4 ms·

I am an NVIDIA employee in an unrelated part of the business and not inclined to comment in depth on these matters and this is not an NVIDIA opinion or official

I am an NVIDIA employee in an unrelated part of the business and not inclined to comment in depth on these matters and this is not an NVIDIA opinion or official but i thought it worth mentioning I don’t see the distinction? NVIDIA is fabless too?? (Heck, so is AMD I think)

fredoralive 5y ago
Nvidia doesn't licence its designs to other companies for incorporation into their own chip designs, ARM does. That is the main distinction I think, and why people want it to be a neutral third party.
- maxwell86 5y ago
  IIUC, part of the deal was that NVIDIA would license its designs via ARM.
  - 310260 5y ago
    They say that now... It could even very well be the case. However, there are plenty of ways to manipulate those licensing agreements to work more in Nvidia's favor than they do today.
    - maxwell86 5y ago
      They've been saying that from the beginning. There are billions of ARM devices, the ability to sell a GPU to each is worth money.
      - kasabali 5y ago
        Ability to forcibly sell a gpu to each is worth even more money (ie. see lawsuit on how Quallcomm bundle their modem on every SoC for a premium)
uxp100 5y ago
AMD spun their manufacturing off as Global Foundries, so yes. And yeah, I’m not sure this distinction the poster is making between arm as like a software company and NVIDIA like a hardware company makes much sense. Arm makes IP (in the chip design sense). NVIDIA makes IP and combines it with IP from Arm and others, and produces some product designs based on the chip, which NVIDIA tests and writes software for. But ODMs make the products and fabs make the chips. I really don’t think calling some of this like software and some like hardware is very explanatory, at least partially because NVIDIA writes a lot of software, and because I think the person reading that description might come away with the impression that arm just produces the architecture, and not actual core designs (which NVIDIA, Denver aside, uses).
- cogman10 5y ago
  They also stopped using global foundries and now use TSMC. Global foundries, like pretty much everyone other than TSMC, fell behind in process shrinks. Their primary business is fabricating secondary chips.
  - chasil 5y ago
    Global Foundries is still used as the northbridge chiplet inside modern AMD processors; only the CPU core chiplets are from TSMC. Global Foundries has sizable operations in Dresden, Germany. Interestingly, this was a major semiconductor supplier of the Eastern Block, prior the fall of the iron curtain. AMD placed their primary foundry in Dresden, likely for the infrastructure and technical knowledge. Global Foundries has decided that more money can be made at 14nm and above, than what is required for smaller process nodes.

fredoralive 5y ago

Nvidia doesn't licence its designs to other companies for incorporation into their own chip designs, ARM does. That is the main distinction I think, and why people want it to be a neutral third party.

maxwell86 5y ago
IIUC, part of the deal was that NVIDIA would license its designs via ARM.
- 310260 5y ago
  They say that now... It could even very well be the case. However, there are plenty of ways to manipulate those licensing agreements to work more in Nvidia's favor than they do today.
  - maxwell86 5y ago
    They've been saying that from the beginning. There are billions of ARM devices, the ability to sell a GPU to each is worth money.
    - kasabali 5y ago
      Ability to forcibly sell a gpu to each is worth even more money (ie. see lawsuit on how Quallcomm bundle their modem on every SoC for a premium)

uxp100 5y ago

AMD spun their manufacturing off as Global Foundries, so yes. And yeah, I’m not sure this distinction the poster is making between arm as like a software company and NVIDIA like a hardware company makes much sense. Arm makes IP (in the chip design sense). NVIDIA makes IP and combines it with IP from Arm and others, and produces some product designs based on the chip, which NVIDIA tests and writes software for. But ODMs make the products and fabs make the chips. I really don’t think calling some of this like software and some like hardware is very explanatory, at least partially because NVIDIA writes a lot of software, and because I think the person reading that description might come away with the impression that arm just produces the architecture, and not actual core designs (which NVIDIA, Denver aside, uses).

cogman10 5y ago
They also stopped using global foundries and now use TSMC. Global foundries, like pretty much everyone other than TSMC, fell behind in process shrinks. Their primary business is fabricating secondary chips.
- chasil 5y ago
  Global Foundries is still used as the northbridge chiplet inside modern AMD processors; only the CPU core chiplets are from TSMC. Global Foundries has sizable operations in Dresden, Germany. Interestingly, this was a major semiconductor supplier of the Eastern Block, prior the fall of the iron curtain. AMD placed their primary foundry in Dresden, likely for the infrastructure and technical knowledge. Global Foundries has decided that more money can be made at 14nm and above, than what is required for smaller process nodes.