> GCN doesn't start until the 7730. [...] there ARE reasons to support cards below
> the HD 7750.

Like the 7730? ;-)

> how much will we get out of GCN? I doubt it will be much.

Me either. That's why I never quoted performance as reasoning for my opinion that giving priority to GCN-based cards would be better.

> our drivers uses a surprisingly small number of graphic primitives (think about it, what
> executes on our highest end cards also works on a Radeon 9200 - or even lower cards).

I'm not sure this makes sense. Programs don't talk to the graphics card drivers directly. There's at least one level of abstraction in between, so I think the graphic primitives can be different between drivers for different GPU generations.

> a well written MorphOS driver for these cards will easily allow our G4 systems to
> step all over Tabor

...except in 3D operations.

> We just need to stop thinking in terms of OS4

*If* the MorphOS team wants OS4 users on Sam460 and X5000 to try out MorphOS, then they just need to *start* "thinking in terms of OS4" ;-)
Hans de Ruiter published the results of his latest OS4 graphics card survey some months back, which shows a clear trend towards GCN-based cards. Of all users of Radeon HD and newer cards (which should be Sam460, X1000 and X5000 users combined), about 75% use HD7xxx/Rx. Unfortunately, we don't have the data to be able to subtract out the X1000 users from this, and we also don't know how many of those HD7xxx cards are HD76xx and lower cards and thus not GCN-based.

>>there ARE reasons to support cards below
>>the HD 7750.

>Like the 7730?

Ah yes...you ARE good at picking up minor mistakes . For my part, I didn't even know there was an HD 7730 until yesterday, I've never seen one.

>> how much will we get out of GCN? I doubt it will be much.

>Me either. That's why I never quoted performance as reasoning for my opinion that giving priority to GCN-based >cards would be better.

Honestly, except for new system buyers, I don't see the logic in using GCN cards, there a lot of 5000 and 6000 series cards at bargain prices right now that will absolutely fly under an NG OS.

>> our drivers uses a surprisingly small number of graphic primitives (think about it, what
>> executes on our highest end cards also works on a Radeon 9200 - or even lower cards).

>I'm not sure this makes sense. Programs don't talk to the graphics card drivers directly. There's at least one >level of abstraction in between, so I think the graphic primitives can be different between drivers for >different GPU generations.

True, most graphics operations run through TinyGL, but our version of OpenGL is so dated that almost all our cards have the commands that have been implemented in this package.

>> a well written MorphOS driver for these cards will easily allow our G4 systems to
>> step all over Tabor

>...except in 3D operations.

Yeah, even if we get AGP R600 and R700 support, 5000-7000 (and later GCN) cards will have a distinct advantage, but I'm beginning to wonder how well Tabor will work with its graphics cards. The P1022 has only SIX SerDes lanes which must drive the NIC, SATA, AND PCI-E interfaces.
Something tells me operations to and from system memory are really going to be really slow.
So, given that we'll have 1.42 GHz G4 PowerMacs going up against a weaker 1.2 GHz P1022 (with its funky spe fpu), there's still a chance that Tabor won't fair well. Its a given that programs that aren't heavily graphically oriented will be much faster on the PowerMac. And graphics programs...we'll just have to see.

>>We just need to stop thinking in terms of OS4

>*If* the MorphOS team wants OS4 users on Sam460 and X5000 to try out MorphOS, then they just need to *start* >"thinking in terms of OS4"
>Hans de Ruiter published the results of his latest OS4 graphics card survey some months back, which shows a clear trend towards GCN->based cards. Of all users of Radeon HD and newer cards (which should be Sam460, X1000 and X5000 users combined), >about 75% use HD7xxx/Rx. Unfortunately, we don't have the data to be able to subtract out the X1000 users from >this, and we also don't know how many of those HD7xxx cards are HD76xx and lower cards and thus not GCN-based.[

Again, good point, I don't know what X1000 owners have upgraded to (but its unlikely they've kept their Radeon HD 4650s).
X5000 owners primarily have Radeon HD 7750 cards or R7 250 cards (both GCN).
I'm not sure either could best a Radeon HD 6870, but it is what it is.

