News articles eventually fall away, so I am starting this thread for Harald 'Geennaam' Kanning's announced Mirari PPC motherboard, as there should be lots to discuss. Some of this may not come from anybody but Geennaam himself, but links to offsite posts still count as answers. A few comments of my own:
1) How much memory does it address? I don't see that in the writeup.
2) My only grumble is a preference for more expansion:
Quote:
The T2081 does not have SATA interfaces. If you need SATA, it must be added with a PCIe card. However, keep in mind that SATA is being phased out, with modern PCs and laptops often using NVMe over SATA SSDs.
I wonder how hard it would be to support an NVMe to SATA adapter board like
The description says it's Linux compatible. Many Amiga users still feel nostalgia for floppy drives, and that goes double for me with optical drives, so SATA (even PATA) is not out.
With an NVMe adapter and networking and sound on the motherboard, there may not be much need for extra cards, but do PCIe splitters often/sometimes work without firmware or drivers? Any such remaining cards might be satisfied with one lane each, and a PCIe 4x splitter like this
I believe the only OS4 SCSI driver is for PCI cards, so I lament that loss, but will survive.
3)
Quote:
Although we have only one memory controller, initial Linux benchmarks show that the single and SMT memory performance is at least on par with the X5020.
The T1042 being based on the same e5500 core as the X5000, if this comment is similar to the topic that Kas1e and Hans recently discussed here
then that's not so impressive. As development proceeds I would be curious if anybody (maybe only Geennaam himself?) could venture a guess as to the nature of the X5000's DMA issues. Because A-Eon and Hyperion aren't talking.
4) In the short term, the Mirari board just might spur some X5000 development, like a UBoot update and lower prices. Competition is good. :)
5) I don't think Geennaam would have started this project without some plans to mass produce the boards. That is, I don't think he would have the resources to fund the manufacturing himself. Which (German?) company might partner with him?
6) Anyone want to guess what year it will go on sale?
@tao I'm not sure this is the right forum to discuss this board but maybe I can guess some of the answers. I have no information about it at all, just read the news and looked at the SoC docs for curiosity so these are just my guesses and what I understood which may be wrong.
Quote:
1) How much memory does it address? I don't see that in the writeup.
I think the SoC is 64bit and can use up to 64GB RAM or so but it does not matter much for AmigaOS which can only use 2GB and the rest only as RAM disk maybe.
Quote:
2) My only grumble is a preference for more expansion:
The T2081 does not have SATA interfaces. If you need SATA, it must be added with a PCIe card. However, keep in mind that SATA is being phased out, with modern PCs and laptops often using NVMe over SATA SSDs.
I think the board was designed for T2081 as an option but only available so far with T1042 which has SATA but one of the two SATA or NVMe ports share the same connection so you always have 1 SATA 1 NVMe and a third port either SATA or NVMe if I got that correctly.
Quote:
I wonder how hard it would be to support an NVMe to SATA adapter board like
Matter of drivers, should work under Linux and it's unclear if AmigaOS will be available on this board but then somebody would need to write a driver for such SATA card to make it work.
Quote:
With an NVMe adapter and networking and sound on the motherboard, there may not be much need for extra cards, but do PCIe splitters often/sometimes work without firmware or drivers? Any such remaining cards might be satisfied with one lane each, and a PCIe 4x splitter like this
Same as above, needs driver and also kernel support. I think this was addressed in the (yet unreleased) pegasos2 kernel so it may be possible to write bridge driver, question is if they are willing to port AmigaOS to this board. Things that are plug and play on Linux are not necessarily work on other OSes. They work on Linux as it has drivers for a lot of hardware.
Quote:
Although we have only one memory controller, initial Linux benchmarks show that the single and SMT memory performance is at least on par with the X5020.
The T1042 being based on the same e5500 core as the X5000, if this comment is similar to the topic that Kas1e and Hans recently discussed here
Considering that X5000 has the fastest PCIe for NG Amigas currently that's not bad if this board can do the same just be more affordable.
Quote:
As development proceeds I would be curious if anybody (maybe only Geennaam himself?) could venture a guess as to the nature of the X5000's DMA issues. Because A-Eon and Hyperion aren't talking.
What does Geennaam have to do with X5000? It's A-Eon's board so they have to find out.
Quote:
5) I don't think Geennaam would have started this project without some plans to mass produce the boards. That is, I don't think he would have the resources to fund the manufacturing himself. Which (German?) company might partner with him?
Why German? You can order PCBs from anywhere, assembling is more difficult and may cost more. AFAIU it isn't Geennaam's project but Skateman's primarily with Geennaam helping with design and development but he does not have to care about how it will be produced or what will happen afterwards.
Quote:
6) Anyone want to guess what year it will go on sale?
