I'm no expert, but from what I recall they do that when a server went back online, a bunch of points got re-calculated, or some other problem got fixed on their end. A good way to make sure is to see if other users all experienced a big spike on the same day.
For someone that has not changed his/her rigs or added any, it could be from a server glitch that has been corrected. Most of the time, in this scenario, the points are very close to zero prior to the massive jump...
The other option... someone just got a quadcore rig up and running... maybe more than one rig...
Mike, what are the system specifications? Let's wring everything out of that box that we can.
It really is incredible
This is my daily use machine.. it is an E8400 dual core (3GHz, 6MB L2). The 8800GTS is the older G80 revision with 640MB of GDDR3. Everything is running at stock clocks right now. Do higher GPU clockspeeds increase production much? I'm only running a single GPU2 client at the moment. Seems like it'll be turning in around 4000ppd by the looks of it.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited November 2008
Damnit, I really wish I would of bought a SLI board now.
Two of my motherboards are SLI. When I bought them, both Gigabyte P35-DS4, it was for the quality of the board and overclocking potential. At the time I remember thinking what a big waste of space the second PCI-e video slot was! Ho ho ho! I don't think that way now. System No. 1 in signature turns out 11500-12000 PPD.
Do higher GPU clockspeeds increase production much?
GPU core and memory - no. Shader clock, yes! A 75MHz boost on shader is worth about 300 or more PPD. But for Folding, if you don't overclock the shader, there's no point in overclocking.
BTW, my comments above refer only to Nvidia GPU Folding. I have no experience with ATI GPU Folding.
Two of my motherboards are SLI. When I bought them, both Gigabyte P35-DS4, it was for the quality of the board and overclocking potential.
Not to Nitpick.. but those boards are Crossfire, not SLI. You cannot run 2 Nvidia cards in SLI on those boards.
You can however, run 2 video cards on any board with 2 or more slots and fold on both, I would assume. Since the folding client runs on any card that is capable, even if they are not linked in SLI or Crossfire
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited November 2008
Shows you how much I know about dual video card gaming!
You can however, run 2 video cards on any board with 2 or more slots and fold on both, I would assume. Since the folding client runs on any card that is capable, even if they are not linked in SLI or Crossfire
Yes. Also, it is necessary in Windows, in some configurations, to force the OS to assign one of the multiple GPU Folding clients to the second video card, otherwise Windows tries to make on GPU/video card process more than one client. With my setup, I set one each video card to -GPU 0 and -GPU 1 in their respective startup shortcuts. I also set the second video card to extend to a second (non-existant) monitor. Some Vista users install a dummy plug on their second video card to simulate a second monitor.
GPU core and memory - no. Shader clock, yes! A 75MHz boost on shader is worth about 300 or more PPD. But for Folding, if you don't overclock the shader, there's no point in overclocking.
Cool.. I'll give it a shot. It has been a while since I overclocked this card, but I believe it'll do close to 700MHz core (the shader will increase along with it, but I'm not sure at what ratio). It sits at 540MHz at default. After two solid days of folding it seems to be doing about 3800ppd. I'll try to overclock it tonight and see what kind of increase I can see.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited November 2008
Right, with Riva Tuner, just check the box, "Link clocks." That will automatically boost the shader clock at the correct ratio.
3800PPD, on an 8800GTS? You should be getting 5000-5800PPD on most work units. There are some out there now that will only process in the mid-3000's, but they are the minority.
Affinities: ensure that you have GPU2 set for one core and your CPU Folding client set for the other core. They should not share.
EDIT: Your GTS, is it the older version "640?" It may have a lower shader clock than the newer G92-based cards.
I just bought an Asus EAH4870 and am very disappointed in my production after reading this thread. I've averaged only 3200ppd in the three days before I upgraded to a Q8200 this past Saturday. My current avg is 2946 running the GPU and the regular client. I have yet to start overclocking but this is obviously subpar performance considering I'm running quad with a much more expensive video card. Going to have to do some checking into that...
Regardless, glad to see you back ll!
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited November 2008
TBonz, no that is not substandard. 4870s are great for gaming, but their driver support for scientific calculations is not developed like NVidia's CUDA. There is nothing wrong with your video card, the drivers just aren't optimized for 3D modeling like CUDA. Sorry, my technical terminology is not well developed in this arena. If ATI gets around to supporting pure processing irrespective of gaming 3D, they ought to be smoking the Folding charts.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited November 2008
After two solid days of folding it seems to be doing about 3800ppd
There are some out there now that will only process in the mid-3000's, but they are the minority.
If you are Folding a Project 5478, mid-3000 is about right right for a 9800GT or 8800GT. I don't know what the clock is on your GTS or what projects you have been folding. Most of the Project 5506s have been 5000-5600 for an 8800GT or 9800GT.
It is an older G80 revision 8800GTS with 96 SPs. An 8800GT is probably about 10-15% quicker than it in this type of scenario.
