The AMD proprietary drivers (fglrx/catalyst) are roughly equivalent to nVidia in terms of stability and upgrade hassle, while the AMD open source driver (Radeon/AMDGPU) is the least hassle experience. Maybe Valve knows this for sure, but my impression is that most Linux AMD users stick with the default open source driver these days because the performance gap has closed up to the point that convenience outweighs it. Personally, I have had zero issues with the AMD open source drivers for several years on a variety of systems and haven't felt any motivation to switch in the closed source drivers, though some people might care if they need the latest OpenGL version right now, or more throughput for some corner case. Basically, reasons for using the closed AMD driver are fading away fast and likely will be gone completely in the current development cycle where we expect Vulkan and OpenGL 4.5 to land in the open source stack.
Is this going to be the new gold standard for the mid-price range?
Dunno. 15% more throughput (according to nVidia) for 25% more money? Looks like AMD still wins the dollars per throughput equation.
In "over two years"? How is 2 years a meaningful period of time for such a headline?
Because it covers the period when it was widely prophesied that the sky would fall on the smartphone boom. Turns out, the sky is only falling on Apple.
Maybe they'll get their head out of their ass next cycle.
Right. For me, no sd slot equals no deal.
Smartphone cameras have been good enough for point and shoot photography for quite a while now.
Good enough to get a great shot of the backside of a baby running away or the place the kitten was one second ago, or great motion blur effects, sure. Good enough for somebody who has never used a DSLR and mainly collects headshots or tourist scenery. Nothing wrong with that, but please don't dismiss actual expertise in photography as snobbery. The optics just aren't there in a cell phone, sensors are far inferior and there is no such thing as an iris, among other issues. And DSLR technology does not stand still, far from it.
Voting for Hillary because she's "not guilty" is like hiring Casey Anthony to babysit your kid.
More like "not guilty" and "not Trump".
I guess I'm stuck using Windows 10 (32-bit) on my vintage 2006 MacBook (Intel Duo Core 32-bit processor).
There are many Linux distributions, some will certainly continue building 32 bit distributions for a long time to come.
Downmodded twice by evil Microsofties. See what I mean?
I don't think Microsoft is even close to the #1 EVIL company now...
Windows 10 malware/spyware. I rest my case.
I'm absolutely NOT saying that the google is the only EVIL corporation out there. I'm just unable to name a more EVIL one
Microsoft. But the gap is closing.
As long as ATI drivers continue to be sketchy, my first choice will be Nvidia.
Well, that's the question, isn't it? ...
I don't see reviewers complaining about driver issues or rendering artifacts, I do see them commenting on the high performance for low price.
They're killing the "Windows is easier than *NIX" argument.
It has always been just an argument, never the truth. What's dieing is the spin.
Perhaps they didn't anticipate the backlash...
Right now a Dota 2 tournament is on hold because a Windows 10 update started in the middle of it.
This is why they're called a "Just in time compiler"
Something of a mischaracterization. It is really a "Better late than never" compiler.
Seriously? This was done in Java Script?
Oh yeah, Google is all full of smart people, don't you know. Saved the $10K of intern salary it would have taken to recode the algorithm in C++ in order to burn $100K of datacenter power.
Given MS can tune specifically for Windows while Vulkan needs to remain somewhat generic then DX will always have a performance advantage even when working with basically the same concepts
Sorry, you lost me here. The entire concept of Vulkan (and DX12, both being derived from Mantle) is to run as close to the hardware as possible, so where does your imagined performance advantage come from?
from what has been seen of benchmarks vulkan gives you great performance cross platform but DX12+ will give better performance on windows
Given that Vulkan and DX12 are both derived from AMD's Mantle, you are blowing smoke. It's a safe bet which one is better designed, also which one will be more widely deployed (hint: handsets outnumber desktops by a wide and increasing margin).
I don't pretend that it's for people who want the highest performance from the latest games or anything approaching that.
But if you look at the Dota2 Vulkan and the Unigine results, it is approaching that. Vulkan on Linux within 1.5% of windows on AMD and 9% on nVidia. Not that OpenGL is any slouch - Unigine OpenGL are within 9% of windows DirectX, showing the huge difference between native support and translation layers. But developers already voted with their allegiance: few had the luxury of supporting two dissimilar rendering platforms and almost all picked the one with the biggest market and modest performance advantage.
That equation changes now. See this this or this if you have any doubts. The new equation is, if you want one platform that delivers top performance across Android and Windows you go with Vulkan, end of story. Desktop Linux is the lucky beneficiary of that.