<snip>

It is all marketing BS.
I really hope NV debug their SETI stuff, like this, i will be giving them a punishment in 2009 :), it will take me an hour to port to my own special project ... it is going to be bloody. I did not think NV would try to port SETI, it is extremely cache sensitive ... I am going to be dancing in my cubicle one more time ... The more they enable for CUDA, the more I ll have fun.

Ps: This is my personal opinion.

There a saying about pots and kettles that springs to mind.

Please let's not have another 6 months of blood-letting over when / whether anything will come of this. Just surprise us all when you overtake the Mark's Frozen Nehi.

F.

---

<snip>

It is all marketing BS.
I really hope NV debug their SETI stuff, like this, i will be giving them a punishment in 2009 :), it will take me an hour to port to my own special project ... it is going to be bloody. I did not think NV would try to port SETI, it is extremely cache sensitive ... I am going to be dancing in my cubicle one more time ... The more they enable for CUDA, the more I ll have fun.

Ps: This is my personal opinion.

There a saying about pots and kettles that springs to mind.

Please let's not have another 6 months of blood-letting over when / whether anything will come of this. Just surprise us all when you overtake the Mark's Frozen Nehi.

F.

Exactly, so, Let's stop saying that CUDA is faster than the CPU, because on SETI, it is NOT in average. On more of the non realistic, corner case claim.

Edit: Here is my machine: http://www.youtube.com/watch?v=wM5jsGzIlK0 , a little proto , but ... 5Ghz ++ ;-) Thanks to Fugger for the tunning ;) (the aircool one)

PS: This is my personal opinion, based on SETI ranking.

<snip>

It is all marketing BS.
I really hope NV debug their SETI stuff, like this, i will be giving them a punishment in 2009 :), it will take me an hour to port to my own special project ... it is going to be bloody. I did not think NV would try to port SETI, it is extremely cache sensitive ... I am going to be dancing in my cubicle one more time ... The more they enable for CUDA, the more I ll have fun.

Ps: This is my personal opinion.

There a saying about pots and kettles that springs to mind.

Please let's not have another 6 months of blood-letting over when / whether anything will come of this. Just surprise us all when you overtake the Mark's Frozen Nehi.

F.

Exactly, so, Let's stop saying that CUDA is faster than the CPU, because on SETI, it is NOT in average. On more of the non realistic, corner case claim.

PS: This is my personal opinion, based on SETI ranking.

Two points where I would take issue on that:

- CUDA running on a high spec graphics card (such as would be used by a keen gamer) *is* faster than anything up to a quaddie CPU. Most people don't yet have the bleeding-edge CPU technology available to you but can boost their performance by using CUDA if they have an appropriate GPU.

- This is not an either / or situation. The holy grail is to have crunching on *both* CPU and GPU (for one or more projects). Even if you do come up with better performance on the CPU, the CUDA will enable more work to be done by hosts that can run it.

F.

---

The holy grail is to have crunching on *both* CPU and GPU (for one or more projects). Even if you do come up with better performance on the CPU, the CUDA will enable more work to be done by hosts that can run it.

F.

As correct as it could be :) GPU is not replace to CPU, GPU is co-processor intended to work along with main CPU simultaneously.
So question what is faster just meaningles. CPU+GPU will be always faster than single CPU ;)

CUDA running on a high spec graphics card (such as would be used by a keen gamer) *is* faster than anything up to a quaddie CPU.

your statement is wrong, let s take the example of CUDA/Badaboom, a main stream Core i7 920 is faster than CUDA at encoding H264, Take Mainconcept Reference 1.0 and run it, you ll see that the CPU is faster.

It is valid for most of the bogus claims that was done with CUDA, it is not faster on SETI (While NV website claims it already http://www.nvidia.com/object/io_1229516081227.html If you read, CUDA is so much faster than you should be on the top of SETI Stats!!! It is not true! )
Most of the claims done by NV are based on super optimized code on the GPU against C code on the processor.

Mainconcept reference 1.0 is a good example , if you compare it to Badaboom, then, strangely ... it is many many time faster than what NV used as reference for their claims, and beat easily the GPU at encoding from DVD to ipod (The most common usuage model ever invented for video conversion)


The GPU is much slower than the CPU in 99% of the case, just because the GPU has no cache, it does not branch well, and it is not compatible with anything.



Ps:This is my personal opinion, AND usually, shameless marketing drive consumer association to get ritch :)

The holy grail is to have crunching on *both* CPU and GPU (for one or more projects). Even if you do come up with better performance on the CPU, the CUDA will enable more work to be done by hosts that can run it.

F.

