Due To A GCC Bug, RDRAND Usage Wasn't As Random As Expected

[pal666]

Intel didn't copy x86-64, they licensed it from AMD, just like AMD licenses x86 from Intel. They literally can't exist without eachother at this point.

they just have cross-license agreement

[SystemCrasher]

1. Intel did it, therefore EVIL CONSPIRACY.
2. AMD copied it: Silence.
3. A year or two after RyZen has been on the market: Look at this cool random number feature that AMD invented single-handed out of thin air! Thank you for being the only innovative company in the history of mankind AMD!!

Sure, AMD is innovative company. After Intel boasted their blasted ME, AMD also implemented their "security" processors, so these days AMD platform isn't anyhow better than Intel. However Intel got grilled far more on their ME backdoor basis. Well, after all Intel also scores major market share.

P.S. CVE in compiler?! Oh dammit, that's what I would call "innovation" for sure. Too bad I'm not a big fan of innovations like this.

[F.Ultra]

Intel didn't copy x86-64, they licensed it from AMD, just like AMD licenses x86 from Intel. They literally can't exist without eachother at this point.

As for IA64, it was a FAR superior architecture to x86 and x86-64. The only reason it lost was the emulation penalty to x86 apps. x86-64 was initially higher performing (no 64 bit apps yet) and so won in the market. We'd be in a much better place if we had a CPU arch that was explicitly parallel, rather then stick with x86 based CPUs. [x86 is a HORRID CPU architecture that should have been killed off decades ago]

Except of course that on launch date Intel:s own x86 (at the time Pentium 4 @1.7Ghz) outperformed the IA64 on both integer performance and memory bandwidth and that was not when the Itanium was running on x86 mode but in native IA64 mode. The reason it lost was not due to it's x86 emulation being too slow but that they never could deliver on the promise of the EPIC architecture since it turned out that superscalar decisions where better done on chip than by a compiler. No one bought them to run x86 software (or very few at least did).

[tildearrow]

Typo:

Using older versions of GCC could lead to RDRAND/RDSEED instrinsics producing non-random results.

[DanL]

* Funny how literally nobody has complained about AMD copying Intel and adding these instructions to RyZen though.

You really are an Intel fanboy, aren't you?

Just your standard first post drive-by anti-AMD trolling, the guy does it at Tech Report too.

Yeah, it's annoying there too. Tries way too hard to be funny or clever or whatever...

[rene]

Intel didn't copy x86-64, they licensed it from AMD, just like AMD licenses x86 from Intel. They literally can't exist without eachother at this point.

As for IA64, it was a FAR superior architecture to x86 and x86-64. The only reason it lost was the emulation penalty to x86 apps. x86-64 was initially higher performing (no 64 bit apps yet) and so won in the market. We'd be in a much better place if we had a CPU arch that was explicitly parallel, rather then stick with x86 based CPUs. [x86 is a HORRID CPU architecture that should have been killed off decades ago]

PS: how is "licensing" not copying? it is not that AMD gave them their CPU blueprints anyways, Intel obviously copied^W implemented the internal details their way, ...

Itanic only was superior in power consumption, and heat generation. The molecule things could rarely be filled optimally by any compiler, and the silicon was an even more over complex monster of transistors, ... There for sure are reasons why each new ia64 silicon architecture over delayed, and under delivered. Been there, done that, did not wanted to keep one even if they would have paid me: https://t2sde.org/architectures/ia64/

I do not believe explicitly parallel would help us, most programs can not be nicely explicitly parallelized on the ISA level, we have this in GPUs, pain to program flexible. We should rather have way more, way more simple cores, without all the AVX vector fluff most normal apps are not touching anyways.

[AndyChow]

Doesn't really matter, since hardware random numbers generators are only used to add some entropy, not define it. I think FreeBSD are the only ones that actually once relied only on the HW-RNG, and, it's not surprising since FreeBSD is has dropped the ball when it comes to security.

[sykobee]

Just your standard first post drive-by anti-AMD trolling, the guy does it at Tech Report too.

[TheBlackCat]

Checked some info on the matter:
VIA's first RNGs should have been in the C3/C3-2 (I still have some). Released acc. to wikipedia early 2003.
Geode LX definitely had a RNG. NS iirc. had the GX and I haven't found RNG infos on that, Geode was sold 2003 to AMD, which continued the GX2 design but quickly developed the LX upgrade.

And for HW-RNGs in general:
https://en.wikipedia.org/wiki/Hardwa...tor#Early_work
(so far you can trust the editors of WP if the did a thorough research, but if you want, every dice is a HW RNG - just not electronic.)

Too...many...acronyms...brain...melting.

[gamerk2]

2nd. Intel copied AMD64, which was way more popular than Intel's failed, expensive, and power hungry IA64 nobody wanted,

Intel didn't copy x86-64, they licensed it from AMD, just like AMD licenses x86 from Intel. They literally can't exist without eachother at this point.

As for IA64, it was a FAR superior architecture to x86 and x86-64. The only reason it lost was the emulation penalty to x86 apps. x86-64 was initially higher performing (no 64 bit apps yet) and so won in the market. We'd be in a much better place if we had a CPU arch that was explicitly parallel, rather then stick with x86 based CPUs. [x86 is a HORRID CPU architecture that should have been killed off decades ago]