New Nvidia core?

[mklvotep]

I've heard that there was a new nvidia core in beta, A member from our team has been testing and the results are a bit disapponting. We have already taken a huge loss in points with these new wu's , now if we're taking another, i think you're going to lose folders or at least interest. I have 20 nvidia cards running apps and all i expect is points. Now the same card is getting less and less points and arent getting cheaper. Some consideration needs to be given to us with the money spent for hardware and electricity, its only fair.

[MoneyGuyBK]

ALL great points, but no use.
I do agree however, the new cores need to be more efficient, as this is not always the case.
Another issue is with NVi and their newer, less reliable Drivers... but that will be for another thread.






Peace

[mklvotep]

The funny thing is the out of balance points, I think ati cards are very capable but their points are awful. I've invested a large amount of money for nvidia cards and to just destroy their efficiency or ppd is a bit unfair. Points arent money, Its not like it has to come out of some account or something. I think they might be trying to cut down on eue's but the eue's arer mostly from the users having an unstable oc. I havent had an eue in 2weeks or so with 20 cards running 24/7. I can accept the wu's as they are right now, but another huge cut in points might bring me to selling off most of my machines and folding on a much lesser scale.

[spazzychalk]

are we forgetting why we fold? all these countless threads about new points systems and so forth. its a motivational system. honestly guys, someone for the love of god tell me what you do with these points that youre coveting so much. anything at all tangible cause id like to do it too. i think its great youre putting all this into systems just to fold, but i also think in a way its against the nature of the project. its not suppossed to change the way you live your life. i found out about this accidently playing with my new ps3. after a week i was thinking about watching a movie or playing a game and i said no i dont want to mess up the folding. WFT?!?!?!?!? i had to check myself. i spent 500$ on something ive wanted a LONG time and now im not gonna use it. hell no. so when you bought all that hardware to fold, what were your rationalizations with the points? maybe an extra turn at carousel? im not being a dick i mean it quite literally. whoever folds the most gets us the closest to the cures that will one day save themselves and their loved ones. so look at WUs done or team rank and not the points. so why sell now that hasnt changed? its actually why i started my own team instead of joining one. i wanted to see how much ive done vicariously and down generations of talk. its a small team, with no special hardware, but in a month and a half we're in the top 8-9% of all the teams ever in existance. with like only 11 or 12 processors... and a mere 250ish WU... all single core cpu and 2 ps3.... thats pretty fulfilling and disappointing at the same time. so anyways, back to point, other than your own life, what do you expect to cash these points in on?

[Amaruk]

...but the eue's arer mostly from the users having an unstable oc.

Tell that to those who can't fold underclocked, much less at stock clocks.

I've invested a large amount of money for nvidia cards and to just destroy their efficiency or ppd is a bit unfair.

A prudent investor would thoroughly research whatever he/she is investing in before handing over their money. Any assessment should include particular attention to risks involved. In this case (loss of PPD) the risk has been documented. For example, here is a link to a news article posted in May, before the Nvidia client was even released.

http://folding.typepad.com/news/2008/05 ... neral.html

Nvidia cards are known to be better with smaller proteins compared to ATI. For example, this was posted shortly after the (beta) Nvidia client was released.

viewtopic.php?f=43&t=3319&start=15#p32155

The Nvidia's advantage running small proteins gave them a points bonus, but this bonus was going to shrink once they started doing some real work with larger proteins. That's exactly what happened when larger WUs were first issued in September.

viewtopic.php?f=52&t=5452

The drop in points, which was explicitly stated, happened because that was the expected behavior.

The funny thing is the out of balance points, I think ati cards are very capable but their points are awful.

Much of that is due to Nvidia running smaller proteins as stated above. Now that they are both running larger ones, it will be interesting to see how they really compare.

[mklvotep]

If some people are having a problem with this core or any , we should be able to chose which ones we use. And i do fold for the cause , but my physical reward is points and yes i care about that. If noboby cares about their points why are they sportin that statistics box in their sig?

I understand about the larger molecule, its the points problem with it. If it takes twice the time as one, it should reward roughly twice as many points, makes sense?

[7im]

If some people are having a problem with this core or any , we should be able to chose which ones we use. And i do fold for the cause , but my physical reward is points and yes i care about that. If noboby cares about their points why are they sportin that statistics box in their sig?

I understand about the larger molecule, its the points problem with it. If it takes twice the time as one, it should reward roughly twice as many points, makes sense?

Why should NV cards get twice the points if larger work units cause a performance bottleneck in the hardware?

NV cards are already making twice the points, or more. The larger work units just bring the performance of NV cards back inline with expected benchmark results. Getting less points doesn't mean it's broken. It just means things are getting fixed, and the project can now do more complex work. NV cards have been getting the equivalent of a free small work unit points bonus. The project has said for a long time that work units would increase in size, and that performance would change. That means the points picnic that NV users have enjoyed is coming to an end.

This is like saying that if my CPU client takes twice as long to complete a large CPU work unit because I don't have enough system memory that I should get double the points. No. I either need to get more memory, or not run big work units, or deal with the fact that certain projects will run slower on my particular type of computer hardware. It would be a mistake for me to expect my hardware to run all types of work units at the same speed or PPD. The work changes, to the PPD will change too.

Same applies to GPUs. Just because NV cards were making 4500 PPD on small work units does not automatically entitle them to make 4500 PPD on large work units, and that has been made clear by the project staff several times. 1500 PPD is the benchmark, so NV cards that are similar to the benchmark GPU shouldn't be making 4500 PPD. And larger work units fix that previous mistake of free extra points. Now they have to earned like all the rest.

