http://www.hexus.net/content/item.php?item=21996

So the Hexus.net labs took nVidia's demo at CES 2010 and pitted against a 5870 test system. They admit that they don't know anything about what test rig nVidia was using, but they used for their tests a "fair"-ish Core i7 975 @ stock with a factory overclocked 5870 at 875 Core/ 4900 Mem, which is a pretty reasonable comparison.

The benchmark was Far Cry 2. The game settings are also posted for reference, although they make no reference to the ATI Drivers.

GTX 285: min 38.4 avg 50.32
GF100: min 64.6 avg 84.05
HD59870: min 48.40 avg 65.84

They then did another benchmark with "Dark Void". Here are the numbers

GTX 285: min 29.28 avg 38.38
GF100: min 53.32 avg 78.32

The lab then comments that this test is skewed heavily towards GF100 because of improved PhysX rendering. Since Dark Void hasn't been released yet, the lab doesn't have numbers, but they will be running a test once they get a copy in a few days using the same ATI setup.

What are your thoughts here?

Is the performance expected/disappointing?
Was the test rig fair? Do you think nVidia have a natural edge because of the test rig, or is an i7 975 @ stock and factory overclocked HD5870 a pretty fair real-world comparison?
Is Dark Void a fair comparison platform? (my gut tells me no, but from what Hexus.net says, Dark Void may end up becoming a benchmark just like Crysis)

Finally, Hexus.net comments that nVidia needs to release a ~$150 card for the Joe Average gamer in the Fermi lineup to be successful. What do you think of this comment? Is the GTX260/GTX250 doing fine for that price range right now, or does nVidia need something better. (I know that the HD5700 cards offer better performance/dollar at that range, but is it a performance difference that the Average Joe will care about?). Do you agree with this statement, or do you think that the selling opint of DirextX11 is a moot point at the lower-end, and that nVidia still has the ability to compete when it comes to the $100-$200 range?

Hardware Canucks did that exact same thing on Sunday. 

What I understand from the Dark Void comparison is that it may come to what I suspected: high-end GF100 cards will not need a PPU!

fermie is right around the corner, i cant wait!

chumbucket843

fermi can switch from graphics to physics 25x faster than gt200. a card dedicated to physx will be faster unless the graphics arent that intensive.

Where did you see 25x?  On Nvidia's site, it says 10x.

http://www.nvidia.com/object/gf100.html

I recognize I am responding to a thread not based off of a released product, but a beta card... Even if they increase the clocks 10%, that is not going to make up the ground from 78.32 to 99.79....

But dang that is nice to see some beta results and not white paper formulas!

luv2increase

I'd be surprised if they would have told you that the cards at CES were indeed running higher clock speeds than the final product and don't expect the final cards to run as well.  Imagine that

Yes, their main goal for the final retail product, is usually to lower it's performance from what they could achieve with a BETA card...

Another gloom and doom prediction for Fermi.

Such a distinct pattern.

Hand the baton off, while there is still time.

Well at we all know that they can't use the excuse about developing the drivers, they have had forever to do it.

MSim

Let only test games that use our PhysX engine, who cares if its not open source where both camps have a level playing field. The only test i would accept is one using open source to show which card performs better.

Open source?  Are the benchmarks used to measure gaming performance open source?  I didn't think they were.

devilstestament

Does this mean the best bet is to wait for the Dual GPU Fermi? And what would something like that run for? I am currently waiting for the first Fermi. Should I be waiting for the Dual Fermi cards? 

I am having a hard time figuring out what you mean.  Can you please clarify?  Why would it be a best bet to wait for a dual-GPU Fermi?  What eluded you to that thought?

fredbsd

MSim

Let only test games that use our PhysX engine, who cares if its not open source where both camps have a level playing field. The only test i would accept is one using open source to show which card performs better.

Open source?  Are the benchmarks used to measure gaming performance open source?  I didn't think they were.

My bad for using open source, i should of said this.
How about vendor neutral so no one has an advantage when it comes to showing how good the performance is. We all know Far Cry 2 and Dark Void use Nvidia PhysX so those games are not vendor neutral.

luv2increase

devilstestament

From what I understood the single-GPU Fermi is good but not better then the 5970 right? Since the 5970 is dual-GPU wouldn't the dual-GPU Fermi logically be a better choice?

Nobody really knows.  A single-GPU GF100 may be faster than a dual-GPU 5970.  We won't really know until the reviews come out after the card is released.  But even then, the performance of the GF100 will go up as the drivers improve.


edit:  MSim, Far Cry 2 doesn't use PhysX.  It uses Havok.

Im hoping that the Dual Fermi comes out and gives the 5970 a run for its money, I would also love to see the Fermi x2 cost no more then $600 like the ATI 5970 does now.... I plan to Buy two Dual GPU's of either ATI or Nvidia so I guess im stuck on making that decision till April comes out...
 
I plan to play Battle Field Bad Company 2 full time on the PC and finally lay to rest my 360, so I want my setup to crush all games I throw at it

ZachA

luv2increase

devilstestament

From what I understood the single-GPU Fermi is good but not better then the 5970 right? Since the 5970 is dual-GPU wouldn't the dual-GPU Fermi logically be a better choice?

Nobody really knows.  A single-GPU GF100 may be faster than a dual-GPU 5970.  We won't really know until the reviews come out after the card is released.  But even then, the performance of the GF100 will go up as the drivers improve.


edit:  MSim, Far Cry 2 doesn't use PhysX.  It uses Havok.

Im hoping that the Dual Fermi comes out and gives the 5970 a run for its money, I would also love to see the Fermi x2 cost no more then $600 like the ATI 5970 does now.... I plan to Buy two Dual GPU's of either ATI or Nvidia so I guess im stuck on making that decision till April comes out...
 
I plan to play Battle Field Bad Company 2 full time on the PC and finally lay to rest my 360, so I want my setup to crush all games I throw at it

I really hope so too I am planning on going Tri SLI