By: Gian-Carlo Pascutto (gcp.delete@this.sjeng.org), August 14, 2019 9:01 am
Room: Moderated Discussions
Adrian (a.delete@this.acm.org) on August 14, 2019 5:06 am wrote:
> This time the AMD CPUs are in a much better shape immediately after
> launch, so such rants are much less likely than in 2017.
>
> Even if there were a few problems with the software compatibility, e.g. with RDRAND and with the initialization
> of the segment registers after reset, I have not heard yet about any hardware problems.
It looks like AMD has reduced the (maximum) boost clocks on each subsequent AGESA update, to the point where the chips can't hit the advertised speed any more. Oops! Loud people on reddit and elsewhere are up in their arms about it. Now I think that's a silly thing to get worked up about, but it begs the question whether this is by accident or if it was needed to improve stability issues.
If it's the latter, there's an ugly question if the launch day benchmarks with higher clocked AGESA are still valid. Luckily we're only talking about <1% perf. But it's not the kind of thing you want to see.
Aside from that, I was hoping the issues with the extreme fickle memory controller from Zen 1 were resolved. As I just had to return a DDR4 kit, I'm not convinced there. I'm hoping it was the RAM at fault, not the CPU.
Another question is if the PMU now works accurately enough that rr works on Zen 2...
> This time the AMD CPUs are in a much better shape immediately after
> launch, so such rants are much less likely than in 2017.
>
> Even if there were a few problems with the software compatibility, e.g. with RDRAND and with the initialization
> of the segment registers after reset, I have not heard yet about any hardware problems.
It looks like AMD has reduced the (maximum) boost clocks on each subsequent AGESA update, to the point where the chips can't hit the advertised speed any more. Oops! Loud people on reddit and elsewhere are up in their arms about it. Now I think that's a silly thing to get worked up about, but it begs the question whether this is by accident or if it was needed to improve stability issues.
If it's the latter, there's an ugly question if the launch day benchmarks with higher clocked AGESA are still valid. Luckily we're only talking about <1% perf. But it's not the kind of thing you want to see.
Aside from that, I was hoping the issues with the extreme fickle memory controller from Zen 1 were resolved. As I just had to return a DDR4 kit, I'm not convinced there. I'm hoping it was the RAM at fault, not the CPU.
Another question is if the PMU now works accurately enough that rr works on Zen 2...