@matthey

Quote:

matthey wrote: Define inexpensive in the PPC AmigaOS 4 market.

Since PowerPC (Power ISA) doesn't have a sub-$500 SBC market, 'inexpensive' is a CPU + Board SKU in the $500 - $1,000 bracket.

The BoM on this board certainly fits in that category. What remains to bee seen is how much overhead + markup will be applied.

Last edited by agami on 06-May-2025 at 03:59 AM.
Last edited by agami on 06-May-2025 at 03:59 AM.

_________________
All the way, with 68k

@matthey

Quote:

matthey wrote: agami Quote: Be grateful if either of these is ever ported to this board. This board is quite a bit "all dressed up and nowhere to go". It's being brought up at a time when Hyperion don't have the capacity to even update existing releases of AmigaOS 4, and the MorphOS team is mostly working with a new ISA. At best this is an inexpensive PPC Linux board for that niche market. Define inexpensive in the PPC AmigaOS 4 market. Yea, this board a decade earlier likely would have attracted some PPC AmigaOS 4 users. Now, it is likely too little too late to change anything. Nice effort but the old silicon now has to compete against RPi hardware. I will do some comparisons starting with a CPU performance comparison. CPU core | fab tech | clock | DMIPS/MHz/thread | DMIPS/thread 68000 3500nm 8MHz 0.16 1.3 68060 500nm 50MHz 1.80 90 PPC601 600nm 66MHz 1.41 93 PPC750 260nm 233MHz 2.25 525 (G3) QorIQ-P5020/e5500 45nm 2000MHz 3.0 6000 (X5000: 64-bit, 2 cores) QorIQ-P1022/e500v2 45nm 1200MHz 2.4 2880 (A1222: 32-bit, 2 cores. no standard FPU) QorIQ-T1042/e5500 28nm 1500MHz 3.0 4500 (Mirari T1042: 64-bit, 4 cores) QorIQ-T2081/e6500 28nm 1800MHz 3.0 5400 (Mirari T2081: 64-bit, 4 cores, 2 threads/core, Altivec) ARM1176JZF-S 40nm 700MHz 1.25 875 (RPi) Cortex-A7 40nm 900MHz 1.90 1710 (RPi 2) Cortex-A53 40nm 1200MHz 2.30 2760 (RPi 3) Cortex-A72 28nm 1800MHz 5.45 9810 (RPi 4) Cortex-A76 16nm 2400MHz 6.60 15840 (RPi 5) When the QorIQ embedded communication chip series switched from PPC to ARM, the ARM cores were Cortex-A7, Cortex-A9, Cortex-A15, Cortex-A53 and Cortex-A72. The Cortex-A72 had more performance at the high end using a similar process where PPC cores had previously had better performance than ARM cores. ARM licensed, copied and improved SuperH (copied from the 68000) for the Thumb ISA with the primary intention of improving code density for the embedded market and ARM looks like it copied and improved the PPC ISA for ARM64/AArch64 with the primary goal to improve performance. It appears they were successful again. The Mirari PPC board has easier expansion than RPi boards. PCIe and NVMe are on the board where only the RPi 5 supports PCIe 2.0 x1 and hats (small boards) are required for it and NVMe drives. The Mirari has PCIe 2.0 x4 in a x16 slot for a graphics card and two other PCIe slots. A more modern PC with PCIe 5.0 x16 with all lanes would have 63 GB/s of bandwidth compared to the Mirari 2 GB/s of bandwidth and RPi 5 0.5 GB/s of bandwidth. The RPi 5 has integrated graphics which have improved but not as much as the CPU performance. Memory performance is important in both CPU and GPU performance. device | memory X5000 DDR3 A1222 DDR3 Mirari DDR3L (L is for low power but not the same as LPDDR3) RPi1 DDR RPi2 DDR2 RPi3 LPDDR2 RPi4 LPDDR4 RPi5 LPDDR4X In general, the newer the generation given with the number the more memory bandwidth but there are many factors. The DDR3L choice to support multiple SoCs could make the memory more expensive and more difficult to source. Modern SoCs often use DDR5 and LPDDR5 memory. One nice feature of the Mirari board is a sizable Trion FPGA. It is intended for audio and glue logic but can likely be used for other purposes too. There have already been questions of whether a 68k Amiga can be simulated or even parts like the CIA timers to improve 68k Amiga compatibility. With enough support, early PPC software for Phase 5 boards may work on it. If audio and with CIA timers, Paula audio could be provided but that was replaced instead of enhanced. Throwing away the 68k and chipset likely loses 95% or more of the Amiga market and a large percentage of Amiga software, including popular retro games. Never look back and fix mistakes that could make hardware more than a micro niche market with micro production for the classes instead of the masses. In conclusion, these are just basic factors and there are plenty I have not covered. The Mirari board looks like a more sensible design than the X5000 and A1222 designs. The PPC SoCs supported make more sense too although it may have been better to support just one SoC better. For example, only supporting the T2080 likely would have been better for the high end and I doubt the price will be low enough for a low end SBC like RPi SBCs with any PPC SoC even though it is good to try to keep the price down. The PPC market is too small for mass production so even a cheap populated board is unlikely to be below 1000 €/$ but may be low enough to kill the A1222 and X5000. The RGL THEA1200 was attacked as it was too much of a threat to the A-EonKit Hyperion alliance. This may be a larger threat financially. The easiest way to reduce sales of the Mirari board is to restrict or delay the AmigaOS for it and threaten business partners as was done with the RGL THEA1200. This will likely backfire in the long run as AmigaOS 4 users convert to MorphOS and AROS and boycott A-EonKit Hyperion alliance products. One thing that can be expected from the alliance is more shenanigans than professionalism.

