It is in line with what we expected.
I bet if overclocked more aggressively the power consumption will reach 290X levels.
Originally Posted by Blameless
With 32 ROPs, that was inevitable.
ATI was able to compete on more even footing with NVIDIA in those days than AMD is now.
They almost certainly did not have the yields to clock the 480s this high out of the gate, no matter what level of power consumption they were willing to accept. Most 580s are also almost certainly going to be more power efficient than most 480s at the same clocks.
Since the RX 480 is only 10% less expensive than the RX 580, I can't imagine recommending the former over the latter.
The last time they were in this situation was probably the 2900XT. GDDR4 had been unsuccessful. With GDDR5 they were able to pull it together with the 4850 and 4870, which were able to fight Nvidia's GTX 280 and 260 pretty well. They had a much smaller die size and could be almost as fast.
I suppose with Vega they could turn this around with HBM2 and the HBCC, but that depends on it being competitive.
The more I think about it, the more I think that the Fury X must have had 64 ROPS on purpose - this RX 480/580 with 2304 SP had 32 ROPs, versus 2816 SP and 64 ROP on the 290X/390X versus 4096 SP and 64 ROPs on the Fury X. I'm wondering with 96 ROPs if the Fury X might have turned out better - maybe even 128 ROPS. At the very least it may have helped at 4k it beat the 980Ti, which could be very useful.
It seems a "fixed" Fury X might have had:
- A much better Hardware Scheduler (They didn't upgrade compared to the 290X)
- Double the RBEs (at 128 ROPS rather than 64 and double the Z/Stencil ROPs)
- Increase the triangle output compared to the 290X (Fury X saw no increase)
- They should have gone with 8 stacks on the Fury X, which would have made it 8 GB (that's plenty)
- The CUs themselves needed to be fixed to address GCN's Occupancy flaws
Vega won't have 8 high stacks, but with HBM2, 8 GB is going to be standard. I think that 8GB will be adequate, especially factoring in how Vega might be able to double effective VRAM on games that support it. WE may see a refresh with 16GB just like how they doubled VRAM on the 390X compared to the 290X.
Originally Posted by LancerVI
I'm tired of waiting.
I've been saying for months that I was waiting for Vega; but I'm getting tired of waiting. The performance is here now with a 1080ti. I don't want to buy it, but AMD is taking entirely too long.
Patience you say??? This is a hobby for me. You're essentially telling me to put my hobby on hold. It's not a need, it's a want. My patience for a "want" is limited and somewhat irrational, but again....it's a hobby.
I want to build now!!!
AMD better release some worth while information quick, fast and in a hurry or people like me are going to say screw it, hold their noses and buy nVidia.
We're expecting information next month. We may see a release in late May or June.
Originally Posted by Mahigan
Not a rebadge. It's able to hit 1400MHz to 1500MHz. It's been re-engineered. Same GPU, made slightly better.
That's not a rebadge.
I just wish they had done something like the 4870 > 4890, which saw improved design, a retimed chip, and modifications to the ASIC power distribution. I think that had they done this, we would have seen perhaps another 15% without too much more power consumption.Edited by CrazyElf - 4/19/17 at 8:35am