I have five Radeon R9 280X cards, only 6 months older than your 750Ti. But yours is OpenCL 3.0 and mine are only 1.2, perhaps Gravity needs a newer OpenCL? Or maybe Gravity uses Cuda on Nvidia cards.

I believe the problem is the onboard memory on each gpu I think gravity tasks take 4096MB, 4GB, of onboard gpu memory and yours only have 3072MB, 3GB. It's an Einstein problem that they keep sending them to cards that can't do the work, it's an ongoing problem at Einstein that they haven't figured it out yet for the BRP4 units either going to older RPi cpu's that can't do those tasks either. They released new tasks with new requirements and failed to figure out how to only send them to pc's that can crunch them.

I believe the problem is the onboard memory on each gpu I think gravity tasks take 4096MB, 4GB, of onboard gpu memory and yours only have 3072MB, 3GB. It's an Einstein problem that they keep sending them to cards that can't do the work, it's an ongoing problem at Einstein that they haven't figured it out yet for the BRP4 units either going to older RPi cpu's that can't do those tasks either. They released new tasks with new requirements and failed to figure out how to only send them to pc's that can crunch them.

But his 750Ti only has 2GB.

mikey:
As far as I can remember, I was told by some very high profile crunchers on EatH, that this 4GB requirement has been eliminated quite a while ago.
Max 2GB will do.

If I remeber correctly, I had a few GTX750ti with 2GB and they used to not work, but then, one day, did.
Sometimes they worked even with only 2GB, because the task didn't need more - so pure luck!
I am using several 4GB NVIDIA GTX750ti.

I just did a search on EatH using "4gb" as argument and found several topics about this solved problem.

Peter,

Sorry, these ones I'm using have 4GB.

mikey:
As far as I can remember, I was told by some very high profile crunchers on EatH, that this 4GB requirement has been eliminated quite a while ago.
Max 2GB will do.

If I remeber correctly, I had a few GTX750ti with 2GB and they used to not work, but then, one day, did.
Sometimes they worked even with only 2GB, because the task didn't need more - so pure luck!
I am using several 4GB NVIDIA GTX750ti.

I just did a search on EatH using "4gb" as argument and found several topics about this solved problem.

I'm going to guess it's the OpenCL version then. Whatever it is, you would think the server could tell before handing them out!

... Same weather here, I'm off to buy some cockatiels!

Nice hobby!

These are the new two:
https://imgur.com/a/OSLSO3G

Peter:
GREAT, lovely ...

Peter:
GREAT, lovely ...

That brings the flock up to 3 cockatiels, 5 lovebirds, 4 amazons, 4 african greys, 7 computers, and a noisy messy house. They're in aviaries, that's not my photo, that's from the woman I bought them from.

Peter:
GREAT, lovely ...

That brings the flock up to 3 cockatiels, 5 lovebirds, 4 amazons, 4 african greys, 7 computers, and a noisy messy house. They're in aviaries, that's not my photo, that's from the woman I bought them from.

Well, at least they have it warm, with 7 computers around them ...

Currently turning some processes off to flush the transitioner backlog. Once that's done I'll turn things back on.

Hi Tom,
was your intervention successful?
On the receiving end of the server nothing changed much, unfortunately. No WUs are being sent (despite that the server status says there are plenty). It seems the server reports numbers that are completely wrong. For example, the server says I have several hundred WUs in progress while, in fact, the last WU my host completed was more than 24 h ago. It seems the server is still a mess and it's a shame that the admins do not care too much about it. I moved my host to other projects entirely now... why should I care about MW@H when even the admins don't.
All best,

Hi Tom,
was your intervention successful?
On the receiving end of the server nothing changed much, unfortunately. No WUs are being sent (despite that the server status says there are plenty). It seems the server reports numbers that are completely wrong. For example, the server says I have several hundred WUs in progress while, in fact, the last WU my host completed was more than 24 h ago. It seems the server is still a mess and it's a shame that the admins do not care too much about it. I moved my host to other projects entirely now... why should I care about MW@H when even the admins don't.
All best,

Be patient! I'm getting more and more work coming through now. The validation queue is shortening, things won't be fully up to date until everything is caught up.

Well, at least they have it warm, with 7 computers around them ...

It's a pity they don't pay the electric bill. More sex please, we're British!

