5 ms·

This is rather misleading. The RTL8366 pictured in the tweet doesn't have an 8051 in it which is why it's absent from the datasheet. I don't see anything in thi

by mbell 7y ago

This is rather misleading. The RTL8366 pictured in the tweet doesn't have an 8051 in it which is why it's absent from the datasheet. I don't see anything in this thread that contradicts this. The RTL8367 does have an 8051 in it, and it is covered in the datasheet. As near as I can tell this entire claim is based on reading a RTL8366 datasheet when the part in question was a RTL8367. I'm not seeing anything nefarious here at all, the RTL8367 is designed for 'smart switches', the 8051 is there to run the webUI. In general, 8051s are _everywhere_, it's one of the most used IP cores in the world. I'd bet the average household has a a few dozen of them in various devises. EDIT: Looking through some older tweets from the same author, I think the claim is based on pinout similarities between the RTL8366 and the RTL8367. My guess would be that these two parts are the same die and the RTL8366 parts are the ones where a defect was found in the CPU core which would be a likely source of manufacturing defects.

joezydeco 7y ago
It's not always a binning thing. I've used similar parts, like bluetooth transceivers, where there are two SKUs: one is a cheaper commodity part and comes with built-in firmware. The other more expensive part allows for modifiable firmware (and, thus, exposes the processor inside). My cynical POV is that the chipmaker knows they'll be spending more internally on SDK development/support and field rep time. So that needs to be recaptured in the cost of the part.
- sharpneli 7y ago
  It is expensive to tape out a different chip. As 8051 originally was about 50000 transistors that’s peanuts nowadays. Likely cheaper to manufacture a single chip and do market segmentation via software than manufacture two if the savings are measly (let’s assume 16nm process) 1/500 mm^2 per chip.
  - bhouston 7y ago
    Intel only taps out about three different dies per generation line I believe. For example the Skylake-SP line just has LCC, HCC and XCC dies, and the the whole multitude of their product line comes from those three. https://investorshub.advfn.com/boards/read_msg.aspx?message_id=141418711 https://investorshub.advfn.com/boards/read_msg.aspx?message_...
  - close04 7y ago
    I think these are actually built on 150nm process. Using a modern process like 16nm would command some hefty premiums, and some redesign work. This is what old fabs end up manufacturing: high volume, low cost ICs. But you're right about the rest. The segmentation is achieved with fuses and firmware.
  - mbell 7y ago
    It's not so much the 'core' as the memory. For example it looks like the RTL8367 has 16KB of program memory, 32KB of data memory, 256bytes of SRAM and a 12KB nic buffer. These memory cells will make up the bulk of the transistors in the 8051 part of the chip and tend to be points of manufacturing failure. In many cases the chip would be designed with more memory than it needs under the assumption there will always be a few failures. So if you need 16KB, you design in 18KB and turn off the rows/banks that fail or test as the worst. If too many fail you sell it as next memory size down, or in this case you zap a couple fuses and sell it as the part without an MCU on board. Of course this doesn't always work, sometimes you need more of the non-perfect parts than you have to fill orders so you just zap the fuses anyway.
  - 0xcde4c3db 7y ago
    There are also chips that support two different feature sets via a mode pin, but instead of documenting the mode pin, the modes are sold as separate products with the mode pin simply labeled as Vcc/ground/NC.
- IfOnlyYouKnew 7y ago
  > My cynical POV is that the chipmaker knows they'll be spending more I don't even see the cynicism. But, in any case, most people would probably consider the standard business school price finding advice to be more cynical: costs are a lower limit only. Beyond that, only the customer's willingness to pay is relevant. Airline fares are good examples, or any of the food products being sold branded as well as unbranded/store-brand/generic/white label versions.
paggle 7y ago
It’s probably just cheaper to include the 8051 in every part and disable it for people who have bought the cheaper part. The way that some Teslas are range handicapped and can be upgraded over the air and given out for free during emergencies.
- zxcvbn4038 7y ago
  Back in the day my father was a service tech for NCR retail systems. For one of the lines they shipped out every system fully loaded with ram and disk - however how much their customers got to use depended on jumper configurations which the service techs would set and verify. I think IBM does the same thing with their mainframes - they ship out fully loaded but you can only use what you pay for, the rest is used as hot stand-by for hardware failures.
  - mehhh 7y ago
    Was this during NCR's bad years where AT&T ran the business into the ground? I don't think NCR really recovered till the NCR 7878 came out, setting the bar for a robust scale/scanner
    - hobojones 7y ago
      Former NCR tech here. I can confirm the parent’s comment - we’d routinely have to validate/ensure jumpers based on entitlement. As for recovery, I worked for NCR once they regained their independence from ATT and they were still running themselves into the ground. By the time I left (around the era of the 7878) things improved slightly.
      - mehhh 7y ago
        NCR seems to be able to survive, but I wonder if the non-Windows point of sale systems will finish eating the low & midrange business. My old employer is prepping for the shift with their software vendor currently, should allow for a much longer period of support & remote debugging...
  - 7y ago
kurthr 7y ago
I'd agree, but there's little reason to speculate. If you can't run it (disabled by fuses), I'm not sure it matters... if you want to confirm the silicon presence, then a little concentrated sulfuric acid to take the package off, and just a 100x microscope and you'll be able to see the memories for the uP. A strange accusation without proof. Fair warning I can't open the Tweet, Twitter is blocking me... or maybe I uBlock them? so I'm making a statement in similar unsupported ignorance.

