You honestly can not make up how funny it is to see gtx 970 owners say how poorly their card performs to the point that they can't even play modern games at 1080p now.
Is there any reason Polaris 10 wouldn't be a good card for crossfire 4k?
I'm talking architecture wise?CF is good if you plan to play games months after release. AMD can't seem to get CF support going quickly in many GameWorks titles. But it eventually gets fixed.
If you're a day 1 gamer, avoid CF.
It appears Ive just discovered first Polaris 11 device in CompuBench database.
Nvidia is replacing the 980ti it seems in May.
so whats AMD response to that?
source
http://www.sweclockers.com/nyhet/21...ti-far-ersattare-i-maj-gtx-970-avloses-i-juni
If polaris 11 is 1ghz and polaris 10 850Mhz we can say goodbye FuryX performance in 1080P.
You honestly can not make up how funny it is to see gtx 970 owners say how poorly their card performs to the point that they can't even play modern games at 1080p now.
Those clock rates are far too low, at least for desktop parts. In fact, they're no different than what we get now with 28nm GCN. Why would AMD switch to a FinFET process (which enables significantly higher clock speeds) and then leave that much performance on the table? We should be getting clock speeds of 1200 MHz or more on the Polaris desktop cards. Both the Apple A9 chip and Nvidia's GP100 got ~40% clock boosts over their non-FinFET predecesor chips. I expect the same for Polaris/Vega. Of course, laptop parts may choose lower clock rates to boost efficiency, but this is nothing new.
There are various ways to design a GPU. You can go wide (more functional units), you can go raw speed (high GPU / memory clocks), or you can do a combination of those. I would much prefer AMD to go back to the 7950-7970 era where they go wide and leave us 25-30% OCing room. This way, they hit their perf/watt targets with stock parts, while satisfying AIBs (higher clocked cards), and giving enthusiasts/overclockers huge headroom for more performance. This is actually exactly what NV did with 980Ti and this strategy demolished Fury X. Since Pascal went to 1.5Ghz, it would surprise me if that chip has 25%+ OCing headroom that 980Ti and 7970 had. That means for Vega, I would much prefer a 6144 shader card with 900-900mhz clocks at 250W TDP, or 5120 @ 1.05Ghz. The way AMD designs their chips is different than NV. AMD's cards' power usage goes way up with higher clocks and voltage. My 390 hits 1165mhz stock voltage but needs a 0.1V boost to go to 1220mhz. Power consumption goes up 67W or so for the last extra 55mhz. That's why I would much prefer a wide lower clocked chip so that people who don't care about 70-100W increase of power usage can OC Vega another 25%+.
But for all of this to make sense, a stock Vega needs to roughly tie GP102 at stock. At the same time, most people do not overclock so I can see AMD raising clocks up because they got criticized by customers and media after low-balling the original clocks on 7950/7970.
I think the average consumer would almost always prefer fast guaranteed performance out of the box. So you are correct that it's probably better for AMD to maximize clock speeds even though I'd much rather them use the TDP headroom towards a larger die (i.e., more functional units). Larger die is more expensive to manufacture though.
I think even the most optimistic predictions on the forum point to Polaris 10 tying Fury X or barely beating it. Surely, that's nowhere near enough to go head-to-head against 1070/1080. I still cannot reconcile what AMD's plan is exactly from the time GP104 drops to Vega? Are they literally going to forfeit the $349-650 market for 6+ months?
sorry but if one thing history has proven again and again, is that AMD has no clue how to make money. Their financial status is the undeniable proof of their incapacity to be profitable...RS since when did you start talking about fictional GPUs like GP102. btw until Polaris 10 and GP104 launches and we know about their performance and pricing I suggest you just cool down. AMD knows where the money is to be made. AMD needs products which sell extremely well and help them gain back market share. Some good examples of extremely popular GPUs - 8800 GT, HD 4870, GTX 670, GTX 970. Most of them sold below USD 349. So what matters is being able to deliver an impressive product in good volume and at the right time. :thumbsup:
sorry but if one thing history has proven again and again, is that AMD has no clue how to make money. Their financial status is the undeniable proof of their incapacity to be profitable...
