SoylentNews

I have run many tests on my new Ryzen 5 3600 with 32GB RAM doing music transcoding from FLAC to MP3 using LAME.

I was interested in finding out just how slow my old SATA spinning rust disks are, so I thought I'd try comparing transcoding time from disk to a different disk (one of two kinds) and from ramdisk to ramdisk, ramdisk to disk and disk to ramdisk.

Creating the ramdisk is very easy. I made one with 16GB or RAM:

mount -t tmpfs -o size=16G ram0 /mnt/ramdisk/

And then I copied my input data onto the ramdisk (from the parent directory containing the input data dirs, and note the trailing / on the rsync source directory):

for i in `cat ramdisk_convert_list.txt` ; do mkdir -p /mnt/ramdisk/ripping/${i} ; rsync -av ${i}/ /mnt/ramdisk/ripping/${i} ; done

The top level script runs the entire encoding test each time for a given number of CPUs or threads ie it iterates from 1 to 12 CPUS/threads.

I used a subset of my input data, 18 albums (7GB of input data, 2GB of output), one of which is a double CD. After running the tests I realised a few problems.

The double album was somewhere down the list of inputs, so depending on the number of threads being used it was often the last to finish transcoding, by a long time, so it was masking any performance improvements from transcoding albums in parallel. To fix that, I made a list of the albums sorted by data size as follows:

for i in * ; do du -s ${i} ; done | sort -rg > sorted_list_of_albums.txt

I also saw a bug in my top level script where it wasn't waiting for all its subtasks (parallel encoding processes) to finish at the end, so the time command was often measuring premature completion, and the next job in the loop was starting while other threads were still running. Sometimes there were two transcoding jobs running on the same album at once! To fix this problem it was as simple as putting a loop at the end of the script counting the number of threads and running wait -n for each iteration to ensure that all the sub-tasks had finished. wait -n retuns if there are no sub-processes running, so if a sub-process finishes before the loop starts, there's no problem.

COUNT=${NCPUS}
while [ ${COUNT} -gt 0 ]
do
        wait -n
        COUNT=$(( ${COUNT} - 1 ))
done

The other thing I did was to run LAME in quiet mode (--quiet). This disables the animated output histogram while encoding, and makes it a bit faster.

Results? Well, I won't bore you with the details, and I haven't drawn any graphs yet, but 1 thread took

Number of threads: 1
real 14m20.573s
user 14m12.824s
sys 0m7.531s
2.0G /mnt/ramdisk/compressed/lossy

And 6 threads

Number of threads: 6
real 2m47.928s
user 15m2.168s
sys 0m7.601s
2.0G /mnt/ramdisk/compressed/lossy

Surprisingly, with 7 threads

Number of threads: 7
real 2m28.698s
user 16m8.499s
sys 0m7.922s
2.0G /mnt/ramdisk/compressed/lossy

But then it started to plateau. In fact some were a few seconds slower. 12 Threads gave

Number of threads: 12
real 2m37.054s
user 19m3.468s
sys 0m8.662s
2.0G /mnt/ramdisk/compressed/lossy

Next I will try the RAM to disk and disk to disk ones again.

Update: here's the data for the ramdisk run in CSV format.

It looks like AMD will launch some Zen 2 refresh CPUs to stave off Intel's Whatever Lake before the Zen 3 launch:

AMD Ryzen 9 3900 XT, Ryzen 7 3800 XT, Ryzen 5 3600 XT ‘Matisse Refresh’ Desktop CPUs Confirmed – Same Core Config, Higher Clocks & Price Cuts For Existing Models

If correct, these should replace the 3900X, 3800X, and 3600X, boosting clocks a bit. Pricing would be the same as the original lineup ($500, $400, $250 respectively), while the older chips will get price cuts.

Separately, Tom's Hardware reported replacements for 3800X and 3700X:

AMD Reportedly Planning Two New Ryzen 3000-Series CPUs

The Ryzen 7 3800X and Ryzen 7 3700X are the two potential Zen 2 candidates that could undergo the refresh treatment. If there is a tiny bit of credence in the rumors, AMD will market the rewarmed parts as the Ryzen 7 3850X and Ryzen 7 3750X with the latter having already surfaced in AMD's Product Master guide since last year. In other words, the Ryzen 7 3850X and Ryzen 7 3750X would just be higher-binned variants of their vanilla counterparts.

"... you ain’t black"

That was weird...

The next thing I need to build is a router. I have my old Phenom II X6 acting as the firewall between the rest of the house and my own LAN. I have to start it up whenever I want to get to the outside world, and I'd rather have something small, quiet and energy efficient.

