site  contact  subhomenews

Observed differences using SATA and NVME SSDs

January 15, 2022 — BarryK

This post is just some observations using SATA and NVME SSDs, not quantified measurements.

I have two 1TB SSDs, that I use to backup my work. I have all my work under folder "bk" on my Lenovo desktop PC, and two backup SSDs. I use rsync to backup to the SSDs alternately, usually about once a week, and keep the SSDs in a safe.

I used to backup to an external HDD, but the SSD is considerably faster and also there is far less electric current being pulled through the USB cable -- or so I thought.

The first SSD that I used for backaup is a Western Digital (WD) Blue SATA 1TB, and I was happy with how fast it is compared with the HDD. It also runs "cool as a cucumber" -- after performing the rsync backup, cannot feel any warmth on the casing.

Recently I purchased a Samsung 970 EVO Plus m.2 m-key NVME 1TB SSD, and bought a cast-alloy housing for it. It is, as expected, fast, but here's the thing -- after performing the rsync backup, the casing feels very hot. Really hot. Here they are:


Two points to make here:

Firstly, I do not notice any appreciable difference in the time it takes to perform the backup. Possibly the usb3 cable will be the main bottleneck in both cases.

Secondly, the heating of the NVME SSD is alarming. It negates the original rationale for using an external SSD, being to reduce current drain over the USB cable.

Of course, the SATA aluminium casing has a bigger surface area, but I don't think that will account for the huge difference in the feel of heat when touching the cases.

My guess is that the NVME SSD is being deliberately run hotter, so as to get more speed, what we would refer to as "overclocking" in the case of a CPU.

Lesson learnt? The WD Blue SATA SSD is far cheaper, and I reckon this is an example where it is better to stay with an older technology.   

Tags: tech

AMD Radeon R5 230 video card ordered

October 18, 2021 — BarryK

Having recently purchased a Nvidia GT218 GPU GeForce 210 video card and experiencing how useful it is to have the actual hardware to experiment with, I decided a cheap AMD card would also be nice to have.

Here are recent posts about the Nvidia card, the nouveau kernel driver, and getting it to work:

I looked at Aliexpress, and found this one, that, apparently, has AU stock:

Here is information on the AMD Radeon R5 230 GPU:

While browsing on Aliexpress, I read the description and specifications very carefully, as I know many vendors can be sneaky. I found some cards without any identification label affixed and advertised as AMD, yet in the specs the GPU is given as "gm107", which is an Nvidia GPU.

The one I have ordered does have a label affixed, but that doesn't necessarily mean anything. But I see also in the photos, "AMD" printed on the printed circuit board. So, seems OK.

Just an extra note, I also found this, almost same GPU, but has a passive heatsink:


...looks nice, but I have already ordered the other one.   

Tags: tech

Google forcing us to use gmail in standard mode

October 09, 2021 — BarryK

Right from the early days of gmail, I have used it in "basic html view". However, recently it has been coming up in "standard view", and despite trying to set it back to basic html view, persists in coming up in standard view.

Gmail UI used to have this option to set default as basic html view:


I got that image from here:

That option is no longer there, nor is there any other way to set basic html view as the default. Now, gmail will automatically come up in standard view if gmail thinks the browser is capable of viewing it.

I even know when Google made the change, as the Brave browser guys found, from August 13, gmail always comes up in basic html view. The opposite problem! Link: seems, gmail looks at the user-agent string, and does not recognise Brave browser. Feedback at the above link says changing the user-agent string in Brave to Chrome* fixes it.

There is a fix, if you want to force basic html view, it can be specified in the URL. A couple of sites give this URL, and I even found it recommended at a Google help site: for me, but some people are reporting that it only works "sometimes".

Here are two more, both work for me:

Some links for more information:     

Tags: tech

GT210 Nvidia card for old workhorse PC

October 08, 2021 — BarryK

I had reason to fire up my old HP workhorse PC recently. It has been gathering cobwebs under the desk since early 2020, when I bought a Lenovo PC.

When I say "HP", it is really a mixed-breed. Bought it second hand from a guy who had put it together as a gaming machine. HP motherboard with i5 CPU, 16GB RAM, UEFI-BIOS. Pre-USB3, so back then added a USB3 adapter card.