I'm not sure that the large quantity but low margin market will help them. They need to create value, they need to maximize profit per product, they need to work on their brand awardness, they need to invest in an ecosystem, they need to push their software development... everything that Nvidia does 10 times better than AMD and that allows green team to sell their products at higher price, thus been profitable.
Don't you think complete command of the consoles running on their hardware and software (Vulkan/DX12) and being easily ported to PC is an ecosystem? Because that's the near future of gaming. Your gaming.
AMD can't make money because Bulldozer is shit and is dragging the company down. It's the same way Tegra drags down Nvidia and Atom drags down Intel. The main difference is that CPU is AMD's main money maker while GPU is Nvidia's.
For now, we still don't know if their console/low API strategy will pay off. What I know is that every time AMD had a new feature, it was Nvidia that took the benefit, like Tesselation. Being an early adopter doesn't guarantee success. So let's see how Polaris and Vega will monetize AMD DX12 advance before making any conclusion...Don't you think complete command of the consoles running on their hardware and software (Vulkan/DX12) and being easily ported to PC is an ecosystem? Because that's the near future of gaming. Your gaming.
Bulldozer was a huge let down, we all agree but graphics division was far from been positive either. They are bleeding money every quarter since too long.AMD can't make money because Bulldozer is shit and is dragging the company down. It's the same way Tegra drags down Nvidia and Atom drags down Intel. The main difference is that CPU is AMD's main money maker while GPU is Nvidia's.
It's far from doom and gloom, 2016 will be a great year for AMD.
I finished it as well, but on 1440p medium to high settings on a HD380. Even then it looks way better than 1080p on high settings.Last AAA game i was played was rise of the tomb raider at 1080P max SMAA.It runs ok on GTX970 1500/8000 but some scenes was bellow 60fps.But with dx12 patch(and VXAO) it runs like crap so i want new 16nm GPU.
Almost no chance of that is there? Zen isn't really going to be factor until quite late into the year, and they're doing badly in GFX stuff now, and are giving up a chunk of the die shrunk GFX card market for the remainder of 2016.
2017 is where they've got a decent chance of having a reasonable year - full, sane, up to date GFX card range for the first time in absolutely ages; Zen actually in some quantity etc.
How decent will depend on how good everything is, but there's at least grounds for hope Then maybe some new console contracts for 2018ish to keep it ticking/staggering(?) along.
True. But 2016 will still be a great year for AMD, better than 2014 and 2015 for sure.
Last quarter, GPU dGPU share rised against NV. I expect that trend to continue in Q1 2016 and Q2.
If Polaris is on track for June, they will be in a much better position to compete. End of 2016, potentially Vega. Otherwise Q1 2017, Vega and Zen. Nintendo NX potentially this year too.
There are various ways to design a GPU. You can go wide (more functional units), you can go raw speed (high GPU / memory clocks), or you can do a combination of those. I would much prefer AMD to go back to the 7950-7970 era where they go wide and leave us 25-30% OCing room. This way, they hit their perf/watt targets with stock parts, while satisfying AIBs (higher clocked cards), and giving enthusiasts/overclockers huge headroom for more performance. This is actually exactly what NV did with 980Ti and this strategy demolished Fury X.
Worth remembering that 'great' for AMD would be say 50 per cent market share in GPU's and a healthy chunk of the CPU market. They've managed that before I think?
Right now it's record low (if slightly improving) dGPU share and horrible CPU share. Even tolerable would of course be a major advance over that.
They might manage that from September 2016 - September 2017, and some chances for the all year 2017.
Well, if you're only starting to find out in April onwards, then that's a huge chunk of 2016 gone!
Worth remembering that 'great' for AMD would be say 50 per cent market share in GPU's and a healthy chunk of the CPU market. They've managed that before I think?
Right now it's record low (if slightly improving) dGPU share and horrible CPU share. Even tolerable would of course be a major advance over that.
They might manage that from September 2016 - September 2017, and some chances for the all year 2017.