With directx12 out, what would be the best gpu to take into consideration? And also mind explaining me what happened with nvidias gpu not gaining any type of boost from directx12.
You probably mean the Fury X and not the 390X. The 390x competes with the 980 and the Fury with th 980 ti. Also the reason why the Maxwell GPUs didn't gain anything from DX12 is because the architecture doesn't support ansynchronous computing however I think NVidia said something about doing asynchronous computing with the CPU but I don't know how well that will work. And with the overall optimisation for DX12 we will have to see how that goes. Right now there's telling who will optimise for it better
according to the 390X, its not realy worth it to grab a 390X over a 390.
Because the performance diffrence between these two cards is just too minor to realy justify the extra cost imo.
I personaly dont realy understand why te 390X even exists.
The Fury would be a better step up from the 390.
At the moment the 980Ti is top of the hill in DX11
How things will change as soon as DX12 is majorly adopted to new games, is only a thing that time will tell.
We have seen just one source that showed us a massive jump in performance on DX12 for AMD cards, and not as much on Nvidia cards.
But ofc just one source is not enough reference, too pull any conclusions from it atm.
1 Like
both are kinda equal at dx12 at this moment.
390x is about the same as 980Ti in DX12. Both deal quite well with dx12.
NV was already running on heavily optimized features in DX11 which enabled low level access in games. (Controlled access points, not free open access) Thus NV doesn't gain almost anything from DX12, as they were already doing those stuff earlier.
As stated above, NV GPU's are optimized for Serialized pipe lines. Serialized pipe lines means data goes in numerical order. (1, 2, 3, 4, 5 .... and so on) for GPU to compute. Asynchronous pipe lining or computing means it goes by without serializing. Means you can actually supply few data executors with different types of compute data, without pre-set order. Thus no need to wait for certain computation to be firstly finished in order to do other compute data that it lighter and still has to be done.
@MisteryAngel
390x is from a better waffle. They tend to overclock higher, run cooler (sometimes), and less likely to break overtime. (most of the time they also get better memory)
This! Also, if you can wait, wait until next years cards. Those will/should be the first to offer 100% feature-set support for DX12. There are no cards out now with 100% dx12 support.
I like waffles...
1 Like
This is how i undestood it:
Normally it is like this: The GPU waits for the CPU to provide the data for the Next Frame. In this Time the GPU is not doing anything.
With DiretX12 it is like this: When the GPU is waiting for the CPU to catch up, it will help the CPU to calculate the data for the next Frame. That improves performance in Games by a fair amount.
The Thing right now is that this feature is currently not supported by nvidias Maxwell architecture.
Yes sure,
But in terms of overclocking, there isnt much headroom left in the Hawai chips anyway.
390X is ofc a better bin, but in terms of performance there is not realy much of a benefit,
unless you are realy lucky to find one which overclocks like a beast.
But thats more like a miracle.
so at the end we should just wait and see right?. Also I have a 980 at the moment, i am going to build a second computer for my girlfriend and i was thinking on which one should i go for. 980ti or 390x, and the furyx looks good too. so i am still meh when it comes to pick one of those 3.
oh gosh, just wait and see realy.
The GTX980 is a great card, upgrading to a 390 (X) will be a downgrade.
Hence even the Fury is neck to neck with the GTX980.
So an upgrade to a Fury wont make sense either.
If core is premium you can oc to 1400~ (much more tolerant to volt oc too and going beyond 1400)
If memory is premium like hynix, samsung or elphida you can clock it (same order) 1700MHz, 1600MHz, 1500MHz.
Most cards stable without additional voltage are able to push 1150-1250MHz (those that are able to go beyond 1150 can go up to 1600MHz core with additional voltage and good cooling)
If that is compared by
core stock 1080MHz core (thats about 70MHz for 1150, and 320MHz for 1400MHz)
in 390/x memory is overclocked to 1500's anyway ~ so with Hynix ones you should be able to squeeze 1700MHz (another 200MHz) and pretty much get around 426GB/s
yeah well, wenn i flash back to the 290 and 290X.
If you got a card with Hynix you where golden.
But the cards with Elpida had allot of issues.
yeah Elpida were the first batch ran with stock cooler and crappy asus, and gigabyte coolers. Thus they broke quite easily. If they had proper cooling they wouldn't die so often and easily. Hynix and Samsung were in premium cores.
Its not that 290 and 390 can not be lucky, they can. They can be the super premium waffle. But note that chance is like below 1% for them. While X series is pretty much golden unless they are mounted with shitty cooler.
(I had that luck with reference 6950 from diamond, I could unlock it to 6970 and clock it better than most of guys could on 6970's even directcu II models which were OP)
Take into account dx12 games won't be the norm for a year or so. So I would say get what's best at the moment.
IMHO R390X is better with 8 gig of RAM for 200 less than Ti pricing, plus 980 and 390X are neck to neck. Unless the ROP count jumps from 64 to 96 or more performance in 4 K @ high FPS is futile. The main reason why I dumped my 5870 crossfire was the 1 gig memory limit on the cards not the FPS. I almost pulled the trigger on a Ti but couldn't justify $650 plus in cost for 1080 gaming. Most say 980Ti barely games 4K
980Ti blows the 390X out of the water, also at 4k atm.
980 vs 390X at 4k is an interessting battle at 4k, i have seen allot of variance in results,
wenn it comes to benchmarks.
But going from a 980 to a 390X doesnt make sense.
In most benchmarks i have seen till now, the 980 seems to be better in certain games.
However with dx12 this might change.