264 works but I need newer then the old power hardware to get 265 working right if not it’s wayyyyyyyyyy slower
I can get kernel 5.7 working on g5 and g4 computers and not my ryzen3 1200
Like I find if you optimize correctly you get way hella faster on ppc then x86
RISC is where it’s at and openpower is a few instructions off from risc v
Also why cards like that giga-accel exist. Purely because its that staggering of speed. I don’t even use my cpu to stream or record video thx to that btw, I have a gtx 970 dedicated to ffmpeg and nvenc xD
What about the brand new H.266 VVC?
It’s literally about to come online with July 2020 as it’s completion date for the standard.
Does anyone have interest in a G5? I was thinking of posting the one I have spare as I only need one. Take a bit of time and do a mirror finish on the outside with a power buffer, install OSX all modded up and streamlined for performance, install Void for everything else, pop it on here or ebay.
Giving first dibs for offers here. I’ll be looking into buffing it soon.
You can’t stop me, the case is scuffed and awful. I’ll get an apple stencil and scorch an apple insignia into the sides if you really really want me to. It’ll take 20 minutes lol.
LMK.
It is old, and slower than today’s stuff and not fully featured, but IBM have opensourced their BlueGene/Q CPU. I am not in this power thing, it makes no sense to me but it is interesting to read casually. You can implement this in an FPGA, modify it or even spin up actual chips if you like apparently.
I cannot find any Power systems being sold by that seller, so those being the leftovers from a CPU upgrade or replacement does seem likely.
Tested by searching for IBM, Google, Zaius, Barreleye, Tyan, Yadro in https://www.ebay.com/sch/i.html?_nkw=<search term>&_saslop=1&_sasl=microint
The part number is strange, the only reference to it appears to be that particular eBay seller. Looking at some of microint’s POWER9 eBay listings, they appear to be selling both modules, with the difference seen in how the pins are arranged. Some packages have a pin/socket split into two halves, others have a single grid with a small gap of absent pins in the middle.
| Part | microint listing | Price | Notes |
|---|---|---|---|
| 00UM648 | eBay item 183967882705 | 99 USD | bisected socket |
| 02CY388 | eBay item 153655284860 | 99 USD | centre gap in socket |
| 02CY396 | eBay item 164156776050 | 99 USD | centre gap in socket |
| 00NJ261 | eBay item 183967855424 | 99 USD | bisected socket |
| 02CY382 | eBay item 183967860999 | 99 USD | centre gap in socket |
Looking at pictures of the Zaius board (used by Google and Rackspace), which is known to use LaGrange modules, my guess would be that the bisected socket is used by LaGrange, and the socket with a centre gap is Monza. Not quite that simple: ERRATA
Additionally, most of IBM’s early marketing seems to be of a module with the centre-gap socket; this lines up with the AC922, which we know uses Monza modules, being the first POWER9 machine sold by IBM.
Other information, HLandau on the RCS Wiki has a table of information on POWER9 machines he knows about. Others here may find it interesting as I have.
odd its too good to be true tho
Well its not gunna be the 22 core its gunna be the 2 core 8 threads or the 4 core 16 threads I’ll bet
A 4-core DD 2.1 Sforza from Raptor is still 500+ USD, this would still be a good deal if you had a board for it. Has anyone seen a Monza/LaGrange machine on the used market?
IMO I’d rather have the low core super high thread chips in my Power machine. Parallel compute ftw.
Are you referring to the SMT8 config? As part of the same discussion where he linked his table of known systems, HLandau mentioned that SMT4 vs SMT8 is mainly a fusing difference, so I am unsure how much of a difference SMT4 vs SMT8 really makes.
It is not quite so simple apparently; looks like @olddellian was asking about this same thing two years ago, and power_gaz on Twitter clearly identified a centre-gap socket module as LaGrange, saying:
These are a pair of LaGrange SCMs, as used in #IBMPowerSystems #S922 #H922 #S924 #H924 #S914 #L922 One is mine and the other belongs to @mr_nmon
There are two chips in the tweeted image, one labelled 02AA547 POWER and one flipped over showing the centre-gap pin arrangement.
this way the L1 Forums posts will also all link back here for those stumbling across them
adi_gangidi reply to @olddellian on Twitter, SMT8 LaGrange and SMT4 LaGrange are physically compatible
(mentioned on L1 Forums) — @q66 (octaforge) and @olddellian discussing cooler mounting compatibility between POWER8 (Turismo?) and POWER9 Sforza:
(mentioned on L1 Forums) — adi_gangidi photo comparing POWER8 Turismo and POWER9 LaGrange, both sizes visible:
(mentioned on L1 Forums) — size comparison between Monza, LaGrange, and Sforza at OZLabs; IHS visible
(Axone or POWER9 AIO/Prime is cancelled as far as I know; I should write the mailinglist to double check…)
My brudda doin gods work over here
D-didn’t they always?
Huh… Cool. Now I just need not T2 SDE on sparc so some sorta linux is useable.
If you check their platforms list, OpenBSD supports macppc and has supported pegasus and socppc in the past.
So the change is probably to 64-bit and supporting OpenPOWER firmware (petitboot and OPAL). Petitboot relies on kexec to switch from petitboot’s Linux kernel to the Linux kernel of the OS, so I wonder what sorts of changes are needed, if any, to kexec into a BSD kernel instead.
There is a sad backstory behind pegasos support being dropped; apparently the vendor cheated an OpenBSD developer out of payment for 3 months.
im intrested
Its the same as the other. Agp, dp 2.0, 4 r’m slots, firmware 5.1.4 so not upgradeable UCH neither is my otherone. Its at fw5.2