While reading your posts over the last months, I came to realise your knowledge in all areas is really impressive! The logic and focus and foresight is even sharper than that of the board designers.
I wish I had you brains!
But then maybe not; I would grow even more frustrated. For by having all that knowledge, wit and confidence but then not being able to actually contribute with tangible products would make me sick.

@thread

More comments on Morphzone

and

More commens on OS4welt

#6

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*

@number6

Looks like there is a lot of positive commentary on Morphzone.

_________________

@AmigaMac

Let's be honest : most (if not all) the negative commentaries can be found here, on AW. It's the essence of this site nowaday (thanks to a few folks who post negative comments about nearly anything).

_________________
PowerMac G5 2,7Ghz - 2GB - Radeon 9650 - MorphOS 3.14
AmigaONE X1000, 2GB, Sapphire Radeon HD 7700
FPGA Replay + DB 68060 at 85Mhz

@K-L

I guess there's nothing bad to talk about. We should all sing lullabies because somehow things will work out differently. The world is watching out on us, it's quite a responsibility on our shoulders...

Shrug

Pa- thanks for your enlightenment

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga

#6 Quote:

More comments on Morphzone

I expect sending a working Mirari board to the MorphOS team would result in timely support which I do not expect from Hyperion for AmigaOS 4. The MorphOS thread mentions an article about hardware with the T2080 SoC with e6500 cores.

NXP dev board T2080 e6500 with Debian sid PPC64 and RadeonHD
https://www.powerpc-notebook.org/2018/07/nxp-dev-board-t2080-e6500-with-debian-sid-ppc64-and-radeonhd/

This is the T2080 I mentioned in my previous post as possibly being better to target. The SoC has more pins by dropping pin compatibility with earlier QorIQ SoCs which could no longer be used. It looks like the T2080 gains more SerDes lanes and may allow the following benefits.

