The Folding@Home GPU Statistics Database

[compdewd]

Hello Folding Community!
This is the update thread for The Folding@Home GPU Statistics Database!

This particular database was designed and is maintained by Patrick Rebsch (compdewd). There are other Folding@Home GPU statistics databases out there and you are free to choose which one you reference for your purposes. In fact, it is suggested that you use all of them that you can find especially if you are going to base a purchase off of the data. (The database here sometimes imports data from other databases, but this only occurs by manual entry and should not be expected.) As with all databases of this nature, the information (such as PPD and TPF) is provided by submissions by others and may be incorrect (see the database disclaimer). This database tries to keep high data integrity using a few different methods:

• Submission of data by others is first entered into a secondary table that is manually parsed and submitted into the main database by Patrick, himself.
• Project data is collected from Stanford once a day to ensure that the PPD for each record is kept up to date. Do note that PPD for each record is not automatically altered to be correct, but is rather calculated and presented on this page.

There are also services that the database offers that you may not find in other databases:

• The ability to see a summary of each graphics card model (per project) rather than all the records individually
• The ability to see one type of graphics card chipset at a time
• The ability to see one type of graphics card model at a time
• The ability to filter by a particular project to compare graphics card models against each other relative to a certain project
• The ability to filter by a range in PPD
• The ability to filter the results by only active Folding@Home projects
• The ability to sort the results by PPD
• The ability to hide some of the columns from view so the page does not venture off-screen on smaller screens

This database strives to offer as many services to the user as possible so it fits the needs of as many users as possible. Feel free to suggest services that should be offered, suggest enhancement to a service, or just comment on how much you like the database!

If you would like to submit data to the database, either post to this thread or use the form found here. You may also email Patrick at [email protected] if you have files such as spreadsheets of your own, or, if you use HFM.NET, the "WuHistory.db3" file. If you post to this thread, please include as much of the following information as possible in your submission:

• Graphics card chipset, model, and driver version
• FAHCore (e.g. 0x15) and core version (e.g. 2.22)
• Project ID that you have worked on or are currently working on
• Average TPF of said project
• Average PPD of said project
• GPU core clock speed and specify if it is overclocked

If you use HFM.NET, you may email [email protected] with your "WuHistory.db3" file and a little bit of extra information. This extra information is the connection between the "Slot Name" and the graphics card. For example, when you go into "Work Unit History Viewer", the "Name" column in the WU History window is the "Slot Name", and the graphics card is whatever card is used on the corresponding Slot Name's number, which is the slot number.

Submitting data of multiple WUs is encouraged. Even if you see an identical or very similar entry in the database already, please still consider submitting your information because if more sources are saying the same thing then the reliability and validity of the information increases. Thank You!



--- Original Post Below ---
I would like to start gathering data on GPUs and their points per day (respective to each project) to make a comparison database. But before I begin asking for information, I would like your feedback on if this is something worth doing and if you think this would be beneficial to many people. I would imagine most of those looking for graphics cards are not looking for them for the primary reason of folding, but rather to game and whatnot, so, would making a large database comparing GPUs' PPD really make a difference to many people when they are looking at graphics cards? I imagine I am probably one of the minority considering buying a graphics card primarily to fold and I know that I would like to have a comparison of GPUs to look at before I buy a card, but would many others?

Also I realize that assembling a database like this is a big task for one person, but if it is worth it to a lot of people, then I'm willing to give my time.

Thank you ahead of time for your responses

[kiore]

The question regarding this GPU vs that one is often asked, but goes out of date fairly quickly, this is likely to be a hot topic once the Kepler series cards are optimized. I agree it is needed to compare work units as these really vary the PPD considerably.
I think it is useful here as many come here for accurate information for folding only or folding mainly GPU choices. Am happy to submit data.
Perhaps as there is an old list this new one could just focus on cards released since then to simply the task somewhat.

[compdewd]

Thanks for your quick response, kiore! I'm glad you support the idea

I found something like an old list: viewtopic.php?p=42917, but my current plan is to add all the data I receive from now on, so I will not be importing data from old sources like from that link from 2008. This may change, but I think I will be receiving enough new data that I can disregard old data like that

[iceman1992]

Yes I think it is definitely something worth doing.
As people have asked many times (and will keep asking) about this GPU vs that GPU, normal PPD for this GPU, etc.
It will certainly be beneficial to many people.

[Jesse_V]

Yes I think it is definitely something worth doing.
As people have asked many times (and will keep asking) about this GPU vs that GPU, normal PPD for this GPU, etc.
It will certainly be beneficial to many people.

