By: Anon (no.delete@this.email.com), August 14, 2019 10:39 pm
Room: Moderated Discussions
Adrian (a.delete@this.acm.org) on August 14, 2019 5:06 am wrote:
> Gian-Carlo Pascutto (gcp.delete@this.sjeng.org) on August 14, 2019 1:33 am wrote:
> >
> > I look forward to reading a rant about the current state of AMD's BIOSes and memory
> > support though. (Goes on to RMA some DDR4 because it won't boot in his 3900X...)
> >
> > --
> > GCP
>
>
> This time the AMD CPUs are in a much better shape immediately after
> launch, so such rants are much less likely than in 2017.
>
>
> While in 2017 the first Ryzen was not able to even compile gcc without errors, now we have
> reports that e.g. a Ryzen 9 3900X can compile the entire DragonFlyBSD distribution with all
> its packages in a little bit more than 24 hours, without any issues, or that an (expensive)
> EPYC 2 server can compile the Linux kernel in 16 seconds, LLVM in 82 seconds and the complete
> gcc with all languages in 730 seconds, also without any problems (see at Phoronix).
>
>
> 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.
While I quite like Ryzen myself (currently setting up a little 3200G for a client.. and its more than enough), I think there is quite an issue touched on above..
For a fairly long time, I have used 90% Intel Chipset/CPUs for projects, and 10% AMD.
I would say the ratio would be much closer to 50/50 except for one area, chipset/driver/bios stability.
AMD has had a long history of good ideas with terrible software support, it seems.
Prior to Ryzen, my last use of them was Kaveri APUs (yes, I often need a semi-competent GPU..) and these were not bad at all, except for a TERRIBLE chipset Raid, and long term driver memory leaks that meant machines seemed to need a weekly (at best) reboot, and a 6 monthly drive format.
Further back, there were almost always issues of one type or another - and almost always to do with chipset..
A similar thing can be seen with Nvidia vs ATI/AMD graphics cards - NVidia was generally (not always) a rock when it comes to driver compatibility - AMD was much much less so.
Now, both areas SEEM to have improved, however history is history - and I must admit some trepidation any time I start evaluating a new AMD platform..
This is a problem with no easy fix, unfortunately. Having said that, so far Ryzen has been good to me - however it is early days.
> Gian-Carlo Pascutto (gcp.delete@this.sjeng.org) on August 14, 2019 1:33 am wrote:
> >
> > I look forward to reading a rant about the current state of AMD's BIOSes and memory
> > support though. (Goes on to RMA some DDR4 because it won't boot in his 3900X...)
> >
> > --
> > GCP
>
>
> This time the AMD CPUs are in a much better shape immediately after
> launch, so such rants are much less likely than in 2017.
>
>
> While in 2017 the first Ryzen was not able to even compile gcc without errors, now we have
> reports that e.g. a Ryzen 9 3900X can compile the entire DragonFlyBSD distribution with all
> its packages in a little bit more than 24 hours, without any issues, or that an (expensive)
> EPYC 2 server can compile the Linux kernel in 16 seconds, LLVM in 82 seconds and the complete
> gcc with all languages in 730 seconds, also without any problems (see at Phoronix).
>
>
> 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.
While I quite like Ryzen myself (currently setting up a little 3200G for a client.. and its more than enough), I think there is quite an issue touched on above..
For a fairly long time, I have used 90% Intel Chipset/CPUs for projects, and 10% AMD.
I would say the ratio would be much closer to 50/50 except for one area, chipset/driver/bios stability.
AMD has had a long history of good ideas with terrible software support, it seems.
Prior to Ryzen, my last use of them was Kaveri APUs (yes, I often need a semi-competent GPU..) and these were not bad at all, except for a TERRIBLE chipset Raid, and long term driver memory leaks that meant machines seemed to need a weekly (at best) reboot, and a 6 monthly drive format.
Further back, there were almost always issues of one type or another - and almost always to do with chipset..
A similar thing can be seen with Nvidia vs ATI/AMD graphics cards - NVidia was generally (not always) a rock when it comes to driver compatibility - AMD was much much less so.
Now, both areas SEEM to have improved, however history is history - and I must admit some trepidation any time I start evaluating a new AMD platform..
This is a problem with no easy fix, unfortunately. Having said that, so far Ryzen has been good to me - however it is early days.