Anyway, the onboard video, Intel GPU, only has a HDMI socket. However, the only spare monitor I have to use with it only has a VGA input. As all of my computers only have Intel GPUs, I thought this is an opportunity to get a video card with Nvidia, AMD, or Radeon GPU, for testing with EasyOS. Only wanted something cheap, so bought this, AU$26.54:


It is a Nvidia GeForce 210 card, with GT218 Tesla GPU, 1GB RAM. This GPU was released by Nvidia in 2009, so it is old. The card is made in China, so don't even know if it is a genuine chip. Holding it in my hand after arrival, it has that look of quality.

Here is the old workhorse:

img2 all over the place, sounds like an aeroplane taking off! Now here's the problem: the card works, but at the point of bootup where the Linux kernel loads the 'nouveau' GPU kernel module, the screen goes blank. Well, there is a message "input signal out of range", then it goes black. The 'nouveau' module is supposed to support the GT218 GPU.

If I prevent the 'nouveau' kernel module from loading, can get to the commandline, but can't run Xorg.

Thinking how to progress from here. But there is another issue: for anyone who boots EasyOS, or one of the pups or *Dogs, and gets the black screen, they need some kind of "video safe mode" in the boot menu -- need to think about this also.

Oh, by the way, regarding those two fake "SSD" drives that I bought, both vendors gave a full refund, without me having to return the drives. I have appended to the blog post:         

Tags: tech

Claimed 2TB USB SSD is only 128GB

October 02, 2021 — BarryK

Initial report on this drive here:

Thanks to Jon, who found some information on the chips, including this:

29F — Intel NAND flash memory
64B — 512 Gbits MLC / 768 Gbits (?) TLC
2a — 2 channel
L — # of Die 1, # of CE 1, I/O common
c — 3.3V Vcc
T — TLC — three-bit-per-cell
H — 3d gen2 (64 Levels)
1 — Product Generation 1

*200827 is the year and week of manufacture*

*listed code of the chip shows chipset is either chinese or russian origin ZB is mostly chinese, but can't see the chipcode listing back to 2008 in russia. Was it the USSR back then?

And thanks David T., who informed me about 'f3probe':

Running it on the "2TB" drive:

# ./f3probe --destructive --time-ops /dev/sda
F3 probe 8.0
Copyright (C) 2010 Digirati Internet LTDA.
This is free software; see the source for copying conditions.

WARNING: Probing normally takes from a few seconds to 15 minutes, but
it can take longer. Please be patient.

Bad news: The device `/dev/sda' is a counterfeit of type limbo

You can "fix" this device using the following command:
f3fix --last-sec=2147483647 /dev/sda

Device geometry:
*Usable* size: 1.00 TB (2147483648 blocks)
Announced size: 1.91 TB (4096000000 blocks)
Module: 2.00 TB (2^41 Bytes)
Approximate cache size: 1.00 MB (2048 blocks), need-reset=no
Physical block size: 512.00 Byte (2^9 Bytes)

Probe time: 9.55s
Operation: total time / count = avg time
Read: 248.4ms / 4213 = 58us
Write: 9.28s / 22706 = 408us
Reset: 0us / 1 = 0us

# ./f3fix --last-sec=2147483647 /dev/sda
F3 fix 8.0
Copyright (C) 2010 Digirati Internet LTDA.
This is free software; see the source for copying conditions.

Drive `/dev/sda' was successfully fixed

...unfortunately, 'f3probe' has got it wrong. I found that the drive is only 128GB. As per the specs on the chip, 512Gbits per chip. So, 512 divide by 8 gives 64GiB.

The "fix" shown above also does not work, Gparted still shows the drive to be 1.91TiB.

The way to determine the correct size of the drive is to create ext4 partitions of different sizes. The ext4 filesystem creates "marks" over the entire partition, and there will be an error if unable to do this.

So, using Gparted, created a 244GiB ext4 partition:


...notice the inverted triangle symbol with exclamation mark in the middle, indicating something wrong. Running a filesystem check:


In Gparted, deleted that partition and created a 122GiB partition:


...notice, no inverted-triangle symbol. Also, a filesystem check passes OK.

