We've been living in that world for a while - what do you think is new here? E.g. http://www.anandtech.com/show/8376/intel-disables-tsx-instru... (and Intel is suspiciously tight-lipped about the (security?) impact of the TSX bug, while AMD's FMA3 bug seems to be a comparatively-harmless hang.)
What security impact of the Haswell/early-Broadwell TSX bug? The support for all TSX instructions where really disabled by the microcode update, and we know that for sure because it crashed the world instantly (in Linux).
Now, working TSX can have a security impact, yes. At least on TSX as implemented on later Broadwell/early Skylake, memory accesses done from inside an speculative region are allowed to bypass stuff they shouldn't. Is that what you meant?
Digging into the history briefly, patchable microcode seems to have existed in commercially released computers as far back as the 1960s.
"...CCROS [microcode] cards on the [IBM System/360] Model 30 could be replaced in the field. This made design modifications easier to accomplish, as well as providing for field diagnosis. ...
The later System/360 Model 25 and models of the later System/370 series were configured with at least one portion of the control store being read-write for loading microcode patches and microdiagnostics."
They've been demoing with nvidia cards for months. It is kind of odd, but nvidia has quite a performance lead at the moment. If you're selling a CPU to people who have a lot of money, and those people are worried about their top-end GPUs being bottlenecked by the CPU, it's a more effective demo. I do expect the demos to change when Vega comes out later this year.
I submitted this with a more informative title, but a mod reverted it to the vague original title. So let me explain what this is about...
The big news is that on the Ryzen launch day, despite benchmarks in general looking outstanding, the gaming ones were somewhat disappointing. AMD said it is due to software problems (eg. detecting the CPU vendor through cpuid and enabling Intel-only optimizations only for Intel CPUs). People seemed to be dubious or overlooked this.
But now AMD proved that with software-only optimizations that took 1-2 weeks they were able to boost the Ryzen gaming performance by 15-30% on the 2 major games they worked on.
Failure to boot, random freezes in Windows and Linux.
That said, I got it sorted out. It turned out to be an incompatibility with both the NVMe disk I was using as well as an incompatibility with the memory I had purchased.
Getting the exact parts listed in your motherboard manufacturer's compatibility list appears to be pretty important for Ryzen currently. It's been rock solid since I swapped out those components.
I'd be interesting in knowing, too. I'm in the market for a new workstation, but not sure I yet trust the Ryzen platform to give me the stability I need.
OT: That social media bar is killing me. It takes up the firsts few columns of letters making it difficult to read, and it keeps reappearing whenever I scroll the page.
Who does stuff like that?
Edit: I actually stopped reading the article because it became so annoying.
23 comments
[ 2.9 ms ] story [ 62.7 ms ] threadNow, working TSX can have a security impact, yes. At least on TSX as implemented on later Broadwell/early Skylake, memory accesses done from inside an speculative region are allowed to bypass stuff they shouldn't. Is that what you meant?
Was this not the same thing? Some microcode update to patch a CPU flaw essentially?
That was almost a decade ago.
"...CCROS [microcode] cards on the [IBM System/360] Model 30 could be replaced in the field. This made design modifications easier to accomplish, as well as providing for field diagnosis. ...
The later System/360 Model 25 and models of the later System/370 series were configured with at least one portion of the control store being read-write for loading microcode patches and microdiagnostics."
https://people.cs.clemson.edu/~mark/uprog.html
The big news is that on the Ryzen launch day, despite benchmarks in general looking outstanding, the gaming ones were somewhat disappointing. AMD said it is due to software problems (eg. detecting the CPU vendor through cpuid and enabling Intel-only optimizations only for Intel CPUs). People seemed to be dubious or overlooked this.
But now AMD proved that with software-only optimizations that took 1-2 weeks they were able to boost the Ryzen gaming performance by 15-30% on the 2 major games they worked on.
maybe they just optimized the game.
https://www.pcper.com/reviews/Processors/Ashes-Singularity-G...
Whatever they did for AoS significantly increased 1800x performance while keeping 6900k performance essentially flat.
That said, I got it sorted out. It turned out to be an incompatibility with both the NVMe disk I was using as well as an incompatibility with the memory I had purchased.
Getting the exact parts listed in your motherboard manufacturer's compatibility list appears to be pretty important for Ryzen currently. It's been rock solid since I swapped out those components.
Who does stuff like that?
Edit: I actually stopped reading the article because it became so annoying.
That's actually a lot for a memory access. All these launch day benchmarks will be obsoleted.