ForgedReality wrote on Feb 25, 2015, 18:14:

Rigs wrote on Feb 25, 2015, 10:19:

ForgedReality wrote on Feb 25, 2015, 10:00:
There's no reason to do that these days though, since AMD can't seem to compete, and nVidia is constantly forcing them to drop their prices.

Psh, bite your tongue! There are ATI/AMD fanboys here...I know because I'm one of them. AMD might not be able to compete CPU-wise, but their videocards still give Team Green a run for their money, no doubt. I'm sure a lot of Radeon users would agree. My HD7850 easily stands up to today's games, even AC: Unity, with only some minor ticks down on the ol' graphics slider...


=-Rigs-=

My GTX580 is two years older and it still outperforms...

http://gpuboss.com/gpus/Radeon-HD-7850-vs-GeForce-GTX-580

Mine is not quite as old, but still outperforms

http://gpuboss.com/gpus/Radeon-HD-7850-vs-GeForce-GTX-670

And uses 90W less in the process.. per hour.. per day.. per year..
AMD builds GPU's with heating function... nice service for winter I guess...

Slick wrote on Feb 25, 2015, 20:00:

RaZ0r! wrote on Feb 25, 2015, 19:34:
Sounds like LucidLogix Virtu MVP

indeed. that was my first thought.

but 2 cards working on THE SAME FRAME? that sounds like a bridge too far. I'd love to see it work, but from what i know ATI and Nvid use very different architectures to achieve the performance relative to each other. It would seem like a real fustercluck to get them to be working on the same frame. does the CPU only spit out the geometry and z-buffer from the bottom half of the scene to one card? and have to re-calculate for the top half? I'd think that there would be a lot of unoptimized shit going on.

still a nifty idea. shows progress towards true parallelization, which to me is the real goal here. However i wouldn't bank on cross-vendor cards working on the same FRAME, that seems like a long bet. the %900 i'm still onboard with though, if a game was %900 more demanding on the CPU side of things using a CPU-optimized API that is...

They may be talking about VR related frame rendering, where technically you could very well render the same "frame" but 2 different viewpoints with different GPU's (but why?????)... and fix the bandwidth problem in 1 fell swoop (2 cards = 2 HDMI cables to VR goggles)

But even if this comes, still easier to just buy 2 identical cards if you really wanted this.. no?

MS is trying to show they doing what they should be goind with direct x 5-10 years ago... yea lazy

RaZ0r! wrote on Feb 25, 2015, 19:34:
Sounds like LucidLogix Virtu MVP

indeed. that was my first thought.

but 2 cards working on THE SAME FRAME? that sounds like a bridge too far. I'd love to see it work, but from what i know ATI and Nvid use very different architectures to achieve the performance relative to each other. It would seem like a real fustercluck to get them to be working on the same frame. does the CPU only spit out the geometry and z-buffer from the bottom half of the scene to one card? and have to re-calculate for the top half? I'd think that there would be a lot of unoptimized shit going on.

still a nifty idea. shows progress towards true parallelization, which to me is the real goal here. However i wouldn't bank on cross-vendor cards working on the same FRAME, that seems like a long bet. the %900 i'm still onboard with though, if a game was %900 more demanding on the CPU side of things using a CPU-optimized API that is...

Sounds like LucidLogix Virtu MVP

uh what is the point? now I have to install two drivers... not seeing the "benefits" of this. Fat chance of this actually working as advertised.

Rigs wrote on Feb 25, 2015, 10:19:

ForgedReality wrote on Feb 25, 2015, 10:00:
There's no reason to do that these days though, since AMD can't seem to compete, and nVidia is constantly forcing them to drop their prices.

Psh, bite your tongue! There are ATI/AMD fanboys here...I know because I'm one of them. AMD might not be able to compete CPU-wise, but their videocards still give Team Green a run for their money, no doubt. I'm sure a lot of Radeon users would agree. My HD7850 easily stands up to today's games, even AC: Unity, with only some minor ticks down on the ol' graphics slider...


=-Rigs-=

My GTX580 is two years older and it still outperforms...

http://gpuboss.com/gpus/Radeon-HD-7850-vs-GeForce-GTX-580

This is such a narrow use case that it's never going to be worth the effort it takes to make it work properly.

Mashiki Amiketo wrote on Feb 25, 2015, 13:29:
That's why it worked perfectly fine right? If it works fine, and people don't have a problem. And then they go out of their way to disable it, all that says is they threw a hissy fit. All they have to do in a support environment is say "we don't support that configuration and will offer no support on it." Problem solved.

Well... not really. The thing is, as an Nvidia support guy, you get a call saying "My GTX970 is giving me all these weird graphical artifacts in games!"

And you spend 2 hours having the guy reinstall drivers and taking out the card and blowing the fan clean etc before they finally admit "Oh yeah, I'm also running an AMD card in there, and its driver is two years out of date."

Everyone who calls for support lies about shit like that, so I can understand from a corporate perspective that disabling it from working is a much, much easier thing to do. (Also, it's Nvidia, who would much rather they were the only GFX manufacturer in the world, and who literally give zero fucks about their customers, so...)