My overclocking venture didn't go well last night. Not sure why, but I used to be able to push this card really far. I kept getting non-responsive driver failures with even small increases. Either I'm doing something wrong or the card isn't what it used to be I'm going to play around with it a bit more later.
I'm only running the GPU2 client so I'll launch a single CPU client as well. What is the best way to get maximum production out of the CPU console client? Also, are there any flags or advanced options that should be turned on in the GPU2 client to get higher-point WUs?
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited November 2008
What is the best way to get maximum production out of the CPU console client?
Ensure that the Config file is set for the client to accept large work units and that Advanced Methods is on. The latter can be set in Config or or manually through a flag.
Also, are there any flags or advanced options that should be turned on in the GPU2 client to get higher-point WUs?
With Nvidia GPU Folding projects, the only open to the public right now that is slow is the 5748. All the others are very close in production potential. Same guidance - Advanced Methods and allow large work units - in config. You do not need to set any flags for a single GPU client.
To reiterate what we wrote before, make sure the CPU and GPU clients each have affinity for a different core. I highly recommend Process Lasso.
Looking even better now! I've added the regular console client to my second core and did a bit of tweaking. Not bad for one machine plus an old P4 doing 'small' WUs somewhere still folding under my name
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited November 2008
It's amazing. And that graph is not the result of inflation of awarded points. It is actual production. GPUs are just incredibly productive in some types of work.
After running it for over a week, I'm seeing some pretty impressive results!
Is it possible to run an additional CPU client (for two regular console clients) along with the GPU2 and single CPU client I'm running today? It seems that the GPU2 client uses less than 10% of one core. Could I run another console client at a slightly lower priority to ensure that it doesn't slow down the GPU2 client?
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited November 2008
Highest number of CPU clients running on a CPU+GPU configuration that I've heard of is 2. That would be two 'normal,' standard clients. For production, the best combo is one SMP client plus multiple GPU clients. But, if you go over two GPU clients, then no CPU client.
Maybe there are more configurations that I haven't been exposed to. Check out FaHWiki and the Folding Forums.
lemonlime
Canada Member
Comments
I could just be talking out of my ass here, too
Fold on dude, fold on.
Too true.
The other option... someone just got a quadcore rig up and running... maybe more than one rig...
On an SLI board with a quad core CPU: 2 X GPU clients plus 1 X Windows SMP client. A Folding farm all in one box!
Way to go, Mike. Spread the word. Spread the enthusiasm. A single $100 8800/9800 GT can bring 4500-5500 points per day.
Mike, what are the system specifications? Let's wring everything out of that box that we can.
Damnit, I really wish I would of bought a SLI board now.
It really is incredible
This is my daily use machine.. it is an E8400 dual core (3GHz, 6MB L2). The 8800GTS is the older G80 revision with 640MB of GDDR3. Everything is running at stock clocks right now. Do higher GPU clockspeeds increase production much? I'm only running a single GPU2 client at the moment. Seems like it'll be turning in around 4000ppd by the looks of it.
BTW, my comments above refer only to Nvidia GPU Folding. I have no experience with ATI GPU Folding.
You can however, run 2 video cards on any board with 2 or more slots and fold on both, I would assume. Since the folding client runs on any card that is capable, even if they are not linked in SLI or Crossfire
Cool.. I'll give it a shot. It has been a while since I overclocked this card, but I believe it'll do close to 700MHz core (the shader will increase along with it, but I'm not sure at what ratio). It sits at 540MHz at default. After two solid days of folding it seems to be doing about 3800ppd. I'll try to overclock it tonight and see what kind of increase I can see.
3800PPD, on an 8800GTS? You should be getting 5000-5800PPD on most work units. There are some out there now that will only process in the mid-3000's, but they are the minority.
Affinities: ensure that you have GPU2 set for one core and your CPU Folding client set for the other core. They should not share.
EDIT: Your GTS, is it the older version "640?" It may have a lower shader clock than the newer G92-based cards.
Regardless, glad to see you back ll!
My overclocking venture didn't go well last night. Not sure why, but I used to be able to push this card really far. I kept getting non-responsive driver failures with even small increases. Either I'm doing something wrong or the card isn't what it used to be
I'm only running the GPU2 client so I'll launch a single CPU client as well. What is the best way to get maximum production out of the CPU console client? Also, are there any flags or advanced options that should be turned on in the GPU2 client to get higher-point WUs?
To reiterate what we wrote before, make sure the CPU and GPU clients each have affinity for a different core. I highly recommend Process Lasso.
I'll be at a million points in no time now
Is it possible to run an additional CPU client (for two regular console clients) along with the GPU2 and single CPU client I'm running today? It seems that the GPU2 client uses less than 10% of one core. Could I run another console client at a slightly lower priority to ensure that it doesn't slow down the GPU2 client?
Maybe there are more configurations that I haven't been exposed to. Check out FaHWiki and the Folding Forums.