So far they said it's a hobby project in prototype phase but all the first prototypes worked which is great result so if no major problems found, the board itself could be made available soon but the question is more when and if AmigaOS would run on it (assuming you want to run AmigaOS on it that's why you brought it up on this forum). Technically it could be easy to port as it already runs on similar SoCs so that means it could be available even in 1-2 years but there may be political issues which is in the hands of the people who own and fight over parts of AmigaOS so it depends on them and not on the board designers and they could not yet even agree on who owns what so AmigaOS support is quite unclear at this point.
I truly hope this gets to market BUT at the rate we're getting updates on existing hardware for OS4.1 I can only imagine how long it will take for OS4.1 to run 'properly' on this new board
Great news and congratulations to everyone involved! I hope the project becomes a reality within the next 12 months. If OS4/MOS is supported, I'm sure I'll be among the first adopters.
Anyway, I still believe that QEMU virtual boards are the best option to ensure Amiga software survives in the future. They will never break, and year after year they will become faster and more accurate clones of the original hardware. Thanks to Zoltan, a future version of QEMU will also include support for a virtual Pegasos I board.
With an NVMe adapter and networking and sound on the motherboard, there may not be much need for extra cards, but do PCIe splitters often/sometimes work without firmware or drivers? Any such remaining cards might be satisfied with one lane each, and a PCIe 4x splitter like this
Same as above, needs driver and also kernel support. I think this was addressed in the (yet unreleased) pegasos2 kernel so it may be possible to write bridge driver, question is if they are willing to port AmigaOS to this board. Things that are plug and play on Linux are not necessarily work on other OSes. They work on Linux as it has drivers for a lot of hardware.
Maybe the manufacturer is expecting a driver module in the Linux kernel? Mmm, hadn't considered that.
Quote:
Quote:
As development proceeds I would be curious if anybody (maybe only Geennaam himself?) could venture a guess as to the nature of the X5000's DMA issues. Because A-Eon and Hyperion aren't talking.
What does Geennaam have to do with X5000? It's A-Eon's board so they have to find out.
I figured the architecture is similar enough that they might eventually stumble on the answer. That is, if they perform Linux DMA tests comparing the Merari with the X5000 and find both are similar, then maybe it's a limitation of the SoC ?
Quote:
AFAIU it isn't Geennaam's project but Skateman's primarily with Geennaam helping with design and development but he does not have to care about how it will be produced or what will happen afterwards.
Oh, I hadn't realized that! Then kudos to them both for undertaking this project!
Maybe the manufacturer is expecting a driver module in the Linux kernel? Mmm, hadn't considered that.
Geennaam has written an NVMe device driver for AmigaOS, and also an audio driver (for the HDAUDIO codec). So he might actually write the needed drivers himself.
Quote:
I figured the architecture is similar enough that they might eventually stumble on the answer. That is, if they perform Linux DMA tests comparing the Merari with the X5000 and find both are similar, then maybe it's a limitation of the SoC ?
There's a decent chance that they're similar, and it would be awesome if someone stumbled on the answer regarding DMA performance.
When the A1-XE was first released, everyone thought that the audio interface was broken, and the audio codec chip and related components were removed from subsequent batches. A few years later someone discovered that the audio interface was just fine; the power saving subsystem just had it powered down by default.
One challenge we've had, is that PPC Linux has no working drivers for RadeonHD/RX drivers. At least on big-endian PPC. The drivers should work on little-endian PowerPC Linux. If we could run the same graphics cards under Linux on the X5000, then we could get some idea of what performance would be possible.
Quote:
Oh, I hadn't realized that! Then kudos to them both for undertaking this project!
Incidentally, both Geennaam and Skateman are from the Netherlands. Geennaam means "no name."
...do PCIe splitters often/sometimes work without firmware or drivers? Any such remaining cards might be satisfied with one lane each, and a PCIe 4x splitter like this
This is the pure hardware solution and it transparent to OS.
I don't believe that. I don't know much about it but AFAIK M2 slot can have PCIe or SATA. If it's SATA it can only attach one device not 6. If it's PCIe then there's a SATA controller chip on the adapter which will appear as a PCIe device to the machine that will need a driver. Maybe the chip used is compatible with something that both Windows and Linux has a driver for so they say it works for all OS (they care for). But if it's not a SATA controller AmigaOS has driver for then it won't work without such driver. Another possibility if it somehow emulates NVMe but I'd expect that to not have multiple ports. 6 ports is usual for SATA controller chips. If you can read what's written on the chip you could search for a data sheet to find out what it is and what it's compatible with and see if there's a driver for it for the OS you want to use. But it may be too early as there is even no OS yet or we don't even know if this will be sold or just made for themselves to play with.
Edit: Oh, you mean the PCIe1x to 4 slots splitter not the M.2 to SATA adapter? That one may also need a bridge that the OS has to recognise but Hans improved this for the pegasos2 kernel to get GPUs work so that's more likely to work.
This one might even have AltiVec, unless T2080FS is not the same as T2080.