That actually goes hand in hand with my argument in support of Tabor.
We need to provide support for the hardware they have, if we want them to try MorphOS.

Finally, as I've mentioned several times, GCN cards will eventually be supported. Mark has (I believe) about three of these.
And I'm not in any hurry, as I want to use an HD 5750 or HD 6750 and dual boot MorphOS with Linux.

Quote:

Very exactly!

Quote:

VERY EXACTLY! Eh?
And complete bullshit.

Mark's has worked VERY hard to complete these drivers and the HD 5000 were close to ready awhile ago.
Why do we need them?
First, these cards are cheap.
Second, the 5000 series is almost twice as capable as the R700/4000 series.

So the R600 and R700 drivers are almost ready, the 5000 and 6000 series drivers are close.
And after that, maybe he'll get around to the GCN cards so SAM460 and X5000 OS4 user can try MorphOS without having to swap their video cards.

Look, you're getting these new drivers and more.
Why not show a little gratitude for my friend and the BEST coder in the MorphOS development team.

You don't want to use superior video cards with these new drivers?
Then why don't you just revert to the Radeon 9200 that was our top supported card just a few years ago and STFU?

Finally, these drivers will be vital to the X64 fork, or were you planning on installing Radeon X1900XT cards in modern X64 hardware?
Sheesh, I don't think I've EVER seen a more pointless, negative, and just plain STUPID group of comment like the last two ever posted here.

[ Edited by Jim 24.09.2017 - 16:39 ]

Welcher Rechner ist denn performanter mit MOS; Der PowerMac G5 Quad oder A5000?

Quote:

Weil MorphOS Quad G5 nicht unterstützt, würde ich den X5000 sagen.

Quote:

MorphOS unterstützt den PCI-E G5, aber der Port ist noch nicht komplett und wir dürfen ihn nicht haben.

Probably totally fubared by translation, but the port lacks network drivers and fan speed control.
Otherwise, the G5 appears to be significantly better.

Außerdem unterstützt MorphOS den X5000 derzeit nicht, alles ist gleich, die G5 ist immer noch das bessere System.

[ Edited by Jim 25.09.2017 - 16:40 ]

Quote:

Well, current version of MorphOS doesn't support the X5000, but we all know that the next version (3.10) will do, as stated multiple times by MorphOS Team members. And we all (?) know that said version won't introduce support for PCIe G5's.

> Welcher Rechner ist denn performanter mit MOS; Der PowerMac G5 Quad oder A5000?

Falls mit "A5000" der X5000 gemeint ist:
Bezüglich Performance hat der X5000 eigentlich nur einen Vorteil: Integer-Operationen. In allen anderen Performance-Bereichen (FPU, SIMD, Speicherdurchsatz, PCIe) hat der X5000 das Nachsehen. Als sonstige Vorteile des X5000 fallen mir ein, dass er neu erhältlich ist, dass er kompatibler zu aktuellen Speichermedien ist (SATA2 vs. SATA1) und dass der Firmware-Prompt auch mit handelsüblichen Grafikkarten erscheint.

und der X5000 ist stromsparender ;)

> except for new system buyers, I don't see the logic in using GCN cards

So you do see the logic for new system buyers. Fine :-)

> there a lot of 5000 and 6000 series cards at bargain prices right now
> that will absolutely fly under an NG OS.

Again, my point is neither about price nor about performance. It's just about what's 3D-supported by OS4 and thus sold with new OS4-compatible systems that happen to be (or will be) supported by MorphOS as well.

>> Programs don't talk to the graphics card drivers directly. There's at
>> least one level of abstraction in between, so I think the graphic
>> primitives can be different between drivers for different GPU generations.

> True, most graphics operations run through TinyGL

I don't think they do. While Ambient's effects (Enhanced Display) do use 3D functions of the GPU, they do not require a TinyGL driver.

http://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=11275&forum=12&start=6

The abstraction layer I was referring to is the 2D one called CyberGraphX/Cybergraphics (they can't agree to one definite name it seems).

