How many ACEs does Maxwell 2 have?

[Pwndenburg]

Don't know, but I hope this does launch AMD right back into the game. Nvidia doesn't have much time to fix a lot of stuff if this data is true. That would be awesome to have competition and sane pricing again.

 

[zlatan]

I remember reading in another Anandtech article that pre-Maxwell 2 HyperQ was not compatible with DirectX 12 Asynchronous Compute, though.

That's not perfectly true. The multi-engine feature is also asynchronous inside the compute engine, so a hardware may not support async copy, and async shader, but it might be able to run two or more asynchronous compute pipeline simultaneously, if there is no graphics or copy pipeline.

 

[PhonakV30]

For the current specs no. Microsoft will need to extend the multi-engine compute with fine-grained preemption support, and the API must support latest data latch for optimal head tracking. Only Mantle support these right now, but a standard will surely evolve in the future.

Can Nvidia use Nvapi for VR?

 

[happy medium]

Since NV controls 70% of the market the developers will adjust ACE usage to NV. This means in principle that the potential more ACE power of AMD will have no measureable effect till NV decides to increase ACE on their hardware. So Business as usual.

This is correct, so I don't understand what all the fuss is about.
and its 76.4% of the market.

 

[happy medium]

Maybe it explains NV being quiet on this front, until Pascal.. then it's all aboard DX12 & VR train!

or they have mothing to worry about and they don't feed the trolls?

 

[Sabrewings]

or they have mothing to worry about and they don't feed the trolls?

I think there's more to it than that. They have been extremely silent on this. I think they were taken back a bit by what happened and they expected a better result. That tells me they think they can make it work and are letting the engineers do their thing to figure out why and how.

It probably doesn't matter either way. By the time AC is that heavily relied upon that a glaring lack will make or break you, the card is probably too old to get much out of anyway.

I think we'll have a clearer picture of what's going on with Maxwell 2 and AC in a month or two.

 

[tential]

or they have mothing to worry about and they don't feed the trolls?

What good could NV have from responding to this issue until they're ready?

So a couple of forum warriors don't purchase their cards due to overreacting before any games are even out with DX12. NV has time to work on solutions and work arounds.

It won't stop people from reading into NV's silence as much as humanly possible.

 

[psolord]

Maxwell 2 has one ace....under its sleeve! 😀

 

[Silverforce11]

or they have mothing to worry about and they don't feed the trolls?

When NV didn't have tessellation: "Tessellation isn't important"
When NV had better tessellation: "Tessellation is very important, use it to the max!"

History, it aint been that long.

 

[Keysplayr]

When NV didn't have tessellation: "Tessellation isn't important"
When NV had better tessellation: "Tessellation is very important, use it to the max!"

History, it aint been that long.

Always have to just give the compost pot just one more stir, eh? Both of you.

 

[Techhog]

Since NV controls 70% of the market the developers will adjust ACE usage to NV. This means in principle that the potential more ACE power of AMD will have no measureable effect till NV decides to increase ACE on their hardware. So Business as usual.

This is correct, so I don't understand what all the fuss is about.
and its 76.4% of the market.

or they have mothing to worry about and they don't feed the trolls?

So, what you guys are saying is that Nvidia is holding back PC gaming.

 

[Vaporizer]

Exactly. Same was true for DX11.1. All said this does not matter (because some vendor) did not support it. It already had for example tiled ressources, which is now (that also some vendor supports it now) considered as relevant.

 

[bononos]

I think there's more to it than that. They have been extremely silent on this. I think they were taken back a bit by what happened and they expected a better result. That tells me they think they can make it work and are letting the engineers do their thing to figure out why and how.

It probably doesn't matter either way. By the time AC is that heavily relied upon that a glaring lack will make or break you, the card is probably too old to get much out of anyway.

I think we'll have a clearer picture of what's going on with Maxwell 2 and AC in a month or two.

Yeah I think they are still shell shocked at this point.
It would be good if this means the older GCN1.0 cards get a free boost in dx12.

 

[TheELF]

You are sure that no iGPU can play The Witcher 3 at acceptable framerates?:sneaky:

https://www.youtube.com/watch?v=WvUu-MHZxMg

With a 30% overclock on the igpu from 720 to 1020Mhz and still only 1024x768@30 all on low.
Sure you can play it but it's not what most people will consider acceptable.