Potential advantages of QorIQ-T2080 SoC board over QorIQ-T2081 SoC board
o PCIe x16 slot with PCIe 3.0 x4 for 3.938 GB/s bandwidth instead of 2.000 GB/s
o working dual SATA
o option to use higher clocked and easier to find DDR3 memory instead of DDR3L memory

The T2080 SoC and accommodating board likely would have been somewhat more expensive and PPC e6500 cores (T208x) may be less compatible than PPC e5500 cores (T1042) already used in the X5000 SoC. Even though the PPC e5500 cores are the same which should make basic support of the CPU easier, other SoC hardware is not so standardized as is common for embedded hardware and some new drivers may be necessary. From the article, the T2080 appears to be a nice upgrade from a PPC G5 not just because of performance but the much lower power used (less heat and fan noise). I believe the SoC uses more power than a RPi4 but not as much as a RPi5 (RPis include integrated GPUs where a discreet GPU could easily push power above a RPi5). The PPC e6500 cores have a clear advantage over the PPC e5500 cores with SMP as there are two threads per core and significant hardware used to provide good multi-threading performance. Without SMP as is the case for AmigaOS 4 and MorphOS, the value is missing.

The max clock speed actually decreased when moving from a 45nm to 28nm chip fab process for newer QorIQ SoCs with PPC cores. Maybe the architects were targeting low power and parallel performance at the expense of single thread performance or maybe the decision to switch to ARM caused development to end early before path delay analysis and pipeline optimizations were complete. This is likely what happened to the 68060 when Motorola decided to switch to PPC as an 8-stage CPU should have clocked to at least 150MHz instead of 50MHz using a 500nm process. At least a newer process usually provides some power reduction even if not fully optimized for the newer process.

#6 Quote:

More comments on OS4welt

There is not much confidence in Hyperion on this German forum except for Hyperion to go bankrupt. There is a new official web page given for the board but then it was taken down for now by Dave "Skateman" Koelman (developed by Harald 'Geennaam' Kanning).

https://mirari.vitasys.nl/

If it was up to Hyperion's MagicSN/Steffen, AmigaOS 4 support would happen and he would have a larger AmigaOS 4 user base to almost justify his AmigaOS 4 game ports. The only people at Hyperion with an understanding and urgency of the need to increase the Amiga user base to survive seem to be Steffen and the 68k AmigaOS developers. If it was up to them, I doubt THEA1200 from RGL would be sabotaged and blocked as it would likely provide more game sales and sooner than the Mirari PPC board. I hope it is not too messy when Hyperion implodes from self inflicted sabotage.

Last edited by matthey on 03-May-2025 at 09:39 PM.
Last edited by matthey on 03-May-2025 at 08:24 PM.

#6 Quote:

Images, description within: Amiganews German Amiganews English

The following new Amiganews German comment is interesting.

Joerg (03-Mai-2025, 19:01) Quote:

@geenaamm > As you could read on os4welt.de, in the thread that was started by you, > Hyperion is eager to support the new board. > But they do not fully own amigaos. I fail to understand why people, especially ones who should know better, still think Hyperion owns ANYTHNG of AmigaOS 4.x. The base source of some small parts, the few usable AmigaOS 0.9-3.1/m68k C sources which could be used for AmigaOS 4.x/PPC (most of it was completely useless m68k assembler code instead and had to be reimplemented from scratch for AmigaOS 4.x, not only Exec(SG), but other large and important parts like dos.library, graphics.library/P96, etc., as well. If at all only the comments in the m68k assembler sources were of any help for that) is owned by Cloanto/Amiga Corp., and for the AmigaOS 3.x/m68k versions additionally by iComp (3.x/m68k versions of P96). For releasing new versions of AmigaOS 4.x/PPC, and especially AmigaOS 3.x/m68k ones, Hyperion's insolvency manager first would have to get a licence from Cloanto/Amiga Corp. ExecSG is, and always was, owned by the Frieden brothers. Trevor/A-EON may have an exclusive licence of it now, but even that doesn't change the ownership. In any case Hyperion can't use it anymore, unless they get a sub-licence from Trevor/A-EON first. Copyright isn't transferable, at least not in Germany and probably not in most other parts of the EU either, even something like "Public Domain" software isn't legally possible here. The ONLY exception is if software is developed in their payed hours by employees of a company, but Hyperion never had any employees doing software development for them (maybe for their games, but definitely not for AmigaOS), only contract workers/companies who still own all of their respective parts of AmigaOS 4.x, and in most cases Hyperion's former licences to use such parts were revoked 10-25 years ago already because of breach of contract. Without getting new licences from the about 3 dozens of AmigaOS 4.x owners Hyperion's insolvency manager can't release something like an AmigaOS 4.2 for current systems, let alone adding support for new hardware.