Max clock speed is 1.8Ghz just like PA-SEMI (X1000), I don’t know if it can be overclocked, but its running at 1.5Ghz now, it has DDR3 support, X1000 only support DDR2.
(NutsAboutAmiga)
Basilisk II for AmigaOS4 AmigaInputAnywhere Excalibur and other tools and apps.
T1042: 4 cores e5500 ( same as X5000 has ), 1.5 GHz - this CPU is on boards presented on project website e5500 core has 3 DMIPS/MHz
low-end version: T1022: 2 cores e5500, 1.4 GHz
powerfull version: T2081: 4 cores e6500 ( AltiVec, two threads per core ), 1.8 GHz e6500 core has 6 DMIPS/MHz ( i.e. 3 DMIPS/MHz per thread ) in dual-threaded mode, 3.4 DMIPS/MHz in single-thread mode T2081 CPU has similar design like T2080 ( CPU from powepc notebook project ), has reduced number of PCIe lines to fit 780-BGA. So T2081 is pin compatible with T1042
For quick and very raw speed comparison (DMIPS/MHz): G5: 2.9 P1022: 2.4 G3, G4: 2.3 PA-Semi 1682M: 2.2 PPC440 core: 2.0
So it looks very interesting and cross fingers!
AmigaOS3: Amiga 1200 AmigaOS4: Micro A1-C, AmigaOne XE, Pegasos II, Sam440ep, Sam440ep-flex, AmigaOne X1000 MorphOS: Efika 5200b, Pegasos I, Pegasos II, Powerbook, Mac Mini, iMac, Powermac Quad
T2080 is three times the price of a T1042 I think.
Yes, these have a normal FPU. Only e500 doesn't have it (up to a certain revision, not sure why they just didn't upgrade to the latest e500 revision for the A1222).
And FPU is optional on some of the PPC400 series.
Looks cool, but a bit late to make any impact. However as being part of a hobby I fully support this
Brainstorming about PCIE above, I recently realized I'd misunderstood the slot descriptions. I was taking
Quote:
PCIe x4 slot with one lane occupied (T10x2 PCIe 2.0; T2081 PCIe 3.0)
as meaning the x4 slot had one lane used ("occupied") for something else, so 4-1 = 3 available lanes. Oops. So the PCIE splitter I listed would never work because it relies on separating the individual 1x lanes. Assuming the term "Serdes" is synonymous with "PCIE lane", the T1042 and T2081, and the other processors in this class, are all a bit starved for support. Only eight lanes each to work with. But I guess that's the definition of "compromise".
I also don't see anything in the specs that mention USB 3.2. Everything is USB 2...and I did browse a lengthy data sheet. What am I missing?
@tao In any case is better to wait for official specs. Still, I will venture a speculation
from amiga-news.de:
PCIe x16 slot with 4 lanes occupied (PCIe 2.0)
PCIe x4 slot with one lane occupied (T10x2 PCIe 2.0; T2081 PCIe 3.0)
PCIe x1 slot with one lane occupied (PCIe 2.0)
In my opinion, it means: PCIe x16 mechanical connector with 4 connected lanes PCIe x4 mechanical connector with one connected lane PCIe x1 mechanical connector with one connected lane
x4 slot has T10x2 PCIe 2.0; T2081 PCIe 3.0, becouse T2081 has 2xPCIe 2.0 and 2x PCIe 3.0 controllers, T10x2 has 4x PCIe 2.0 controllers. One of T2081 PCIe 3.0 one is used for PCIe x1 connector. So there is possible another one PCIe 3.0 device, probably one of NVMEs.
Next, all CPUs has 8 serdes lines, so all NVMEs, SSDs, PCIe lines, USB3.0 and NIC is connected to eight lines. Therefore there is 4line PCIe-switch in board design ( exact type is visible on photos )
But again, let's wait for Skateman and Geennaam. And let's wish them the best of luck and strength
AmigaOS3: Amiga 1200 AmigaOS4: Micro A1-C, AmigaOne XE, Pegasos II, Sam440ep, Sam440ep-flex, AmigaOne X1000 MorphOS: Efika 5200b, Pegasos I, Pegasos II, Powerbook, Mac Mini, iMac, Powermac Quad
It's nice that there's new PPC hardware, but I'm sure it'll be the same as always with these, everyone will be surprised that it was supposed to be low-cost hardware, but the price tag is well over 1000€ for just the motherboard, if that's enough, I'll keep my fingers crossed that it won't be.
News feed
OS4 Feedback forum
OS4Depot Feedback forum
AmiCygnix forum
ABC shell forum
AmiKit forum
Cinnamon Writer forum
CodeBench forum
Digital Universe forum
Dopus 5 forum
E-UAE forum
Gnash forum
Ibrowse forum
JAmiga forum
Odyssey forum
OWB forum
Qt forum
SmartFileSystem forum
Timberwolf forum
TouchDevice forum
TuneNet forum
Unsatisfactory Software forum
Donate