I am accustomed to Chinese vendors exaggerating, expected it, but this is taking exaggeration to a whole new level.

It is really bad news for users. If someone thinks of archiving their family photos, once they exceed the 128GB actual capacity, the filesystem will be corrupted, and they might not be able to recover their files.

I will be asking for my money back. Had to scratch it somewhat, opening it up, so cannot return it (and have them sell it to another sucker). It's an AU vendor, so they have legal obligations in this country.

Next up, will test the "4TB" drive.

EDIT 2021-10-03:
Wow, it gets worse: the 4TB drive has only 128GB.

I mentioned in the previous blog post, that it is two separate "2TB" drives internally. There is a very good reason for this -- they have msdos partition tables, that cannot handle drives any bigger than 2TB. These drives cannot have a GUID partition table (GPT), which can handle drives over 2TB, because that creates a backup GPT at the physical end of the drive, and this causes an error with these drives, due to them not being the reported size. Also, only msdos fat, exfat and f2fs filesystems can be used, that do not mark out the entire partition.

So, the "4TB" drive is actually two "2TB" drives. I ran 'f3probe' on one of them:

# ./f3probe --destructive --time-ops /dev/sdb
F3 probe 8.0
Copyright (C) 2010 Digirati Internet LTDA.
This is free software; see the source for copying conditions.

WARNING: Probing normally takes from a few seconds to 15 minutes, but
it can take longer. Please be patient.

Bad news: The device `/dev/sdb' is a counterfeit of type limbo

You can "fix" this device using the following command:
f3fix --last-sec=124084223 /dev/sdb

Device geometry:
*Usable* size: 59.17 GB (124084224 blocks)
Announced size: 1.91 TB (4096000000 blocks)
Module: 2.00 TB (2^41 Bytes)
Approximate cache size: 31.00 MB (63488 blocks), need-reset=no
Physical block size: 512.00 Byte (2^9 Bytes)

Probe time: 15'14"
Operation: total time / count = avg time
Read: 2.62s / 131442 = 19us
Write: 15'11" / 702863 = 1.2ms
Reset: 1us / 2 = 0us

...very interesting, it is saying the real size is only 59GB!!!

So, tested creating ext4 partitions with Gparted, first, a 122GB partition:


...notice the inverted-triangle symbol. ran a filesystem check:

img5! Then created a 61GB ext4 partition:


...success. Also a filesystem check was OK.

For this drive, 'f3probe' got it right. I will be contacting the vendor and asking for a refund.  

EDIT 2021-10-07:
Both vendors have refunded in full, without requiring return of the drives. This is not something that they want to go to eBay's arbitration!

I do not intend to contact eBay and report these fake drives. There are lots of vendors selling them on eBay and Aliexpress. Also, there are other products, such as smartphones, that are also fake products (or at least, highly misleading specifications). This is going on in such a massive scale, and for so long, eBay and Aliexpress must be aware that it is happening. They are not policing their vendors, so I am not going to raise one small voice just targeting a couple of vendors.      

Tags: tech

Ultra-cheap USB 4TB flash drives

October 01, 2021 — BarryK

I have become suspicious about the memory capacity. I changed the filesystem to ext4 and the 'mount' utility crashed when tried to mount the partition. In theory, it is possible to hardwire a smaller block of memory so that it repeats over the 2TB -- need to find out if this has been done. ext4 marks out the entire partition, and that would explain its failure. Doing some more tests.

A 4TB (4 terrabyte, 4,000GB) SSD flash drive for AU$51.88 including postage, too good to be true? That's about US$30. These cheap "SSDs" are being sold all over the place, on eBay, Aliexpress and Amazon. Most intrigued, I had to buy a couple, to test and find out whether they are telling porkies ("telling porkies" is British slang, meaning telling lies).

Found a couple of vendors on ebay with AU stock, bought these two, first one claimed 2TB, second 4TB:

The 2TB drive cost me AU$36.66 and the 4TB drive AU$51.88 including postage. They are even cheaper if bought from Aliexpress, from China. Here is the "2TB" drive:


And here is the "4TB" drive that I purchased:


...370MB/s write speed, surely they jest!

