nvidia rtx-2080

[Mark Rose]

Originally Posted by petrw1

Over 3,600 GhzDays per day with TF 74-75.

How many watts is it using to do that?

[petrw1]

Originally Posted by Mark Rose

How many watts is it using to do that?

Seems to me it was in the 260 range.

[Mark Rose]

Originally Posted by petrw1

Seems to me it was in the 260 range.

Not bad!

[chalsall]

Originally Posted by petrw1

Seems to me it was in the 260 range.

[Valleyspeak]: "Like, wow! Like, totally!

Mark, what do you dissipate for your ~3,150 GHd/D? Over how many cards?

While the capex isn't small, over the long term it might make sense (all consumables in).

[petrw1]

I fiddled with various values for:
NumStreams
GPUSievePrimes
GPUSieveSize


Only 1 made a noticeable improvement; I maxed this one:

GPUSieveSize=128

and my GhzDays/Day increased from about 3,650 to almost 4,000.

[Chuck]

I have been running a GTX 1080ti for several months and it is producing about 1590 GHz-days/day. I have been climbing the ranks of trial factoring producers, now slowly advancing from position 24. I expect this card will top out at about 590,000 GHz days per year, around position 10 or 11.

Above position 8, the gaps begin increasing dramatically, around 300,000 to step up one position.

I could switch to a 2080 ti and get 3600 GHz-days/day, which would be an annual total of 1,314,000. This would be about position 4. (My PSU could support both cards, but there would be too much heat generated in the case to run both.)

Other than the useless vanity of being in the top 5, does the project actually need this additional amount of trial factoring, or are we comfortably keeping abreast of LL work with the existing capacity?

[chalsall]

Originally Posted by Chuck

Other than the useless vanity of being in the top 5, does the project actually need this additional amount of trial factoring, or are we comfortably keeping abreast of LL work with the existing capacity?

While we're always interested in more throughput (we'll likely start going to 77 "bits" soon for LLTF), I would say don't invest in new kit unless you /really/ want to. The cycles/watt stats are impressive, but the ROI is probably too long to make it really worthwhile for most participants.

[Mark Rose]

Originally Posted by chalsall

[Valleyspeak]: "Like, wow! Like, totally!

Mark, what do you dissipate for your ~3,150 GHd/D? Over how many cards?

While the capex isn't small, over the long term it might make sense (all consumables in).

Obviously if you're doing TF getting the RTX 2000 series is the way to go from an energy efficiency standpoint. But lots of crypto miners are dumping GPUs on eBay right now, including the pretty decent GTX 1000 series.

For the last while, I've been running, most of the day:

2 x 1070 at ~120 watts each, for about 1600 GHd/D total.
1 x 1060 at ~105 watts each, for about 500 GHd/D.
3 x 580 at ~ 250 watts each, for about 1200 GHd/D (when running 24/7).

Originally Posted by Mark Rose

Obviously if you're doing TF getting the RTX 2000 series is the way to go from an energy efficiency standpoint. But lots of crypto miners are dumping GPUs on eBay right now, including the pretty decent GTX 1000 series.

For the last while, I've been running, most of the day:

2 x 1070 at ~120 watts each, for about 1600 GHd/D total.
1 x 1060 at ~105 watts each, for about 500 GHd/D.
3 x 580 at ~ 250 watts each, for about 1200 GHd/D (when running 24/7).

I am selling 7 of my Radeon RX580. Who is interested can contact me with a private message.

[GP2]

For factoring purposes, does it matter if it has 8GB memory or 11GB? Does it matter if it's the "Ti" version or the plain version? Is it really just as fast as a Titan V (for factoring)?

And how noisy is it?

[kriesel]

Originally Posted by GP2

For factoring purposes, does it matter if it has 8GB memory or 11GB? Does it matter if it's the "Ti" version or the plain version?

Factoring only requires a fraction of a GB. The Ti version of a given model number is faster (and uses more power), although it may be less efficient per watt. There are huge tables of such data available at mersenne.ca.

c:\Users\Ken\Documents>"C:\Program Files\NVIDIA Corporation\NVSMI\nvidia-smi.exe"
Thu Nov 22 18:17:36 2018
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 378.66                 Driver Version: 378.66                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 1070   WDDM  | 0000:03:00.0     Off |                  N/A |
| 89%   85C    P2   124W / 158W |    226MiB /  8192MiB |     94%      Default |
+-------------------------------+----------------------+----------------------+
|   1  Quadro 2000        WDDM  | 0000:1C:00.0     Off |                  N/A |
|100%   93C    P0    N/A /  N/A |     87MiB /  1024MiB |     99%      Default |
+-------------------------------+----------------------+----------------------+
|   2  GeForce GTX 105... WDDM  | 0000:28:00.0     Off |                  N/A |
| 49%   80C    P0    68W /  75W |    154MiB /  4096MiB |     98%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID  Type  Process name                               Usage      |
|=============================================================================|
|    0      9996    C   ...ocuments\gtx1070-mfaktc\mfaktc-win-64.exe N/A      |
|    1      1992    C   ...ments\mfaktc-quadro2000\mfaktc-win-64.exe N/A      |
|    2      8504    C   ...\documents\tf-gtx1050ti\mfaktc-win-64.exe N/A      |
 +-----------------------------------------------------------------------------+

Quadro 2000:

Starting trial factoring M628318517 from 2^83 to 2^84 (3117.74 GHz-days)
 k_min = 7696260968951940
 k_max = 15392521937909134
Using GPU kernel "barrett87_mul32_gs"
...
Date    Time | class   Pct |   time     ETA | GHz-d/day    Sieve     Wait
Nov 22 18:10 | 2952  64.1% | 3763.4  15d00h |     74.56    82485    n.a.%

GTX1070:

Starting trial factoring M147499657 from 2^76 to 2^77 (103.76 GHz-days)
 k_min =  256128947220300
 k_max =  512257894443200
Using GPU kernel "barrett87_mul32_gs"
Date    Time | class   Pct |   time     ETA | GHz-d/day    Sieve     Wait
...
 Nov 22 18:29 | 2423  52.4% | 14.470   1h50m |    645.35    82485    n.a.%

GTX 1050 Ti:

Starting trial factoring M147498487 from 2^76 to 2^77 (103.76 GHz-days)
 k_min =  256130978906880
 k_max =  512261957818688
Using GPU kernel "barrett87_mul32_gs"
Date    Time | class   Pct |   time     ETA | GHz-d/day    Sieve     Wait
...
Nov 22 18:39 |  528  11.6% | 31.030   7h19m |    300.94    82485    n.a.%