Since when does Performance Per Watt make my game run better? He did not mention a stat that literally means nothing when gaming. He said that it has competitive performance with the 1060, which it does.
Not looking too competitive to me. Efficiency in this range isn't such a huge deal for desktop usage, but extending into SFF, notebooks, and scaling up the architecture without needing extreme amounts of power are characteristics of "competitive" and "good" architectures. RX 580 is a little faster than a GTX 1060, but if Nvidia wanted to re-release and over-volt GP106 like AMD did with RX 580, I'm sure it'd have no problem catching back up and still consume less power. Nvidia's architecture is simply in a class of it's own being able to do the same amount of work with less transistors, less bandwidth, and less power consumption.
Nope, still wouldn't work. Even in fairy tales of everything scaling 100% linearly (which it never, ever does), outright doubling both RX 480 and RX 580 come up short in GPU limited situations vs 1080 TI, let alone Titan Xp. Furthermore, 2x RX 480 with true 100% scaling would be consuming 320+ watts while RX 580 would be consuming 400 watts. Back to the drawing board!
Not looking too competitive to me. Efficiency in this range isn't such a huge deal for desktop usage, but extending into SFF, notebooks, and scaling up the architecture without needing extreme amounts of power are characteristics of "competitive" and "good" architectures. RX 580 is a little faster than a GTX 1060, but if Nvidia wanted to re-release and over-volt GP106 like AMD did with RX 580, I'm sure it'd have no problem catching back up and still consume less power. Nvidia's architecture is simply in a class of it's own being able to do the same amount of work with less transistors, less bandwidth, and less power consumption.
Since when does Performance Per Watt make my game run better? He did not mention a stat that literally means nothing when gaming. He said that it has competitive performance with the 1060, which it does.
You quoted me talking about DX-12 gaming and this is from Techpowerup,
This is not a DX12 game, not to mention that even in this NV game RX480 is doing just fine against the GTX1060.
- Average (Gaming): Metro: Last Light at 1920x1080 because it is representative of a typical gaming power draw. We report the average of all readings (12 per second) while the benchmark is rendering (no title/loading screen). In order to heat up the card, the benchmark is run once first, without measuring power consumption.
- Peak (Gaming): Same test as Average, but we report the highest single reading during the test.
Power draw doesn't magically drop on DX12 or sponsored games from either camp. Look at hardocp, anandtech, Tom's, hardwarecanucks, techspot, and techreport. All those sites use different testing methods and games to find performance and power draws, and all those sites - shockingly - come to the same conclusion. The RX 480 and RX 580 are very competitive in performance but consume more or way more power than the competition. Everything else I said about scaling the architecture up and down, doubling RX 480 or 580 still not equaling a 1080 TI, etc. remain as well.
I said Polaris 10 is very competitive against Pascal in DX-12 games, i didnt talk about power draw ONLY. Performance, Power draw, die size, memory capacity, price etc etc are all making Polaris 10 very competitive against Pascal (GP106). Unlike what we see with VEGA 10 against Pascal now.
Also, i said double Polaris 10 + HBM2 would be much better than VEGA 10 against Pascal GP104 GTX1080, not GTX1080Ti. Double Polaris 10 with HBM 2 would be smaller than VEGA 10, with higher performance and lower Power draw than what we seem to have with VEGA 10 today.
To which I said "for desktop use..." but you apparently missed that. The more versatile a product is, the more use cases it has, the better it can scale up and down, the undeniably better it is. Those attributes might not affect you much, but they have a direct correlation to how successful the product can be.
During "desktop" use both cards are going to be sitting at idle using almost no power. An RX480 that I own actually sits there with its fans off during day to day use.
Again power usage does not mean much in day to day work or gaming. Sure it could change the PSU required, but that is typically only true if comparing cards of different classes.
Sent from my iPhone using Tapatalk
Polaris is only competitive on performance. It's 15% larger die with 20-25% more transistors and it's 30-80% higher power draw make it a very inefficient design to maintain competitiveness. As I already said many times over, Polaris 10 is a one trick pony on it's last leg that is being propped up by mining. It's nearly entirely absent in notebooks, it scaled down to be considerably worse as Polaris 11 vs. GP107, and the refresh is so power inefficient that nearly all of Nvidia's LAST GEN, PREVIOUS NODE parts are more efficient.
