p7620 and 7621 kFactor

[rwh202]

Hi,

These two core_15 projects are showing a kFactor of 3.0 in the project summary, which I guess is being picked up by FAHControl to give some quite spectacular PPD estimates.

Base credit is 5187 with an estimated credit of 77239 and 323955 PPD on a stock 780Ti. It would be nice if core_15 was really able to give bonus points but I suspect that's just wishful thinking...

[P5-133XL]

I confirmed that psummary does now list a K-factor for these projects. I've also confirmed that the v7 clients are estimating numbers that imply QRB. I do not know if the servers are giving the QRB points.

The researcher in charge of the projects has been notified.

[jadeshi]

Sorry about the confusion. This was an oversight on my part. I've fixed 7620 and 7621 such that they give a k-factor of 0 like all other Core15 projects.

[P5-133XL]

Will people that received projects with QRB get the credit for QRB now that it is off?

[jadeshi]

I've put 7620 and 7621 on accept mode temporarily, so all WU's already sent out will get the erroneous k=3.0 bonus, but future assignments will use the correct k-factor=0.

[P5-133XL]

Perfect.

[rwh202]

Thanks for looking into and sorting this so promptly.

I'm not sure given the recent stats downtime, but it doesn't look like a bonus was actually awarded, but I might be wrong.

[P5-133XL]

I agree, I'm not seeing those points in EOC either even though the client says so. So I checked the database:

05:56:45:WU00:FS00:Sending unit results: id:00 state:SEND error:NO_ERROR project:7620 run:433 clone:0 gen:160 core:0x15 unit:0x000000af664f2dd14e42f5d52579c412
05:56:45:WU00:FS00:Uploading 798.58KiB to 171.64.65.10505:56:48:WU00:FS00:Upload complete
05:56:48:WU00:FS00:Server responded WORK_ACK (400)
05:56:48:WU00:FS00:Final credit estimate, 58982.00 points

Hi P5_133XL (team 10047),
Your WU (P7620 R433 C0 G160) was added to the stats database on 2014-04-17 23:14:26 for 5187 points of credit.

No Joy! So much for QRB on Core_15...

[OVV]

I'm currently folding a 7620 WU and the expected reward is 5187 points that is about one third of all the other 762x WU (14093 pts)
My opinion is that the 7620-7621 should have the exact same value of all other 762x including point and deadlines due to same number of atoms.

Thanks

[Rel25917]

Take a look at the deadlines for 7620/1 and the other 762x units. There is a reason the points are different. I'm guessing the 7620/1 finish in about a third the time.

[bruce]

K-Factors and bonus points may vary. Benchmarking only considers the baseline points before the bonus is added.

Too many people compare total PPD after the bonus points are added and say then complain that they're not benchmarked correctly. You need to compare BASELINE POINTS per FRAME, not Bonus Points or Total Points.

[OVV]

The TPF for WU 7620 is exactly the same as every other 762x WU (with my Nvidia quadro 2000 about 16 min), number of atom is the same so the most easy assumption is that also the reward should be same.

[7im]

The TPF for WU 7620 is exactly the same as every other 762x WU (with my Nvidia quadro 2000 about 16 min), number of atom is the same so the most easy assumption is that also the reward should be same.

The points on the benchmark GPU can be assumed to be the same across all these projects. However, if your hardware is different, that assumption does not always hold true. Differences in memory size, speed, or core counts can translate in to points differences from the benchmark. This is most common when a project uses resources right on the borderline of core counts or memory. That may not be the case here, just that points won't always be consistent with the benchmark.