AMD Sapphire R9 390 8GB OR NVIDA GeForce 980 4GB which one and why?
I can get em for the same price
My brother tells me that the 390 is getting better than his 980 with new games, etc.
Pic related it's the one I'm getting or the sapphire from my nephew the two of them used but for $150 bucks, I'm joining the Pc master race /g/ help me out.
(He's gonna buy the 1080 and give the 980 to me or his son and his son gives me the 390)
>>54744558
>GTX 1070 being released
>GTX 980 gonna get gimped
Buy the 390X senpai
>>54744558
does your mom pay the electric bill?
>>54744558
just wait for polaris
>not waiting for the 10 series and picking up a 980ti for da low low
it's like you fucks want to get jewed
>>54744558
390 is a rebranded 290 released in 2013
390 doesn't support opengl 4.5, only 4.3 and gets shit fps in doom
you can get the 980, try it out, resell it, and have enough money left over for a new 390 and lunch.
>>54744949
OP here, I'm getting it for just $150 bucks both of them, I don't know much about PCs, so that's why I'm asking wich one should I take of the two of them, just someone tells me wich one is better.
Thanks in advance.
Thanks for your fucking help /g/.
>>54744949
>390 doesn't support opengl 4.5, only 4.3 and gets shit fps in doom
all of this is an nvidia lie
>>54744558
The 390 is comparable to the 970, 980 is superior.
stock 980 smokes the 390. stock 980 is slightly better than the 390x, and smokes that and fury non-x with an overclock. but what you should really be doing is waiting 2 weeks and buying the 1080.
>>54745336
he's partially correct. AMD eventually implemented the extensions added to core in 4.4 and 4.5 but never implemented the changes made to GLSL, so they still break spec and only technically support up to 4.3 because of this.
>>54745400
1080 is meme tier garbage.
>GDDR5X in 2016
Get real. Vega soon, friend.
>>54745434
>Get real. Vega soon, friend.
>just wait 6 more months. i'm sure it will get better then!
>>54745037
150 is highway robbery for a 980. buy it immediately.
i doubt even in 2018 we won't have gpu with the same power as the 980 for $150.
>>54745472
GDDR5X is already outdated compared to HBM, HBM2 is going to shit all over it.
>>54745495
Get out of my thread fucko
>>54745495
HBM is just a shitty stopgap until HMC is ready, which is going to be 16x the density and bandwidth of HBM if you structure your memory accesses properly.
>>54745413
you mean the retarded nvidia only extentions?
>>54745522
>retarded nvidia only extentions?
all opengl extensions are open and can be implemented by anybody, plenty of extensions authored by employees at nvidia, amd and intel (sometimes even collaboratively) have been implemented on all of their GPUs.
>>54745543
DELET THHIS
>>54745520
HMC isn't going to be incorporated into a GPU until at least 2020.
>>54745615
2018.
>>54745629
chuckled at 2017 until i realize the current year
http://www.geforce.com/hardware/10series/geforce-gtx-1070
>>54744558
This story sounds suspicious.. are you sure there's not a wife's son or a don't speak to me that you left out?
Anyway why doesn't he give his son first choice
>>54745876
Because its his brothers' wife's' son
>>54745876
Cuz my nephew it's fucking 13 years old, why the fuck he's gonna do with a 980? I'm 23 not retarded enough to break it or something, also my nephew it's fine with the 390
>>54744945
>implying 980 ti won't be gimped
>mfw 1060 will outperform it
>>54744687
>gimped
No, maxwell and pascal are basically the same architecture.
>>54748540
see
>>54748581
There is no 'gimping' Nvidia just stopped making any new performance enhancing drivers for kepler because they're too busy making money selling maxwell cards. Now Pascal is basically Maxwell die shrink you're still going to get the benefits.
AMD gets 'better' because they're still trying to sell 3 year old cards to you as new. Don't believe me? Try using anything pre-GCN and see if they made any new drivers for it.
Kek
>>54748581
never stopped nvidia before
>>54748613
>never stopped nvidia before
>980ti comes out
>970 still getting performance enhancements
Hurr durr hurr nvidia gimps older cards hurr
>>54748607
>last driver for the HD5870 released in 2009 was in late February
thats bretty good desu
>>54744558
if you can get both for 150 each... buy both and sell the 980, winning?