And as I already countered, doubling Polaris wouldn't solve AMD's problem. Even if it scaled 100% linearly in performance (which it wouldn't) power draw would also scale 100% and would be higher than Vega, while die size would still be similar to Vega (maybe 20-30mm2 smaller but largely inconsequential). GCN needs to die; its now extremely outdated when stacking up all the metrics.
Again you dont get it, double Polaris 10 + HBM 2 would be way more efficient (close to 250W-275W TDP) than 345W TDP VEGA 64. Unless something is not working correctly or not working at all, today VEGAs Gaming performance and performance/watt is way way worse than Polaris 10, unlike what Fury was vs Tonga at 28nm. Actually Fury Nano was more efficient (perf/watt) in DX-12 than GM200 GTX980Ti unlike what we see from VEGA vs GP102 today.
Not sure how that works. One Polaris is easily pulling 200 watts while gaming, two of them could easily pull down 400 watts.
Not sure how that works. One Polaris is easily pulling 200 watts while gaming, two of them could easily pull down 400 watts.
It's useless. People are going to believe self-made fantasies that defy reason even in an onslaught of irrefutable facts.
Polaris 10@ 1243 MHz is pulling out 137W of power, and the GPU alone pulls 101W. I am talking about Radeon Pro WX7100.Not sure how that works. One Polaris is easily pulling 200 watts while gaming, two of them could easily pull down 400 watts.
Where did Vega ended? I have not seen any reviews of RX Vega, yet.But using prior estimations, two Polaris would be 180% of GTX 1080 in DX12 and still cost less!
Certain posters have been selling AMD fantasies for years. I've just stopped taking them seriously, with Vega ending where it did, I'm surprised some of these posters don't just stop.
EDIT: These are the same posters that expect AMD to deliver better performance or near equal performance as rock bottom prices.
Where did Vega ended? I have not seen any reviews of RX Vega, yet.
I have written "wait and see", or "where are reviews of RX Vega"?Just stop. You're one of the posters I'm referring to. You aren't going to win any arguments with this "wait and see" nonsense when you were selling Vega as >GTX 1080 Ti long before AMD's own numbers deflated that balloon.
It's fine for you to guess and make estimates but when we do it "proof."
I have written "wait and see", or "where are reviews of RX Vega"?
So, how low do they actually plan to clock Nano? They did say that at 150W perf/w is similar to GTX 1080. So wouldn't that probably be somewhere between GTX 1060 and GTX 1070? Or more like very similar to Fury X.Polaris 10@ 1243 MHz is pulling out 137W of power, and the GPU alone pulls 101W. I am talking about Radeon Pro WX7100.
The rest is consumed by GDDR5 memory and the board inefficiencies. 2 stacks of HBM2 are pulling, at 400 GB/s, around 8W of power, and total memory subsystem(Memory controllers+ HBM2) should pull not more than 15W of power. So you have around half of what GDDR5 consumes.
Funniest part is that 4096 core GCN5 chip clocked at around 1.2 - 1.35 GHz should not pull more than 175W of power, under load.
Where did Vega ended? I have not seen any reviews of RX Vega, yet.
If that is the case, it means two things. Vega is bringing actual performance decrease versus Polaris 10.So, how low do they actually plan to clock Nano? They did say that at 150W perf/w is similar to GTX 1080. So wouldn't that probably be somewhere between GTX 1060 and GTX 1070? Or more like very similar to Fury X.
I'm not sure if that is a good or a bad thing... I guess it all depends on the price.
Not sure how that works. One Polaris is easily pulling 200 watts while gaming, two of them could easily pull down 400 watts.
Ask AMD, they said Vega 64 trades blows with the 1080, their words not ours.Where did Vega ended? I have not seen any reviews of RX Vega, yet.
My RX480 does NOT pull 200W when gaming. That measured at the wall comparing idle, pegged CPU but no gpu, and pegged gpu you but no CPU. You can certainly get them that high with overclocking, but they aren't that high stock.
The number provided there is Package power, as in GPU + memory, without the board inefficiency.Just stop. You're one of the posters I'm referring to. You aren't going to win any arguments with this "wait and see" nonsense when you were selling Vega as >GTX 1080 Ti long before AMD's own numbers deflated that balloon.
It's fine for you to guess and make estimates but when we do it "proof."