You can only really compare benchmarks done in the same area. Same principle as was being pointed out RE: the Battlefront Vega demo. We don't know where TPU does their bench run.
This is such a basic concept that it's hard to believe this is truly a technical discussion forum sometimes. We need an apples to apples benchmark before drawing any performance conclusions. The only things we can do is guesstimate based on the architecture improvements we currently know about.
some people wont believe it even when compared apples to apples.
a fun fact, NH and sweclockers (swede websites) compared the same gpu technology and achieved different results.
One was Nvidia based and one was amd based.
so even when compared apple to apple, results dont make sense
and btw a bit off topic but I had to respond to the mod in charge,
be good at your job, dont suck at it as you currently do.
If you suck at it, then resign and make place for someone who understands what people write.
Shame on you ElFenix to go off topic on a GPU forum.
This is the collected nVidia discussion from the Radeon Vega Architecture Preview thread. The rule is simple but you just can't help yourselves. Shame on you all.
AT Moderator ElFenix
Odd thing to wonder about when the other GPUs have EKs including the Radeon Pro Duo. If you were just trying to get conversation started again... =DWhen Vega finally appears I wonder if EK will make a waterblock?
Vega has some power improvements and is in general an improved architecture clearly. So you don't need much energy savings. Hbm energy savings should be enough alone right?Partially because Polaris's power draw is such that scaling it to competitive performance levels would break the bank in terms of the power budget for a big consumer GPU.
They'll probably manage to improve things a bit of course (HBM2 definitely will a bit in power terms), how much we'll have to see.
It's not definitive by any means, but Fury X's performance per watt was about 30% better than a 380X 4GB at 4k according to TPU. There's a lot more that goes into it, but there should be a substantial power savings moving to HBM2.Vega has some power improvements and is in general an improved architecture clearly. So you don't need much energy savings. Hbm energy savings should be enough alone right?
I disagree with this statement. If we take RX480 and GTX1060 as reference, Pascal has a 20% perf/watt advantage over Polaris. This is not a showstopper for a big consumer GPU so why would it break the power budget?
If you look at it in terms of FLOPS and you factor in the wider memory bus and the extra 2GB GDDR5, the power is somewhat similar.
The problem would be gaming perf vs TFLOPs and that's where Polaris loses. How much better can Vega do, remains to be seen.
HBM2 and many other things will help but better utilizing the available compute overall could be the biggest gain in gaming.
Edit: This is a GCN3 vs. GCN4 comparison with same CUs/frequencies from hardware.fr:
Vega should only add to this.
[...] Vega with 4096 sp and 2 HBM2 stacks is around 500 sq mm on 14nm FINFET. Fiji with 4096 sp was 600 sq mm on 28nm. [...] The fact that Vega 10 is almost 50% larger with same sp count seems to hint there is major changes in the architecture.