As correct as it could be :) GPU is not replace to CPU, GPU is co-processor intended to work along with main CPU simultaneously.
So question what is faster just meaningles. CPU+GPU will be always faster than single CPU ;)

Actually, for the moment, the Cuda version is slower than the CPU! (CPU + GPU < CPU yet)

your statement is wrong, let s take the example of CUDA/Badaboom, a main stream Core i7 920 is faster than CUDA at encoding H264, Take Mainconcept Reference 1.0 and run it, you ll see that the CPU is faster.

Why would I want to encode H264 using CUDA? And I did accept that the 4 cores of a quaddie (e.g. the i7) would be faster than a single GPU (even top-of-the-range GPU). But it isn't either / or. However fast or slow CUDA is, its crunching is as well as whatever you can provide with the CPU and is also available to those who have to put up with e.g. P4's, C2D's or (dare I say it?) AMD CPU's. So more science gets done and the project wins.

Actually, for the moment, the Cuda version is slower than the CPU! (CPU + GPU < CPU yet)

From what I have read here on the boards, a gtx280 can knock out WU's at least 4 x faster than any single core of my overclocked Q9450.

F.

---

your statement is wrong, let s take the example of CUDA/Badaboom, a main stream Core i7 920 is faster than CUDA at encoding H264, Take Mainconcept Reference 1.0 and run it, you ll see that the CPU is faster.

Why would I want to encode H264 using CUDA? And I did accept that the 4 cores of a quaddie (e.g. the i7) would be faster than a single GPU (even top-of-the-range GPU). But it isn't either / or. However fast or slow CUDA is, its crunching is as well as whatever you can provide with the CPU and is also available to those who have to put up with e.g. P4's, C2D's or (dare I say it?) AMD CPU's. So more science gets done and the project wins.

Actually, for the moment, the Cuda version is slower than the CPU! (CPU + GPU < CPU yet)

From what I have read here on the boards, a gtx280 can knock out WU's at least 4 x faster than any single core of my overclocked Q9450.

F.

on some units, on others, it is much slower ... Shameless marketing, speaking about the good, and only the good. They can't even beat a Phenom if you put a Core 2 Duo with the G280

In average, CUDA is slowing down the calculus, yet ...
If it was that fast , it would generate a RAC of 24 000 ... I don t see it, and trust me, I tried very hard to reproduce.

Bottom line: One more lie.

The greatest value of Cuda is to nVidea..........

They sell the freakin' cards, and pushed this all through...and don't nobody try to tell me otherwise......'cuz I knows the facts........been told under the table....

Kitty New Year................

---

"Time is simply the mechanism that keeps everything from happening all at once."

Shameless marketing, speaking about the good, and only the good.

**!!!!!*****

I'm speechless :)

F.

---

Shameless marketing, speaking about the good, and only the good.

**!!!!!*****

I'm speechless :)

F.

And some others should be..............

Ya know.....if it could be said the nVidea donated a quarter of a million dollars to Seti.........I might be quiet about this whole shenannigan.....but they have done it out in the light of day...open to all who have their eyes open....
Gee whiz......what a sad moment for Seti........a scientific project.......not supposed to be a marketing project..........

---

"Time is simply the mechanism that keeps everything from happening all at once."

I am still waiting to see anything impressive from CUDA ... and then, it is my turn ...

It is going to be fun ...

I expect Who might be talking about Larabee, engineering samples are supposed to be available by end of 2008.

Claggy

I don't have any graphic board, only an Opteron 1210 running Linux at 1.8 GHz. I would like, however to present some data taken from Folding@home statistics:
Nvidia active GPUs average 0.11 TFLOPS
ATI active GPUs average 0.098 TFLOPS
PS3 (CELL) active average 0.028 TFLOPS
MAC Intel active average 0.003 TFLOPS

This simply by dividing the total TFLOPS by the number of active CPUs as given by Folding@home.
Tullio

---

I am still waiting to see anything impressive from CUDA ... and then, it is my turn ...

It is going to be fun ...

I expect Who might be talking about Larabee, engineering samples are supposed to be available by end of 2008.

Claggy

Got nothing to do 'bout Who..you.......
Just know what I'm talking about...so do you?

The truth has been told.....maybe not so to you.......
'Bout the birth of the Cuda....and what's told untrue.........

Ya can't fool the kitties.....they sniff it out here and there.
And the stink of the Cuda carries far and awhere.....................

---

"Time is simply the mechanism that keeps everything from happening all at once."

The holy grail is to have crunching on *both* CPU and GPU (for one or more projects). Even if you do come up with better performance on the CPU, the CUDA will enable more work to be done by hosts that can run it.

F.

As correct as it could be :) GPU is not replace to CPU, GPU is co-processor intended to work along with main CPU simultaneously.
So question what is faster just meaningles. CPU+GPU will be always faster than single CPU ;)

Actually, for the moment, the Cuda version is slower than the CPU! (CPU + GPU < CPU yet)

Just wrong.
CUDA takes ~3-4% of CPU and performs as ~2 cores of my 9450 quad.
That is CPU+GPU gives <~6 cores while CPU alone gives only 4 cores.
Experimental fact.