Note about that second one: they packed the wrong USB cable. Minor annoyance.

I did see some feedback on Amazon, can't find it now, that they are cheating on capacity claim. First, tested the 2TB drive...

The 'fdisk' and 'gparted' utilities do show size as 1.91TiB, single exfat partition. I wondered if they were fudging capacity somehow, so decided to use 'dd' to write zeroes, until it ran out of drive:

# dd if=/dev/zero of=/dev/sda bs=1M conv=fsync status=progress oflag=sync
470956376064 bytes (471 GB, 439 GiB) copied, 81754 s, 5.8 MB/s
449139+0 records in
449139+0 records out
470956376064 bytes (471 GB, 439 GiB) copied, 81755.5 s, 5.8 MB/s

...notice that time, 81755.5 seconds -- that's 22 hours and 42 minutes. I just left it running overnight, but then realised how long it was going to take to get through all the 2TB. So, skipped ahead:

# dd if=/dev/zero of=/dev/sda bs=4M conv=fsync status=progress oflag=sync,seek_bytes seek=1950G
3351248896 bytes (3.4 GB, 3.1 GiB) copied, 663 s, 5.1 MB/s
dd: error writing '/dev/sda': No space left on device
801+0 records in
800+0 records out
3355443200 bytes (3.4 GB, 3.1 GiB) copied, 664.265 s, 5.1 MB/s

...seems like it is 2TB, but using 'dd' does not really confirm that it is 2TB. If a small memory is just mapped repeatedly at higher addresses, to fill the entire 2TB address-space, 'dd' won't know that.

That sustained write speed, 5.1MB/s, is down amongst the cheapest flash sticks that you might buy in the bargain-bin at your local supermarket.

I put an f2fs filesystem onto it, and wrote a file to it:

# dd if=/dev/zero of=/mnt/sda1/dummyfile bs=4M count=256 conv=fsync status=progress oflag=sync
1073741824 bytes (1.1 GB, 1.0 GiB) copied, 213 s, 5.0 MB/s
256+0 records in
256+0 records out
1073741824 bytes (1.1 GB, 1.0 GiB) copied, 213.047 s, 5.0 MB/s

For the record, 'fdisk' reports:

# fdisk -l /dev/sda
Disk /dev/sda: 1.9 TiB, 2097152000000 bytes, 4096000000 sectors
Disk model: SSD
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: dos
Disk identifier: 0xedc4bd24
Device Boot Start End Sectors Size Id Type
/dev/sda1 32768 4095967231 4095934464 1.9T 83 Linux

I did the same for the 4TB drive, that is the red one in above photo, put f2fs onto it, and did a file write test:

# dd if=/dev/zero of=/mnt/sda1/dummyfile bs=4M count=256 conv=fsync status=progress oflag=sync
1061158912 bytes (1.1 GB, 1012 MiB) copied, 105 s, 10.1 MB/s
256+0 records in
256+0 records out
1073741824 bytes (1.1 GB, 1.0 GiB) copied, 105.965 s, 10.1 MB/s

Interesting, the 4TB drive is internally two separate 2TB drives, so shows up as sda and sdb:

# disktype /dev/sda
--- /dev/sda
Block device, size 1.907 TiB (2097152000000 bytes)
DOS/MBR partition map
Partition 1: 1.907 TiB (2097118445568 bytes, 4095934464 sectors from 32768)
Type 0x83 (Linux)

# disktype /dev/sdb
--- /dev/sdb
Block device, size 1.907 TiB (2097152000000 bytes)
DOS/MBR partition map
Partition 1: 1.907 TiB (2097118445568 bytes, 4095934464 sectors from 32768)
Type 0x83 (Linux)

The write speed of 10.1MB/s is much better! This is getting above the bargain-bin flash drives. My SanDisk Ultra drives give about 15MB/s sustained write speed, and my SanDisk Extreme is up around 90MB/s. SanDisk do sell cheaper ranges.