@ Peter

Regarding your five Radeon R9 280X cards, how do you like them? Sounds like you are pretty well pleased. What is the per card power draw under MW@Home? And how many WUs can you do per card in an hour? I have 2 PCIe 16 slots available, and am thinking of replacing a quadro P1000(6 WU per hour) with one of these, and populating the second slot also. Another option is to go new, with Radeon 6600 XTs. But the way you went sounds like more bang for the buck. My power supply is 750W, so hopefully wont need to upgrade that. CPU is a Ryzen 9 3900x.

ps. we have a cockatiel like your 2 new ones. He'll be 20 this year, and active as all get out. His name is Cheeks, and he's an alcoholic. Will post a picture when i figure out how!

[/img]

Regarding your five Radeon R9 280X cards, how do you like them? Sounds like you are pretty well pleased.

Love them. One of them I play games on.

What is the per card power draw under MW@Home?

Less than the 250W TDP, I think about 180W.

And how many WUs can you do per card in an hour?

If I run enough at once on the card to make the % usage sit in the high 90s (this depends on processor speed), they do one per 38 seconds per card.

I have 2 PCIe 16 slots available, and am thinking of replacing a quadro P1000(6 WU per hour) with one of these,

280X are way faster than that, that's 45Gflops on double precision (what MW uses), the 280X is 1024, 23 times faster! Only 3 times faster for single precision (every other project).

and populating the second slot also.

You can run huge numbers of cards off 1 slot. Get some risers designed for bitcoin, thus: https://www.ebay.co.uk/itm/363533746057 - that one is a single, but you can get ones with 4 sockets in one. And you can daisychain them - I tested a 4-way with a 4-way plugged into it, so 16 cards off one slot. And the slot can be 1x.

Another option is to go new, with Radeon 6600 XTs. But the way you went sounds like more bang for the buck.

Work out the electricity cost also.

My power supply is 750W, so hopefully wont need to upgrade that.

Is it a good 750W, or a Chinese 750W? Cheap supplies (eg. CIT) don't like going over 50% power. Good supplies (eg. Corsair) will do the rated power. I run most of mine of a 1kW 12V LED lighting power supply. A lot cheaper than PC power supplies.

CPU is a Ryzen 9 3900x.

Snap! (Almost) - I have a Ryzen 9 3900XT. Wonderful CPU, except AMD's virtualisation implementation sux - if you run anything with Virtualbox (eg. LHC, Rosetta, Cosmology), then the interface gets sluggish, but if it's just for Boinc and you're not trying to interact with it, then it's fine. Otherwise, limit the number of VB tasks.

ps. we have a cockatiel like your 2 new ones. He'll be 20 this year, and active as all get out. His name is Cheeks, and he's an alcoholic. Will post a picture when i figure out how!

Alcoholic!? I've never tried giving mine alcohol, I do homebrew at 23% alcohol (don't tell the taxman) and consume 5 times the recommended daily allowance, which is apparently perfectly fine. 4 times the RDA is apparently the same as being T-Total - less than that and you're actually healthier. My 16 parrots are all chocaholics though, it's apparently bad for them, but they manage to steal some sometimes.

I haven't had a single credit from validation for the past 3 weeks. Will be nice if they ever get the servers working correctly again.
My RAC continues to plummet.
I've stopped the project until it can catch up with my past pending production.

---

I think the RAC from past work (pending validation) is beyond salvation. The hope is that, at some point, the server would start to operate normally and we'll be able to build up RAC again. Although, I'm not very optimistic given the only thing the admins tend to do is to switch off and on services and hope for the best.

I haven't had a single credit from validation for the past 3 weeks. Will be nice if they ever get the servers working correctly again.
My RAC continues to plummet.
I've stopped the project until it can catch up with my past pending production.

You might aswell put more in to get a nice big bonus.

I think the RAC from past work (pending validation) is beyond salvation. The hope is that, at some point, the server would start to operate normally and we'll be able to build up RAC again. Although, I'm not very optimistic given the only thing the admins tend to do is to switch off and on services and hope for the best.

Why would it disappear? Whenever the server gets your tasks validated, you will get credit.

You only not gain credit for any returned results that either error or are inconclusive from my experience.
If they're returned they can probably be counted.

I've got 600 results waiting a quorum partner, 700 valid 100 inconclusive.and 11 errored.

so those 600 could go either way but I could have another 500 results to add to the tally

You only not gain credit for any returned results that either error or are inconclusive from my experience.
If they're returned they can probably be counted.

I've got 600 results waiting a quorum partner, 700 valid 100 inconclusive.and 11 errored.

so those 600 could go either way but I could have another 500 results to add to the tally

As long as the results are useful to science I don't really care, the tiny pittance from Gridcoin I won't miss.