Nvidia Fermi GF100 working desktop card spotted on Facebook

[STEvil]

Given the physics enabled materials available in the engine without creating custom content that is the best "house" that can be built. Given the rigid body physics used for the structures and how long ago the engine was programmed it is impressive. Show me a better PhysX vid.

edit

I should mention that the engine has limited capabilities of crush and angle/vector physics. If you put a 200 Ton weight on top of a board it wont crush it. If you put a weight off-center the item holding it up will not deform/snap.

[Heinz68]

Nvidia gimped Havok? Please explain.

NOTHING to do with Havok. I NEVER said anything about Havoc in my reply to you. Go back and read Piotrsama reply in post #356. To me it looked like he was talking about Batman and I just continued on that subject.

Since I never said anything about Havok in my reply to you I have no idea what is Warboy trying to spell out for me.

[Heinz68]

Well, for people who really cares about games with GPU PhysX, you can always do something like this:




hmm, HD5870 + GTX275 faster than 2xGTX285 SLI at a GPU PhysX game.

Can you please post a link for the above benchmark.
Thanks Heinz

[Blacky]

Too bad it's buggy as hell.

lets see if they fixed it on the new cryengine *waits crysis 2*

[trinibwoy]

Heinz, are you even following your own posts? I asked what Nvidia has to do with what's possible on a CPU and you responded like this:

I didn't realize that Nvidia could dictate what's possible on a CPU. What's stopping somebody else from doing it? Why doesn't Havok have great fluid and cloth effects running on the CPU if it's so easy?

Well not everybody is investing big money in the TWIMTBG

What does TWIMTBP have to do with coding a physics engine? Or were you just not able to answer my question?

2. Gimped

So there was no mention of PhysX or Batman anywhere. Now how exactly does Nvidia or TWIMTBP gimp Havok's ability to run complex physics on the CPU?

[Heinz68]

Heinz, are you even following your own posts? I asked what Nvidia has to do with what's possible on a CPU and you responded like this:

So there was no mention of PhysX or Batman anywhere. Now how exactly does Nvidia or TWIMTBP gimp Havok's ability to run complex physics on the CPU?

I hope this is the last time I need to explain it.

As I said already "go back and read Piotrsama reply in post #356" here it is:

If Nvidia keeps PhysX artificially running on a single CPU thread to kill performance when no NV GPU is present, then you are right.

Batman is the only game I know about where where CPU use is limited to single thread when PhysX is enabled. Read Tom'sHardware and many others. Therefore I believed that was what Piotrsama was referring to.

When I did reply to your #357 post i highlight this part of your post "I didn't realize that Nvidia could dictate what's possible on a CPU." just to make sure I'm not talking about Havoc.

As I said before I never said anything about Havoc in my reply to you, neither was I interested to discus which is better Havoc or PhysX, but if you really want to know, for me Havok is 100% better since it works on my PC setup.

[trinibwoy]

Yes please explain it some more. I fail to see what Batman has to do with what's possible on a CPU (your bolding). Batman is just one game.

[Heinz68]

Enough time wasted on you.

[trinibwoy]

That's too bad. I was really looking forward to your explanation on how Nvidia is limiting the level of physics that's possible on a CPU. I see people making these claims all the time and at the end they can never back it up....and that trend continues.

[Humminn55]

That's too bad. I was really looking forward to your explanation on how Nvidia is limiting the level of physics that's possible on a CPU. I see people making these claims all the time and at the end they can never back it up....and that trend continues.

Are you really that dense?

CPU use is limited to single thread when PhysX is enabled.

There....happy? nVidia has the game devs write the coding so that physics is limited to running on a single thread in the cpu, eschewing multi threaded running capabilities, when a non-nVidia video card is found in the system.

Is that so hard to understand? I guess so........

[Chumbucket843]

why would nvidia make physx run fast on a cpu? they would get more much speed and sales by using gpu's. physx is also designed for high throughput processors, not cpu's so that would probably waste a lot of time and money and defeat the purpose of physx altogether. the last time i checked physx was using 60% of quadcore systems too. core i7 is misleading because you have to understand how hyperthreading works. 4 threads have the same amount of execution resources 8 threads do when HT is on so the speed is invisible to task manager. the speed up is from reduced memory latency and pipeline bubbles.

[trinibwoy]

CPU use is limited to single thread when PhysX is enabled.

Sigh, speaking of dense.....

I'm not sure why english is so difficult to understand but what does PhysX have to do with the ability of other people (i.e NOT NVIDIA) doing better physics on the CPU? I missed the announcement that Nvidia has a monopoly on physics code.

[highoctane]

Sigh, speaking of dense.....

I'm not sure why english is so difficult to understand but what does PhysX have to do with the ability of other people (i.e NOT NVIDIA) doing better physics on the CPU? I missed the announcement that Nvidia has a monopoly on physics code.

Don't you know Nvidia locks developers in their secret dungeon lair and forces them to do their bidding, developers have no choice but to comply.