"Hyperion's insolvency manager"?

@matthey

The Mirari website only functioned on a regional basis at the time.

np. It's comiing back for all after some tidying up. (publicaly stated in discord)

#6

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*

@matthey

I'll need some time catching up with other postings. Been busy all day proving google's experimental AI is just that "experimental" heh

Interesting post from Joerg. The only part I question is the same term that you question.
Not sure whether he knows something new, or has misunderstood something, or is just hoping for a reply disputing the term.

The balance of what he posted...I'll let others better versed in the topic comment if they so chose.

*shrug*

#6

Last edited by number6 on 03-May-2025 at 09:40 PM.
Last edited by number6 on 03-May-2025 at 08:55 PM.

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*

@DWolfman

Quote:

DWolfman wrote: @number6 Definitely sounds promising. If the price works out right, I may look to getting one when OS 4 is available for it. Since the network chip on my old AmigaOne G4 board got fried, I haven't had a working physical Amiga of any kind. Stuck running WinUAE on my AMD Ryzen 9 5900XT system. It's fast enough, but only when I run AmigaOS 3.x. PPC emulation on it is not as fast as the real AmigaOne was with a 1 GHz G4 CPU.

For WinUAE's QEMU PPC beating AmigaOne G4 @ 1Ghz, it needs to be Intel AlderLake/RaptorLake or AMD Zen 4/Zen 5 level.

I don't have access to Intel ArrowLake.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@matthey

The main claim to fame for RPi 5 is affordable price, but it's nearly crossing into PC's entry-level AM4 or LGA 1700 platforms.

This topic's PPC solution is reasonable. There's no bad hardware, just bad pricing.

https://www.jeffgeerling.com/blog/2023/forcing-pci-express-gen-30-speeds-on-pi-5
RPi 5 can support PCIe Gen3. The problem is with the quality of RPI HATS, hence PCIe Gen2 default setting. This web link NVMe PCIe Gen3 example reached 900 MB/s.

BCM2712 SoC has six PCIe Gen3 lanes in total.

At least this topic's PPC solution has a standard PPC FPU.

---------------------------

A PCIe 5.0 x16 slot offers up to 128 GB/s of data transfer since it's full duplex i.e. 64 GB/s per direction. PC games are mostly one-directional, while GpGPU with host CPU can be two-directional.

AM5 socket processors offer 28 PCIe lanes e.g. my MSI B850 Edge Ti WiFi motherboard has PCIe Gen5 16X and two NVMe slots at Gen5 4X (via the SoC). Fast NVMe helps with 4K video NLE. AMD / ASMedia hasn't released a PCIe Gen5 southbridge.

Last edited by Hammer on 04-May-2025 at 12:03 AM.
Last edited by Hammer on 04-May-2025 at 12:00 AM.
Last edited by Hammer on 03-May-2025 at 11:49 PM.
Last edited by Hammer on 03-May-2025 at 11:45 PM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

#6 Quote:

Interesting post from Joerg. The only part I question is the same term that you question. Not sure whether he knows something new, or has misunderstood something, or is just hoping for a reply disputing the term.

Joerg said in his Amiganews German Comment, "I fail to understand why people, especially ones who should know better, still think Hyperion owns ANYTHNG of AmigaOS 4.x."

Perhaps the 2009 Amiga Inc v Hyperion VOF settlement agreement is the source of the misunderstanding of AmigaOS 4 ownership?