I was looking at fanless mini-PCs. There are quite a few to choose from these days, and they seem quite powerful. I don't want an ARM system, because I'd like to be able to run x86 binaries natively on it. Many years ago I worked for a company making network storage appliances (32-bit x86). Our OS was based on RedHat and was stored in a compressed file on a flash disk. When it booted, the root filesystem was decompressed into RAM and run as a RAM disk. I put the whole thing together, and I'd like to do something similar with my router. What I'm saying is it doesn't need a huge amount of disk space, but plenty RAM would be good. In fact, it could even run off an SD card or similar. It would need at least two ethernet ports. WiFi wouldn't be essential since I have a newish WAP which works quite well.

The other thing I need to do is to retire my printer server. It needs to be virtualised. I have a Xerox colour laser printer over a decade old. The print quality is superb and I got it very cheap. The only problem is it's a Winprinter. Fortunately, it's a Xerox-re-engineered FujiXerox engine, and there was a binary Linux driver for it (32-bit, x86) which I have been running on my 2003 vintage Athlon XP 2000+. It's noisy and takes up a lot of space. So I thought I could try to make a virtual machine to run the software using maybe quemu or VirtualBox or one of those. Then I could run the VM on my new router.

Buying a new printer would be an admission of defeat :-)

Does anyone have any recommendations for such a machine?

Hot Chips 32 (2020) Schedule Announced: Tiger Lake, Xe, POWER10, Xbox Series X, TPUv3, Jim Keller Keynote

Online-only conference, August 16th to August 18th.

IBM, USCB, and Google's quantum computing presentations should be interesting in light of: Some Serious Drama Went Down on Google’s Quantum Computing Team and IBM and Google disagree on quantum computing achievement

Other interesting presentations:

Marvell ThunderX3 ARM server CPU
2nd generation of the Cerebras Wafer Scale Engine
Alibaba RISC-V
ARM Cortex-M55
DeepMind and Google's 2nd and 3rd generation tensor processing units (TPUs)
Manticore: A 4096-core RISC-V Chiplet Architecture for Ultra-efficient Floating-point Computing
Baidu Kunlun AI processor
Alibaba Hanguang 800 NPU

The democrats have quite the sense of humor:

Sen. Gary Peters (Mich.), the top Democrat on the panel, warned that the "extremely partisan investigation is pulling us apart."

Role reversal is so cool!

Watch for updates...

UPDATE:

So, do the Biden tapes exonerate him? Doesn't matter, they could be fake, they, and the girl will still occupy his entire campaign. The guy is so damaged. The DNC/GOP is very happy with Trump. They will be just as happy with Biden. *Back to Normalcy*, right?

Man who called coronavirus ‘fake crisis’ gets infected, wife in critical condition

Florida man who downplayed coronavirus as ‘fake crisis’ gets infected, warns others after ending up in ICU with wife

Florida man who called coronavirus ‘fake crisis’ gets infected, warns others

He didn’t wear a mask and called the coronavirus ‘a fake crisis.’ Then he was infected

Man who called coronavirus ‘fake crisis’ gets infected, issues warning

Quoting first article:

A Florida man who worked as a rideshare driver and refused to wear a mask out of skepticism of the coronavirus pandemic has been hospitalized, along with his wife, who is now facing a serious threat from the disease.

Brian Hitchens said was a self-proclaimed COVID-19 skeptic about a month ago, WPTV reported.

“I thought it was maybe the government trying something, and it was kind of like they threw it out there to kinda distract us,” said Hitchens.

On Facebook, Hitchens said he believed the pandemic was “blown out of proportion” and said he was putting his faith in God.

“I’d get up in the morning and pray and trust in God for his protection, and I’d just leave it at that. There were all these masks and gloves. I thought it looks like a hysteria,” Hitchens explained.

Now, Hitchens and his wife are in the hospital after contracting the virus.

“I don’t want to see anyone go through what I went through,” Hitchens said. “This wasn’t some scare tactic that anybody was using. It wasn’t some made-up thing. This was a real virus you gotta take seriously.”

Hitchens is showing signs of improvement, but says his wife’s condition deteriorated and she had to be sedated and placed on a ventilator 3 weeks ago.

“After 3 weeks I have come to accept that my wife may pass away and the peace I have about it is that I know without a shadow of a doubt that she will be going home to be with the Lord but I also do believe in miracles and I’m holding on to the chance that she may get healed but if not I am thankful for her I know we’ve been married for 8 years,” he said.

“Looking back I should have wore a mask in the beginning but I didn’t and perhaps I’m paying the price for it now but I know that if it was me that gave it to my wife I know that she forgives me and I know that God forgives me,” he said.

^{_<no-sarcasm>}
My first almost instinctive reaction, upon reading the headline was the Nelson "haha". But when reading the first paragraphs, this story is nothing but sad.

