AMD Vega MegaThread! FAQ and Resources

Page 6 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
When I say developed with I mean the hardware implementation.

It's basically Direct3d 12 baked into the GPU Command Processor.. you guy's are probably right. I guess it is probably wishful thinkin on my behalf.
You make a good point though an It looks like they may be getting themselves into a similar situation with the VR Wifi. They will be the only other company developing all the other pieces of the puzzle.

I wonder will they do this for Vulken too ?

If this is baked into a Polaris GPU it may be baked into Vega, probably also wishful thinking but we live in hope eh.
 
i don't know about Vega but if Polaris have something similar then AMD would have mentioned it. AMD ACE for example that used in AMD Async compute implementation we know that they exist since the very first GCN.
 
I think that new "unit" is something akin to the Cell design. You have one strong core that is in charge of dispatch and smaller ones to do specific tasks. Well, not the *exact* same, but smells like they are doing an "orchestra" design. I'm not fond of those TBH. They are very short lives and have horrible cross-compatibility.

That being said, you don't need to have changes in the GPU nor CPU components, since it should be "transparent" for them. Or at least, just get a few instructions telling them how to react to issued instructions to other components. But this is just a *really* general view on how it should be. The devil, as usual, is in the details.

Cheers!
 


I think the Polaris in Scorpio is the first GPU on the market to have this baked in..does anyone know when AMD started work on it.. What I mean to say is would they of had time to implement it in Vega...

Unless of coarse this is part of the recipe for their "secret sauce" :)

It's possible I guess (that it is in Vega).. that's a given..

I can hope against hope that we don't have to wait for ages to see this in a PC... !

Unless of coarse as I say they include it in the secret sauce recipe for vega that is... would we know about it if they just did it an didn't tell everyone how they achieved the numbers they are getting ?

Edit:
Now that I think about it, I guess technically AMD is already licence'd to use Directx 12 an all it's runtime component's... Does it really matter weather they implement it in software via a driver or in hardware via the GPU Command Processor..
I would of thought that legally it would be up them how they took advantage of it.
Unless they have signed some sort of a binding contract that ties their hands on how they use Direct3d 12.. Otherwise I think that it would be up to them entirely, how they implement it in 'their' hardware or software for that matter...
 
we have to remember this is custom design made by AMD and MS. this solution is not done by AMD only. it is possible that this new "GPU command processor" end up being xbox exclusive. since MS have their hands on developing this new hardware unit i don't think they will simply going AMD to have it as their own.
 


I'm not saying just AMD will have rights to use it Although it is possible if they halped develope it this could happen I wouldn't rule it out..
They are defo the first anyway. I would presume Nvidia could do similar solution maybe... Unless MS want to sell it seperate to Directx12's ULA.. Otherwise it is covered already in each company's rights to implement Direct3d however they see fit, in hardware or in software..(unless AMD or MS or both own IP on this tech).. Whats the diff both companys are licence'd to use this already..

Also it does seem like the next evolution of the technology.. an they next logical step.. It would be nice if PC users got this.. who's to say it will not be included in the "secret sauce" in Vega (or the next GPU) !! We live in hope...

We Shall See.. in time I guess.

Edit:
Also if they (again AMD or Nvidia or Both) not aloud legally to add Direct3d 12 to the GPU in hardware they will simply do the very same thing but instead Vulken will be used (an as a result Vulken would become far superior to Directx)... An I can't see Microsoft letting this happen as Vulken would likely become the Standard API being used in games across the board in PC's...

This reason alone should force Microsoft to allow it to happen in Directx first !!

No matter what way I look at it.. this is the next step for sure.. just a question of which API gets in there first with this solution an becomes the standard.

I'd say this will happen in both Directx and Vulken and more than likely in that order as well..
 
when it comes to low level vulkan definitely better than DX12 in that regards. we see massive performance boost with doom but it doesn't mean we going to see the same improvement in every other vulkan based games. remember doom is coming from developer that has dedicated decades of their time using OpenGL. if there is anything that can sway game developer from using vulkan it will be the same reason why many developer avoiding OpenGL in the first place.
 


ROFL!!
 
The first article you linked is based on the second one. But obviously it hasn't launched ot Tom;s and everyone else would have huge articles about it.

But there are issues on the specs as well. They put GCN 1.3, the RX 480/580 is GCN 4, so Vega is supposedly an older architecture?

However, based on that supposed performance, it would be the Vega 11 and designed to beat the 1070, which they surmise it does by 11%.

Looks like it was an April fools joke.
 
Hopefully your right..

As the number's don't look great.. So I hope it is a hoax or the Vega 11.. The first article say's 20 hours ago.. so pretty recent ?
 
The first article is commentary on the TPU page. The TPU page was probably an April 1st thing.

