Linux Kernel Expectations For AMD Threadripper 2

[phoronix]

Phoronix: Linux Kernel Expectations For AMD Threadripper 2

If you have already pre-ordered your AMD Threadripper 2990WX processor or just planning to be an early customer of that high-end desktop processor or the Threadripper 2950X, you may be wondering about Linux requirements from these new high-end AMD CPU offerings. Here's the gist of the Linux support state of AMD Zen+ CPus for those wanting to get ready for Threadripper 2...

http://www.phoronix.com/scan.php?page=news_item&px=Threadripper-2-Kernel-Bulletin

[chuckula]

It would be great if AMD finally got around to reporting the temperature accurately instead of playing stupid games like this.

+27 degrees is even stupider than the +10 degrees from the first RyZens.

[tildearrow]

Minor typo:

of AMD Zen+ CPus for those wanting to get ready for Threadripper 2

[hiryu]

So far I haven't been able to find anything on the net about when the embargo passes other than "next week". Which could mean any day of the week. Does anyone have a hard date?

Also, is it known if we're getting a new chipset? Perhaps we won't as the old motherboards are still compatible. When I attempt to search on this topic, I just find X399 boards.

My requirements for Threadripper 2 don't need gaming performance, etc to meet or exceed Coffeelake... Just Skylake-X. I know things were a bit behind with the first iterations of Ryzen and Threadripper (and possibly still for Ryzen 2, just less so?), but it would be nice to make the switch. I bought/built a Skylake-X system late November/early December... Only to have it announced we needed a performance killing mitigation for meltdown on Jan 1st. I was and still am livid about this to the point where I'm already considering ripping out my mobo/cpu to go with AMD. While KPTI, etc hasn't really hurt gaming performance... I use my machine for more than gaming, and some of loads are system call heavy.

Anyway, I'm definitely excited about this launch to see where the numbers ultimately end up being, and because I bought AMD shares early in the year and I'm already coming out quite well.

[tildearrow]

It would be great if AMD finally got around to reporting the temperature accurately instead of playing stupid games like this.

+27 degrees is even stupider than the +10 degrees from the first RyZens.

Maybe their reason for their flawed temperature reporting is to decrease the range (the limit for an 8-bit signed integer is 127) and to please liquid nitrogen cooling overclockers?

[Tomin]

So far I haven't been able to find anything on the net about when the embargo passes other than "next week". Which could mean any day of the week. Does anyone have a hard date?

Anandtech said August 13th, the same day as they are available. I think it was the same with Ryzen 2.

https://www.anandtech.com/show/13123/amd-threadripper-2-teaser-pre-orders-start-today-up-to-32-cores/5

My CPU-focused review, using our newest benchmark suite, will be posted on August 13^th.

Also, is it known if we're getting a new chipset? Perhaps we won't as the old motherboards are still compatible. When I attempt to search on this topic, I just find X399 boards.

There won't be a new chipset. Older boards will work with WX processors (the ones with 250 W TDP) if they have enough headroom and manufacturer allows it. I think X processors (180 W TDP) are fine on all older motherboards with updated firmware. Anandtech also mentions this:

https://www.anandtech.com/show/13123/amd-threadripper-2-teaser-pre-orders-start-today-up-to-32-cores/2

-- these processors should go straight in to all motherboards currently on the market without a BIOS flash, although once installed, an updated BIOS is recommended for enhanced memory and feature support.

So they somehow made it work even without BIOS updates if I understood correctly but you should update anyway.

[andreano]

Maybe their reason for their flawed temperature reporting is to decrease the range (the limit for an 8-bit signed integer is 127) and to please liquid nitrogen cooling overclockers?

Is it even intentional? Why keep changing it then?

I assumed it was design flaws, but is designing a small thermistor thermometer so impredictable as to overshoot by a constant 27°C (thermistors are nonlinear)?

[bridgman]

I suspect the different per-product offset is related to something like thermal resistance between junction and sensor.

[lucasbekker]

I suspect the different per-product offset is related to something like thermal resistance between junction and sensor.

I heard that the offset had something to do with its affect on the fan speed of the cooling and the various turbo states.

[kobblestown]

But why wasn't this offset communicated earlier? It seems like a really simple and non controversial change that could be accepted in the kernel even without real hardware to test with. Big fail for AMD in my opinion. I say this not because I hate them but because I like the Zen architecture and I want it to succeed.