thats what i am curious about

The only way to see would be to have a 5870 run this benchmark on Windows Vista with DX10 installed rather than DX11.  That would force the benchmark to run in DirectCompute 10 mode with Shader Compute 4.0.  That is what DX10 cards are running this benchmark with.  DX11 cards with DX11 API installed on Vista/Win7 are running this benchmark with DirectCompute 11 with Shader Compute 5.0.

The results cannot be correlated between a DX10 card and a DX11 card.

luv2increase

The only way to see would be to have a 5870 run this benchmark on Windows Vista with DX10 installed rather than DX11.  That would force the benchmark to run in DirectCompute 10 mode with Shader Compute 4.0.  That is what DX10 cards are running this benchmark with.  DX11 cards with DX11 API installed on Vista/Win7 are running this benchmark with DirectCompute 11 with Shader Compute 5.0.

The results cannot be correlated between a DX10 card and a DX11 card.

yes they can. it makes no difference here. this is just a not so great benchmark to begin with. your cards supporting dx11 doesnt make them as special as you want to think. we are seeing fairly sporadic scores from different cards so we cant reach any conclusions.

luv2increase

chumbucket843

luv2increase

The only way to see would be to have a 5870 run this benchmark on Windows Vista with DX10 installed rather than DX11.  That would force the benchmark to run in DirectCompute 10 mode with Shader Compute 4.0.  That is what DX10 cards are running this benchmark with.  DX11 cards with DX11 API installed on Vista/Win7 are running this benchmark with DirectCompute 11 with Shader Compute 5.0.

The results cannot be correlated between a DX10 card and a DX11 card.

yes they can. it makes no difference here. this is just a not so great benchmark to begin with. your cards supporting dx11 doesnt make them as special as you want to think. we are seeing fairly sporadic scores from different cards so we cant reach any conclusions.

Cough cough...  Please take a look at my sig.

I gave you my "analytical" reasoning for thinking why I believe what I believe about the issue.  I am waiting for you to do the same.  The bold quote above is not "analytical".  You must provide a hypothesis (educated guess) for what you believe.

I state that they are using "different" kernels due to different APIs.  Can you counter that with something educated?

i am pretty sure everyone on this forum agrees with me that you cant stop talking about your 5870s and its getting old. i could dig up some posts to prove it. there right there in your avatar for starters.

 why would he make a completely new bench for CS 5.0? that would be pointless. its more than likely they are the same. kind of like running a dx9 game on a dx10 card.
you are the problem. no one else here gets in as much arguments and fights as you to need to do that. if you think my post is uncivil then that is just crazy. dont read a fragment of my post and make judgments. the benchmark is acting really strange with scoring so we cant reach a logical conclusion yet. if you tried version .15 cpu's beat the gpu's. maybe a future version will fix the scores.

chumbucket843

luv2increase

chumbucket843

luv2increase

The only way to see would be to have a 5870 run this benchmark on Windows Vista with DX10 installed rather than DX11.  That would force the benchmark to run in DirectCompute 10 mode with Shader Compute 4.0.  That is what DX10 cards are running this benchmark with.  DX11 cards with DX11 API installed on Vista/Win7 are running this benchmark with DirectCompute 11 with Shader Compute 5.0.

The results cannot be correlated between a DX10 card and a DX11 card.

yes they can. it makes no difference here. this is just a not so great benchmark to begin with. your cards supporting dx11 doesnt make them as special as you want to think. we are seeing fairly sporadic scores from different cards so we cant reach any conclusions.

Cough cough...  Please take a look at my sig.

I gave you my "analytical" reasoning for thinking why I believe what I believe about the issue.  I am waiting for you to do the same.  The bold quote above is not "analytical".  You must provide a hypothesis (educated guess) for what you believe.

I state that they are using "different" kernels due to different APIs.  Can you counter that with something educated?

i am pretty sure everyone on this forum agrees with me that you cant stop talking about your 5870s and its getting old. i could dig up some posts to prove it. there right there in your avatar for starters.

why would he make a completely new bench for CS 5.0? that would be pointless. its more than likely they are the same. kind of like running a dx9 game on a dx10 card.
you are the problem. no one else here gets in as much arguments and fights as you to need to do that. if you think my post is uncivil then that is just crazy. dont read a fragment of my post and make judgments. the benchmark is acting really strange with scoring so we cant reach a logical conclusion yet. if you tried version .15 cpu's beat the gpu's. maybe a future version will fix the scores.

You started the flame war.  I never flamed you, yet it is never ending with you.  When you don't agree with me, you don't professional give any counter arguments, but you'd rather stoop so low as to flame me.  Now, whom is the mature one?

I am going to ask you again since you failed to do so the first time to give me a "civil" argument against my belief on the subject.  I will add some information for "everyone" just to see how exactly the "inexperienced" programmer coded the program.