I would say to the guy, "God gave you a brain, use it."

Maybe this is a lesson learned too late and the hard way.

But I think the real blame goes deeper than just laughing at or blaming this working guy. I doubt he hatched his misconceptions on his own. He probably heard them in various right wing echo chambers he visits. This is where the real blame lies. (pun intended)

How is it in the 21st century we have half the population listening to echo chambers that outright deny science, have a disdain for education and intelligence, pick the worst human beings to be their representative (not the best and brightest), and blame the other half population for their chosen representative? How is it we have echo chambers that openly and brazenly spew outright lies? Alternate facts? Calling the free press the enemy of the people, without seeing the irony of 1984. This is where I think the true problem lies. I'm not absolving this guy of all blame, but I think he was convincingly led into dangerous (maybe fatal) misinformation, by someone who knew that they were lying. Someone made up these lies out of thin air. They didn't just get on the air by themselves.
_{^{</no-sarcasm>}}

Due to popular demand.
On your grocer's shelves...
Clorox Bleach now in Grape and Orange flavors! Traditional unflavored Clorox now available in convenient pre-loaded syringes.

I sometimes create burner accounts when I want to participate in some activity anonymously. For this purpose, I usually sign up for a new Google, Yahoo, or MS Outlook/Hotmail account.

They're making it more difficult to stay anonymous. And I understand that they don't like having their services used for sending out spam. I don't like that either. But the measures they are taking do more than cut down the spam. They now insist that you "verify" your account by giving them a phone number that can receive a text. They resort to sneaky lies and false claims about what is really going on, maybe saying vaguely that it's "for your protection" without going into any details about why and how, or maybe claiming it is to reduce spam.

This time it was MS Outlook/Hotmail. I log into a burner account, and their system announces that my account has been locked. I must give them a phone number so they can verify me. That's a lie. They want a phone number they can connect to what is likely a real person. What if I have only a landline, and cannot receive texts? What if I don't want to give out my cell phone number? Also, I especially dislike how expensive texting is. The price is too obviously not based upon the amount of data transmitted. Or, I've already used my number for another account, and they won't let numbers be used for multiple accounts? On that last one, how would they know, unless they are indeed tracking that bit of data? I could also go buy a burner phone, but that's far too much trouble and expense for a lousy burner email account.

There are some online services that can receive SMS texts for you, using their own phone numbers. I tried it. As I expected, that did not work. Outlook claims that they cannot sent a text to that number. Bull! They mean that they won't send a text to such a number, because they have a blacklist, and those numbers are on it.

* * *

I also ran into an old trick a day ago, the old "Give us your credit card number to start your 'Free' trial. You can cancel anytime before the trial period ends and not be charged!" The site in question is a genealogy site, myheritage.com. As expected, reviews of the site talk of "mistakes" being made with charges nevertheless being charged after cancellation.

I thought of trying a fake credit card number again. (I tried it several years ago for something else, and it didn't work.) You can't just make up any old 16 digit number. Like with Windows license keys, only a small subset of the full range of values is valid. However, there are credit card generators. Didn't work of course. The site checked with the credit card issuer, which rejected the fake. But that does tell me that they do check the number right away, and do not wait until the first payment is due.

I've also had to improve fake street addresses. Used to make up something completely fake. Now I pick a real address, of a church or used car dealership or some such, in a city I've never visited.

Cybergang behind Lady Gaga data dump leaks hacked law firm's emails and contracts mentioning President Trump, says it is posting 'most harmless information' first

The unidentified cybergang who dumped Lady Gaga's data after the hack of a prominent entertainment law firm have dumped what they say is data about the president.

The cybergang dropped links to "the first part of data, with the most harmless information," on the dark web site. A sampling of the files reviewed by Business Insider were were legal emails, documents, and contracts mentioning the president, from an "Apprentice" costar's emails seeking to land a new show during his campaign, to the contracts negotiating legal use of his videotaped interviews.

The ransomware gang has stolen data from businesses and leaked it on a dark web site in the past, and a ransomware expert who has followed past data dumps said it's difficult to know what they data they have stolen – but said the crime adds to a wave of highly public ransomware extortion attacks.

Criminal group that hacked law firm threatens to release Trump documents

Hackers Release Dozens of Law Firm’s Emails Citing Trump, but There’s No ‘Dirty Laundry’

The criminals’ claims to have revealing info on Trump in connection with their hack of Grubman Shire Meiselas & Sacks were puzzling, given that the law firm has never represented Donald Trump or the Trump Organization.

The hackers claimed the emails they released Saturday contain “the most harmless information” about Trump — apparently trying to imply they are holding back more compelling material. But considering that what they published contained nothing remotely interesting, it seems likely that the cybercriminals are vastly exaggerating the value of the data they’ve stolen.

