How to increase read time



  • Hi all,

    I am just mining with 54 TB and I got reading speeds like 290 MB/s with USB 3.0. I mine on my ubuntu machine with jminer. I use my Geforce GT 1030. When I just connect one 8 TB drive I achieve the same reading speeds. I saw some guys posting about 1000 MB/s. How can I speed things up or is this limited to the geforce gt 1030?

    0_1506114989841_Bildschirmfoto 2017-09-22 um 23.09.25.png



  • @ciscler , you get more speed with added capacity. It takes one of my machines 32 seconds to scan one 8 TB drive. I takes the same machine 33 seconds to scan eight, 8 TB drives. an eightfold increase.



  • I mine with the same if not 3tb less. Use the cpu I get 400mb to start and goes down to 140mb. On my server I get 30mb. What this means is their is a bottle neck somewhere. This is when you purchase a hard drive or usb controller.
    The server getting 30mb has usb 1 ports which I'm trying to take a usb 3 external harddrive though it.

    The newer computer giving me 400mb per second because four of the ports are usb 3. The rest are usb 2.



  • Hmm. I thinkthe bottleneck is the GPU. First I had a Geforce GT 730 and reached only about 250 MB/s and with the GT 1030 I got the 280 - 290 MB/s. But I am not sure...



  • @ciscler , like I said, you need to add more drives, if the speeds don't increase then there is a bottleneck. or disconnect half of your drives and report the new times.



  • @ciscler said in How to increase read time:

    Hmm. I thinkthe bottleneck is the GPU. First I had a Geforce GT 730 and reached only about 250 MB/s and with the GT 1030 I got the 280 - 290 MB/s. But I am not sure...

    @ciscler if you download something like MSI Afterburner or equivalent you can watch the GPU usage to see if it's hitting high usages or not, that will tell you if it's the GPU



  • @ciscler try also playing around with ChunkPartNonces property in Jminer. Use a 2^n number up to a value where you don't see any improvement or Jminer crashes :)



  • @ciscler said in How to increase read time:

    Hmm. I thinkthe bottleneck is the GPU. First I had a Geforce GT 730 and reached only about 250 MB/s and with the GT 1030 I got the 280 - 290 MB/s. But I am not sure...

    those two cards are almost identical in specs, the 1030 being just slightly better, which might explain your slight results, or it might be something else... one way you COULD try to test this, is when you see a new block for your pool, let your miner run through it, then stop your miner, and restart it so it runs the same block again... this time you should be using cached results and see an improvement in read speeds... now this will vary slightly depending on how large your HDD's cache is, and if you have some sort of ubuntu (or windows for others) caching going on, but you SHOULD see an increase since it was the last accessed data its still sitting in your buffers...

    I've done this myself to test settings in jminer and see how much it can choke down with my GPU... on a new block my 4 drives would take roughly 15 seconds give or take a few for occasional variables, but when I stop the miner, then relaunch it and mine the same block again, I can reduce my times down to 3 seconds since all the data is in buffers so its purely the GPU/bandwidth of the drive/controller at play... it really helped shine a light on some things as to whether or not I was maxing out on hardware, software, or what specifically within the hardware...



  • @Darkbane when I mine the same block again I have exactly the same speeds.



  • @ciscler said in How to increase read time:

    @Darkbane when I mine the same block again I have exactly the same speeds.

    do you have any disk benchmark software that will work with whatever little space you have left on the drives, and perform the test on all the drives at once? so you can find out if this is a USB bottleneck rather than a GPU issue?

    P.S. or as someone else suggested, get something like GPU-Z to see what percent of your GPU is saturated and if there is any room left to put more data through?