ADDON: BTW, I use only 9600GSO. More powerful GPU can give even best results. And sure even poor cheap 8500GT will outperform these 3% of CPU it takes for feeding.

Shameless marketing, speaking about the good, and only the good.

**!!!!!*****

I'm speechless :)

F.

And some others should be..............

Ya know.....if it could be said the nVidea donated a quarter of a million dollars to Seti.........I might be quiet about this whole shenannigan.....but they have done it out in the light of day...open to all who have their eyes open....
Gee whiz......what a sad moment for Seti........a scientific project.......not supposed to be a marketing project..........

@Mark
We are agreed that the CUDA App was released with indecent haste - for whatever reason (and I guess that there will never be an official explanation of that reason). However, the genie is out of the bottle and we have to live with it now (particularly all the set'n'forget boxes that it has been downloaded to).

You have always been a staunch supporter of the project and the science and, once the bugs have been ironed out of it, the CUDA app will enable my overclocked quaddie to almost double its output (given the right graphics card which is much cheaper than duplicating my set-up) - with a correspondingly much greater effect for those with slower CPUs than mine. So the science wins out - where's the beef?

The process was flawed; you don't like how we got to where we are. But this is where we are and the future is more crunching power for S@H.

If a CUDA card can produce more valid results than the 4% of one CPU that is used to service it then the project wins.

F.

---

Shameless marketing, speaking about the good, and only the good.

**!!!!!*****

I'm speechless :)

F.

And some others should be..............

Ya know.....if it could be said the nVidea donated a quarter of a million dollars to Seti.........I might be quiet about this whole shenannigan.....but they have done it out in the light of day...open to all who have their eyes open....
Gee whiz......what a sad moment for Seti........a scientific project.......not supposed to be a marketing project..........

@Mark
We are agreed that the CUDA App was released with indecent haste - for whatever reason (and I guess that there will never be an official explanation of that reason). However, the genie is out of the bottle and we have to live with it now (particularly all the set'n'forget boxes that it has been downloaded to).

You have always been a staunch supporter of the project and the science and, once the bugs have been ironed out of it, the CUDA app will enable my overclocked quaddie to almost double its output (given the right graphics card which is much cheaper than duplicating my set-up) - with a correspondingly much greater effect for those with slower CPUs than mine. So the science wins out - where's the beef?

The process was flawed; you don't like how we got to where we are. But this is where we are and the future is more crunching power for S@H.

If a CUDA card can produce more valid results than the 4% of one CPU that is used to service it then the project wins.

F.

Oh....Fred...my man....
No, there will never be an 'official' explanation for what was done it haste....NDA with nVidea will forever prohibit it.....

I suspect they may be in contact with me in the next few days to offer a few vid cards to keep my mouth shut....

This was nothing but a marketing ploy........
If the apps and Boinc manager ever are able to handle it properly...maybe it will some day benefit the science........maybe....

But outta the gate.....it was nothing but an attempt to sell more vid cards........

Why else would you need an NDA in place to protect code? Or to speak about it? So ATI couldn't figure out what they were doing?

Laughing my buttocks off over this one.....

---

"Time is simply the mechanism that keeps everything from happening all at once."

I don't know/understand why you are so negative about CUDA?

If the SETI@home-Enhanced-CUDA-app is bug-free and the BOINC support will be well also..
You will see some rigs with GTX 280/285 and/or GTX 295 in the top #10 host list..!



My guess..
GTX 280/285 = Core2 Quad (~ 5 Cores)
GTX 295 = Core2 Quad (~ 10 Cores)


If you will have a Core i7 with 4 x GTX 295.. woohooo.. it will be the #1 host.. :-D

But of course he will eat the wattage like nothing.. ;-D



BTW.
Why use some 'science-organizations' or companies GPU-power for their work? ;-)

---

I don't know/understand why you are so negative about CUDA?

I have promised to keep quiet about that............
It's not the reality of it...it's how it came about........

'Nough said.....'

Done discussing it.......end of line.

---

"Time is simply the mechanism that keeps everything from happening all at once."

CUDA is Fast on the right GPU..
My GTX 260 core 216 would be top 10 IF the app was stable... (current clocks are 700/1510/1070)

60 credit wu takes 10 min 24*60/10 = 144 * 60 = 8640
14 credit wu takes 3.25 min 24*60/3.25 = 443 * 14 = 6203

So Depending on the work assigned I would be from top 10 to top 50...
Again this is IF the app were stable.

~Bob

---

Do you Good Search for Seti@Home? http://www.goodsearch.com/?charityid=888957
Or Good Shop? http://www.goodshop.com/?charityid=888957