6 ms·

It's hard to imagine how one gains confidence in this kind of product. How does this vendor have the resources to qualify this chip? Intel can already barely

It's hard to imagine how one gains confidence in this kind of product. How does this vendor have the resources to qualify this chip? Intel can already barely qualify theirs. The last seven generations of AMD parts shipped with serious defects that were discovered by customers in the field. Centaur and Via are far smaller. Is it really possible to be scrappy and scrappily produce a bug-free assembly of billions of transistors?

wmf 7y ago
Centaur has already built several generations of x86 processors so you need to make a stronger argument than that.
- seriesf 7y ago
  Sure but what have they built that’s this complicated? The recent AMD flaws have been things like frequency scaling doesn’t work (zen). The more features you have, the harder it becomes to validate the product.
myself248 7y ago
Is it necessary to be bug-free? AMD still sells chips...
- kick 7y ago
  There's some sort of irony in calling out the x86_64 vendor who's been the least-touched by all of the recent vulnerabilities
  - saagarjha 7y ago
    To be fair, I am not sure people have been looking particularly deeply at Centaur processors.
    - loeg 7y ago
      I guess technically the Via Nano (2008) was an x86_64 ISA chip, but their previous cores were all 32-bit and they haven't actually productized the chip in the article yet. The Nano was functionally obsolete long before Spectre and Meltdown cropped up. So I don't believe GP was considering Centaur in the category "x86_64 vendor," which is maybe pedantically incorrect but probably broadly reasonable given the context.
  - myself248 7y ago
    To be clear, I was calling back to PP's example of AMD's errata. I just built a Ryzen machine myself and like their chips just fine. That was my point. Even if they have bugs, they sell and they work. Centaur's chips will have bugs. Intel's chips will have bugs. They all work, and they all sell. Being bug-free is not necessary, and an unfair bar for PP to set for Centaur when nobody else in the market meets it either.
ghettoimp 7y ago
Any processor in any product you have has bugs. For instance, according to Intel's own documentation[1] there are 137 errata for "8th and 9th Gen Core" processors, including unpredictable behavior, hangs, security key exposure, etc. It just has to be good enough to sell to folks. [1] https://www.intel.com/content/www/us/en/products/docs/processors/core/8th-gen-core-spec-update.html https://www.intel.com/content/www/us/en/products/docs/proces...
asimpletune 7y ago
I had a professor in college who contracted for them doing formal verification of their CPUs, in case you’re wondering.

wmf 7y ago

Centaur has already built several generations of x86 processors so you need to make a stronger argument than that.

seriesf 7y ago
Sure but what have they built that’s this complicated? The recent AMD flaws have been things like frequency scaling doesn’t work (zen). The more features you have, the harder it becomes to validate the product.

myself248 7y ago

Is it necessary to be bug-free? AMD still sells chips...

kick 7y ago
There's some sort of irony in calling out the x86_64 vendor who's been the least-touched by all of the recent vulnerabilities
- saagarjha 7y ago
  To be fair, I am not sure people have been looking particularly deeply at Centaur processors.
  - loeg 7y ago
    I guess technically the Via Nano (2008) was an x86_64 ISA chip, but their previous cores were all 32-bit and they haven't actually productized the chip in the article yet. The Nano was functionally obsolete long before Spectre and Meltdown cropped up. So I don't believe GP was considering Centaur in the category "x86_64 vendor," which is maybe pedantically incorrect but probably broadly reasonable given the context.
- myself248 7y ago
  To be clear, I was calling back to PP's example of AMD's errata. I just built a Ryzen machine myself and like their chips just fine. That was my point. Even if they have bugs, they sell and they work. Centaur's chips will have bugs. Intel's chips will have bugs. They all work, and they all sell. Being bug-free is not necessary, and an unfair bar for PP to set for Centaur when nobody else in the market meets it either.

ghettoimp 7y ago

Any processor in any product you have has bugs. For instance, according to Intel's own documentation[1] there are 137 errata for "8th and 9th Gen Core" processors, including unpredictable behavior, hangs, security key exposure, etc. It just has to be good enough to sell to folks. [1] https://www.intel.com/content/www/us/en/products/docs/processors/core/8th-gen-core-spec-update.html https://www.intel.com/content/www/us/en/products/docs/proces...

asimpletune 7y ago

I had a professor in college who contracted for them doing formal verification of their CPUs, in case you’re wondering.