Mashiki Amiketo wrote on Feb 25, 2015, 13:29:

Shineyguy wrote on Feb 25, 2015, 11:27:
To be fair, their statement about it was that they couldn't support such setups as that so they disabled it in their drivers.

From a support standpoint, I can completely agree with what they did. I would not want to have to try to support a competing vendors software and it's interaction with my own.

That's why it worked perfectly fine right? If it works fine, and people don't have a problem. And then they go out of their way to disable it, all that says is they threw a hissy fit. All they have to do in a support environment is say "we don't support that configuration and will offer no support on it." Problem solved.

Do you know what kind of support call volume that nVidia had received before they disabled this in their drivers? I remember seeing plenty of forum posts about people having issues with it (as well as some posts about people having it working.)

As a person that works in a support capacity in a large retail chain I.T. department, I would never want to have to support a product that I have limited visibility on. It's a nightmare and it costs companies a lot of money to have to devote resources to a product that they do not support.

jdreyer wrote on Feb 25, 2015, 12:31:
I don't see why they would. They'd still be selling cards to people that use AMD primarily. Wouldn't a sale in that case be better than no sale at all?

Nope cause if you had a high end AMD card for primary, you'd only need a 2 or 3 generation old Nvidia card for PhysX (for example). You'd probably buy used and not new so Nvidia wouldn't make any money.

This sounds like one of those checkbox marketing items that breaks when anyone really tries to use it.

I want to believe, but...

Creston wrote on Feb 25, 2015, 14:33:
Maybe DX12 will have some amazing magic in place that would just allow you to parallel boost performance like that, but this bit does not seem to indicate so: Lots of the optimization work for the spreading of workloads is left to the developers.

I really don't think this will wind up having much, if any, effect at all.

I'm willing to bet you are correct. I don't see this as a significant market anywhere near worth what the engineering cost to make it workable in a given title would entail.

Like you say, if DX12 fully abstracted all GPU resources and the dev was completely unaware of the cards perhaps. But that seems the opposite of the "closer to the metal" approach touted. The quote you used does indicate where the expected effort lies.

It's already a massive pain to get a title working on all the normal single GPU cards out there, harder still to get it working well in SLI/CF, and sweet jesus save us all if we have to support Any/Any.

{PH}88fingers wrote on Feb 25, 2015, 13:38:
on Nvidia 970: I laughed https://www.youtube.com/watch?v=spZJrsssPA0

I laughed too, but at the electricity bill for the ATI GPU's being a selling point even for a gimped 970

jdreyer wrote on Feb 25, 2015, 12:33:

Creston wrote on Feb 25, 2015, 11:15:
I'll believe it when I see it, but it would be pretty awesome.

I think few devs would ever bother with putting in the work required, though, (even if it's not difficult) since it'd be a pretty rare occurrence where people are using a hybrid system like that.

I think a lot of people have extra vid cards hanging around they'd throw in their system if they thought they could get extra performance from it.

I don't think you really would gain anything. If you SLI or Crossfire a fast card with a slow card (if the drivers even allow it, I don't think they even do), you'd only run at the speed of the slowest card because otherwise one card would already be rendering the end credits while the other one is still working on the intermission cutscene.

Maybe DX12 will have some amazing magic in place that would just allow you to parallel boost performance like that, but this bit does not seem to indicate so: Lots of the optimization work for the spreading of workloads is left to the developers.

I really don't think this will wind up having much, if any, effect at all.

I foresee no problems whatsoever from mixing vidcards.

on Nvidia 970: I laughed https://www.youtube.com/watch?v=spZJrsssPA0

Shineyguy wrote on Feb 25, 2015, 11:27:
To be fair, their statement about it was that they couldn't support such setups as that so they disabled it in their drivers.

From a support standpoint, I can completely agree with what they did. I would not want to have to try to support a competing vendors software and it's interaction with my own.

That's why it worked perfectly fine right? If it works fine, and people don't have a problem. And then they go out of their way to disable it, all that says is they threw a hissy fit. All they have to do in a support environment is say "we don't support that configuration and will offer no support on it." Problem solved.

It was always possible to do this with certain motherboards and you could even fake drivers in some cases. Performance isn't fantastic as it's basically AFR at the slowest card's speed (similar to a RAID) but you generally get at least 60% out of them. Been capable for many, many years now so this is just a formality, like audio driver modding I guess most people aren't aware you can get a lot more out of a strange system if you know where to look.

Dagnamit wrote on Feb 25, 2015, 12:42:
Does anyone know if the XBone could support hardware upgrades? Seems to me like with the API you could sell an upgraded Xbone or an upgrade kit with out changing out the primary SoC that's already being sold. Shift the costs of upgraded games to devs, who need to code the support into the game, who can then turn around and charge $80.00 for the HD version.

It can't support any real upgrades outside of storage. They would need a whole new hardware revision for a new SoC. Hardware upgrades on consoles defeats the whole purpose of a stable hardware platform anyway, no one has really done it with notable success.