Does anyone have "X1000 with RadeonHD 7770" to test cow3d?
So far it seems: Sam460cr: 120FPS X5000: 230FPS X1000: ???FPS
- Kimmo --------------------------PowerPC-Advantage------------------------ "PowerPC Operating Systems can use a microkernel architecture with all it�s advantages yet without the cost of slow context switches." - N. Blachford
The cow3D seems to be affected by something other than the video card model because each NG model seems to end up with the same result when "b" is pressed. 144 for the Sam460 and 95 for the X1000 .
The interesting result is why is my Sam460ex Quake 3 result a few fps's higher than other Sam460's ?
While newer drivers with additional optimisations are being beta tested, I feel the need to address this:
Quote:
The old Warp3D drivers never be any good aswell, but considering the results that one (for RadeonHD) seems even less and less optimized ...
If you're expecting performance to increase linearly with GPU processing power then you're always going to be disappointed. It just doesn't work this way. There are various possible bottlenecks that can limit performance and GPU performance is only one of them. The CPU also places limits on how many render ops and vertices can be processed per second. In our case we're also dealing with an old API that puts extra load on the CPU, and yes, that most certainly does get in the way.
Warp3D has no hardware Transformation Clipping and Lighting (TCL). This means that all of those calculations have to be done on the CPU (by MiniGL). There's no way that any driver optimisations can eliminate this overhead.
Slower CPUs (like Sam4x0 machines) will have greater trouble with the additional overhead of software TCL than faster machines, and this is exactly what we're seeing in the results posted to this thread. Just compare the Quake III fps achieved by A1-X1000s vs Sam460s with the same card. Don't forget that the game itself also needs CPU processing power for game logic, sound, rendering, etc.
There are other factors at play (e.g., no GART support), but I think that I've said enough to prove my point.
So, while a more optimised driver is coming, please consider the limitations that we're working with, and adjust your expectations accordingly. The drivers are doing quite well considering the limitations it has to work within.
If you're expecting performance to increase linearly with GPU processing power then you're always going to be disappointed. It just doesn't work this way.
Yep but considering that one of the purposes of supporting Warp3D on the newest cards were about gains some speed ...
Quote:
There are various possible bottlenecks that can limit performance and GPU performance is only one of them. The CPU also places limits on how many render ops and vertices can be processed per second. In our case we're also dealing with an old API that puts extra load on the CPU, and yes, that most certainly does get in the way.
Surely CPU is an important factor and i agree, however under the same machine, a Sam460 with a lower gfx and with a similar operative system like MorphOS 3.8, this system can provides results four times higher with Quake 3 !
Quote:
MorphOS = 100+ FPS <---> AmigaOS4 = 28.0+ FPS
Does MorphOS have any support for TCL or Gart in its video drivers ? Honestly i doubt ... but even assuming that TCL is already supported in MorphOS, how can we justify a difference of four times under the same machine ?
I'm not blaming anyone of course, only I would like to understand from where it comes all that huge difference between our 3D, especially from a pay driver an average user would expect much much more !
Quote:
So, while a more optimised driver is coming
Ok this is good news, but only if not only a facade improvement, after all that years of wait users start expecting a significant improvement to justify the expense ...
Aniway changing argument (but not too mutch) I was wondering if eventually other programmers have the rights to access to the latest Warp3D's sources, a few days ago I spoke with Massimiliano Tretene (m3x), he was willing to help out, even only for look at ... but he was never be able to do so ... can you help him ?
Yep but considering that one of the purposes of supporting Warp3D on the newest cards were about gains some speed ...
The purpose of the W3D_SI driver was to provide people with PCIe machines and Radeon HD cards usable 3D today, rather than to keep them waiting for Gallium3D.
Quote:
Surely CPU is an important factor and i agree, however under the same machine, a Sam460 with a lower gfx and with a similar operative system like MorphOS 3.8, this system can provides results four times higher with Quake 3 !
Newsflash! MorphOS has hardware TCL! Considering that they designed their own private 3D driver API when hardware TCL was a standard feature, they'd be fools if they didn't use it.
Quote:
... but even assuming that TCL is already supported in MorphOS, how can we justify a difference of four times under the same machine ?
Hardware TCL really is a big deal, so yes, that alone is a major factor. Plus, MiniGL isn't exactly the fastest. Those who choose to use Warp3D directly and do their own software TCL achieve faster results.
NOTE: I'm not saying that lack of hardware TCL is fully to blame; only that it alone could be enough.
Quote:
Ok this is good news, but only if not only a facade improvement, after all that years of wait users start expecting a significant improvement to justify the expense ...
Now you're just being insulting...
I've now explained twice why you can't expect too much from a Warp3D driver. If you continue to expect miracles, then I can't help you.
Quote:
Aniway changing argument (but not too mutch) I was wondering if eventually other programmers have the rights to access to the latest Warp3D's sources, a few days ago I spoke with Massimiliano Tretene (m3x), he was willing to help out, even only for look at ... but he was never be able to do so ... can you help him ?
That's up to A-EON. However, if m3x really wants to help, then I suggest that he looks at MiniGL's source-code. Redesigning MiniGL's internal render pipeline would probably help a fair bit,** although I'm not sure that it's worth the effort (after all, MiniGL + Warp3D is set to be replaced).
Hans
** For starters, it should try to deliver large vertex arrays to Warp3D rather than cutting them up (and even emitting single triangles) during clipping.
Does MorphOS have any support for TCL or Gart in its video drivers ? Honestly i doubt ... but even assuming that TCL is already supported in MorphOS, how can we justify a difference of four times under the same machine ?
Add on top of all that their 3d system designed well from scratch, while our warp3d too old, even for amigas, and Hans only make a driver for what we have, and not whole new subsystem with new internal APIs. For sure its better than waiting for Gallium another 15 years ?:) Sure perfomance wise some may be disappointed after years of waiting, but such a life, we should be thankful that amikit spend their own money and Hans his time, on making that "stop-gap" stuff, so at least there possible to use 3d.
Once there will be TCL, GART and all that stuff, and minigl will make use of it, then we can for sure compare things. But for now it's even will be good when we will reach half of perfomance of quake3 at least with all current limitations and shortcomings. And probably we will: latest beta vesion of w3d_si noticable faster than 1.8/1.9 versions. But you can't expect miracles, at least while there is no TCL, which is shitload of work, just morphos devs already done with it, while we are not.
@Hans Quote:
after all, MiniGL + Warp3D is set to be replaced
:) Probably after years of waiting of that gallium thing (remember that topic on aw.net, where few years ago i say that it will take in the best case 2 years, but maybe more, and everyone was jump on me ?:) ) Now how many years passed .. 3 , 4, 5 , do not remember. But what i trying to say, is that probably it is worth to improve minigl. As when all that mambo-jambo with gallium will see light no one know, and even when it will, no one know what kind of speed it will give for os4.
So, making today w3d drivers and improve minigl are 100% good and right thing. Probably even impove w3d whole subsytem are worth too , if someone will pay for it of course and someone will be in interest to do so :)
News feed
OS4 Feedback forum
OS4Depot Feedback forum
Amigabounty forum
OpenAmiga 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
