5970 Reviews

[Tempered81]

It's not a 'bug' it's an inherant drawback to all Multi-GPU AFR setups, 2 cards, 1 card (2GPUs), both vendors.

+1. this is a perfect explanation. I watched a video of 5850xfire vs. 5970 vs. gtx295 vs. 5870.Makes me want to have a single gpu even more. I dont care if its 50 fps multi-gpu vs. 35 fps single-gpu, if the 35 fps is smoother.

Here is the video i watched comparing the gameplay experiences:

http://www.youtube.com/watch?v=emG7ZNIsxw8

(watch in HQ. the 5870 looks smoother even with lower fps.)

 

[Grooveriding]

+1. this is a perfect explanation. I watched a video of 5850xfire vs. 5970 vs. gtx295 vs. 5870.Makes me want to have a single gpu even more. I dont care if its 50 fps multi-gpu vs. 35 fps single-gpu, if the 35 fps is smoother.

Here is the video i watched comparing the gameplay experiences:

http://www.youtube.com/watch?v=emG7ZNIsxw8

(watch in HQ. the 5870 looks smoother even with lower fps.)

I'm going to have to disagree with this. My 5870 vs when I got the 2nd one is a better overall experience. Although there are only two games that couldn't already be maxxed out on just 1 card.

 

[Tempered81]

Some folks want multi-gpu for the faster FPS. I've had multi-gpu and single-gpu setups. My GTX260 SLI was very fast and actually quite hassle-free, but I never had a GTX285 OC/SSC to compare to. I was always a proponent of multi-gpu, but I'm slowly becoming a fan of fast single gpus these days. Maybe it's just simpler and I enjoy that.

There are a lot of people on this forum that still prefer single-GPU rigs.

 

[SlowSpyder]

I am not confused. Ageia PPU was a PCI card released 4 years ago. It was meant to offload physics calculations away from CPU. Now it can't handle the load of newer games so whether new driver should still support it is debatable as it can't handle the load of newer games anyways. However, Ageia PhysX PPU still works.

It's not a question of if it's powerful enough. Nvidia DID remove it's functionality as well (I'm not sure if you are aware of that, but Nvidia 'turned off' Aegia PPU's just the same as they did with their graphics cards as PPU's when an AMD graphics driver is detected) A piece of hardware that was sold with only that purpose, Nvidia removed it's ability to run Physx on AMD hardware. There won't be much uproar over it as I'm sure very few people use an Aegia card with their AMD part, and at this point not having Physx isn't a game changer.

Nvidia video card is being sold as a video card that support physX, not being sold as a PCI-E card that offloads PhysX from the video card. Yes you are use Nvidia card as a PPU, so the mindset is Nvidia does support it. Now you know it doesn't. On the surface, those who use cross-over configuration are screwed. In practice, it still works after modding.

I wish to stop here as we can't get to the bottom of this. We have different point of views, and I clearly see your point and I don't disagree with your POV. This topic itself is very arguable.

That's fine, we can agree to disagree. But the one other point I'll make is that I think Nvidia did indeed sell their parts as optional stand-alone Physx cards. In the driver you have the option to dedicate a card to just Physx, so that would seem like a function they meant for their cards to be used as. Otherwise why would you be able to choose to use just a 9800GTX+ as your Physx card to be used with your GTX280? (just using two cards as an example) Nvidia gladly sold parts knowing that they were being used for that function with AMD and other Nvidia cards, than took away that functionality. If they didn't want their cards used as a stand-alone Physx part, why did they put that option in the driver? It seems to me that they did in fact purposely sell their cards with that functionality in mind, but never specified that you had to use an Nvidia card only system.

Let me answer your question by asking you a question.

4000 series is capable of supporting DirectCompute, but is disabled. Why doesn't this seems to be a problem as users are being forced to buy 5000 series for Dx11, but are very upset about AA and PhysX?

