Chapter-2

Black-Edges Sorcery Chapter-2: Hardware or software problems (EEPROM and soldering)

“I see where this is going. Letting a lead developer do full-time tutoring. But I must say, N-gram is a thing of the past, and so is CNN, and TensorFlow deep learning. Learning them now is no different from learning Adobe Flash in web development today,” Dong Yang said in a self-mocking tone.

“There is still some difference…” Zhang Ming wanted to explain, but suddenly he didn’t know what to say to Dong Yang’s argument.

“Any computer skill is better than knowing nothing about computers. The same goes for AI — any foundation in AI is better than knowing nothing about it,” Zhang Ming said after some thought.

“Ah…” Dong Yang sighed, then continued, “It seems that you have been away from the tech industry as a whole. How big was our team back in 2015? Twenty brilliant people. And how many people does it take to train an AI currently? At least a few hundred. We couldn’t even get our hands on the cards needed to do the training, let alone start training an AI like we did back then. Things have changed; we are not moving forward. Technology has left us in the dust.”

“Unfortunately, I am unable to bypass the whitelist on the Dell BIOS — the card is a year too old to be registered in the whitelist Dell wrote,” Dong Dong’s voice suddenly came from the wall speaker.

“You may need to locate the whitelist yourself and add the ID to the software. The Dell Precision uses an AMI BIOS,” Dong Dong said.

“Can you find and download the required firmware for the iFix RT809, plus the hex editor and compilers? Move them to my desktop. Also check if the EEPROM on the Dell needs a 1.8V adapter,” Dong Yang said, standing up to look for the programmer on the shelf.

“Zhang Ming, don’t just stand there; help me clean up the soldering table. The soldering irons and all the required tools should be on top; anything not needed, just dump it on the floor below the table,” Dong Yang called over to Zhang.

“Is not this a software issue? Dad said it’s a whitelist issue. Why do we suddenly need the soldering iron?” Xinyi asked her dad, looking puzzled.

“A software issue does not always mean the fix will be software. A software issue like this one will need a hardware fix,” Dong Yang said.

“The Dell computer will need an SPI NOR flash chip in a WSON8 package. It is a 256-megabit chip — 32 megabytes — and it runs on 3.3 volts from the motherboard. All tools have been downloaded to the BIOS tools folder on the desktop,” Dong Dong suddenly said.

“I found the EEPROM chips, but there is an issue — they are all mixed up. Most of them are from GigaDevice. So we need to find a GD25B256 in this box. This one is a GD25B64 — 8 megabytes, too small. The SOP8 package is already a pain to read; WSON is even worse,” Dong Yang said, turning on all the lights in the room.

“Can’t you use the clip method? I saw clips over there — the ones with the probes and the DuPont cables you have over there,” Zhang Ming said, pointing at the rainbow-colored cables in the box Dong Yang had just brought out of the cabinet.

“We could try. I never seem to make them work as advertised. Those things are tricky. My assistant has gotten them to work in the past, though. Are you able to use them reliably?” Dong Yang said while looking at Zhang Ming.

“I haven’t tried with the WSON ones, but I’ve gotten SOP8 clips to work before,” Zhang Ming said.

“Well, you can get it connected, while I find the whitelist and get the ID to the program.” Dong Yang stopped looking for the WSON chip.

“What are those chips? Isn’t software something you download and install?” Xinyi asked, looking curiously at the stack of chips Dong Yang had left on the table, finally brave enough to ask.

“They are actually called SPI NOR flash memory. For a long time I thought they were EEPROM — I was silly back then too, since I had no formal education in electronics. Well, it’s similar to how everyone calls UEFI ‘BIOS.’ They used to be called EEPROM and the naming stayed, but EEPROM and SPI NOR flash are completely different tech,” Dong Yang said while connecting the USB cable to the PC and the microprogrammer.

Zhang Ming looked at his daughter, who had more questions on her face, and explained, “Erm, BIOS is the software that is inside the EEPROM chips. Well, technically it’s called UEFI now. It’s the software that starts when you press the power button, and then loads the operating system from the disk.”

“Why can’t we just download, edit, and install this BIOS in the operating system? The system can read and run it, right?” Xinyi asked.

“Although BIOS is software, it’s treated as hardware — because it’s permanently stored on the motherboard. So you can take it as hardware; you need to do surgery: remove the chip, burn new software into it, and solder it back onto the motherboard,” Zhang Ming said while removing the Dell motherboard from the case.

“Well, you actually can update them without going through the so-called surgery. But you will have to download the official BIOS image and use the official tools to update the BIOS — but that won’t solve our whitelist issue. Also, we are working with bin files, with hex. So those tools are not available to us,” Dong Yang said while pulling open the hex editor. The bin file filled the screen — rows of hex that no human eye could read. For a moment he just stared at the first lines, counting the old way, trying to guess where each component lived. “I used to do this line by line, you know. Took me days just to map one BIOS.” He glanced at the wall speaker. “Dong Dong, you can read it, right?”