[trinibwoy]

Oh, thanks I missed that development. Not surprised though...evil bastards

[STEvil]

Troll banned for a week.

[postumus]

+rep to whomever delivered that ban, guy never has anything constructive to add imo. Was about to place him on ignore anyway.

[Lokinhow]

Can you please post a link for the above benchmark.
Thanks Heinz

http://xbitlabs.com/articles/video/d...x_8.html#sect2

I hope this is the last time I need to explain it.

As I said already "go back and read Piotrsama reply in post #356" here it is:


Batman is the only game I know about where where CPU use is limited to single thread when PhysX is enabled. Read Tom'sHardware and many others. Therefore I believed that was what Piotrsama was referring to.

When I did reply to your #357 post i highlight this part of your post "I didn't realize that Nvidia could dictate what's possible on a CPU." just to make sure I'm not talking about Havoc.

As I said before I never said anything about Havoc in my reply to you, neither was I interested to discus which is better Havoc or PhysX, but if you really want to know, for me Havok is 100% better since it works on my PC setup.

there is this one too

http://techreport.com/articles.x/17618/13

Ok, so it's hard to see, but Task Manager is showing CPU utilization of 14%, which means the game—and Nvidia's purportedly multithreaded PhysX solver—is making use of just over one of our Core i7-965 Extreme's eight front-ends and less than one of its four cores"

[***Deimos***]

Wow... look at all the violence I started...

I never mentioned Havok, or wanted to start a big argument.

My point was simply that the "you need to buy nVidia video card to get them" special PhysX effects in Batman were done LONG before. (ie smoke/fog effects in Aquamark3 - its NOT new but gullible younger gamers wont know otherwise).

And in the tomshardware link I posted and others, they clearly show CPU under utilization and little effect when PhysX enabled.

As Chumbucket843 pointed out, this is simply to make sure GPU PhysX looks/performs as good as possible by crippling/not-optimizing it for CPU.

EDIT: STEvil, that Crysis thing was running 0.1fps and sped up for youtube. I dont care if I get 1 000 flying boxes with CPU, or 60 000 flying boxes with nVidia GPU. Problem is nVidia is marketing all these common simple special effects as being only possible on nVidia GPUs.

[ajaidev]

I did read a jolly good article about the fermi/GF100/GT300 in a local Indian Mag. They focused more on the super computer application and cache design rather than anything else.

They positioned AMD's evergreen against Nvidia's Fermi, they started of saying both support IEEE 754 2008 with FMA but fermi has ECC while evergreen is limited to ECD "A little bit on EDC vs ECC in a real environment, etc".. The interesting part was the cache sub sytems.

Lets start of with L1, Fermi has 16 SM's with 64KB memory attached. This 64KB can be configured as 16KB shared memory and 48KB L1 cache or vice verse. Now comes the evergreen it has 32KB shared and 8 KB L1, this can not interchange. Now comes the L2 Evergreen has 4*128kb blocks meaning total of 512kb L2 and fermi has 768kb L2. To this mix Evergreen also has a 64kb data share.

Extract:


Total Shared cache in Fermi- 256kb or 768kb "16kb or 48kb each"
Total L1 cache in Fermi- 768kb or 256kb "48kb or 16kb each"
Total L2 cache in Fermi- 768Kb "Shared"

Total Shared cache in Evergreen- 640kb "32kb each"
Total L1 cache in Evergreen- 160kb "8kb each"
Total L2 cache in Evergreen- 512kb "4 blocks of 128kb each"
Global data share 64kb "Connected to L1 and L2"

This means fermi has a big advantage in terms of L1 and L2 flexibility. The L2 fermi has is big enough to store overhead for retracing, etc. The evergreen is also not a bad machine at all but it seems to me that a fermi may perform better in GPU-CPU operations that need flexibility than Evergreen.

[RaZz!]

Someone wake me up when something actualy happens.

this is fermi rendering the morocco track from dirt2 in dx11!!!11

[Manicdan]

this is fermi rendering the morocco track from dirt2 in dx11!!!11

your right!!

look at how good they handle AA on those power lines!!
and the Physx of that rolling dustweed, its incredible!!
and the AF logic is beyond perfect!!

too bad the bump mapping sucks, they could have atleast done a better job with the potholes, the whole road just looks to flat....

[Vivi]

your right!!

look at how good they handle AA on those power lines!!
and the Physx of that rolling dustweed, its incredible!!
and the AF logic is beyond perfect!!

too bad the bump mapping sucks, they could have atleast done a better job with the potholes, the whole road just looks to flat....

now that i look at it, they've nailed the look on those power lines, almost no pixels visible!

[STEvil]

EDIT: STEvil, that Crysis thing was running 0.1fps and sped up for youtube. I dont care if I get 1 000 flying boxes with CPU, or 60 000 flying boxes with nVidia GPU. Problem is nVidia is marketing all these common simple special effects as being only possible on nVidia GPUs.

Yup.

[saaya]

this is fermi rendering the morocco track from dirt2 in dx11!!!11

[LiquidReactor]

So...any more news on when Fermi will actually hit the market?