It seems odd to populate all four dies and then disable two. It would have to be the same two disabled all the time which is not going to allow selling an Epyc with failed cores as a consumer part. If it is the case, I would have rather AMD pulled an Intel and used a smaller cheaper socket for...
All the AMD cards I have owned do not down clock all the way when you plug in more than one monitor, so I normally plug one into the onboard and one into the GPU.
If any of these ASICs have the memory built in, increase the memory requirement beyond what they have and kill the device. IIRC LTC memory requirement is rather small so that it would fit into a CPU's cache.
2.4GHz Quads with HT. IIRC I was on the lower end of BigAdv on Linux a couple of years ago so I am guessing SMP only. If there is a windows vs linux penalty, it might not matter much if I'm stuck with SMP.
Looking at getting getting a box running again. I reacquired one of my folding rigs from a few years ago wondering if how it will fair. Last I was here, running linux under VB was giving better performance than Windows native, is that still true? The box is going to double as a HTPC so looking...
OpenSM needs to be on a single node only. Since the VM does not see the physical IB hardware, it will need to be the Windows box.
The IPoIB interface does not need a default gateway, dns, or wins. Just IP and subnetmask
Don't change your GbE interface settings.
As for trroubleshooting the...
RMA'd 3 2TB EARS drives in the last few weeks. The first one, I got back a 2TB EARX, ok fine. Two weeks later, I RMA two together and receive 2 2TB EFRXs... All three were new drives and I did advanced replacement.
You definitely need them on a different subnet if you are going to have Ethernet and IB. I had a similar experience with passthrough not working, but I think it was before the update that fixed PSOD.
Try this:
IPoIB on Windows w/ IP: 192.168.2.2 255.255.255.0 (/24)
CX-2 added as uplink on new...
Either use the server's IPoIB IP when mapping or browsing the network share or put a entry in the client's hosts file that gives a name to the server's IPoIB IP.
I went with HGST for a build I am working on because of availability and price. 3TB SAS2 7K3000. Power consumption difference between the ES.3 and Ultrastars is not big.
The last place I worked, some of the buildings were fallout shelters built in the 50s and 60s. One could move a few feet through a bulk head and lose all wireless reception.
I have an Ideal 30-496 I like; it is small and light. I also have the more expensive monoprice crimp and use it more often. Also, in my experience the ratchet gets in the way.
Also, to echo zero2dash, I buy premade cables when I can.
I use spideroak for my desktop(Linux), though I have a small dataset. My father uses carbonite with ~2TB of data and it is slow as molasses. I have seen it take about a week for 10 JPEGs to get backed up, he has 25/25 fios and the system is on 24x7. Jungledisk looks like an interesting option...
Reds also are only 1E14 error rate, equal to consumer Toshiba or Seagate drives, but worse than any Ent. drive.
For OP, if this is a home server, not going to have data you will lose money when you cant access, and you don't have a client/boss to answer to when the array is down, consumer...
Biggest power differential is 5.16W at start up. Lets use that value for worst case.
5.16W * 24 drives = 123.84W - Not a small amount, maybe could go with a smaller PSU
How about running cost. I pay ~$0.13 to ~$0.45 US per KWH, so worst case again, $0.45. Also, do running 24x7 with a 94%...
Even with the Mellanox stack installed you cant pass the card through as far as I know. IIRC, it is something to do with how the firmware deals with soft resets. I tested with Linux, which wont reboot, but you will see errors in dmesg.
I didnt see good speeds until -b 4096 -c 4096. It is a MicroServer related problem and not badblocks settings. Reran testing on host1 with a LSI2008 and stock settings gave me better performance.
Results on MicroServer:
Test Value | Avg. Read (MB/s) | Avg. Write (MB/s) | Test Duration Wall...
Edit: Ignore, hardware problem. See post #501
Adding to the badblocks discussion, I got 3 more Seagate 3TB. Using -b 4096 -c 256 vs -c 2048 saw a doubling in instantaneous throughput. After 4096 only 10's of MB/s were added with each -c doubling.
If you are looking at Ultrastars, there are also Enterprise Western Digital and Seagate drives.
ST4000NM0033 - Seagate ES3 4TB SATA
ST4000NM0023 - Seagate ES3 4TB SAS
WD4000FYYZ - Western Digital RE 4TB SATA
WD4001FYYG - Western Digital RE 4TB SAS
I have already changed my setup so this is not 100% comparable, but from my daily computer, E3-1245 V2 Kernel 3.2.0-23, to Host1, G34 6128 Kernel 3.5.0-19, datagram mode gets a max of 13.9Gbps ConnectX-2 to ConnectX-2. So slower, but still better than 10GbE.
Edit: Ignore, hardware problem See post #501.
When using badblocks on 3x 3TB Segates I noticed an increase in instantaneous speed from -c 128 to 2048, past that gave little increase. Disk throughput was viewed using iotop. I did not look at average speed because I ended up killing badblocks...
I have some kind of hardware problem with my 20Gbps ConnectX card and my G34 platforms so I haven't tested ConnectX to ConnectX. So far I have 10Gbps => 7.85Gbps(ConnectX - Infinihost III Ex MemFree), 20Gbps => 11.63Gbps(Infinihost III Ex both ends), 40Gbps => 17.8Gbps(ConnectX-2 both ends)...