>>> a well written MorphOS driver for these cards will easily allow our G4 systems to
>>> step all over Tabor

>>...except in 3D operations.

> Yeah, even if we get AGP R600 and R700 support, 5000-7000 (and later GCN) cards
> will have a distinct advantage

Also here I was not talking about performance but about the existence of 3D support, which OS4 provides for GCN1 (Southern Islands) GPUs.

> I'm beginning to wonder how well Tabor will work with its graphics cards.
> The P1022 has only SIX SerDes lanes which must drive the NIC, SATA, AND
> PCI-E interfaces.

Yes, and when you think of it, PCIe x4 + 2x SATA + GbE would require 7 SerDes lanes, wouldn't it? See there for a more detailed elaboration of Tabor's P1022 SerDes assignment conundrum ;-)

>> Unfortunately, we don't have the data to be able to subtract out the
>> X1000 users from this

> I don't know what X1000 owners have upgraded to

It's irrelevant as MorphOS won't support the X1000 :-) The reason I'd want to subtract out the X1000 owners is to see what the Sam460 owners have (X5000 owners all use GCN-based cards, no doubt).

> I'm not sure either could best a Radeon HD 6870

To me it's not about "best", but about when the GPUs used in MorphOS-supported machines (that aren't currently used with MorphOS by their owner) will be supported by MorphOS. The sooner the better I'd say. And to a lesser degree it's about 3D support :-)

> We need to provide support for the hardware they have, if we want them to try MorphOS.

Exactly! :-) And this includes the GPUs, of course.

> GCN cards will eventually be supported.

Yes, I'm confident they will. I'm just not sure every generation before them must be supported ;-)

> the Radeon 9200 that was our top supported card just a few years ago

The Radeon 9200 was never the fastest R200-based card ;-)

> these drivers will be vital to the X64 fork

All the more incentive to skip some generations I'd say :-)

Quote:

Yes, actually if you want to split hairs, it's not even an R200 card.
Although I lump everything from the 8500 to the 9200 in that category/generation.
And that would leave you're best choices as the 8500 or 9100.

As to skipping earlier generations, I like having the option of using cards that came before the GCN cards.
They offer a good value for your money.
And they're easier to get working under Linux.
But, we should see GCN eventually, nothing is ever on a timetable, so we've got to be patient and wait at least 'two weeks'.

> actually if you want to split hairs, it's not even an R200 card. Although
> I lump everything from the 8500 to the 9200 in that category/generation.

As do I (including the 9250). "R(V)2x0" seems too awkward ;-) And Wikipedia does it too, after all.

> that would leave you're best choices as the 8500 or 9100.

Yes, with the 8500 as the "top supported card just a few years ago". The 9100 is just a renamed 8500 LE.

lmao.. Jim you sound as if you're entitled to something.. "Jim wants this.. Jim wants that.".. pick up a fuckin book, educate yourself and start coding. lmao.

Quote:

The only drivers that are "vital" for ab X64 port are for those "Radeons" AMD puts inside the CPU these days, cos that the only way we would be able to use MorphX64 on laptops and miniITX systems.

If I had to use an ATX for some other reason, I wouldn't mind stuffing something like a 9600 into it.......

Quote:

You know, I haven't even checked the specs on the latest integrated CPUs, but the older gpus were all Terascale.
So maybe driver support for gpus before the GCN cards makes sense in light of that.
But I haven't checked the specs of the Ryzen CPUs with integrated graphics, because while they have been announced, they aren't going to be available any time soon.
AMD just released a Socket AM4 CPU with integrated graphics, but it uses their one of their older CPU designs.


[ Edited by Jim 27.09.2017 - 11:27 ]

Quote:

I've never insisted on anything, I just go with the flow.
Personally, I was in favor of a shift to ARM, but the developers are going with X64, so I'll follow - they lead.
And MorphOS isn't documented enough (in a book or via electronic documentation) to just start coding.
And even if it was, I don't think I'll ever be able to code at the level some of the developers are at.
If you think you can, you try it.