2009 Amiga Inc v Hyperion VOF Quote:

1. Grant. (a) Hyperion acknowledges that the Amiga Parties are the owners of the Software, without prejudice to any third parties with rights in said Software. The Amiga Parties acknowledge that Hyperion is the sole owner of AmigaOS 4 (with the exception of the Software), without prejudice to any third parties with rights in said software.

It would seem that Amiga Parties were coerced into acknowledging something else that was not true. How much of the contract has to be fantasy before it becomes fraud?

#6 Quote:

The balance of what he posted...I'll let others better versed in the topic comment if they so chose.

Geennaam commented to Joerg's comment. I'll post together.

Joerg (03-Mai-2025, 19:01) Quote:

@geenaamm > As you could read on os4welt.de, in the thread that was started by you, > Hyperion is eager to support the new board. > But they do not fully own amigaos. I fail to understand why people, especially ones who should know better, still think Hyperion owns ANYTHNG of AmigaOS 4.x. The base source of some small parts, the few usable AmigaOS 0.9-3.1/m68k C sources which could be used for AmigaOS 4.x/PPC (most of it was completely useless m68k assembler code instead and had to be reimplemented from scratch for AmigaOS 4.x, not only Exec(SG), but other large and important parts like dos.library, graphics.library/P96, etc., as well. If at all only the comments in the m68k assembler sources were of any help for that) is owned by Cloanto/Amiga Corp., and for the AmigaOS 3.x/m68k versions additionally by iComp (3.x/m68k versions of P96). For releasing new versions of AmigaOS 4.x/PPC, and especially AmigaOS 3.x/m68k ones, Hyperion's insolvency manager first would have to get a licence from Cloanto/Amiga Corp. ExecSG is, and always was, owned by the Frieden brothers. Trevor/A-EON may have an exclusive licence of it now, but even that doesn't change the ownership. In any case Hyperion can't use it anymore, unless they get a sub-licence from Trevor/A-EON first. Copyright isn't transferable, at least not in Germany and probably not in most other parts of the EU either, even something like "Public Domain" software isn't legally possible here. The ONLY exception is if software is developed in their payed hours by employees of a company, but Hyperion never had any employees doing software development for them (maybe for their games, but definitely not for AmigaOS), only contract workers/companies who still own all of their respective parts of AmigaOS 4.x, and in most cases Hyperion's former licences to use such parts were revoked 10-25 years ago already because of breach of contract. Without getting new licences from the about 3 dozens of AmigaOS 4.x owners Hyperion's insolvency manager can't release something like an AmigaOS 4.2 for current systems, let alone adding support for new hardware.

Geennaam (03-Mai-2025, 22:08) Quote:

@Joerg I fully understood what you are saying. After all, my licenses were not respected either. The issue here is that nobody knows which components are affected except those directly involved. I know that some parts are owned by you, as you have stated on numerous occasions. That, along with ExecSG, is what I was referring to when I said "...not fully own...". To make matters worse, there is a lot of contradictory information floating around, possibly further distorted by a game of Chinese whispers. For example, I was always led to believe that the kernel was bought and is now owned by Trevor Dickinson, and that Thomas Frieden is acting as a consultant since, as the original coder, he possesses the most knowledge to assist the ExecSG team. But now you say that Trevor Dickinson merely holds an exclusive license. Similarly with P96, you claim that it is co-owned by Cloanto and iComp. However, the iComp page states: "A pre-existing license between the P96 authors Tobias Abt & Alexander Kneer and Hyperion has been turned into a co-ownership of P96: Hyperion owns and maintains the PPC port of it, which is part of AmigaOS4 and its successors. This means that P96 in Hyperion's PPC-based OS does not require separate licensing." This is why I want to stay away from politics surrounding AmigaOS4. Unfortunately, the announcement of this mainboard forced me to post on websites and Discord again. Still, I can say that I thoroughly enjoyed the past year when I had no interaction with anything Amiga-related except for the Mirari team.

Is that Hyperion not respecting licenses again?

