28-Mar-2024 11:49 GMT.
UNDER CONSTRUCTION
Anonymous, there are 61 items in your selection (but only 11 shown due to limitation) [1 - 50] [51 - 61]
[News] Details on ATI Radeon support in OS4ANN.lu
Posted on 16-Jan-2002 16:48 GMT by Teemu I. Yliselä61 comments
View flat
View list
Bill Toner talked about issues regarding the support of ATI Radeon chipsets in AmigaOS4 on the AmigaOne mailing list. Read the message below.

ATI Radeon support will happen, but I'm not sure when it will be completed. Permedia 3 support has been announced as well. This will initially support the original Radeon chipset, RV100 that has been around a while, and will at some point support the newer Radeon chip, the RV200 as well. My business partner and I are the "other party" Hyperion mentioned in their OS4/Radeon/Permedia3 announcement a while ago for the Radeon part. We hope to have something usable soon, but it certainly won't be done by their announced OS4 release date in Feb. But it will most definitely be done for OS4.x, and I have had an AmigaOne on preorder as well for some time... Prometheus will be the first bridge supported, as that'a all that I have right now. We do intend to support Mediator if we can, but Elbox's sense of PR has kept me disgruntled with them and I have not signed/returned their NDA form yet, so we can't yet support Mediator but I'll do it soon, plus Mediator 1200 models at least will be more work due to their small memory window access point. Especially since some boards apparently work with 8MB window setting and some other older boards only work well with the smaller 4MB window setting. Elbox PR said something about the trashed graphics in 8MB mode were some hardware dongle feature or something like that, but I find that hard to swallow... But we'll work it out at some point and find a Mediator owner to test it with. When I originally contacted DCE about supporting Grex a year ago or so now, I still have yet to receive any response at all from them. So I'm not sure they wanted Radeon drivers for their card which I thought was odd, but oh well. Perhaps I should try them again... AmigaOne will also be supported via the AGP slot. Regarding Pegasos, my NDA with ATI does not have "Morphos" written on it anywhere. But I got this NDA back before I knew of the Pegasos project at all and I run AmigaOS on my Amigas, not anything else, so Morphos at the time didn't have much meaning to my machines as none of them even have PPC cards. AmigaOS 3.9 and AmigaOne/AmigaOS 4.x will be what we target while we consider if the effort in renegotiating the NDA is worth the returns from MorphOS users.

But please, don't run out and buy a Radeon card just yet. We may dongle-ize our drivers to some tag we set in Radeon cards we sell, in an effort to reduce P96 piracy. Elbox screwed the P96 guys, as not all Mediator users will pay the P96 "shareware fee" Elbox talked about. We are going to include fully licensed P96 (and possibly warp3d if they don't have licenses already with all PCI bridge companies) with Radeons we sell to guarantee both P96 team, warp3d team,  and my company get return on our work. I don't intend to trash talk anyone, but piracy is a fact of life. We intend to circumvent that a bit, in the same way that PicassoIV did. PC store-bought Radeons may not work, just a warning.

As for the porting of Mediator/Grex drivers to OS4.x, I wonder how this will be done. Surely if OS4 runs on Elbox's SharkPPC they will port Mediator drivers to that OS4 implementation, but I'm not sure they'll let their drivers work on hardware other than Mediator+SharkPPC. DCE may not have Grex drivers work under OS4 on hardware other than Grex. Hopefully they will support more bridges/motherboards, especially Elbox since they charge extra for their driver CDROM for Mediator anyway. We'll have to wait and see, but I'm not getting any hopes up. We'll be supporting any PCI bridge we can with Radeon drivers.

--
Bill Toner
Forefront Technologies, Inc. (previously Progressive Data Systems)
bill@prodatasys.com

