Mantiz Venus - different SSD performance using enclosure ports versus MacBook ports?
I took out the GPU to isolate the SSD
even for 22G, 30mb/s is really slow isn't it? The drive starts out fine 100+ but then falls and stays at 30-40.
I havent done any benchmarks for the gpu yet (MSI GeForce GTX 1060 AERO ITX 6G OC) because I've been trying to figure out the hard drive and disconnection problem first.
As @Mymantiz_john said, you're copying over a lot of small files. The initial large spike of speed you see is twofold: first, you're saturating the cache of your drive, which could be 256MB to 512MB, which gets filled very fast and always shows a large speed in the beginning. Second, one of the more longer-taking tasks is to allocate a new entry in the lookup table of a disk for every new file. If you have many larger files, they can sequentially write large amounts of data, while smaller files each need a new entry, and write just a bit of data. So the performance you're seeing is normal behaviour. I'm not saying Mantiz Venus in itself couldn't also form a bottleneck, but I think you will see a similar behaviour pattern when copying those same files to an external USB drive or hard disk.
So the drive controller, USBs and Ethernet are all USB hub-based?
What bandwidth allocation does the enclosure USB-C 3.1 controller get with the new firmware? USB-C 3.1 is rated up to 10Gbps. Does this new firmware give any priority to the USB-C 3.1 attached devices to draw up to 10Gbps of the 32Gbps (31.2%) bandwidth on-demand (dynamically)?
The design is USB 3.0 Gen I, not GEN II 10GB.
Me too thought it was 3.1 Gen2...
So, even with a USB 3.0 Gen1 card, actually you loose 10Gbps instead of 5Gbps: weird Intel static allocation!
When you will resolve this static allocation with a dynamic one, any plan to make the addon card a USB 3.1 gen2?
Any luck testing for the sata connection speed problem? I tested again with one big file and it has the same pattern.
Plugging my external usb drive gives me 100mb/s or more the whole way through for the same transfer (even with multiple files).
I also have a strange problem where the unit disconnects and reconnects (thunderbolt says the unit is attached but the drive does not show up), this happens at random intervals (sometimes an hour sometimes many disconnects within a few minutes) . Is is possible that I have a defective board? (My thunderbolt hub and adapters don't have this problem)
Try a speed benchmark tool like Crystal DiskMark and post the pic result here.
Will do that after i get home from work, the bigger issue for me atm is that the drive disconnects at random, (so sometimes the Venus is shown as connected, but no drives show up.) I'll put the gfx card back in to see if that changes anything, but I'm guessing from the strange speeds and disconnects that I might have a faulty board. Waiting to see what John can suggest.
It would be useful also someone else making a CrystalDiskMark on its SSD on Mantiz (specifying TB3 2x or 4x): so we can compare some results on different units.
Nothing related with the SSD or USB, and it is not an issue either. A new FW is going to be released soon and everybody will get it with 22xxMiB/s
On a TB3 system, they'll get 22xxMiB/s as long as the other ports are not being used concurrently (SATA, USB 3.0, LAN).
Can please you disclose what the max bandwidth for the USB-C controller is in this new firmware?
Without that info, we'll assume it's 10Gbps resulting in 31.2% reduction in eGPU bandwidth when USB-C is fully 10Gbps saturated . Intel guarantee's 10Gbps for their USB-C 3.1 off a TB3 controller so let's see if this will be the case.
Users will be able to measure the bandwidth apportioned to the USB-C controller once this new firmware is released. Done by running a CUDA-Z performance test concurrently with a USB 3.0 & SATA SSD sequential read, taking the resultant CUDA-Z H2D value and subtracting it from 2254 (100% eGPU bandwidth).
The posted Mantis Venus comparison against Sonnet Breakaway Box 550 looks at the value of the Mantiz Venus' extra ports.