Folding Forum

On a new install, there are no cores automatically installed. The cores come when a new WU is assigned that needs a new core. The WU will have a minimum version required. If you do not have a previous core (from this being a new install) or the core you have is older than the minimum required then the client will download and install the most recent version which at this time for FAHCore_15 is v2.25. As a default, it would seem to break the common sense rule to automatically download an old version in preference to the newest version that exists. It is just annoying and unfortunate that the newest version happens to be one that most people dislike unless you have a Kepler. This is not something new that is being forced upon you, rather this is how the clients have operated ever since they first started automatically updating cores.

Also Kepler owners don´t like v2.25, but we simply have no choice.

Heiko

that the newest version happens to be one that most people dislike unless you have a Kepler.

Oh the irony... this core doesn't work properly with kepler either. It offers some expanded functionality out of the box, but doesn't utilise the full potential of the card like the beta flag does. From all aspects, without confirmation that it's doing something behind the scenes that we don't know about, it's a failure for both fermi and kepler architechture when compared to what we had prior to it's release.

PPD (for Fermi) is the ONLY aspect that could be called a failure, not all aspects. v2.25 works just fine, and actually produces more science (Keplers would be getting 0 PPD without it). You just don't like the PPD it produces on Fermi. Kepler PPD is way better than before.

The fact that PG has released this fahcore to the public is all the confirmation we should need. Yes, I would like more details, as World+Dog has already stated before, but I don't require those details to continue folding. If the details come, great, if not, the science is still getting done.

As always, past folding performance is no guarantee of future performance. Stuff changes. Fold on.

7im wrote:The fact that PG has released this fahcore to the public is all the confirmation we should need.

Ah yes, the ol' PG infallibility argument. Makes me giggle every time!

Yes, I would like more details, as World+Dog has already stated before, but I don't require those details to continue folding.

1st corollary:

Since effective communication is consistently abdicated, FOLDERS really shouldn't desire it.

7im wrote:PPD (for Fermi) is the ONLY aspect that could be called a failure, not all aspects. v2.25 works just fine, and actually produces more science (Keplers would be getting 0 PPD without it). You just don't like the PPD it produces on Fermi. Kepler PPD is way better than before.

The fact that PG has released this fahcore to the public is all the confirmation we should need. Yes, I would like more details, as World+Dog has already stated before, but I don't require those details to continue folding. If the details come, great, if not, the science is still getting done.

As always, past folding performance is no guarantee of future performance. Stuff changes. Fold on.

Most of the sites I visit the Kepler folders are still not running there cards not enough PPD to justify the expense is the reason most often given the 2.25 core is a fail all the way around it lowers the production on Fermi cards and does not have enough production on Kepler to entice people to fold with them. I myself delete the 2.25 core every time it is downloaded and replace it with 2.22, It is either that or shut the cards down, the 2.25 was not developed because 2.22 was producing bad science, it was devolved as a crippled 2.22 core (changed the environmental variables to a longer pause to deal with screen lag). All of the current WU's work fine with the 2.22 core on Fermi and produce good science. 2.25 should not be forced upon those who do not need it. 2.25 is a step backwards not forwards

sswilson wrote:It offers some expanded functionality out of the box, but doesn't utilise the full potential of the card like the beta flag does.

You're as bad as a politician from the "other" party. You really should get your facts straight before you say things.

Previously, the beta flag downloaded V2.25. Now the Adv flag downloads V2.25. The use of "full potential of the card" is identical when running V2.25 with beta and when running V2.25 with Adv.

bruce wrote:

sswilson wrote:It offers some expanded functionality out of the box, but doesn't utilise the full potential of the card like the beta flag does.

You're as bad as a politician from the "other" party. You really should get your facts straight before you say things.

Previously, the beta flag downloaded V2.25. Now the Adv flag downloads V2.25. The use of "full potential of the card" is identical when running V2.25 with beta and when running V2.25 with Adv.

Well, maybe it was just the luck of the draw, but the WUs I was pulling on the new V7 (without using the beta flag) were only running at 72% power, so yes, I'd call that not utilising the full potential of the card. In hindsight, maybe I got my facts wrong WRT what core it was using, but again it points to the lack of communication on this rollout.

edit: And on that note, I'm outta here. I'll check back when GPU QRB is fully enabled.

I am getting lower scores on my GTX 560/GTX 560 Ti just like everyone else with 2.25. But they run at 99% GPU utilization according to GPU-Z, and at the same temperatures as before insofar as I can see (Project 8018). My machines are dedicated to avoid the screen lag problem (along with heat and fan noise problems), so it does not do either me or PG any good to suffer a reduction in points for that purpose.

However, PG has an incentive to get as many results as possible from my equipment, and I am inclined to think that it has something to do with the science. The optimizations that work for Fermi may not work for Kepler, and they might produce different results; not good. Furthermore, the optimizations may limit the type of problems they can work on, though that is of course speculation on my part. But the point is that the PPD are meaningless without the science. We are here to produce results, not generate random numbers.

p7622-p7626 just got a big boost in points from 5187 to 14093 and associated PPD! You may want to do a "download projects from Stanford" (tools menu) in your copies of HFM.net to locally update its database.

Seriously PG, if you are going to update these projects -- Why not p7620 or p7621?

Psummary wrote: 7620 171.64.65.105 p7620 1980 12.60 17.90 5187.00 100 OPENMMGPU Description tjlane 3.00
7621 171.64.65.105 p7621 1980 12.60 17.90 5187.00 100 OPENMMGPU Description tjlane 3.00
7622 171.64.65.105 p7622 1980 37.94 48.78 14093.00 100 OPENMMGPU Description tjlane 0.00
7623 171.64.65.105 p7623 1980 37.94 48.78 14093.00 100 OPENMMGPU Description tjlane 0.00
7624 171.64.65.105 p7624 1980 37.94 48.78 14093.00 100 OPENMMGPU Description tjlane 0.00
7625 171.64.65.105 p7625 1980 37.94 48.78 14093.00 100 OPENMMGPU Description tjlane 0.00
7626 171.64.65.105 p7626 1980 37.94 48.78 14093.00 100 OPENMMGPU Description tjlane 0.00

7620 and 7621 had their K factor increased from 0 to 3...

OK.

Then why are they being treated differently? Two projects getting bonus points, while the others remain with the old benchmark system but with far more points. It seems confusing.

These projects are general release/advanced. Is this the start of re-benchmarking standard WU's to the new QRB benchmarking system? I would have expected a statement from PG about the status of the testing before that officially happened.

Not that I'm complaining, I'm just confused and surprised.

Just noting the difference.

I wouldn't care to speculate, so am looking forward to some sort of directive from someone who knows

After updating HFM, I get the following PPD:

• 460@1600 21.5k
• 560Ti 384 cores@1900 25.8k
• 570@1600 31.3k

Nice, I no longer dislike the PPD on these