More Far Cry confusion? (nv45 anand's review)

[PP]

No AA this time.

I suppose same time demo is used in both test (at least is what they say). 6800 Ultra scores doesn't match:

http://www.anandtech.com/video/showdoc.aspx?i=2102&p=5

http://www.anandtech.com/video/showdoc.html?i=2096

A 20% increase with 1.1 patch?

 

[karlotta]

my understanding is the first graph is ATs demo , and the second one is from the nvda demo. both are using the 1.2..?

 

[PP]

AT say:

Level Analysis: mp_airstrip
This is our default benchmark here at AnandTech. We've been using this demo to benchmark with FarCry ever since we first included it in our benchmark suite.

Then, I think they benched nv45 with this demo and patch 1.1, and later the same in the patch 1.2 review.

 

[christoph]

they are using different systems in those tests, but it seems odd though

 

[Galduta]

A 20% increase with 1.1 patch?

A 25% increase .. 87 fps ------ 109fps with

Intel Pentium 4 3.4GHz EE Socket 478
RAM: 2 x 512MB Micron DDR2 533
2 x 512MB Corsair 3200XL
(Samsung 2-2-2-5)

 

[jvd]

remember patch 1.1 for 6800s regarded them as nv3x and thus they ran on nv3x path. Which should be faster as it was using fx 12 when possible and p.s 1.4 or under shaders .

The drops seen at hexus for the radeons is wierd though.

THey loose about 10fps from going to patch to patch in which nothing should have changed for them

 

[Veridian3]

The patch sent to the press by nvidia had several issues on the Radeon, a score drop doesnt surprise me.

You can see the radeon issues in detail here : http://www.driverheaven.net/articles/farcrypatch/

 

[Evildeus]

There was also a drop on 6800s...

 

[jvd]

There was also a drop on 6800s...

as i said , could be going from nv3x to sm 2.0 path.

 

[Evildeus]

It could, and it could not. Until someone ask, no quick stance please.

 

[jvd]

It could, and it could not. Until someone ask, no quick stance please.

Its not quick stance. The drop is consistant with forcing the r3x0 path on the 6800ultras .

That is why i believe its from going from nv3x path to p.s 2.0 path.

Unless you believe they gimped the ps 2.0 path to make p.s 3.0 look better

 

[PP]

The drop is consistant with forcing the r3x0 path on the 6800ultras .

Path SM 3.0 is intended to perform more (or at least equal) than nv3x path in 6800 series. There are only performance improvements, no new effects, and it keep combining fp16 and fp32. In other tests you can see there isnt a 25% loss when comparing patch 1.1 and 1.2 results.

IMO, 6800 scores in nv45 preview are wrong.

 

[jvd]

The drop is consistant with forcing the r3x0 path on the 6800ultras .

Path SM 3.0 is intended to perform more (or at least equal) than nv3x path in 6800 series. There are only performance improvements, no new effects, and it keep combining fp16 and fp32. In other tests you can see there isnt a 25% loss when comparing patch 1.1 and 1.2 results.

IMO, 6800 scores in nv45 preview are wrong.

why would they be wrong .

IN patch 1.1 the 6800s were regarded as nv3x parts and thus ran the nv3x path.

Which they ran faster than patch 1.2 p.s 2.0 path . They ran patch 1.2 p.s 2.0 path right about the same speed as 1.1 forced into r3x0 path .

patch 1.2 ps 3.0 path is the fastest of them all for the 6800s .

I don't see why there are any problems

 

[jimmyjames123]

Its not quick stance. The drop is consistant with forcing the r3x0 path on the 6800ultras .

I'm not so sure that this statement is really accurate. There are 6800 owners at NVN who have tried running their cards in both NV3x mode and the R300 mode, and the differences in fps between the two modes seems to have been relatively small for them. See some of the FarCry threads there for references. Also remember that the 6800 cards already have very fast PS 2.0 speed, and there have been some synthetic tests showing that in some instances, speed is as fast or faster using PS 2.0 vs using PS 1.1 shaders.

I suggest emailing Derek Wilson at Anandtech to get more clarification on this

 

[jvd]

jimmy i don't care with what forum users say. Checking out websites that have done testing like guru3d shows the drop. there is another german site i believe that has done it too

 

[jimmyjames123]

LOL

Do you know the details about how Guru3d tested? Do you know what drivers they used?

Basically, you are just picking and choosing what you want to believe. Nothing new here really

There really is no good reason to believe that very powerful cards like the 6800U will incur a significant performance hit when run in the R3xx path, but you are free to believe what you want to believe I guess.

 

[jvd]

Why would it not see a drop from fx 12 , fp 16 with p.s 1.4 shaders to fp16/32 / p.s 2.0 shaders ?

 

[jimmyjames123]

FaryCry primarily uses PS 1.1 shaders, with PS 2.0 used for the lighting effects.

 

[jvd]

FaryCry primarily uses PS 1.1 shaders, with PS 2.0 used for the lighting effects.

Okay jimmy . Whatever u say

 

[digitalwanderer]

FaryCry primarily uses PS 1.1 shaders, with PS 2.0 used for the lighting effects.

You just continue to post until you embarress yourself too much and then you go away for a while before coming back, let us guess which time it is now....