Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Dec 26, 2018.
How come the Fury X does so well at higher resolutions then?
right this is geting crazier
Well, thing is that killing all cards with 4GB VRAM is easy as long as you forbid games from using lower texture LOD. But on 1080p, there is point from which higher texture resolution makes no difference. And that is before 4GB cards start to suffer. Therefore 4GB cards will continue to do just fine.
What I would like to see is actually improved IQ through 32x AF. (This bar should have been risen long time ago.)
Higher texture IQ is instead done via increased rendering resolution which is much more wasteful in computational resources. (That's one of reasons why people actually use downsampling.)
4gb cards are fine but why is nvidia still trying to push 3gb variants is beyond me. Even at 1080p you will be vram limited.
4gb cards are a bad investment. You pray to developers to not use lots of textures.
To all of you above just check that here. 590 rx is the ultimate mid-range gpu at the moment, it has proper power to utilize more than 4gb vram.
Shadow of War uses great amount of Vram 'till today. Learn from it. It clearly shows for example FuryX is weaker in comparison to smaller 590.
This is a bit of a loaded answer.
The only time you're going to see a major difference is when individual textures, or even multiple immediately required textures for a single frame, are going to overload the frame buffer to an extent that it becomes impossible to get by with less memory. That's simply not going to happen, so increasing LOD quality and even anisotropic rates may be of some value, but it will just look all out of place unless all the textures are matching in quality, and at that point, the issue is going to be the GPU's capabilities and the memory bandwidth to move that data. Simply put, and yes, we've heard this a million times before, no point in sticking 8 GB on a lower end card, as it'll make no difference. Sames goes for bandwidth, which is the Achilles heel of the RX 4/5 series.
So, to answer my own question, it's because the Fury X is capable of processing data faster due to a higher spec GPU and HBM memory.
SOTTR is also a good comparison how even a 6gb is easily used. I would think twice before buying a 399$ 2060 6gb if you plan to keep the gpu 2 or more years.
Allocated does not equal needed.
Still if you dont have enough vram game will try to compensate by using more system memory. That cant be a good thing, right?
Not if it doesn't need it.
It clearly shows how much drivers have improved in the last year or so.
It can cause stutter in twitch shooter game if you are quickly/frenetically turning around. (But hey, I am yet to see one with such requirements on VRAM.)
As of actual situation of that SotTR. 1080p/1440p/2160p benchmarks have RX470(4G), RX570(8G) + Fury X(4G), RX580(8G) to compare.
On all mentioned resolutions RX570 is ahead by 8/10/13% from RX 470, but that difference falls within rounding error as fps on 1440p is not exactly OK, and 2160p performance is far from playable.
For RX580 and Fury X, RX580 is ahead by 5/-3/0% and again that's within rounding error for given fps.
So, it is great example showing that all those allocated gigabytes are actually not needed to render each frame.
The more the merrier
i used to be not that stingy about vram on cards i brought my gtx 770 2gb on release and what i thought was that it was not worth it because it will not become a factor , considering that that card played decently anything i could play till the R.I.P. with my motherboard ... i would say i was spot on ... mind you that the card had a delayed retire plan since the time i was planning to update to a 480 back then was the time etherium mining exploded .... now that i have 1060 6gb i got the 6gb because i thought the 3gb are kind of already limiting and i felt 6gb will be safe to not affect me when i retire my 1060 eventually ...now for a 2060 ....6gb... i would have been more confident if there was an 8gb model .
I feel pretty good about my 1060 3gb. In all the games I have played it has never used 6gb of memory!
I'm pretty sure we only need two 2060 models - one for desktops and one for laptops.
So Nvidia have finally realized that their 2080 and 2080ti are so stupidly expensive that only very few enthusiast will bite, so they blow up the 2060 range beyond proportion.
hmmm what we have here:
-RX 590 > RX 580 (a small bit only)
-in it's time the Fury X was fighting in beween GTX 1060 and GTX 1070 (even with "boost" of recent driver (i call that a "fix" not a "boost") ).
-the RX 580 is at GTX 1070 level
-4go is too weak over 1080p now, Fury have 4go and RX 590 have 8go
Then i would say the RX 590 is in the right place but that the Furry X might step on same level in some case.
When I overclocked my RX-580 to RX-590 levels, It was not exactly impressive by itself. But I have 2360MHz GDDR5 with tight timings and that allowed it to kind of match Fury X.
But even then Fury X has 24% higher shader power. What goes for RX-590 is 1.5GHz which is 43% faster than Fury X clock. Therefore all those parts of GPU which are not multiplied benefit from clock. But that requires proportionally quicker access to data.
I guess it's better than just having one, but honestly, I think this will cause more confusion than anything.
Anandtech's review makes a notion of how 4GB is becoming too limiting in modern games. The GTX 980 gets less than half the performance of the GTX 1060 6GB and R9 390 in Wolfenstein 2 at 1080p.