[EvilAlchemist]

I understand about the larger molecule, its the points problem with it.

Here is a good way to look at it.

Project A takes 4 hours on the benchmark gpu and worth 100 points when completed.
If your gpu takes 4 hours to complete Project A, you earn 100 points every 4 hours.
If your gpu takes 2 hours to complete Project A, you earn 100 point every 2 hours.
If your gpu takes 10 hours to complete Project A, you earn 100 point every 10 hours.

Your hardware might be able to complete some work units faster and some might be slower.

If it takes twice the time as one, it should reward roughly twice as many points, makes sense?

Here is the flaw in this logic. I will use the classic client as an example.

Lets say I have a Intel C2D 3 Ghz & a Pentium 3 @ 1 Ghz.
The 3 ghz system completes a work unit every 5 hours, but the Pentium 3 takes 4 days to complete the same unit.
By your statement, the Pentium 3 should get more Points Per day then the C2D because it took longer to complete the same work unit.

[mklvotep]

I understand about the larger molecule, its the points problem with it.

Here is a good way to look at it.

Project A takes 4 hours on the benchmark gpu and worth 100 points when completed.
If your gpu takes 4 hours to complete Project A, you earn 100 points every 4 hours.
If your gpu takes 2 hours to complete Project A, you earn 100 point every 2 hours.
If your gpu takes 10 hours to complete Project A, you earn 100 point every 10 hours.

Your hardware might be able to complete some work units faster and some might be slower.

If it takes twice the time as one, it should reward roughly twice as many points, makes sense?

Here is the flaw in this logic. I will use the classic client as an example.

Lets say I have a Intel C2D 3 Ghz & a Pentium 3 @ 1 Ghz.
The 3 ghz system completes a work unit every 5 hours, but the Pentium 3 takes 4 days to complete the same unit.
By your statement, the Pentium 3 should get more Points Per day then the C2D because it took longer to complete the same work unit.

Not what i meant at all. I'm saying same card or gpu, bigger unit bigger points or longer unit bigger points.
Cause if these wu's were benchmarked with an nvidia card instead of ati it would work that way. Nvidia wu's should be benchmarked with nvidia cards.

[Amaruk]

Nvidia wu's should be benchmarked with nvidia cards.

First, ALL GPU2 work units can be run on either ATI or Nvidia.

Second, everyone who complains about points dropping on their Nvidia GPUs and suggests the above as a solution has no clue how benchmarking works.


Let's say we use an Nvidia 8800GT to benchmark for all Nvidia cards.

That 8800GT folds the 1392 atom WU in 3 hours, 39 minutes, 51 seconds.

It folds the 973 atom WU in 2 hours, 8 minutes, 57 seconds.


The formula for benchmarking is:

Points = 1500 * (DaysPerWU)

1392 atom WU = 1500 * 0.152673611 = 229 points instead of 511.

973 atom WU = 1500 * 0.089548611 = 134 points instead of 384.


The result?

ALL Nvidia GPUs lose 55% of their current points folding the 1392 atom WUs.

ALL Nvidia GPUs lose 65% of their current points folding the 973 atom WUs.


By comparison, the recent Nvidia points drop many are complaining about is only 15 - 35% on my 8800GTs.

[EvilAlchemist]

Thank you Amaruk ...

I had something very similar typed up, but didn't have time to do all the PPD calculations like you did.

Excellent explination of that topic!

[mklvotep]

I guess i would rather not have that happen... Thanks for the reply.

[toTOW]

It looks like you don't have all the details about the situation ... the core is still in beta, and you shouldn't run it (and of course, you can not conclude without all the details).

We (or PandeGroup) will post an announcement with all the detail when the core will be released.

[Indignity]

Can I interject & ask what was wrong with the previous cores like the ones that were being pumped out 1-2 months ago?

I can tell you what was right with them. The nVidia GPUs handled them very well once you got your core/shaders dialed in. It was cool watching a card putting out 4000PPD suddenly jump to 5000PPD+ by a simple overclock of the cards, plus they weren't in jeopardy of reaching their thermal limits constantly. I don't know what's better for the manufacturers, sending out units that stress the contributor's hardware to the point of failure or letting the contributors burn up their own hardware by overclocking their equipment in an attempt to generate more points. AMD/Intel & nVidia are the winners in the end.

Seriously, I don't understand what happens to the "old" WUs & why the hell the cores are forever in constant change!!!!

Blah, blah, blah points........ Well, yes, they ARE important regardless of what anyone says. It's a measure of each individual's contribution and how they can make an educated decision on purchasing more equipment to contribute to this "Project". I won't go on a diatribe but I wish there wasn't a "secret society" behind this.

VJ Pande & the rest of the group needs to take a serious look at how WCG is run from top to bottom and consider making some changes. Until then, this project will continue to lose contributors because you are basically telling them that their contribution means nothing. I know because I have slowly regressed in my F@H status because of all of the frustrations & now my equipment is going towards a cause that is very well documented to be doing exactly what they say it is.

[mklvotep]

It looks like you don't have all the details about the situation ... the core is still in beta, and you shouldn't run it (and of course, you can not conclude without all the details).

We (or PandeGroup) will post an announcement with all the detail when the core will be released.

Don't look at me I didn't run it.lol. My whole point was, i had 16 cards avging 80000ppd now with 20 im avging 73000ppd. then a member on our forum tested the beta and it looked as if another huge ppd drop was on its way. Some people are just upset. I'm dedicated to this project , In less than a year I have over 8million points and 17000 wu's. It's just concern you know..