Agreed.

[compdewd]

Well 3 for, 0 against: I think I will start collecting information now.

You may post to this thread, private message me, or email me at [email protected]. I will post a link to the database/datasheet when I have an acceptable amount of data

Edit: Sorry, forgot what to specify Please refer to the original post for what kind of data to submit

[bruce]

Make that 4 to 0. I think this is a fine idea.

A couple of comments, though.
1) Although you don't plan to import the old data, we do get questions about almost every supported GPU so it might make sense to leave a future opening that can incorporate data from as far back as NVidia's G80 series an AMD's HD5000 series. Don't let that slow you down with your primary goal, though.
2) GPU overclocking is common, so you're going to see considerable variation even within data for identical hardware.
3) Some time back, VijayPande mentioned that they'd like to incorporate a bonus program for GPUs into the FAH projects. I have seen no recent statements about it and there's no prediction when, if ever, it will be rolled out but if a QRB system does come to pass, it may seriously impact the data you're able to gather. Rather than gathering data in PPD, perhaps there's another fundamental unit that can be gathered such as TPF and baseline Points Per WU that won't need to change if a bonus is later added.

[compdewd]

Thank you, Bruce, for the suggestions and support!

I think I will adopt your suggestion of TPF rather than my idea of PPD. I won't require people to specify baseline credit since that's something I look up myself.

Now I just need to get myself together and state what I want!

Edit: Please see bolded portion of the original post for details

[Dark Pulse]

A question: Does this extend to hardware that could be considered "beta" class as well, such as GeForce 6-series stuff?

That may be a bad thing right now, though, considering right now there's no "public" support for them at this time. If it's permitted, though, I could certainly share numbers, but I'll wait to hear both from the OP and on a mod/admin's okay before doing so.

[kiore]

OK some test results to see how this goes:

GPU: GTX 560 ti 384 shader version.
Overclock, yes 900MHz
Driver version: 301.42 .
Client Version: V 7. 1.52
Fah core 0x15
Project: 8008 .
TPF: 2m 07s
est PPD: 17075.

#2 card same as above on same project:
TPF: 2m 06s
estPPD 17211.

Note, to confuse things there are 3 different GTX 560 ti's with different shader counts.
I have rounded up the PPD numbers.

edit I have added in client and core version.

[compdewd]

Dark Pulse, of course, I will do whatever an admin thinks, but I think that beta hardware should be included in this because it may give people an idea of what the hardware will be like when it is publicly supported and if they should wait to buy or not. Also, considering the beta forums are open to the public, I don't see a problem with having beta hardware TPFs available to the public. Then again, I'm not an admin

[kiore]

Dark Pulse, of course, I will do whatever an admin thinks, but I think that beta hardware should be included in this because it may give people an idea of what the hardware will be like when it is publicly supported and if they should wait to buy or not. Also, considering the beta forums are open to the public, I don't see a problem with having beta hardware TPFs available to the public. Then again, I'm not an admin

If it is allowed then a separate 'Beta', or 'unsupported' would flag this.

[bruce]

When a FahCore is being developed/optimized/fixed, there can be significant TPF variations from one version to the next. (That's implied in the beta team rules of engagement.) Once a FahCore is released to the general public, variations can still happen but they're much less likely and/or are more likely to be quite small.

If you choose to collect beta information, you'll probably need to monitor it for changes and figure out a way to purge data associated with an earlier version of the FahCore, should that be necessary. I suppose collecting the FahCore version number (not the FahCore number) with the other data would help you manage that well enough to keep your results accurate.

(@kiore: you didn't report that information.)

[JimF]

MSI GTX 560 Ti (384 shaders); 833 MHz; not overclocked
Driver: 301.42
Client: 7.1.52
Core: 0x15:Version 2.22
P7626
TPF: 5 min 54 sec
PPD: 12660

==================================
Radeon HD7770 (640 processors); 1 GHz, not overclocked
Driver: 12-8
Client: 7.1.52
Core: 0x16:Version 2.11
P 11293
TPF: 4 min 57 sec
PPD: 5338

[compdewd]

I don't want everyone to get too excited about this yet. I can only handle so many different types of data. I was planning on this being a simple spreadsheet to compare GPU A TPF of project X to GPU B TPF of project X. I realize that there are many more factors to comparing GPU A to B, so I will record all the information I receive, only actually using what I can, and hopefully someday the database can be expanded to something greater based on the extra information I record. Therefore, extra information is now encouraged.