The POWER and PowerPC General Discussion / News Thread

Ah, makes sense.

That’s a neat use case. Not really a fan of death and destruction on that scale, but I never really thought they’d need those sort of simulations.

I don’t have proof besides me calling into Raptor support and asking about it. They mentioned they had supply chain delays, and I figured that was as a result of COVID. I mean it’s great news that the flow of goods are mostly back to normal, but I’d imagine that more niche operations like Raptor’s would still be delayed anyway.

Perhaps my language was somewhat hyperbolic in retrospect. That wasn’t the intention.

2 Likes

It’s all good, I could have been overly critical too. I’ve been putting up with people (idiots) at my company losing their heads running around like it’s got a 90% death rate and every time I see “because covid” I have an autonomous response to reee.

I won’t be surprised if things are delayed, that’s what happens when you work on a small volume project with China. Additionally, there are small delays being caused by covid. It’s not like the “we’ll run out of components by April” that organizations like LTT are claiming, but it’s definitely a slight delay.

2 Likes

I appreciate the response from you. :slight_smile:

You are absolutely spot on about the hysteria around COVID. If people take reasonable precautions as they would during a normal flu season, this would almost be background noise in the grand scheme of things. But alas, somebody has to profit off the panic. :stuck_out_tongue:

2 Likes

Its big memes

2 Likes

So I attended the recent OpenPOWER coffee chat session hosted by Hugh Blemings. Seems like this will be done a few more times at least, and if anyone wants to check this out, here’s the link.

https://openpowerfoundation.org/openpower-virtual-coffee-calls

2 Likes

Slightly off topic note, but I recently went looking through the ATX and SSI EEB specifications, trying to understand the standoff layout on the Talos II, which is marketed as EATX. This was largely prompted by Gamers Nexus’ article about EATX being undefined (or video if you prefer), as Raptor’s website and even wiki makes no mention of the SSI EEB specification that mostly describes the form factor that Talos II actually uses.

I started a thread about this on the RCS forums if you want more details, but it seems like Talos II is using SSI EEB-compliant mounting locations (some of which are alternatives to the ATX locations), but then replaces the use of an EEB-required location C with a mounting location that not used by SSI EEB or ATX.

This mounting location is labelled B in the microATX specification, but if you were using SSI EEB conventions it would probably be called C′. I think this might date back to the IBM AT board, but I cannot find a specification for that.

Regardless, it makes for an interesting situation where you probably need a chassis/case that is both microATX and SSI EEB compliant to have all the necessary standoff holes predrilled.

@wendell did you have any issues with standoffs?

1 Like

When it says EATX, they mean SSI EEB. EATX is more marketable, but is not a term nor standard.

1 Like

The issue is that they do not mean SSI EEB, since their B (or perhaps C′ if you prefer) mounting location is only compliant with versions of SSI EEB older than 3.5; a modern chassis that is SSI EEB compliant is not guaranteed to have that mounting location.

SSI EEB v3.51 is copyright 2003, so this B/C′ hole has probably been obsolete for at least 16 years.

1 Like

https://www.winehq.org/pipermail/wine-devel/2020-April/165261.html

The first of about a dozen patches from André Hentschel to add ppc64le support to Winelib.

1 Like

Apart from being inherently awesome due to the “we do it because we can” factor; what is the goal of this port?
What does André plan to test against?

  • Were it for Big-Endian Power, I would assume your first test case could be applications from the old Windows NT port to PowerPC.
  • You could strap this to QEMU and perhaps run x86 Windows applications that way; I think Darwine was working on this in the Macintosh PowerPC era
  • some other usage I am not imagining?

I was expecting to see a Darling port before Wine, but I suppose stranger still is the choice to port to Little-Endian first; but maybe too much has changed in winelib to make Darwine’s work usable?

I think the big thing here is to allow native open source ports of MS Windows only software to the PPC world. Also maybe for bug and regression testing?!

I suppose I cannot actually recall any programs I know using winelib, open source or otherwise. I always assumed that along with the OpenGL DirectX-implementation, WineD3D, its use outside of Wine was more a curiosity or proof of concept. Is winelib being used currently by well known Linux software?

In more general POWER9 info, tle from the Raptor forums has a YouTube channel where he has tested different software on his Blackbird (two videos linked below); also, Borley from the same forums did a brief bit of testing comparing the different steppings (DD2.2 v. DD2.3) of a 4-core chip, though we aren’t sure if the DD2.3 test had the proper firmware; you can visit the thread here,
https://forums.raptorcs.com/index.php/topic,103.msg919.html


H.264 & H.265 decoder testing

Diablo I (via devilutionX project),

1 Like

I was curious if anyone else with a Power9 system has any data on latency.

I haven’t tried to measure it but it just “feels” as if the Power9 responds faster than a Ryzen 3900. Both of them are running SSDs. The 3900 has an Optane 905P, actually.

It seems like the Power9 opens SSH connections quicker and command-line programs finish faster.

Maybe it’s mostly the SSH? Does Power9 have some kind of acceleration, or lower latency encryption than Ryzen? How would I objectively quantify that?

I mean… its risc, yeah, thats kinda what happuns when you go from CISC to RISC.

Theres reasons the amiga was the goto video workstation back in the day, even tho ibm was the top.

2 Likes

How does transcoding video go power9 vs amd 3000?

I suppose this sums it up?

1 Like

Ah man I can’t wait for Power 10 and see how it performs with 7nm. Hopefully there is a larger push for software optimization.

3 Likes

Uh well converting video with ffmpg on power (970mp /power4-5) is faster then my 1200 ryzen 3 by minutes

Is that h264? Which codec?

1 Like