Notice from the photo, they are claiming up to 370MB/s write speed. It may be true, because this is where they are being sneaky. Flash drives usually have an internal cache, and a small number of bytes written to the drive will first go into the cache, then it will fill, and then writing to the flash media will occur, much slower. So, the "sustained write speed" that I have measured, is more accurate.

So, were they worthwhile purchasing, did I waste my money?

I have opened up the "2TB" drive to see what chips are inside. Two chips, no way is that 2TB! Can't find a datasheet, the chip has a "i" on it, pretending to be an Intel chip. I found a photo of the chip elsewhere with "AMD" on it. Numbers on each chip:
29F64B2ALCTH1 200827 ZB170528
I am conducting a test to find out just how much memory there actually is.
Will do the same to the "4TB" drive. Stay tuned.

EDIT 20211002:
The actual capacity of the "2TB" drive is only 128GB, see test here:    

Tags: tech

7 inch tablet with 4G LTE

August 05, 2021 — BarryK

Strangely, such a tablet is very hard to find these days, a 7 inch tablet with 4G LTE bands that suit Australia.

I own a 7 inch tablet with 3G support, brand name OrientPhone, think was purchased in 2014, from China. Wonderful phone, drew admirers whenever I used it in public, due to its large proportions compared with normal phones. Problem was, the 3G bands no longer work, and besides, the phone died a few years ago -- just stopped working.

Ever since then, I have sometimes browsed online looking for a replacement. However, the 7 inch tablet market has also died. Yes, there are modern phones up to about 7 inches, but they are skinny things, with screen ratio not suited to playing 720p or 1080p videos -- such videos do not fill the screen.

I travel often by train, and like to browse the Internet. Also when hiking want to watch videos at night. A 7 inch tablet is ideal, as good screen area, and can still fit in a pants pocket. Also, they are very light.

Just now did a search for information on my OrientPhone, can only find information on a later model:

...looks like mine, except has a higher screen resolution, mine is 1280x800. The OrientPhone brand is no longer available.

However, today discovered a fantastic site for comparing phones and tablets:

...the level of detail is amazing, and it automatically knows that I am in Australia and filters the 3G, 4G and 5G bands on a device for suitability for Telstra, Optus and Vodafone.

Brilliant, I was able to find one tablet that is just like my old OrientPhone, but with 4G LTE, and even supports the Telstra B28 band, which is very hard to find. It is the Alldocube iPlay 7T. And, it is available from the Alldocube factory outlet via Aliexpress:


...yes, very old specifications! But, will do what I want. And, it weighs only 224g, about the same as my Huawei smartphone (216g).

Only 103 Australian Dollars including postage (+GST), cheap, but I wouldn't want to pay big bucks for such modest specs.

Also noticed the 8T, 8 inch:


...not much higher cost, and also supports all the Australian 4G frequencies. Ha ha, imagine answering calls with this:


The 7T is smaller, but still a big thing to hold in the hand. Found this video made by "Friendly Horse", of the 7T showing it being held:

...but why is he wearing a mask in his own home? I like eccentric characters!

Friendly Horse also owns the 8T, and here is a snapshot from another of his videos, showing them alongside:


...immediate impression, the 7T looks OK for holding in the hand, slipping into a pocket, etc. Not so for the 8 inch -- you would need a handbag, backpack, or something like that to carry it in.

The 7T is what I have been looking for, ordered it. Will probably do a write-up, suitability for carrying around on train, etc., and using instead of a phone, and practicality on a hike.   

Tags: tech

Passive heatsink for Pi4

January 12, 2021 — BarryK

As reported recently, I bought a plastic case with fan for my new Pi4:

Which is fine, it runs very cool, but I would rather not have the noise.

Right now. the Pi4 is compiling SeaMonkey, with "make -J2", and the tray temperature applet is showing 45 degC. Even though it is mid-summer here, ambient temperature today is very low, I would guess about 20 degC.

However, there is a passive heatsink that does a pretty good job, as this guy on YouTube has reported:

It has an opening for the wifi signal, but how well that works remains to be seen.

Anyway, ordered one off eBay:


These are available locally, from Altronics and others, but at twice the price. I don't mind waiting while it comes from China. 

Tags: tech