[ Edited by Jim 27.09.2017 - 11:28 ]

> I haven't even checked the specs on the latest integrated CPUs, but the older gpus
> were all Terascale.

GPUs in AMD APUs are GCN-based since 2013/2014 (Jaguar/Steamroller with GCN2). Current Excavator-based APUs use GCN3-based GPUs, and announced (Q4/2017) Zen-based Ryzen Mobile (Raven Ridge) APUs are going to use GCN5-based Vega GPUs.

https://en.wikipedia.org/wiki/AMD_Accelerated_Processing_Unit#Graphics_Core_Next-based_GPU
https://en.wikipedia.org/wiki/List_of_AMD_accelerated_processing_unit_microprocessors#Feature_overview

> So maybe driver support for gpus before the GCN cards makes sense in light of that.

Only for APUs released in 2012/2013 (Piledriver) and earlier.

> they aren't going to be available any time soon.

Q4/2017 doesn't sound too far away.

> AMD just released a Socket AM4 CPU with integrated graphics, but it uses
> their one of their older CPU designs.

Yes, the Excavator-based A6-9550 with GCN3-based GPU.

Quote:

Well that shows you how much I paid to that announcement, DDR4 enabled Excavator CPU, pretty pointless.

So...in light of the shift to GCN gpus, one more point for your argument, BUT your still getting R600 up to the 6000 series first. 😁
Because it's Mark's call.

You know the reason I don't push on features anymore?

Because I'm suspicious that part of the reason we haven't received 11,2 support is that I pissed off Frank about it a couple of years ago misinterpreting his statement that it wouldn't take long to do, as a statement that it would happen.

It could just be hubris on my part, that assumption, but I did piss him off.
It could very well be that the primary reason is supporting too many model.
But if I screwed this up for the rest of you...
Anyway, no suggestions, no questions, I just take what's offered.

Part of what kind of teed me off about that comment about learning to code.
I DON'T insist on anything, ocassionally politely ask, but I'm as clueless as the next guy about the future.
And I don't ask questions anymore.

[ Edited by Jim 27.09.2017 - 17:13 ]

Quote:
While I don't think everyone must be a coder there are quite some ways to code for MorphOS. Not for every taste and language (I failed with MUI and C++ when I tried), but with C MorphOS is pretty well documented, albeit not necessarily comfortable.
Or Hollywood - simple, well documented with brilliant support. Not suited to develop drivers or such, but you can develop really nice and different things with it. .


[ Editiert durch Zylesea 28.09.2017 - 00:49 ]

> I'm suspicious that part of the reason we haven't received 11,2 support is that I pissed
> off Frank about it a couple of years ago misinterpreting his statement that it wouldn't
> take long to do, as a statement that it would happen. It could just be hubris on my part,
> that assumption

Yes, I very much think it is ;-)

> if I screwed this up for the rest of you...

I doubt you did.

Quote:

I doubt it too, primarily because I don't think Frank Mariak would consider something like that (AND it inflates my place in the general universal scheme of things).
But I take it to heart when I disappoint people.

BTW - Thanks for the info on the Ryzen w/ integrated graphics release.
I thought that wasn't happening until Spring '018, at the earliest.

And that bridge cpu is an ugly thing when AMD has something better in its toolbox.

> Thanks for the info on the Ryzen w/ integrated graphics release.
> I thought that wasn't happening until Spring '018, at the earliest.

It seems there's going to be two different Zen-based Raven Ridge APUs, one for mobile and one for desktop (AM4). The mobile one is due Q4/2017 and what I referred to, while the desktop one is due Q1/2018 and probably what you referred to.

http://www.pcgameshardware.de/screenshots/original/2017/09/Roadmap-pcgh.png

Damn, I was getting psyched about that.
Oh well, we could see some Ryzen based laptops soon, and that will mess with Intel's dominance of the laptop market, so things could get interesting.