Because the 4000 series may support some functions of DX11, but were never DX11 compliant. A better analogy would be AMD removing DX10.1 functionality from the Radeon 4xxx cards when they are used in an Intel system. If from day one, Nvidia did not allow their cards to be used as PPU's with AMD cards that would be one thing. But the fact it that it worked fine, Nvidia gladly sold GPU's with that functionality and never said that it was not supported or would not be supported, took your money, than changed the spec after they sold parts.

 

[Seero]

That's fine, we can agree to disagree. But the one other point I'll make is that I think Nvidia did indeed sell their parts as optional stand-alone Physx cards. In the driver you have the option to dedicate a card to just Physx, so that would seem like a function they meant for their cards to be used as. Otherwise why would you be able to choose to use just a 9800GTX+ as your Physx card to be used with your GTX280? (just using two cards as an example) Nvidia gladly sold parts knowing that they were being used for that function with AMD and other Nvidia cards, than took away that functionality. If they didn't want their cards used as a stand-alone Physx part, why did they put that option in the driver? It seems to me that they did in fact purposely sell their cards with that functionality in mind, but never specified that you had to use an Nvidia card only system.

The coding is very simple apparently. "Disable PhysX when there exists a non-Nvidia's video card present."
We both know why it is in place, though we have different opinions on the reasoning behind. I don't have cross-over configuration so I really don't feel the pain. I do however understand that people on that boat feel let down. Yes, the disable of function is harsh, but necessary to stop people from using cross-over configuration.

Now the ageis PPU disabling is a bug and Nvidia have to fix it. However, the product itself is 4 years old...

Let me say it your way. Nvidia did hurt those who use cross-over setup and Nvidia will try to illuminate the configuration for good.

Because the 4000 series may support some functions of DX11, but were never DX11 compliant. A better analogy would be AMD removing DX10.1 functionality from the Radeon 4xxx cards when they are used in an Intel system. If from day one, Nvidia did not allow their cards to be used as PPU's with AMD cards that would be one thing. But the fact it that it worked fine, Nvidia gladly sold GPU's with that functionality and never said that it was not supported or would not be supported, took your money, than changed the spec after they sold parts.

This more or less summarizes your stand.

In my words:
If any user found any methods to use a program that are not supported, then programmer must continue to support these new methods forever because customer did paid for it.

If Nvidia allows user to use a Nvidia video card as a PCI-E can handles PhysX only, then why can't user use Non-Nvidia Video card to handle display?

Why does Nvidia disallow something that they never said is not allowed?

Now it is your turn to summarize mine. :biggrin:

 

[nitromullet]

+1. this is a perfect explanation. I watched a video of 5850xfire vs. 5970 vs. gtx295 vs. 5870.Makes me want to have a single gpu even more. I dont care if its 50 fps multi-gpu vs. 35 fps single-gpu, if the 35 fps is smoother.

Here is the video i watched comparing the gameplay experiences:

http://www.youtube.com/watch?v=emG7ZNIsxw8

(watch in HQ. the 5870 looks smoother even with lower fps.)

I'm going to have to disagree with this. My 5870 vs when I got the 2nd one is a better overall experience. Although there are only two games that couldn't already be maxxed out on just 1 card.

This isn't exactly what he is saying. I think he is making the comparison between a multi-gpu setup vs. a single gpu setup with roughly the same performance capabilities. Sure, if you take a single gpu that is already on par with a competing dual gpu setup and use two of them you will get better performance. That's a no-brainer because there isn't really anything out there that pushes 5870 Crossfire.

 

[akugami]

I'll openly admit ignorance here to the details or the backstory...but the first question that popped into my mind when reading this portion of your post is do we know whether or not:

1. TWIMTBP games make Nvidia look better because they, presumably for sinister reasons, have been optimized to perform poorly on AMD architecture?
2. TWIMTBP games make Nvidia look better because they, presumably for acceptable reasons, are better optimized for execution on Nvidia's architecture?

