GPU to 72 status...

[kladner]

I have another "hung" DC: completed on PrimeNet, still showing in GPU72.
https://www.mersenne.org/report_expo...4500411&full=1

[kladner]

There are now four (I think) DC completions not showing in GPU72.

[petrw1]

Originally Posted by chalsall

Yes. Looking at James' GPU cross-over analysis, it makes sense to TF from 45M and up to 73 bits on most cards.

Where do I find this?

[petrw1]

Originally Posted by chalsall

Yes. Looking at James' GPU cross-over analysis, it makes sense to TF from 45M and up to 73 bits on most cards.

What about 50-59M exponents?

Should they go beyond 73 bits?

[James Heinrich]

Originally Posted by petrw1

Where do I find this?

For example for a GTX 1080:
http://www.mersenne.ca/cudalucas.php?model=684

[chalsall]

Originally Posted by petrw1

What about 50-59M exponents? Should they go beyond 73 bits?

Yes. 57M to 59M should go to 74 bits. But since the ranges we're currently working are in the Cat 2, 3 and 4 ranges I figured we should (mostly) finish those off first, and then move up there early next year.

[petrw1]

Originally Posted by chalsall

Yes. 57M to 59M should go to 74 bits. But since the ranges we're currently working are in the Cat 2, 3 and 4 ranges I figured we should (mostly) finish those off first, and then move up there early next year.

This https://www.gpu72.com/reports/estima...tion/primenet/ makes it look like all the 5xM are in the queue.

[chalsall]

Originally Posted by petrw1

This https://www.gpu72.com/reports/estima...tion/primenet/ makes it look like all the 5xM are in the queue.

No more so than 41M through 44M. The Total column shows "Completed".

I actually need to update that report to reflect the work needed in the high 5xMs (and add a 74 bit column). I have done so for the Current TF Factoring Depth report already. I'll work on adding that this week.

Edit: Actually, I just updated that report.

[petrw1]

Originally Posted by chalsall

No more so than 41M through 44M. The Total column shows "Completed".

I actually need to update that report to reflect the work needed in the high 5xMs (and add a 74 bit column). I have done so for the Current TF Factoring Depth report already. I'll work on adding that this week.

Edit: Actually, I just updated that report.

[chalsall]

So, one of the drives on the GPU72 server is failing.

It is scheduled to be replaced in the next hour or so. It's "hot-swappable", and one of the drives in a RAID1 set, so there should be no downtime.

But... Crap happens...

P.S. I seriously hate Seagate. In this case I have no choice, but when I do I refuse to buy Seagate drives.

# smartctl -a /dev/sda

=== START OF INFORMATION SECTION ===
Model Family:     Seagate Barracuda 7200.14 (AF)
Device Model:     ST1000DM003-9YN162
Serial Number:    S1DA97XG
...
SMART Self-test log structure revision number 1
Num  Test_Description    Status                  Remaining  LifeTime(hours)  LBA_of_first_error
# 1  Short offline       Completed: read failure       10%      7476         244751313
# 2  Extended offline    Completed: read failure       90%      7462         244751313
# 3  Extended offline    Completed: read failure       90%      7459         244751313
# 4  Extended offline    Completed: read failure       10%      5487         1953513120
# 5  Extended offline    Completed: read failure       10%      5460         1953513120
# 6  Extended offline    Completed without error       00%      5446         -
# 7  Extended offline    Completed: read failure       10%      5373         1953512248

[James Heinrich]

Originally Posted by chalsall

one of the drives in a RAID1 set, so there should be no downtime.
...when I do I refuse to buy Seagate drives.

I also avoid Seagate like the plague. If I had RAID1 with two Seagate drives and one failed, I would expect the second one to also fail before the array could be rebuilt.
I haven't bought a Seagate drive in 8 years now because about 80% of the Seagate drives I've ever had have died on me, sometimes getting replaced 3 times in the warranty period.
Not quite as bad as the Deathstars, but I still don't trust them to hold my data.