by mbell 7y ago

joezydeco 7y ago

It's not always a binning thing. I've used similar parts, like bluetooth transceivers, where there are two SKUs: one is a cheaper commodity part and comes with built-in firmware. The other more expensive part allows for modifiable firmware (and, thus, exposes the processor inside). My cynical POV is that the chipmaker knows they'll be spending more internally on SDK development/support and field rep time. So that needs to be recaptured in the cost of the part.

sharpneli 7y ago
It is expensive to tape out a different chip. As 8051 originally was about 50000 transistors that’s peanuts nowadays. Likely cheaper to manufacture a single chip and do market segmentation via software than manufacture two if the savings are measly (let’s assume 16nm process) 1/500 mm^2 per chip.
- bhouston 7y ago
  Intel only taps out about three different dies per generation line I believe. For example the Skylake-SP line just has LCC, HCC and XCC dies, and the the whole multitude of their product line comes from those three. https://investorshub.advfn.com/boards/read_msg.aspx?message_id=141418711 https://investorshub.advfn.com/boards/read_msg.aspx?message_...
- close04 7y ago
  I think these are actually built on 150nm process. Using a modern process like 16nm would command some hefty premiums, and some redesign work. This is what old fabs end up manufacturing: high volume, low cost ICs. But you're right about the rest. The segmentation is achieved with fuses and firmware.
- mbell 7y ago
  It's not so much the 'core' as the memory. For example it looks like the RTL8367 has 16KB of program memory, 32KB of data memory, 256bytes of SRAM and a 12KB nic buffer. These memory cells will make up the bulk of the transistors in the 8051 part of the chip and tend to be points of manufacturing failure. In many cases the chip would be designed with more memory than it needs under the assumption there will always be a few failures. So if you need 16KB, you design in 18KB and turn off the rows/banks that fail or test as the worst. If too many fail you sell it as next memory size down, or in this case you zap a couple fuses and sell it as the part without an MCU on board. Of course this doesn't always work, sometimes you need more of the non-perfect parts than you have to fill orders so you just zap the fuses anyway.
- 0xcde4c3db 7y ago
  There are also chips that support two different feature sets via a mode pin, but instead of documenting the mode pin, the modes are sold as separate products with the mode pin simply labeled as Vcc/ground/NC.
IfOnlyYouKnew 7y ago
> My cynical POV is that the chipmaker knows they'll be spending more I don't even see the cynicism. But, in any case, most people would probably consider the standard business school price finding advice to be more cynical: costs are a lower limit only. Beyond that, only the customer's willingness to pay is relevant. Airline fares are good examples, or any of the food products being sold branded as well as unbranded/store-brand/generic/white label versions.

paggle 7y ago

It’s probably just cheaper to include the 8051 in every part and disable it for people who have bought the cheaper part. The way that some Teslas are range handicapped and can be upgraded over the air and given out for free during emergencies.

zxcvbn4038 7y ago
Back in the day my father was a service tech for NCR retail systems. For one of the lines they shipped out every system fully loaded with ram and disk - however how much their customers got to use depended on jumper configurations which the service techs would set and verify. I think IBM does the same thing with their mainframes - they ship out fully loaded but you can only use what you pay for, the rest is used as hot stand-by for hardware failures.
- mehhh 7y ago
  Was this during NCR's bad years where AT&T ran the business into the ground? I don't think NCR really recovered till the NCR 7878 came out, setting the bar for a robust scale/scanner
  - hobojones 7y ago
    Former NCR tech here. I can confirm the parent’s comment - we’d routinely have to validate/ensure jumpers based on entitlement. As for recovery, I worked for NCR once they regained their independence from ATT and they were still running themselves into the ground. By the time I left (around the era of the 7878) things improved slightly.
    - mehhh 7y ago
      NCR seems to be able to survive, but I wonder if the non-Windows point of sale systems will finish eating the low & midrange business. My old employer is prepping for the shift with their software vendor currently, should allow for a much longer period of support & remote debugging...
- 7y ago

kurthr 7y ago

I'd agree, but there's little reason to speculate. If you can't run it (disabled by fuses), I'm not sure it matters... if you want to confirm the silicon presence, then a little concentrated sulfuric acid to take the package off, and just a 100x microscope and you'll be able to see the memories for the uP. A strange accusation without proof. Fair warning I can't open the Tweet, Twitter is blocking me... or maybe I uBlock them? so I'm making a statement in similar unsupported ignorance.