Hm, maybe two HBM2 chips aren't enough bandwidth? Or the new cache architecture is either not any good or not working well atm.
@ 4K though it's falling flat on its face for no known reason. This should be it's strength, not it's weakness. There are instances where the card in its current forum is barely faster than a Fury X @4K that are head scratching. What is going on here?
The bandwidth is actually poor to be using HBM2. Vega FE got 480GB/s.Hm, maybe two HBM2 chips aren't enough bandwidth? Or the new cache architecture is either not any good or not working well atm.
The bandwidth is actually poor to be using HBM2. Vega FE got 480GB/s.
Take Volta V100. It got HBM2 too, but the bandwidth is 900GB/s.
Nvidia is beating them in their own game.
Bet Titan V which I guess will be out this year, will feature the same memory as Volta V100.
They were just waiting for AMD to finally get the Vega FE out so that they can go up a generation and have the fastest card again.
Not long since we saw the leak of the Titan V in a testbench. Its just a matter of months imo.
Hm, maybe two HBM2 chips aren't enough bandwidth? Or the new cache architecture is either not any good or not working well atm.
Nvidia's delta color compression and tile based rendering allow it to gain significant performance over the limited bandwidth compared to AMD. Polaris has better color compression which is passed onto Vega (and improved?), and Vega is also supposed to introduce draw stream binning to its rasterizer which should help with bandwidth gains as well, when it's applicable. This new feature could take significant driver work though to truly make differences, and based on tests so far it doesn't seem to be doing anything yet.480GB/sec isn't stellar, but it's as much as the 1080 Ti has.
Could 8-Hi HBM2 have really bad latency?
Raja says die size is closest perfect square to 500mm² = 484mm² die size (22²):
https://twitter.com/GFXChipTweeter/status/881370308694822913
PCPer and GamersNexus both overestimated the die size...
For comparison:
Polaris 10 = 232mm² (14nm)
Fiji = 596mm² (28nm)
Hawaii = 438mm² (28nm)
480GB/sec isn't stellar, but it's as much as the 1080 Ti has.
Could 8-Hi HBM2 have really bad latency?
Rx580 is memory bandwidth bottleneck and have same bandwidth as GTX1070 whitch is like 30-40% faster.So NV delta color copression + tile rasterizer is 30-40% better.Vega need atleast 700GB/s to compete with 1080TI.Nvidia's delta color compression and tile based rendering allow it to gain significant performance over the limited bandwidth compared to AMD. Polaris has better color compression which is passed onto Vega (and improved?), and Vega is also supposed to introduce draw stream binning to its rasterizer which should help with bandwidth gains as well, when it's applicable. This new feature could take significant driver work though to truly make differences, and based on tests so far it doesn't seem to be doing anything yet.
We'll see I guess....
I think the Titan V will use a GV102 chip though. GV100 got 5376 cores and out of those it have a bunch of Tensor cores which I think GV102 will be without. So that should reduce some space.480GB/sec isn't stellar, but it's as much as the 1080 Ti has.
Could 8-Hi HBM2 have really bad latency?
Fury X @Stock vs Vega FE @1052MHz and 800MHz HBM2. Less bandwidth (don't know why his HBM2 is only running at 800MHz) but same core clock. There is definitely something not right with Vega, even with less bandwidth the difference between the Fury X and Vega FE at the same core clock is way too big.
https://www.forum-3dcenter.org/vbulletin/showpost.php?p=11419198&postcount=7679
http://www.planet3dnow.de/vbulletin...giemessungen?p=5162951&viewfull=1#post5162951
How is that unfair?That comparison seem a bit unfair.
Fury X rig got a 1800X while the Vega FE got the 1700X. Not sure how the benchmark scale with CPU speed, but perhaps Vega FE needs a CPU with high clock too? Meaning a 4-core CPU at 4.5GHz might be better. Meaning the benchmark is CPU bottlenecked on the Vega FE while not on the Fury X.
I dont know. Just throwing theories out there
Fury X @Stock vs Vega FE @1052MHz and 800MHz HBM2. Less bandwidth (don't know why his HBM2 is only running at 800MHz) but same core clock. There is definitely something not right with Vega, even with less bandwidth the difference between the Fury X and Vega FE at the same core clock is way too big.
https://www.forum-3dcenter.org/vbulletin/showpost.php?p=11419198&postcount=7679
http://www.planet3dnow.de/vbulletin...giemessungen?p=5162951&viewfull=1#post5162951
It does if the Vega is more powerful (which it is).How is that unfair?
1700X and 1800X are basically the same. 3.4GHz/3.8GHz (3.9 with XFR) vs 3.6GHz/4.0GHz (4.1GHz). That's only a 200MHz difference in clock speeds. Doesn't make any sense a CPU bottleneck for Vega but none for the Fury X.
This would explain the rumored poor mining performance480GB/sec isn't stellar, but it's as much as the 1080 Ti has.
Could 8-Hi HBM2 have really bad latency?
Can you imagine selling a 550mm2 card lower than what Nvidia sell a 314mm2 GTX 1080 for? Think about that for a moment. Nvidia can almost make 2 GTX 1080s for the same cost as manufacturing 1 Vega RX.
Are you suggesting amd didn't know the gaming performance was bad when they first debuted the card and have been doing damage control since?
And since when did we take everything amd says at face value.
From all I've seen ok the Nvidia side anyway, pro gpus still game well.
.
Amazing that so many people here can look at Nvidia's Pro vs Consumer GPUs and ignore the massive difference the drivers have but ignore it for AMD's side.
Why do you look at gaming benchmarks, and not see the difference in compute performance benchmarks?[omg]
http://cdn.wccftech.com/wp-content/uploads/2016/12/NVIDIA-Pascal-Quadro-P6000_Hitman-2016.png[/img]
Too bad the actual data begs to differ. You're wrong on this matter.
Because this driver does not have the gaming optimizations for the Vega architecture. As for pcper lets say they have no knowledge in the subject of Vega GPU architecture design or Vega GPU driver design.
Have you considered that Vega FE will also get the driver updates? Why would that piss off buyers of Vega FE?
The speculations here are insane, let's stick with facts instead. Here is a dose of reality.
The one caveat to this is that the Vega architecture itself is still unoptimized in the driver. But this would affect gaming and non-gaming workloads most of the time.
Woof so much to catch up on. That die is HUGE!!!! Pretty sure I remember, and other's have said, Raja mentioned it was smaller?
Vega Crashes during blender benchmark Hundreds of people piling in with their incredibly witty and original observations about 'poor volta, wait for navi, another fail for AMD' etc etc
After this particular 'truth' has gotten the whole way around the world, this appears:
Crashes are caused by a bug in the benchmark, not the gpu/drivers, fix available
The point of the post is that a p6000 is keeps up with a titan xp in gaming despite being a pro card.Why do you look at gaming benchmarks, and not see the difference in compute performance benchmarks?
Which was exactly the bloody point of post you quoted.
Here, help yourself.
https://www.pcper.com/reviews/Graph...B-Air-Cooled-Review/Professional-Testing-SPEC
How come 1792 CUDA core P4000, is only few % slower than Titan Xp, which should twice as fast?
So then you are responding to the post with completely different thing.The point of the post is that a p6000 is keeps up with a titan xp in gaming despite being a pro card.
Stop making excuses because the card is a pro card.
Doss that make sense now?
So then you are responding to the post with completely different thing.