If AMD assists a game developer and the resultant game does not demonstrate some kind of performance bias (owing to optimization made by AMD for better code exectution on AMD's architecture) in favor of AMD's own products then I really have to question the value of AMD's program in terms of ROI for the shareholder equity that the program is consuming.

I think the second is more likely due to the efforts of nVidia who provide more in the way of developer support in the form of having coders look at the code and optimize it for their architecture. Granted we do have a little grounds for suspicion that there has been a little hanky panky going on considering the Assassin's Creed DX10.1 removal as well as the Batman AA issue. I'm more inclined to think it is the second and that the developer just work less on optimizing it for ATI rather than anything sinister.

Keep in mind that I don't condemn nVidia for any of this. As long as a company keeps things within the letter of the law, they can do a lot of stuff that would be deemed dirty. A sound business decision may not be the best for consumers but people who are going to comment on it need to be aware of both sides of the coin. I have never, from a business standpoint, thought that nVidia and their Batman AA shenanigans was a bad idea. It's just the consumer in me who feels it is bad.

AMD's program, called Get In The Game (GITG), seems to provide funding more than anything and they do get a cool splash screen on the game much like nVidia's program. I think the developer has some leeway to use the monies for general development and you'd be hard pressed to convince me some of it isn't used to optimize the game for nVidia cards. It is quite difficult once you give the money away to properly track what is being done with it. Heck, IBM took Sony's money that was meant to be used for developing Cell and used the money to work on the CPU for the Xbox 360. So yes, I'd agree that the way AMD is going about their program also makes me question the ROI. It probably isn't the best use of AMD's limited resource but we have some members here still claiming AMD does not help developers or have a program similar to TWIMTBP even after they have been told about the GITG program.

AMD's GITG program should also be the arm that is helping developers on DX11. Many of them probably won't officially be GITG games but it behooves AMD to help developers get the games out ASAP when nVidia does not have a DX11 part out yet.

 

[lopri]

AMD tossed out that GITG program a long times ago, iirc. They must have something else but I don't know what it's called.

 

[PingviN]

They must have something else but I don't know what it's called.

"WHDX11HWYD"

We Have DirectX11 Hardware, You Don't

 

[NIGELG]

"WHDX11HWYD"

We Have DirectX11 Hardware, You Don't

 

[T2k]

+1. this is a perfect explanation. I watched a video of 5850xfire vs. 5970 vs. gtx295 vs. 5870.Makes me want to have a single gpu even more. I dont care if its 50 fps multi-gpu vs. 35 fps single-gpu, if the 35 fps is smoother.

Here is the video i watched comparing the gameplay experiences:

http://www.youtube.com/watch?v=emG7ZNIsxw8

(watch in HQ. the 5870 looks smoother even with lower fps.)

This is nonsense. My 4850 X2 2GB is a LOT faster and smoother than any 4850 or 4870 I have ever tried.

 

[nOOky]

I'd have to agree, I had crossfired 4870's and it was much better than a single card. I can't imagine lower fps causing reduced smoothness, unless it's a driver bug.

 

[Tempered81]

Well sure it's faster on the 4850X2. However, take for example 5870 vs. 5970 or GTX285 vs. GTX295. Some people would rather have the 5870 and 285, and some would rather the dual-gpu cards.

If you watch this video (http://www.youtube.com/watch?v=emG7ZNIsxw8) in HD, pay attention to the twenty seconds between 1:20 - 1:40. The 5970 is probably running 50-60 fps while the 5870 is probably running 35-40 fps, but single gpu rendering looks a LOT smoother than the AFR.

 

[Munky]

If it worked just fine meaning that it works on previous drivers, then you don't need to update it as that driver is fine. If the new driver causes problems, don't use it.

The cross-over setup works on old drivers, but it was never a supported configuration. They stopped this configuration because it is not supported. You may still complain if the new Nvidia driver cause graphic corruption with that configuration, but Nvidia's driver never need to work with ATI's hardware to begin with.

