Keep 290X or upgrade to 390x?

[Anti-grandbean]

My current build uses an XFX R9 290x with 4 gigs of vram and a clock speed of 1000mhz. I run at 1080p currently with plans to move up into 1440p soon, when DX12 utilizes Vram and the like better.

I am really confused still, after hours of reviews and scouring the web, about how exactly much performance you get with the 390(x) vs the 290x? Will the 290x be a better card once DX12 becomes the new standard or will the 390(x) series do much better?

Anyway - my current specs are as follows:

CPU - AMD FX8350 slight oc to 4.5 ghz
Cooler - Corsair H80i GT
GPU - XFX R9 290X 4gig
PSU - Seasonic M12II 850w fully modular
Mobo - MSI 990FXA Gaming
Ram - G. Skill Ripjaw X 16 gigs DDR3 1600 (4x4)
HDD - Seagate 2tb
SSD - Not yet installed - Samsung 850 EVO 256Gig
Case - Coolermaster HAF X Full tower 4x 200mm fans + 1 140mm
OS - Windows 10 64bit

 

[Gingerbread]

By the time DX12 becomes the standard (probably at least 2 years more) There will be better GPUs out there. Might as well wait for them.

 

[jdwii]

Keep your current GPU if anything if you are running on a older monitor upgrade that you will notice a bigger difference

 

[Gillerer]

AMD Radeon R9 390(X) is pretty much a rebranded/tweaked 290(X) with twice the memory, so there is absolutely no sense in this upgrade path.

If you're currently on 290(X), you should wait until the new GPUs come out next year from both AMD and Nvidia, that use the new TSMC 16 nm FinFET(+?) process node - the first such process shrink (for GPUs) in years. Also, both AMD and Nvidia high-end cards will use 8 GB of HBM memory (290(X)/390(X) still use GDDR5), which could boost performance in some cases.

(As a side note, there have been reports - at least on the MSI Afterburner forum - that Windows 10 has vastly improved GPU memory management, in that if a GPU runs out of memory, memory is no longer paged between GPU and system memory but the required data is instead streamed directly from system memory. This massively reduces hitching and bad performance in situations where an application requires more VRAM than is available in the system. So VRAM size really shouldn't be that much of problem anymore, since you're already running Windows 10.)

EDIT: Added the stuff about VRAM and corrected the process node from 14 nm to 16 nm; added "+" (Samsung has 14 nm).

 

[renz496]

Keep your current card. If DX12 end up using more VRAM then developer need more optimization to do.