>>54748712
the hd 5000 series is the same as the hd 6000 series
>>54748682
Nvidia gimps cards that are replacements
980 Ti isn't a 970 replacement
970 was a 780 replacement
now suddenly the old HD7970 that competed with the 680 is now handing the 780 it's ass.
>>54748740
HD6970 launched in 2010 so thats still bretty good, 5 and a half years of support
>>54748682
oh now i see now, you're completely right mr guy. I guess Kepler and Fermi didn't exist after all.
>>54748744
No, they don't 'Gimp' cards, they stop performance related improvements through drivers. The latest drivers benefit mainly the latest architecture. The old architecture cards stay stagnant. There is no gimping. There is no loss of performance, The newer cards get faster and the old cards stay the same. AMD's cards still improve because they're still on GCN and the drivers improve GCN across the board because they're still on old as fuck architecture and rebranding old cards for sale.
>>54748763
>support
They already had unified drivers by then dumb dumb. I doubt you'd have seen any improvements on them like the GCN cards have been.
>>54748779
>old architecture
in 1.1 they completely redid the schedulers, clock control system, and added various new ASIC's, while shader layout also got massively changed to allow much much denser packing
in 1.2 they completely redid the memory hub twice, did completely new media ASICs, and improved tessellation immensely. A few instructions were added as well.
Want to know the change from Kepler to Maxwell?
Way way more cache, increased shader granularity from 192 to 128 shader groupings, and removal of part of the scheduler to make it so there was only one scheduler per grouping, rather than all schedulers having access to all groupings. Again a few new instructions were added. Everything else was software.
Honestly, Nvidia had their last big shakeup back with Fermi to Kepler, where they had to redo the shaders completely by dropping the hotclock feature.
>>54748851
You're a moron and it shows
http://www.anandtech.com/show/8526/nvidia-geforce-gtx-980-review/4
>>54745262
You already asked in /v/ you fucking faggot
>>54748905
No, I asked here first but no one answered so I asked /v/ and this thread didn't died.
Err so should I go for the 980?
>>54748914
do your own research
>>54748914
No
GTX 1070 has better performance, HEVC Main12/Main10/Main hardware decoding, VP9 hardware decoding, HEVC Main10 hardware encoding, DP1.4, HDMI 2.0b & 8Gbps memory
There's no reason to get a GTX 980 at all
>>54748946
Does the GTX 1070 cost $150 like the 980 he's buying from his brother?
Amd is announcing its new lineup in less than a week if you want to be a waitfag
>>54748914
First step: go on www.google.com
Second step: click on the search bar
Third step: write the phrase "980 vs 390 games benchmark"
Fourth step: read the title of the links
Fifth step: open the links that interest you (for example, a game that you want to play)
Sixth step: see what card wins
Seventh step: buy the card that wins
Eighth step: give your dog away to someone you can take care of it and kill yourself
>>54748965
[citation needed]
>>54748959
No, but it's well worth the $379 for all the features and 8GB of memory
>>54749000
I don't have the money for, I still got to buy stuff, etc. So I have to choose between these two.
>>54749000
Does it have more than twice the performance of a 980? Because it costs more than double than the option OP has
>>54748880
>Maxwell 1 was for all intents and purposes an optimized Kepler architecture
>Maxwell 2 is almost a half-generational update on its own
now point by point
>architecture to support the 16x raster coverage sampling required for Target Independent Rasterization and UAVOnlyRenderingForcedSampleCount, Maxwell 2 will support the more advanced Tier 2 tiled resources, Rasterizer Ordered Views, Typed UAV Load, Volume Tiled Resources, and Conservative Rasterization
updated instruction, previously was 8x raster coverage sampling; hardware support for breaking textures into smaller parts and loading individually, 2D and 3D, mostly microcode has been done in software since 2011 anyways
notably all of these are ROP improvements, a single part
>while the VXGI greatly benefits from the hardware features built into Maxwell 2, it is not strictly reliant on the hardware and can be implemented through more traditional means on existing hardware
ROP again
so yeah basically they improved the ROP's
>>54744949
I'm getting 77 to 110 fps in Doom on my 390 cuckboy
>>54745629
This isn't even remotely accurate. Vega is coming in October of this year.
>>54748988
>not looking at the price difference
dude, the contender to the 980 is the 390x
>>54748581
They can check the HWID
