Folding Forum

All of this year, I do not recall ever seeing a PPD estimate from V7 for my GPU slot, mainly because the credit estimates are always "Unknown". V7 provides an ETA and a TPF, but it doesn't seem to know how much the WU is worth. IIRC from earlier posts, this is because the servers are not telling the client this information, and that an update in the server software will fix this issue. Is this correct? If so, I'd really like to see an update, because its weird to have V7 display PPD estimates, but have the sum all wrong because there's no PPD estimate for the GPU slot. This didn't seem to occur on my desktop, but on my laptop with it's Nvidia 240m, it does. Currently the laptop is assigned to the 171.67.108.21 Work Server and the 171.67.108.26 Collection Server. This is not a problem that cuts down on productivity, but I think it's more important than just a cosmetic issue. I believe that there's a possibility that some folks are put off by the seemingly low PPD sum, and don't know that their GPU is actually producing a lot of points, but they just can't tell. I realize that nobody wants to update a server that is working, and that updating a server would most likely require a bit of downtime, which is disruptive. But perhaps they will gain more in the long run. At least V7, and perhaps even F@h as a whole, would look better for it.

Jesse_V wrote:All of this year, I do not recall ever seeing a PPD estimate from V7 for my GPU slot, mainly because the credit estimates are always "Unknown". V7 provides an ETA and a TPF, but it doesn't seem to know how much the WU is worth.

See this viewtopic.php?f=61&t=22020

iceman1992 wrote:

Jesse_V wrote:All of this year, I do not recall ever seeing a PPD estimate from V7 for my GPU slot, mainly because the credit estimates are always "Unknown". V7 provides an ETA and a TPF, but it doesn't seem to know how much the WU is worth.

See this viewtopic.php?f=61&t=22020

Right, only mine doesn't show a PPD estimate at all. ETA = 1 hour 22 mins, Base Credit = Unknown, Estimated Credit = Unknown, Estimated PPD = Unknown, TPF = 1 mins 48 secs.

Yeah I know, mine too. Just curious, what GPU do you have? I have the GTX260 which uses fahcore_11.
I'm wondering whether PPD estimates are currently limited to fahcore_15 servers? Or OPENMMGPU servers, whatever that is?

Recall harder.

Recall a PG mention that GPU would get QRB? Recall that all WS would be upgraded to suport V7 features? So, yes, there are plans to update. Also recall PG doesn't give ETAs.

iceman1992 wrote:Yeah I know, mine too. Just curious, what GPU do you have? I have the GTX260 which uses fahcore_11.
I'm wondering whether PPD estimates are currently limited to fahcore_15 servers? Or OPENMMGPU servers, whatever that is?

Nvidia GT 240m. FahCore 11.
https://simtk.org/home/openmm/

7im wrote:Recall harder.

Recall a PG mention that GPU would get QRB? Recall that all WS would be upgraded to suport V7 features? So, yes, there are plans to update. Also recall PG doesn't give ETAs.

I also recall that there's a "soon" and "not soon" category. I never asked for an ETA, but how about if I ask instead, any ideas on which category is the update is in?

Sure, go ahead and ask for a non-ETA ETA.

I'm willing to venture a guess. It's strictly unofficial -- based on what might or might not happen, so don't shoot me if I'm wrong. I predict it will be "soon." I think I already went out on a limb in a previous post and suggested it would be "by the end of the year" (mostly to put the proper perspective on the word "soon"). Of course I'd be very happy if it was "by the end of the month" instead, but who knows.

In the past, "soon" was reserved for things that are either being worked on or things for which a work plan has been made and "not soon" is reserved for things that are beyond those stages.

The example 7im mentioned is a good one because it shows how various plans interlock. To add QRB to GPUs, all the GPU servers have to be updated. When QRB was added to SMP, all the servers had to be updated to some minimum server code version number. To support OpenCL any V6 clients had to be updated to V7 which means the GPU servers that distribute OpenCL assignments had to be at least some (different) minimum version number. Since FahCore_a4 runs on either SMP or on the Uniprocessor, servers that distribute A4 WUs had to be at least at some (other) minimum version of the server code so some servers that distribute work to Uniprocessor clients may inform a V7 client of the Baseline Points data and other servers that distribute uniprocessor work may not (specifically FahCore_78). Moreover, both the V6 and V7 client has to be smart enough to work with servers that do and do not distribute Baseline Points data.

Oh, and there's also a question of the active projects on that server. Will the active projects on Server Version N simply migrate to Server Version (N+1) or do they need to be modified somehow? ....I could obviously go on and on.

In any case, my point is that deciding which feature can be updated without causing widespread difficulties is a complex process. Certainly updating the code on that server (and perhaps a couple of others) has some obvious benefits from our point-of-view but it would be a shame to do that an then find out that there's a new version of the server code due next week which will be needed "soon." Only updating it once is obviously preferable that updating it twice.

thanks that gives me some hope for QRB for GPU. lets hope its soon.

Yup, GPU QRB - the sooner the better, but I appreciate that it arrives when it arrives and that will be soon enough

(This coming from a user with a HD4770, so not a GPU folder!)