“In general, you don’t need to update them if the BIOS on the PC works. If the BIOS on the PC suddenly doesn’t work, I would say 99% of the time those tools will not work, so it requires an SPI programmer. So a BIOS problem is a hardware problem,” Zhang Ming added.

“Right, it became a paradox. Updating a failing BIOS directly needs a working BIOS — so why update the BIOS if the BIOS is working? If the BIOS is not working, it won’t be able to update itself,” Dong Yang explained with a sense of humor.

“Speaking of BIOS updates — remember about two years ago, in 2024, the Raptor Lake 13th and 14th generation cores? Intel tried to fix hardware degradation with software updates to the BIOS, but ultimately failed,” Zhang Ming said.

“Ah, the 13th and 14th generation degradation issue. Those were magical times. Never in my life did I think I would see board manufacturers ship untested BIOS with a warning label claiming that the current BIOS version is in beta or still in development,” Dong Yang recalled with a smile.

“Were you affected by that incident?” Zhang Ming asked.

“Affected? The whole industry was shaken. You know how BIOS updates go — one accident at the power plug, and boom, the whole motherboard is dead for good. There was an update every month back then, and there was literally no one doing EEPROM swaps in Ezhou. I’m not a repair shop, but many people found their way to me and brought their boards,” Dong Yang said with exaggeration. “I don’t make any money from this, but there are many people I can’t possibly reject. I did get some dead boards for cheap online, with broken BIOS.”

After a moment of pause, Dong Yang looked at Zhang Ming and asked, “Somehow I was wondering what exactly went wrong. There was an urban legend saying that some additional oxygen atoms got in at some stage of manufacturing, making some batches less durable. Some said Intel screwed up the SA/IO voltage — but there would be way more dead CPUs if that were true. Some said Intel screwed up the Vcore… Like, seriously… Vcore? Vcore should literally take any beating as long as core temps are kept controlled — that’s how liquid nitrogen overclocking works.”

“I did follow the issue. It’s complicated. The oxygen part is urban legend. And the SA/IO thing — that was real, but it was 2023, and it was the board makers pushing the memory controller too hard, ASUS especially. Different story.”

“So what was it, then?” Dong Yang asked.

“The core voltage,” Zhang Ming said. “There was a bug in a microcode algorithm — eTVB, they called it. The CPU asks the motherboard for voltage all the time, hundreds of times a second, that’s normal. But this bug made it keep asking even when the cores were already hot. Months of that, and the silicon quietly degrades.”

“So it really is a software BIOS fix for a hardware problem? I wasn’t expecting that. I can totally see how that ended up. Intel first pushed the blame to board partners, then asked them to set a cap on Vcore voltage and hoped the issue would go away,” Dong Yang said.

“Exactly. What is physically broken can’t be fixed with a patch of new code. A BIOS is often treated like hardware, but it’s still software code. Intel then gave the affected CPUs a five-year extended warranty after all that drama,” Zhang Ming said.

“I never trusted the CPU myself to begin with. I always ran my CPU at a fixed frequency, and set the Vcore to the lowest voltage that passed the P95 cache test. Fixed frequency and voltage, no matter what the CPU wants. Now all my recent CPUs are ES chips, where my only options are adjusting the Vcore offsets — the lowest offsets that pass the P95 cache test,” Dong Yang said proudly.

“And that’s why yours never died. ES chips — we used to keep track of and collect those ES chips. Aren’t they collected now? I guess Intel laid off those as well,” Zhang Ming said with a sad smile.

“Dad, why doesn’t the BIOS update help Intel solve the problem with the CPU? And why would the motherboard die when updating the BIOS they provided?” Xinyi asked, as she had been holding all her questions for too long.

“Why? Because a BIOS update can only control the motherboard from sending too much voltage to the self-destructing CPU. It cannot fix the broken CPU — it can only prevent the CPU from further self-destructing. As for why the motherboard dies? You should treat BIOS as hardware here. When a BIOS update stops halfway, it’s like swapping an EEPROM chip halfway. The motherboard will have no way to continue, as it’s missing chips. So it becomes a hardware issue,” Dong Yang answered, before Zhang Ming could.


Author’s note — for readers who don’t understand the story

  • There is a meme called the BIOS paradox: updating the BIOS needs a stable system, but getting a stable system needs a BIOS update — thanks to the 13th and 14th gen Intel saga.
  • People in the tech industry love legacy terminology and industry shorthand. Some of the terms are inaccurate, and some name completely different technologies — like “EEPROM” and “BIOS” in this chapter.