DX12 Benchmarks are out, time to trade in the 970?

That seems like an extremely premature conclusion considering this is a single example using beta game code that's probably unrepresentative of your typical DX12 game (in workload terms) and using non final drivers. On the other hand, if the double memory is worth the extra £30 to you anyway, then you can't really go wrong.

 

4GB has proven to show it's limitations in some corner cases and I've no doubt they will become more prevalent over the next few years so there's certainly a benefit to be had. The extent of that benefit is debatable though. But yes, with the extra general performance it might be worth making the swap anyway. I doubt this DX12 performance gulf between the two IHV's will carry through to normal games but there's always a chance and if it does then you'll have obviously made a smart move. Assuming you intend to keep the GPU for the 2-3 years it's going to take to start seeing reasonable numbers of DX12 games of course.

 

That's a good question to ask once DX12 benchmarks are actually out.

 

But when the real world DX12 benchmark are out and 970 is outclassed by R9 390 by a lot, then 970 2nd hand price may plummet. Hopefully not... hopefully it's just the driver.

 

I did not expect such a premature thread on dx12 performance "conclusions" here already.

 

It's too early to conclude anything. All the isolated performance previews didn't try to compare within hardware stacks, and some of the conclusions you can draw after mushing them together don't make much sense.
If all the tests are thrown together, there's no clear gain beyond 4 Intel cores, and no wide margin between the 980, 980 Ti, 290, 390, and Fury X. Since Nvidia's DX12 performance is a tick below its DX11 in a number of cases, you can go to the next conclusion that DX11 is the same as DX12, so I don't think anything is settled.

The only clear results are that you shouldn't get anything below an Intel quad core, and don't get AMD for DX11 or CPU.

 

We can't answer your question precisely because of the premature nature of things. We don't have crystal balls.

 

Just shiny ray-traced ones.

 

You should do it.
You correctly identified that the 390 has double the VRAM. Which is very important.
As you say, cards that get rekt and have only 4GB of VRAM aren't very future-proof. And AMD knows this.
You've made the calculation and the math checks out. Now you should follow through.
I think it's great that the 390 is getting some sales.

 

If you really feel the need to do this now then do it, it seems you have your mind made up already. The only thing I can think of is can your power supply handle a 390x? IIRC it has a higher wattage rating than a 970...

http://www.tomshardware.com/reviews/amd-radeon-r9-390x-r9-380-r7-370,4178-10.html

haven't found one for the 970 yet.

 

Btw, can someone benchmark the power consumption on DX12 and if can, compare it to the same game/benchmark app in DX11?

 

@TomRL
R290 is a pretty nice GPU, if a bit power hungry (although as a high-end gaming enthusiast, this should not be #1 on your list of priorities anyway). R390 is nicer still (not sure about its power usage compared to its predecessor though, but again - who the eff really cares, right?! )

Still, I wouldn't run out and buy a whole new GPU over some very sketchy benchmark figures cooked up with early drivers. It's too soon for that.

Of course, if you're made of money, then why the hell not. Just Do It, if that's what you desire; you're a grown man, or so I assume anyway. You can buy whatever you want.

 

Frankly, the results from these sites so far do not support that, either. Mostly because the results don't make much sense.
If you focus on the 290 catching the 980 Ti, sure, that seems huge.
However, in the context of other reviews showing very close margins between all the cards from 290 through Fury and the 980 and 980 Ti, it's basically rolling dice.

If the somewhat lateral move to the 390 does not affect the time window to a future GPU purchase with a more fully-featured DX12 GPU at the 16nm generation or later, then whatever.
Otherwise, I would question whether either current non-halo GPU is going to buy you that much time over the other before they start to show their age against GPUs that will have the benefit of a manufacturing jump and time to implement the various DX12 feature tiers that current products implement somewhat inconsistently.

 

I haven't seen any evidence of that, and that forum post that seems to be doing the rounds presently certainly doesn't qualify as more than speculation based on what appears to be a faulty premise.

 

What exactly do you mean by this?