Grubman Shire Meiselas & Sacks refused to pay the cyberthieves’ initial $21 million ransom, after which the hackers doubled their demand to $42 million. Now that it’s clear the law firm, which has brought in the FBI to conduct a criminal investigation, will not even negotiate with the criminal ring, the hackers are claiming they will auction off the client data they stole on the dark web.

The attack on the Grubman Shire Meiselas & Sacks network, which appears to have occurred on or around May 7, allegedly resulted in the theft of documents on multiple music and entertainment figures. The hackers have claimed those include Lady Gaga, Madonna, Nicki Minaj, Bruce Springsteen, Mary J. Blige, Ella Mai, Christina Aguilera, Mariah Carey, Cam Newton, Bette Midler, Jessica Simpson, Priyanka Chopra, Idina Menzel and Run DMC.

What is the name of the group?

A computer with 12 penguins.

After about eight years, I have built a new PC. The venerable AMD Phenom II X6 1045T is still going strong with 6TB of disk and 8GB of RAM (and an nVidia GTX650) but it was time for something new.

I bought an ASUS X570 Plus motherboard ("Tuf Gaming"), AMD Ryzen 5 3600 (6 cores/12 threads) and 32GB of unbuffered ECC Crucial (Micron) RAM and built it yesterday. It's got an nVidia GeForce 1650 graphics card that I bough a couple of weeks ago and was in the Phenom II. I bought a not very expensive Fractal Design case and a 700W power supply.

My budget didn't extend to any new (or indeed newfangled SSD) disks. I had two 2.0TB Western Digital green drives lying about and a 650GB Western Digital blue drive that I extracted from the cast-off intel Core 2 Quad that I acquired.

On Friday I downloaded the latest Slackware-current and everything's up and running. The physical build, including extracting the old hard disks, took a couple of hours. The hardest part was figuring out how to persuade the UEFI BIOS to boot off of the SATA drives (hint: they're "legacy").

I installed Slackware from a USB memory stick. The only problem I had was that I forgot to format the UEFI boot partition when I partitioned the disk, and so it didn't install the ELILO bootloader stuff in the right place. I had to mount the partition and copy it manually. The next problem I had was getting it to see the ELILO stuff. Apparently, this firmware (unlike my Acer laptop) can't attempt to boot from arbitrary file names. The boot stuff has to be in the EFI/boot subdirectory and the bootloader file has to have the standard name: bootx64.efi.

When you do a Slackware installation, one of the boot options is Memtest86+ (rather than the Slackware installer). However, when you do a UEFI installation from USB stick, that option apparently isn't there. It's definitely there on the DVD installation. Fortunately I have a USB DVD writer and was able to boot off that to run a memory test. I didn't do the complete suite of tests (32GB takes a long time) but after about 10 tests there were no errors.

Over the years I have ripped all my music CDs to FLAC so that I can transcode them to Ogg/Vorbis listen to on my Android phone. Unfortunately my car only knows MP3 format, and I had 128kbps encodings of most albums and they sound terrible. A while ago, I re-encoded everything as top-quality MP3s. They sound a bit better but still not as good as Ogg/Vorbis. It took about 2.5 hours to re-encode them using LAME on all six cores of the Phenom II. Today I will try using the 12 virtual cores on the Ryzen and see how long it takes.

Note that the Phenom II has faster disks in it (Western Digital Blue, 3TB) vs. the Ryzen (Green, 2TB). I noticed about a 15% speed improvement when I went from green to blue a few years back in the Phenom II. However, I was encoding to and from the same drive, Today I will read from one drive and write to the other. I might try it on both machines.

If it isn't faster, I will cry.

Update: Performance comparisons.

I ran my multi-threaded FLAC to MP3 transcoding scripts on three machines to transcode 135 albums (from CD). One or two were double or triple albums. I have pairs of hard disk in each system, and I used one for the reading and one for the writing of the compressed data.

AMD Ryzen 5 3600 6 cores/12 threads, 3.6GHz nominal
32GB Unbuffered ECC RAM

bash-5.0# time recode_list mega_convert_list.txt
Number of threads: 12

real 14m47.234s
user 107m3.341s
sys 1m33.449s

Phenom II X6 1045T 6 cores 2.7GHz
8GB DDR2 800, Unbuffered ECC

bash-5.0# time recode_list mega_convert_list.txt
Number of threads: 6

real 50m27.843s
user 279m55.120s
sys 3m52.876s

Intel Core 2 Quad Q8300 4 cores 2.5GHz
8GB DDR2 800

bash-5.0# time recode_list mega_convert_list.txt
Number of threads: 4

real 71m50.033s
user 256m19.096s
sys 5m42.453s