Hammer Quote:

The main claim to fame for RPi 5 is affordable price, but it's nearly crossing into PC's entry-level AM4 or LGA 1700 platforms. This topic's PPC solution is reasonable. There's no bad hardware, just bad pricing.

Mostly true. The RPi 5 provides good CPU performance/$ but the integrated GPU lags behind, the PCIe discreet GPU support is very weak and far from plug and play, NVMe on a hat connections are a weak point and inconvenient and the SoC runs hot raising the final price for cooling. Low end PC hardware is competitive in value at a little higher price. The Mirari board may have been more convenient than the RPi 5 board if the T2080 SoC had been used with PCIe 3.0 x4 in a x16 slot for GPU, SATA worked and DDR3 memory was supported. Requiring a discreet GPU and requiring a larger case are less convenient and drive up the price but the larger board is worth the cost for convenience and without hats. The only way to get the price down into RPi territory is with an integrated GPU. SiFive with RISC-V was not having much success with more expensive boards using discreet GPUs until integrating a GPU and dropping the price below $100 USD with the StarFive VisionFive 2. The Mirari board has the same problem as early SiFive boards with too weak of CPU performance for the high price. The weak CPU performance would be fine with a good integrated GPU as the price can be lower. The Mirari board also suffers the same problem as RISC-V boards which is too small of market as the much larger 68k Amiga market was left behind.

Hammer Quote:

https://www.jeffgeerling.com/blog/2023/forcing-pci-express-gen-30-speeds-on-pi-5 RPi 5 can support PCIe Gen3. The problem is with the quality of RPI HATS, hence PCIe Gen2 default setting. This web link NVMe PCIe Gen3 example reached 900 MB/s. BCM2712 SoC has six PCIe Gen3 lanes in total.

If I understand correctly, the T2081 also supports PCIe 3.0/Gen3 but only a single lane in the PCIe x4 connector if my understanding is correct. It looks like the T2080 could have supported PCIe 3.0 x4 for the x16 GPU slot according to the article I linked earlier. Yea, a hat connection or any kind of connection for very high speed NVMe or PCIe SerDes lines/lanes is not a good idea.

Hammer Quote:

At least this topic's PPC solution has a standard PPC FPU.

The A1222 is right up there with the A600 mismanagement. Instead of both projects being cancelled due to high cost, they were both doubled down on. The Mirari board threatens to make Trevor eat his A1222 boards unless the A-EonKit Hyperion Alliance can sabotage the Mirari board like they did THEA1200 from RGL.

Hammer Quote:

A PCIe 5.0 x16 slot offers up to 128 GB/s of data transfer since it's full duplex i.e. 64 GB/s per direction. PC games are mostly one-directional, while GpGPU with host CPU can be two-directional. AM5 socket processors offer 28 PCIe lanes e.g. my MSI B850 Edge Ti WiFi motherboard has PCIe Gen5 16X and two NVMe slots at Gen5 4X (via the SoC). Fast NVMe helps with 4K video NLE. AMD / ASMedia hasn't released a PCIe Gen5 southbridge.

SerDes lanes are used for the highest speed SoC I/O but they are also expensive. PCIe for a discreet GPU uses too many lanes for a low cost SoC/SBC. Integrated GPUs can have more performance at a lower cost without wasting SerDes lanes on the GPU. SerDes saved can be used elsewhere for more performance. Standard GPUs have an advantage. RPi, AMD, Apple, Sony and Nintendo understand the advantages but the PC and Amiga world are full of dinosaurs.

