Margins on a $300 1080 would be very close to zero.
Margins on a $300 1080 would be very close to zero.
Margins on a $300 1080 would be very close to zero.
Any evidence?Margins on a $300 1080 would be very close to zero.
Any evidence?
Because nvidia released the very similar die size 560 Ti (334mm2) at $249 at on the then brand new 40nm process in 2011 and STILL had great margins that year. So you're saying an entire year later nvidia wouldn't make any money on a $300 330mm2. Sorry but you and I both know you're dead wrong.
Same memory interface width..
Just like many have pointed out there is a GOOD REASON nvidia has taken record profits from consumers.. They have jacked up prices of all their chips up an entire segment.
Margins on a $300 1080 would be very close to zero.
60% margin for the entire company, which sells cards that have next to nothing margin wise at the bottom, and they do so by the millions.If at $549 NVIDIA enjoys a gross profit margin of 60% (probably on the high side honestly...), then that implies a COGS of $220.
Margins on a $300 product with that same exact cost structure would be $80/$300 = ~27%, which would be awful.
Now, I know this might seem decent (it's not...), but also remember that the cost of the entire finished AIB (PCB, VRAM, cooling shroud, box, etc.) is going to be relatively fixed -- the only part that would come down in price in this case would presumably be the GPU. So, NVIDIA's margins on the silicon die itself would probably be much lower than the 27% that I just mentioned.
The whole point of the demo was to show off Threadripper and its I/O capabilites, not Vega. As such you would want to add enough GPUs to the point where there is no GPU bottleneck, and then show that Threadripper is still capable of keeping up. It's still a crappy demo mind you, as it should instead have been done as a comparison with a slower CPU, like they did with the DOTA2 streaming demo, which didn't feature FPS numbers either, but still did a perfect job of getting the message across.
Either way though this is all irrelevant. I just found this 60FPS recording of the demo, and it quite clearly shows that the refresh rate was 30 Hz, not 60 Hz, so all of the number I mentioned above should be cut in half.
So unless crossfire scaling was terrible, a single Vega FE is basically the same speed as Fury X (possible 20-30% faster if they were running the 1.2 version of Prey). I can only assume that these drivers must be extremely premature, otherwise it just doesn't make any sense*.
*another possibility is of course that the frame with 2.7 tears, was very unrepresentative of the performance.
Not sure if it was the image used for the calculation but it was the only image I saw here, and when I first saw it it was definitely not impressive with the math some people churned out:
Basically a door, some shadows and one enemy.
Watching that video you linked, I don't know how anyone could walk away with it with any confidence. I'll just assume the drivers aren't ready - but if that were the case, AMD should have not shown that live footage.
AMD gonna do what AMD gonna do. I'm still giving AMD the benefit of the doubt, especially after Ryzen.
Care to show that scene on a 1080 / 1080 Ti @ 4k?
Send me a Steam key and I'd gladly reproduce it for you
Haha I don't own it either. I tried looking online for youtube videos of it but haven't seen any and without playing the game I don't even know what to really look for. I'm assuming its near the end of the game since it's a big fight scene.
If at $549 NVIDIA enjoys a gross profit margin of 60% (probably on the high side honestly...), then that implies a COGS of $220.
Margins on a $300 product with that same exact cost structure would be $80/$300 = ~27%, which would be awful.
Now, I know this might seem decent (it's not...), but also remember that the cost of the entire finished AIB (PCB, VRAM, cooling shroud, box, etc.) is going to be relatively fixed -- the only part that would come down in price in this case would presumably be the GPU. So, NVIDIA's margins on the silicon die itself would probably be much lower than the 27% that I just mentioned.
Not sure if it was the image used for the calculation but it was the only image I saw here, and when I first saw it it was definitely not impressive with the math some people churned out:
Basically a door, some shadows and one enemy.
Watching that video you linked, I don't know how anyone could walk away with it with any confidence. I'll just assume the drivers aren't ready - but if that were the case, AMD should have not shown that live footage.
AMD gonna do what AMD gonna do. I'm still giving AMD the benefit of the doubt, especially after Ryzen.
I'm not really sure how the content of the scene in this frame (a door, some shadows and one enemy) would have any influence on the quality of the math (which is based on calculating the height of the screen tears)?
Obviously one could argue that the frame in question isn't the most representative of overall performance, but that doesn't really have anything to do with the math itself (only the conclusions drawn from said math).
Also I agree that the demo doesn't exactly inspire confidence in Vega, but then again it wasn't really supposed to (the demo was about showing off TR's I/O capabilities, not Vega)
That won't stop people from trying to extrapolate performance metrics from the demo, form opinions on apparent high/low performance and initiate countless click-bait articles on Vega.The scene is about halfway through the game and it's completely obvious from Su the demo is about the threadripper platform and not an "OMG look at Vega" demo.
Send me a Steam key and I'd gladly reproduce it for you
If it was the scene used for the math, it isn't a very demanding scene.
Would have been more of a showcase for Threadripper if they showed 2xVega while content creation or streaming was going on in the background. Or, at least something that would tax more than 32 lanes and 8 cores. It showed nothing of Threadripper's capabilities, and only showed a poor demonstration of CFX Vega.
That won't stop people from trying to extrapolate performance metrics from the demo, form opinions on apparent high/low performance and initiate countless click-bait articles on Vega.
Buy it, benchmark it, return it within 2 hours.
It's a pc game designed around 4 core 8 thread machines and at 4k very high is completely gpu bound anyway. How is that ever going to demo threadripper in any meaningful way?The scene is about halfway through the game and it's completely obvious from Su the demo is about the threadripper platform and not an "OMG look at Vega" demo.
Also the game is fantastic.
I believe it was more about the PCIe lanes available in Threadripper allowing massive crossfire setups.It's a pc game designed around 4 core 8 thread machines and at 4k very high is completely gpu bound anyway. How is that ever going to demo threadripper in any meaningful way?