A community for everything relating to the GNU/Linux operating system (except the memes!)
Also, check out:
Original icon base courtesy of lewing@isc.tamu.edu and The GIMP
Probably depends on the speed of your usb pen and port. If you know for a fact that everything USB side is 3.1, USB will be faster. Otherwise eMMC will be faster. If everything is USB3.1 gen 2, USB looks to be around 3x as fast. Gen 1 is about 25% faster.
What I found in a quick search:
If any part of the USB chain is 2 or lower, it's slower.
That’s still theoretical speeds, I doubt any drive will be that fast.
Device ships with good emmc storage, it’s all pretty dire. The surface go my friend has got about 150MBps writes and 250 reads. That’s what I’d expect from a “good” emmc machine. I bought some $100 Walmart special a year ago and it gets like maybe 150 reads and 50 writes. Hard drives are fast than this.
Some USB drives are actually pretty good. I have a drive that sustains over 350MBps reads and 150 writes, and bursts almost up to full 400MBps.
That that drive is the exception, not the norm. Those micro center flash drives with white labels? You may get 1 gig of writes in before they crater to 20MBps or less. And the newer black ones? I’ve seen single digit MBps transfers. Putting an OS on there is suffering. Shit just writing the iso on there is bad.
Emmc will be consistently mediocre. If OP had an EMMC laptop then I’m going to guess they didn’t pay extra for a fancy flash drive so the experience is going to be DIRE. If you want to play with the live environment it’s fine.
And the easiest way to figure out the USB version is to look at the USB port and connector. If there is white or black plastic piece inside of them, it’s USB 1 and 2 respectively. If the plastic part is blue, red or some other color, it’s USB 3 or better.
have you seen how atrocious the flash chips put on removable media is
I suspect the usb storage has a much worse or at least much more variable latency profile. Control frames on usb3 are a whole .125ms apart
Other than memory speed, there is one more blocker - your cpu (ish). live usbs do not store the raw image uncompressed, they would be much larger in size. instead, the file system used (usually squashfs) is compressed (usually zstd (default level 3), but could be lz4, or xz, etc). whenever a file is loaded, it is first uncompressed, and if you have enough memory, you can try the load to ram (or memory, wording may differ) option, where, important parts of image are fist uncompressed and stored in memory, resulting in better performance. Now most cpus are fast enough to decompress, so limiting factor still is likely your storage (usb x.y standard) read speed (and if it stably runs that speed, or is thermally throttled), but if you are on a faster underlying source, it can make a difference.
Anecdotely, I use squashfs to compress most things i keep, and it is fast enough for most purposes, but i have observed that for benchmarks, especially single threaded, there is a significant difference. for geekbench 6, my singlecore score was close to 0.6 times of the actual score, when read from uncompressed, or from memory. for all core, score was nearly 0.85 times of the uncompressed/memory score. Would you realistically feel the difference, no imo. I even have a file system level compression (btrfs, zstd, level 3), and i do not feel a significant difference.
I have an Intel Celeron N3060. Is this the main reason for the bad loading times in the live boots?
this may very well be the case. I can not say for sure, my guess is that most cpus should be able to do something like 200mb/s zstd(3) uncompress, but you can try to benchmark uncompress speeds (there is also a zstd benchmark command), and if it is statistically significantly lower than your storage (use hdparam/dd to benchmark them), then it is indeed your cpu.
Installed will be faster.
But you should note that emmc systems are not intended for heavy use. They have a much shorter wearout lifespan than SSDs. If you or your OS do a lot of writing to disk, you might end up with a failure.
Just be aware of that possibility, and have backups.
I know that, during my own move from Windows to Linux, I found that the USB drive tended to lag under heavy read/write operations. I did not experienced that with Linux directly loaded on a SATA SSD. I also had some issues dealing with my storage drive (NVMe SSD) still using an NTFS file system. Once I went full Linux and ext4, it's been nothing but smooth sailing.
As @MagicShel@lemmy.zip pointed out, performance will depend heavily on the generation of USB device and port. I was using a USB 3.1 device and a USB 3.1 port (no idea on the generation). So, speeds were ok-ish. By comparison, SATA 2 can have a transfer rate of 2 GB/s. And while the SSD itself may not have saturated that bandwidth, it almost certainly blew the transfer rate of my USB device out of the water. When I later upgraded to an NVMe drive, things just got better.
Overall, load times from the USB drive is the one place I wouldn't trust testing Linux on USB. It's going to be slower and have lag compared to an SSD. Read/Write performance should be comparable to Windows. Though, taking the precaution of either dual booting or backing up your Windows install can certainly make sense to test things out.
You could install a hot mount SATA drive slot too if your using a workstation. I actually have 3.
It would be faster, but note that eMMC is the worst kind of solid state drive out there and you should replace it as soon as possible, or at least don't store anything critical on it.
For me, even if it wasn't better, I'd be willing to sacrifice launching times, in order to get rid of windows