Just because it works doesn't mean it is supported. I can submerge my PC into oil and it will work, but it will still void all the warranties doing so. I can lap the surface of my CPU for better cooling, but it will still void the warranty. You found that you can have ATI card for display while having a Nvidia card for physics, it works, but not supported.

Haha, that's like saying Mac OS is not supported on x86 cpu's unless it came bundled with an Apple logo. Anyone with a bit of sense can tell that there's no technical reason why it's not supported, it works just fine, except that Nvidia want's you to keep buying their cards even when those are doing something completely unrelated to physx.

 

[nitromullet]

I'd have to agree, I had crossfired 4870's and it was much better than a single card. I can't imagine lower fps causing reduced smoothness, unless it's a driver bug.

Again, I think some of you are taking this out of context. Yes, two cards are generally faster and smoother then a single card in situations were the single card can maintain decent FPS on its own. However, when you compare a single card with similar performance to a competing gpu setup in a situation were both setups are stressed, generally the single card setup will feel smoother even if the fps are similar.

 

[Seero]

Haha, that's like saying Mac OS is not supported on x86 cpu's unless it came bundled with an Apple logo. Anyone with a bit of sense can tell that there's no technical reason why it's not supported, it works just fine, except that Nvidia want's you to keep buying their cards even when those are doing something completely unrelated to physx.

Is it wrong?

I don't see MS running on Mac.

 

[cmdrdredd]

"WHDX11HWYD"

We Have DirectX11 Hardware, You Don't

Show me the games, GOOD GAMES. That is all for now.

 

[cmdrdredd]

Dude. I was showing how Wreakage selectively quoted a paragraph go look at it at [ H ].

It doesn't matter. Half this thread is bitching about the program. Devs CHOOSE to enter that program because it's basically free help to make their life easier. You might not like it, but that's the breaks.

 

[nitromullet]

Is it wrong?

I don't see MS running on Mac.

Windows runs on a Mac just fine now that Apple has switched to Intel.

http://www.apple.com/macosx/compatibility/

...even Windows 7 (officially supported by Apple) soon.

http://support.apple.com/kb/HT3920

 

[Seero]

Windows runs on a Mac just fine now that Apple has switched to Intel.

http://www.apple.com/macosx/compatibility/

...even Windows 7 (officially supported by Apple) soon.

http://support.apple.com/kb/HT3920

If AMD is willing to pay for PhysX, and QA on their parts, then we can all enjoy this new tech. I am not saying they should or should not, i'm saying they have not.

 

[GaiaHunter]

If AMD is willing to pay for PhysX, and QA on their parts, then we can all enjoy this new tech. I am not saying they should or should not, i'm saying they have not.

Why should AMD pay so NVIDIA can sell their cards has PhysX accelerators?

NVIDIA with this move just decided it doesn't want to sell PhysX accelerators - they want every bit of ammunition to fight ATI in the graphics accelerator department.

If you are so inclined to believe that NVIDIA are angels and the main reason to drop support is QA and what not, it is up to you.

QA is a valid excuse, but only that - an excuse.

Those of us that understand that NVIDIA is just a company that is driven to make a profit can see the true reason behind it - they are atm 2nd place in performance.

AMD also has valid reasons not to support PhysX - they don't want to pay their competitors and depend on them to control a feature of their cards.

For the consumer, the best is that they come with a solution so accelerated physics (both on CPU and/or GPU) can become something that it isn't at the moment - relevant and game experience changing.

 

[Munky]

Is it wrong?

I don't see MS running on Mac.

You mean you never heard of Bootcamp that allows a Mac to run Windows? It's just business as usual for Nvidia, but you'd have to be either a fanboy or just ignorant to buy obsolete, inferior video cards from Nvidia just because they want to lock you in using gpu-accelerated physx as leverage.