This is part of the response from the author of the program when asked if it gives favor to ATI or Nvidia cards:
http://www.ngohq.com/news...enchmark-released.html

1) I don't favour any vendor, the codepath is one and the same for ATI/NV any any other DirectCompute enabled device that could be istalled in the system. It looks like this:

benchmark(compute shader code) -> DirectX11 API (compiler) -> DirectX11 API (run) -> drivers -> GPU.

You see that?  Compute Shader code?  On DX10 hardware, it would have to default to version 4.0 even if DX11 is installed (in software).  On DX10.1 hardware, it would default to Compute Shader 4.1.  On DX11 hardware, it would only be "LOGICAL" to assume it would default to Compute Shader 5.

Can someone give their input on this? chumpbucket, I am trying to get down to the bottom of this.  Your inflammatory remarks are hindering the progress.  If you want to give any more input, please be kind.  Thanks.

PS  This isn't a matter of ATI vs Nvidia so the fanboy or whatever people call it nowadays needs to stop.  This is about DX10 vs DX11.  The scoring stuff is irrelevant.  I'm just trying to figure out what kernel the program is using on what hardware.

Just in case anyone doesn't know, it is "impossible" for Compute Shader 5.0 to run on DX10 or DX10.1 hardware.  That is a given.

This benchmark is BS.

Darius92388
well if your gonna suggest this benchmark doesnt suck

then the 5870s sure suck even with dx 11 it should be performing better score wise since this is a dx 11 based benchmark

You are correct.  Any way one looks at it, this benchmark has some serious coding issues.

It doesn't correlate at all to this; therefore, something is fishy with it.

http://images.anandtech.c...092209122344/20136.png

Darius92388

perhaps you should rma your 5870s perhaps they are broken? maybe rma the drivers

What is your reasoning?

Darius92388

well even with a poor driver and perhaps poor crossfire scaling and poor coding this guy used a ati card to design it your score should of at LEAST equaled 60k+

Somebody help me...

I don't think you've been assimilating the information given from the posts in this thread.  I beg you to please go back and re-read it.  On the hand, you may just be being sarcastic.  If that is the case, I will ask you to go back and re-read it again as well because I had some things to say about people acting in that manner when this is supposed to be a serious discussion amongst adults.


edit:  Let's not forget who started this thread.  In my eyes, you are in a "trolling" status.  Let's change that, shall we?

moose517

yup, i would say 5800 series are fail, even on my half working rig i got this score

Oh my goodness.  Quit trolling my thread people.  Can't we be serious for once???  This is beginning to get irritating.

Darius92388
so simply put ati=FAIL

On Nvidia's own DirectCompute benchmark/tech demo, ATI=Won.  The "validity" is much more with regards to that.
http://images.anandtech.c...092209122344/20136.png




Also, I know you are trolling the thread now, and I hope the moderators do something about it.  I am requesting it.  I say this because by your own admission, you stated the benchmark was basically FALSE.

Darius92388
i condem this benchmark for being unusable and say that the scores are not countable

I'd say it is time to quit the kiddie stuff Darius.  It is immature and making you appear immature to other members.

i have to agree with darius this test is terribly done, and shows no worthwhile data


and are you seriously asking for help?  I for one will not come to your aid, you have done nothing but bicker on this forum and i think its about time you get some of your own medicine

If you are being serious, well it appears all 5870s score in the 3xxxx range, but we both know you were being sarcastic.  You can't say the test is bogus one minute and then say ATI sucks the next and "base" that opinion on poor results of a test of which you said was bogus in the first place.

I never called any immature either.  Don't put words in my mouth. 


I've been attempting to be technical about this discussion, and all that you are doing is saying ATI sucks. Why?  It makes no sense.  That is being a fanboy.  You never want to get technical about anything thus far in this thread.  Why is this?

moose517
hmmm, looks like you did.....

Huh?  Moose517, I suggest that if you want to start trouble and a flame war to take it to another thread. 

Do you have anything "constructive" to say with regards to DirectCompute, OpenCL, DX10, DX11, Compute Shaders, GPGPU and/or Pixel Shaders?

Darius92388

Luv you really shouldnt tell people to stop bashing you to then turn around and bash them

an eye for an eye leaves the world blind

your op you should stay on topic

anyway as i asked where is the link for this chart?

plus you should take it down if it is copyrighter material as it is against tos

Proof?

The link has been on this page for a few hours now.  I guess you failed to read the link I said everyone should if they want a better understanding.  And here I thought you read the same thing I did so we could have a professional discussion.  I know what your motives are.


edit:  I am now going to abandon my own thread.  The discussion has gotten extremely off-topic due to a select few individuals.  I forgive them though.  The power of forgiveness is AWESOME!  Sera sera... What will be will be...

MODS, can you please lock this thread.  Thank you.