For sale is my mini-HPC cluster setup. I built it for running CFD and astrophysics simulations, but don't have any room for it in my current place. I know roughly what I could part it out for, but it has some value as a ready-to-go cluster, so I'd like to try to sell it as a whole first...
AHA! It worked! I have FDR speeds now!
You can burn mellanox .bin firmware directly to EMC SX6005 switches almost exactly like you do with rebranded HCAs. You need the LID of the switch from "ibswitches", and you need to install MFT. Then the command is:
flint -d lid-X -i fw.bin...
Ok, made a little progress on this. I installed MFT. Then I did "ibswitches" to find the LID of the switch. I then did "flint -d lid-<switch lid> query full". This returned the firmware version (9.3.2000, so not latest), as wel as the PSID, which was "EMC1260110026". Now, if you look at the...
Good call. Mine has two psus, so there's a better chance it's an FDR switch than FDR10, but EMC could have some special configuration, so who knows.
I'm going to install MFT and try some of those tools to figure out what's going on.
Yes. There are two types of SX6005, one with a "T" in the part name and one with an "F". The former is FDR10 and the latter is FDR. However, the EMC switches just have "SX6005" and then a purely numerical part number on the pull out tab.
Basically, I'm trying to figure out if the switch is...
I know the mellanox ones do. The "T" means FDR10, and the "F" means FDR, but the EMC part numbers are just numbers. Do you know how I could look up my EMC part number to determine whether it's an FDR10 or FDR?
I bought one of these EMC SX6005 switches. My IB cards and cables will show 56 Gbit/s link speed when connected back to back, but the link speed drops to FDR10 when connected through the switch. The switch is unmanaged, so I don't think I can check its port speeds. Since it's EMC, the part...
I wouldn't say "completely abandon"... The MPSS for these is fully working and still available. They just aren't going to update it anymore and have ceased production.
For sale are 80+ Intel Xeon Phi Co-processors. The available model numbers are 7-series ES (engineering sample), 71S1P, and 7110P. I also have many of the extension brackets and pcie brackets. Conditions range from tested/good to for-parts/broken. Prices range from $15 to $179. See THE TABLE for...
Awesome! Thanks!
The pro firmware link is broken, but the normal one worked. I mirrored it, along with all of the cx354/3a bin's from the Mellanox website, on my google drive.
Is the mlx file and ini necessary, or can just a bin work? It looks like izx's instructions require the mlx and ini, while kiteboarder's go straight to the bin. It looks like Mellanox has removed all of their custom firmware downloads from all sites now :( , so I can't find the mlx and ini...
Answer to question 1 here. So the answer is yes. In fact 25.6 Gbit/s is the fastest expected bandwidth for QDR. The key is this: PCI_LANES(8)*PCI_SPEED(5)*PCI_ENCODING(0.8) = 32 Gb/s (that's where that number comes from). 32*PCI_HEADERS(128/152)*PCI_FLOW_CONT(0.95) = 25.6 Gb/s. My system is...
System:
Desktop (headnode): I7-5960X, GA-X99-SLI motherboard
Server (slave nodes): 6027TR-HTR (four X9DRT-HF). Only using one node for this testing. 2x E5-2667 V2
OS: CentOS 7.5 with Infiniband Support and various other packages. No custom OFED installs.
HCAs: One Sun/Oracle X4242A QDR...
Here is my guide for flashing firmware to these cards. It pulls from various sources and covers flashing firmware with and without psid changes. Also summarizes everything I've done about this specific issue.
So here the ODM is Mellanox, and the OEMs are Sun/Oracle, IBM, HP, Dell, etc?
Nice catch with HP increasing the BAR size. I wonder if 32MB would work without turning on above 4G decoding. I bet most motherboards have a default upper limit on BAR size.
So the versions on the broadcom/emulex...
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.