@matthey

Quote:

matthey wrote: Hammer Quote: The main claim to fame for RPi 5 is affordable price, but it's nearly crossing into PC's entry-level AM4 or LGA 1700 platforms. This topic's PPC solution is reasonable. There's no bad hardware, just bad pricing. Mostly true. The RPi 5 provides good CPU performance/$ but the integrated GPU lags behind, the PCIe discreet GPU support is very weak and far from plug and play, NVMe on a hat connections are a weak point and inconvenient and the SoC runs hot raising the final price for cooling. Low end PC hardware is competitive in value at a little higher price. The Mirari board may have been more convenient than the RPi 5 board if the T2080 SoC had been used with PCIe 3.0 x4 in a x16 slot for GPU, SATA worked and DDR3 memory was supported. Requiring a discreet GPU and requiring a larger case are less convenient and drive up the price but the larger board is worth the cost for convenience and without hats. The only way to get the price down into RPi territory is with an integrated GPU. SiFive with RISC-V was not having much success with more expensive boards using discreet GPUs until integrating a GPU and dropping the price below $100 USD with the StarFive VisionFive 2.

Around $100 means that it's the target of a Celeron-based MiniPC, which might also include a Windows license, and that is overall better (especially performance-wise).
Quote:

The Mirari board has the same problem as early SiFive boards with too weak of CPU performance for the high price. The weak CPU performance would be fine with a good integrated GPU as the price can be lower. The Mirari board also suffers the same problem as RISC-V boards which is too small of market as the much larger 68k Amiga market was left behind.

The new board also suffers from another, more important, detail: the price, which is missing. I haven't found even a price range (but I haven't read the latest comments)

If the price could be lowered by only supporting the e6500 cores and a single type of memory (DDR3), then that's the way to: e5500 and LDDR3 are meaningless for the purpose of this board / computer.
Quote:

Hammer Quote: https://www.jeffgeerling.com/blog/2023/forcing-pci-express-gen-30-speeds-on-pi-5 RPi 5 can support PCIe Gen3. The problem is with the quality of RPI HATS, hence PCIe Gen2 default setting. This web link NVMe PCIe Gen3 example reached 900 MB/s. BCM2712 SoC has six PCIe Gen3 lanes in total. If I understand correctly, the T2081 also supports PCIe 3.0/Gen3 but only a single lane in the PCIe x4 connector if my understanding is correct. It looks like the T2080 could have supported PCIe 3.0 x4 for the x16 GPU slot according to the article I linked earlier.

Yes, it uses less lanes for the PCIe. Evidently it hasn't so many available, so it's only mechanically compatible with the PCIes slots, but the bandwidth is much lower than the expected.
Quote:

SerDes lanes are used for the highest speed SoC I/O but they are also expensive. PCIe for a discreet GPU uses too many lanes for a low cost SoC/SBC. Integrated GPUs can have more performance at a lower cost without wasting SerDes lanes on the GPU. SerDes saved can be used elsewhere for more performance. Standard GPUs have an advantage. RPi, AMD, Apple, Sony and Nintendo understand the advantages but the PC and Amiga world are full of dinosaurs.

But RPi, AMD, Apple, Sony and Nintendo haven't standard GPUs: they have SOME GPUs, exactly like PCs. Because there's no "standard GPs".

Which is quite normal on PCs, essentially since the beginning. And those projects / companies are using PCs' technologies (SOME of them, like I've said).

@DWolfman

Quote:

DWolfman wrote: @number6 Definitely sounds promising. If the price works out right, I may look to getting one when OS 4 is available for it. Since the network chip on my old AmigaOne G4 board got fried, I haven't had a working physical Amiga of any kind. Stuck running WinUAE on my AMD Ryzen 9 5900XT system. It's fast enough, but only when I run AmigaOS 3.x. PPC emulation on it is not as fast as the real AmigaOne was with a 1 GHz G4 CPU.

The problem is NOT the hardware (a Ryzen 9 5900XT has A LOT of single core/thread performance, which is the only thing which matters when talking about Amiga / post-Amiga emulation), rather the software.

OS4/MorphOS emulation relays on WinUAE, which carries the entire chipset emulation.

But even just using QEmu, it's still not fast enough because PowerPC support isn't that mature. It requires more effort to improve.