 

[TheELF]

So, what you guys are saying is that Nvidia is holding back PC gaming.

No, nvidia can utilize 100% of the async computing potential of their cards with only one "ACE" ,AMD needs 8 of them to get up to 100%

 

[Glo.]

No, nvidia can utilize 100% of the async computing potential of their cards with only one "ACE" ,AMD needs 8 of them to get up to 100%

Not exactly. In DX12 the GPU is 100% loaded regardless if you have Asynchronous compute or not. ACE help managing the pipeline with out-of-order commands. That is all what ACE does.

Everything depends on that how well the application is coded. It still can not utilize all features of GCN, even while doing 100% of work on the GPU. Nvidia did good job here. Maxwell architecture is pretty simple to code for. With GCN its not the case.

But if you will understand what you are doing, with what hardware, the results will pretty much mirror the differences in compute power of GPUs. So 8.6 TFLOPs Fury X will be quite faster than 6.7 TFLOPs Titan X. Only - if you will understand what you are doing with what type of hardware.

From my perspective, best way would be both worlds in one design. Easy to code, extremely complex architecture with a lot of compute power and graphics capabilities.

 

[gorobei]

the financial reporters are estimating the first gen of hardware at roughly 1M cv1 units for oculus, and a hodgepodge of htc vive, newly renamed sony PS VR(formerly morpheus), hololens, gear vr for a total number just under 10M.

most people's first experience will likely be gearVR or google cardboard.

the hardcore VR gaming contingent will go rift or vive. certainly all the Elite and SC space jocks will be gaming on release day.

the wildcards will be ms hololens and sony psvr. they could make or break the inbetween markets, which may sour people on the vr experience.

castAR may be a darkhorse in the casual market. a group of friends sitting around a table in a club with a shared 3d game could really change perceptions.

 

[Skurge]

So Gears of War is going to use Async Compute. Not sure to what extant though, but since it's a game coming from Xbox One it might not be that much.

So that makes all the currently announced DX12 titles.

 

[zlatan]

Can Nvidia use Nvapi for VR?

Not just can, they must use it. But NVAPI is not a real API, it's just a services library for the existing graphics APIs. It won't allow such a low-level optimization what AMD can do with Mantle, with API-to-API interoperability.
The primary reason why AMD don't use AGS Lib for LiquidVR, is the natural limitations of these kind of services extensions. It won't allow them to expose Latest Data Latch for example.

 

[flopper]

So, what you guys are saying is that Nvidia is holding back PC gaming.

Have been like that for years.
Nothing new from them.

AMD innovate.

 

[Erenhardt]

No, nvidia can utilize 100% of the async computing potential of their cards with only one "ACE" ,AMD needs 8 of them to get up to 100%

Technically your right, nvidia can utilize 100% of their async computing potential with 0 aces. But it may have something to do with their async potential equals 0.

Not doing something when you can't do it is doing it to 100% of your potential. Hard to argue with that line of thinking.

 

[TheELF]

Technically your right, nvidia can utilize 100% of their async computing potential with 0 aces. But it may have something to do with their async potential equals 0.

Not doing something when you can't do it is doing it to 100% of your potential. Hard to argue with that line of thinking.

That is the same point everybody else is making though,amd has so much compute potential because it does not get used otherwise,they have 50% of (otherwise) uselless stuff and argue that you gain 50% if you use this (otherwise) useless stuff.
How is it better to have useless stuff on your card?

 

[Headfoot]

That is the same point everybody else is making though,amd has so much compute potential because it does not get used otherwise,they have 50% of (otherwise) uselless stuff and argue that you gain 50% if you use this (otherwise) useless stuff.
How is it better to have useless stuff on your card?

If you're trying to say there is 0 idle time in an nVidia gpu while gaming, you're wrong. Only their engineers truly know how much time is spent idle across most games

 

[flopper]

Nvidia, not enough.

 

[TheELF]

If you're trying to say there is 0 idle time in an nVidia gpu while gaming, you're wrong. Only their engineers truly know how much time is spent idle across most games

No I'm not, I'm only talking about compute hardware (the actual parts that do the computing) ,but there are plenty of other areas that are sitting idle and will be utilized by dx12.