Details on ATI Radeon support in OS4 : Comment 51 of 61ANN.lu
Posted by priest on 18-Jan-2002 06:21 GMT
In reply to Comment 44 (Christophe Decanini):
But not on Amiga ????????????
Or... 8)
Details on ATI Radeon support in OS4 : Comment 52 of 61ANN.lu
Posted by priest on 18-Jan-2002 06:24 GMT
In reply to Comment 46 (Anonymous):
>The CSPPC/MK3 has 40MB/s UWSCSI.
Ok, ok.
Now I start to believe that past 10MB/s is possible on Amiga. ;)
Details on ATI Radeon support in OS4 : Comment 53 of 61ANN.lu
Posted by José on 18-Jan-2002 10:14 GMT
In reply to Comment 41 (Alkis Tsapanidis):
Yeah right... even Amijoe was to have 100mhz.
Details on ATI Radeon support in OS4 : Comment 54 of 61ANN.lu
Posted by Alkis Tsapanidis on 18-Jan-2002 10:39 GMT
In reply to Comment 49 (Mike Veroukis):
The stats are somewhere. It's not a G5... It's a G4/1.2GHz.
By February the newer 1.6 models will ship to retailers:)
The PPC now has enough raw power:)
It's FSB is 133.
Details on ATI Radeon support in OS4 : Comment 55 of 61ANN.lu
Posted by Amifan on 18-Jan-2002 11:08 GMT
In reply to Comment 54 (Alkis Tsapanidis):
Isn't the 1.6 GHz version called G5? If so then it has (rumour) a 400Mhz fsb.
Details on ATI Radeon support in OS4 : Comment 56 of 61ANN.lu
Posted by Alkis Tsapanidis on 18-Jan-2002 11:19 GMT
In reply to Comment 55 (Amifan):
The G5 is a different story.
Details on ATI Radeon support in OS4 : Comment 57 of 61ANN.lu
Posted by Alkis Tsapanidis on 18-Jan-2002 11:20 GMT
In reply to Comment 53 (José):
It has been confirmed.
Details on ATI Radeon support in OS4 : Comment 58 of 61ANN.lu
Posted by Anonymous on 18-Jan-2002 13:44 GMT
In reply to Comment 50 (Amifan):
>>>You want to transfer the gfx data from main memory to the GFX card.
>>>They have both the same theoretical maximum transfer speed.
>>>
>>No, you are wrong - AGP2X has twice the speed of AGP1X.
>>
>No you are wrong ;-) I was talking about system memory -> gfx memory.
>The AGP transfer speed can only go as fast as your main memory goes.
>
The theoretical maximum AGP1X bandwidth is 264MB/s.
The one of AGP2X is 528MB/s.
Memory is 1067MB/s (and you even admit it should be 350-400MB/s in real
world tests).
I hope you now see your statement "they have both the same theoretical
maximum transfer speed" is plainly wrong.
Details on ATI Radeon support in OS4 : Comment 59 of 61ANN.lu
Posted by priest on 18-Jan-2002 15:23 GMT
In reply to Comment 58 (Anonymous):
people sometimes forget the bus width ;)
Correct me if I'm wrong (theorethical maximums):
IDE 8bit UDMA66 (66Mhz) - 66MB/s (From IDE drive's cache to computer system bus.)
PCI 32bit 33Mhz - 132MB/s (this is the standard PCI of PCI standard) ;)
PCI 32bit 66Mhz - 264MB/s (66Mhz works with some motherboards/cards)
PCI 64bit 33Mhz - 264MB/s
... (is there any 66Mhz capable 64bit cards/motherboards?)
(most PCI cards work only in 33Mhz limitting the whole bus to 33Mhz)
AGP1x 32bit 66Mhz - 264MB/s
AGP2x 32bit 66MhzDualPumped - 528MB/s
AGP4x 32bit 66MhzQuadPumped - 1056MB/s
... (8x is not really used yet, AGP's main feature is direct access to system memory)
memory bus performance is degraded by RAM chip capabilities, so I leave the calculation to someone else...
OCS, ECS and AGA access chip ram at 7Mhz clock. True or false?
OCS and ECS machines have 16bit chip RAM bus (theor. 14MB/s).
AGA machines have 32bit chip RAM bus (theor. 28MB/s).
020ec on the A1200 motherboard can access Fast RAM at 14Mhz clock. True?
On modern motherboards (x86, PowerMac, AmigaOne, Pegasos) the system memory bus is 64bit wide.
PC66 528MB/s (THEORETHICAL bus speed, chip latencies take it down...)
PC100 800MB/s
PC133 1064MB/s
DDR200 1.6GB/s (THEORETHICAL bus speed...) (100Mhz dualpumped)
...
Z2 is a 16bit 7Mhz bus (14M).
Z3 is a 32bit 7Mhz bus (28M).
PCI daughter cards on Z3 slot are limited by 28MB/s theorethical Z3 bandwith to CPU card memory.
( real life speed is around 10MB/s, true? )
PCI cards on the PCI daughter can access each other at full PCI bus speed (if the daugtercard supports it).
Does the bus from Mediator PCI daughter card on A1200 expansion slot have a 14Mhz 32bit bus (56MB/s) to CPU card memory?
(Or is it only 7Mhz or is it same as CPU clock?)
How about mediator A4k?
(CPU cards on A1200 accelerator slot or A2k/A3k/A4k CPU slot can have any memory bus width and speed to the RAM, so...)
Blizzard680x0 accelerators have 32bit buses at CPU clock (for example 50MHz).
BlizzardPPC has a 32bit memory bus, but what is the Mhz multiplier setting? Is the bus clock 33Mhz or 66Mhz or what??
BlizzardPPC has a 32bit gfx card bus, 33Mhz?
CyberstormMK1 and 2 have a 32bit bus at CPU clock
CSMK3 has a 64bit memory bus, even though the CPU has 32bit bus to bus ctrlr?
CSPPC has 64bit memory bus, but again, what's the multiplier setting/Mhz?
CSPPC has also 64bit gfx card bus, 33Mhz?
I think there exists 4 and 8 bit clockports on Amigas. What is the Mhz? How fast is the clock port "bus" when you add soundcard/USBcard/etc. there?
And, modern CPUs have superb fast buses to L1/L2 and L3 memory... but that one needs another day...
Details on ATI Radeon support in OS4 : Comment 60 of 61ANN.lu
Posted by Alkis Tsapanidis on 18-Jan-2002 17:29 GMT
In reply to Comment 58 (Anonymous):
You "forgot" to mention that the AGP 1x itself is a 66mhz PCI slot.
The "theoritical" rates you mention can only be achieved by special 2x
(burst AFAIK) transfer methods. The same holds true for AGP 4x.
Most (all?) the current chipsets do not implement all the methods.
IIRC only 2x & 4x GFX > Memory access works with the current chipsets.
CPU > GFX 2x & 4x isn't implemented, or too buggy to work.
Details on ATI Radeon support in OS4 : Comment 61 of 61ANN.lu
Posted by Christophe Decanini on 18-Jan-2002 18:23 GMT
In reply to Comment 52 (priest):
>Now I start to believe that past 10MB/s is possible on Amiga. ;)
Anonymous, there are 61 items in your selection (but only 11 shown due to limitation) [1 - 50] [51 - 61]
Back to Top