Even better it would be to only emulate the user-land of OS4/MorphOS, transparently running their PowerPC applications on the host system. This will furtherly reduce the emulation overhead and squeezes the most from host system.
So, exactly like the "virtualizer" idea for the Amiga OS (68k, of course) which I've shared since several years.
BTW, Petunia & Trance will be replaced with native x64 versions in this case (there's no sense, at all, for an x64 to emulate a PowerPC which emulates a 68k itself: just shortcut and directly let it emulate the 68k!), furtherly and greatly improving the emulation speeds.

In short: efforts are required from the software side to get much better performance, because the hardware has already plenty of power (which isn't used that good).

@matthey

Quote:

How much of the contract has to be fantasy before it becomes fraud?

What’s the value of a settlement agreement when the signing parties cannot agree on what they have settled.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@matthey

I suggest that you take Joerg's comments with a grain of salt, because he:
- Is NOT privy to all agreements, nor is he a lawyer (e.g., my understanding is the Trevor outright owns ExecSG, and Hyperion's original agreement to use that code still stands)
- Parted ways with Hyperion on very bad terms, which taints his opinions. I don't know exactly what happened, but I'd say that the bridge between him and Hyperion was burnt to the ground. Burnt so badly that he did indeed revoke Hyperion's permission to use his code (and he had the right to do so)

If all AmigaOS 4 developers behaved the way that he did, then Hyperion would indeed be in a difficult position. However, other developers of AmigaOS 4 have NOT behaved in a similar fashion.

I can only speak of myself. When I did contract work for Hyperion (once, small contract), I made sure to get paid before handing over the source code. Any other changes I made to the code outside the contract, I donated to the code base (I see no value in claiming ownership over reordering lines of existing code or adding/tweaking small fragments of code here and there).

So, there was no way that we'd get into a fight over unpaid work. No rage quit. No claiming copyright over parts of the code-base and revoking license to use said code. Nothing to renegotiate licensing for.

Mind you, my contributions to the OS4 code base are rather small. I mostly just made tweaks/fixes to the graphics.library & Picasso96 so that modern GPUs could be used more efficiently.

Hans


P.S., It looks like Geennaam has discovered why the Friedens stopped posting or reading Amiga forums years ago...

Last edited by Hans on 04-May-2025 at 12:48 PM.

_________________
Join the Kea Campus - upgrade your skills; support my work; enjoy the Amiga corner.
https://keasigmadelta.com/ - see more of my work

@matthey

Quote:

Joerg said in his Amiganews German Comment, "I fail to understand why people, especially ones who should know better, still think Hyperion owns ANYTHNG of AmigaOS 4.x." Perhaps the 2009 Amiga Inc v Hyperion VOF settlement agreement is the source of the misunderstanding of AmigaOS 4 ownership?

This is most likely a combination of three things:

1. Joerg revoked Hyperion's license for his contributions to OS4 due to outstanding payments and removed his code from the central OS4 repository about 1.5 decades ago. The code was simply reinstated later on, after Ben Hermans claimed the problem had been solved (it hasn't, according to Joerg). Joerg knows he's not the only one who claims Hyperion is pirating his code.

2. Hyperion never "owned" much regarding OS4, since all of the code used to be licensed from external developers, who retained ownership. In most (all?) cases, these licenses contained quite a few restrictions (e.g. "binary only", "PPC only"). That changed later, when developers were pressured to sign NDAs containing clauses that assigned copyright to OS4 contributions to Hyperion - but by then development had slowed down considerably.

3. https://www.amiga-news.de/en/news/AN-2024-04-00037-EN.html

Quote:

Is that Hyperion not respecting licenses again?

IIRC, Geennaam's issue was with ACube, not with Hyperion.

Last edited by cgutjahr on 04-May-2025 at 01:26 PM.

@Hans

Quote:

P.S., It looks like Geennaam has discovered why the Friedens stopped posting or reading Amiga forums years ago...

I don't think using the Friedens as an example is a good idea. Let's put it this way: I've met people who could handle not being praised all the time better than those two.

Plus: You're a developer, and you seem to handle it just fine. Another clue that the problems were not the forums, but the Friedens?