I'm not trying to be negative, just trying to keep things in perspective. I expect Vega to at least be competitive. I'd love AMD to come out swinging with a GPU that crushes the 1080ti, but after getting into the hype in the past, I've learned to not get to wound up. People keep over-hyping AMD's products and end up disappointed when they are comparable instead of crushing. I'll get excited after the benchmarks come out and there is something to celebrate.
 


it could be. i still remember back in 2012 that TPU make some joke about 7990. they show the card spec on GPU-Z (manually edited by Wiz) and in less than 24 hours various websites picking it up as evidence that 7990 is exist and might as well in reviewers hand.
 

No, AMD simply never released a versioning scheme for their iterations of Graphics Core Next - they usually refer to it by its code name (Southern Island, etc.).
Going by the point scheme, Polaris is indeed the fourth revision:
GCN "1.0": Southern Islands (Pitcairn, Tahiti etc.) - the original.
GCN "1.1": Sea Islands (Bonaire, Hawaii) - a major revision that introduced the ACE and the Graphics Command Processor, allows Async Compute to be very efficient.
GCN "1.2": Volcanic Islands (Tonga, Fiji) - better performance with tessellation and texture compression.
GCN "4": Arctic Islands (Polaris) - process shrink only.
Vega is actually the 5th generation.

 


You pretty much proved my point, Vega is not GCN 1.3 as the article indicated. But, Polaris was not just a smaller node, it's a significant architecture revision of Tonga.

Vega should be GCN 5, it will be interesting to see what it brings.
 
A bit off-topic, but interesting none the less!

Today I got the "want to participate in the Steam survey?" question and it crossed my mind to actually take a look and see how things are for both camps.

Surprise, surprise. The trend is not what I expected, like at all. But at the same time, it's not really surprising:

http://store.steampowered.com/hwsurvey/videocard/

Now, this is just Steam and all arguments you want to make saying it's not the whole world might be valid, but it is undeniable that this is a very valid source of information and it is very important as well.

I wonder if Vega will make a dent if things keep going like that.

Cheers!
 


for every 10 cards being sold out there seven of them is nvidia? will be interesting to see the market share for Q1 this year. right now AMD and nvidia are entering new round of stalemate. the last time there is noticeable change in market share was Q2 2016 where AMD gain quite significant share back. polaris was supposed to continue that momentum but it seems nvidia are successful in their effort from letting AMD to gain more market share by launching GTX1060.
 


I don't understand what you quoted nor said. Can you clarify, please?



There's no stalemate IMO. Cruel reality is AMD really needs a hands-down winner. Being on par with nVidia is not enough it seems (talking about RX480 and Vega next).

Since they can't get into a price war like they did with the HD4K series, they'll have a tough time it seems. Brand perception is giving them a beating like never before.

They won't get a hands-down winner if they keep on pushing GCN or at least alleviate it by making new APIs more predominant. It's such a hard balancing act on where to put the little resources they have, it's scary. Making a better GPU getting rid of GCN like I would like would be the most expensive approach and not even necessarily the best one; pushing for better usage of the new APIs where their GCN uArch is strongest seems like the best path forward, but brand recognition would still be present there; that is to say, nVidia just needs to be on par there (giving you the point with the 1060) to sell a bunch more.

It's really sad, even more because as of late, AMD has been making really good improvements all around their Software ecosystem; not only in the hardware department.

Cheers!
 
Thing is, with how DX11 is made, Nvidia is pretty much impossible to beat with its current chips - AMD built Mantle and GCN together to show how much of a dead end DX11 (and OpenGL) were, with wasted resources and bottlenecks all around, and is now pushing GCN and DX12/Vulkan to remove as many of them as they could.

I'm impressed with AMD for managing to change the playground so much (they moved both Microsoft and Intel over to their side with their low-overhad APIs and managed to force Nvidia to leave their comfort zone) and actually manage to provide worthwhile products in an environments where juggernauts with ten times their budget just lug their weight around to win developers over. Managing to duke it out in GameWorks titles and kicking ass on games targeted at their architectures (Doom, DE:MD) is quite an accomplishment when they don't have half the R&D and production capacities their competitors have.
 


Sorry that was the percentage of DirectX 12 compatible GPU's on steam over the last few months. Showing it's growing.
 
i mean stalemate in marketshare. before it was 65/35. they were stuck in that position for quote sometime until AMD market share dip down to 28% for the first time at the end of kepler era. then AMD start gaining back the lost share but the momentum stop once AMD reach 30% of market share. now the position is 70/30 for two quarters straight already. some people try to look it into more positive for AMD by saying AMD able to keep that 30% share despite nvidia have the high end for themselves for 6 months. but if you look it from the other angle in two consecutive quarters for every 10 cards being sold 7 of them is nvidia card. not bad as before but still.....look quite bad isn't it? the fact that we see massive deal on RX400 series recently that's most likely to encourage sales. if such pricing really did affect nvidia sales i think we probably would see the same deal on those 1060s. 1060s pricing did not change much probably point to good sales that further price drop is not needed. as much as many people want to see price war happening between nvidia and AMD i don't think that would be very good for AMD when we think about it for the long term.
 
Status
Not open for further replies.

Latest posts