Powervr will limit Ent&L to KYRO2SE !
- Thread starter Mephisto
- Start date
Hi Mephisto,
well, enT&L is touted as a new feature of the Kyro-II SE core, too, so I'm not really surprised it might not work on the regular Kyro chips.
Apparently, there have been some smallish changes in the core logics rather than simpy releasing the exact same chip with a 25MHz speed bump. But perhaps Kristof can tell us more?
ta,
.rb
________
TS100
well, enT&L is touted as a new feature of the Kyro-II SE core, too, so I'm not really surprised it might not work on the regular Kyro chips.
Apparently, there have been some smallish changes in the core logics rather than simpy releasing the exact same chip with a 25MHz speed bump. But perhaps Kristof can tell us more?
ta,
.rb
________
TS100
Last edited by a moderator:
From Beyond 3d interview:
Q: The T+L engine, is it a software or hardware T+L engine? As there is a lot of confusing online what it exactly is.
A: When you read the press release (which I wrote), you see a clear statement that it is a software T+L engine.
If it's software T&L engine what stops them from doing it to in Kyro 2 drivers?
Q: The T+L engine, is it a software or hardware T+L engine? As there is a lot of confusing online what it exactly is.
A: When you read the press release (which I wrote), you see a clear statement that it is a software T+L engine.
If it's software T&L engine what stops them from doing it to in Kyro 2 drivers?
Joe DeFuria
Legend
I can only see 1 of 2 answers:
1) Even though it is software based T&L, it may require a small, yet specific, hardware feature or characteristic for proper implementaion. (Some hardware feature that was specifically added to the new KII SE has that old KII lacks.)
or
2) Purposely withholding software T&L from KI/II solely for marketing purposes. (Won't be able to advertis a see too much perofmance difference between KII and KII/SE without artificially handicapping the KII.)
Hopefully, it's number one....
1) Even though it is software based T&L, it may require a small, yet specific, hardware feature or characteristic for proper implementaion. (Some hardware feature that was specifically added to the new KII SE has that old KII lacks.)
or
2) Purposely withholding software T&L from KI/II solely for marketing purposes. (Won't be able to advertis a see too much perofmance difference between KII and KII/SE without artificially handicapping the KII.)
Hopefully, it's number one....
The way i understand it, they can't state that it's hardware TnL or hybred of the sorts... pr reasons...maybe?
Anywho.... i think it's called software tnl no not hybred at the moment.... is that it's still all done in the drivers...however, it's taking a load off the cpu BUT it's not exactly doing it in hardware on the card either.... it's being preoptimise and then sent to the kyro 2 se which has it's own speical agrorithlms to do that "TnL" part.... being much much more effeicent then a standard dx software tnl engine...
Anywho.... i think it's called software tnl no not hybred at the moment.... is that it's still all done in the drivers...however, it's taking a load off the cpu BUT it's not exactly doing it in hardware on the card either.... it's being preoptimise and then sent to the kyro 2 se which has it's own speical agrorithlms to do that "TnL" part.... being much much more effeicent then a standard dx software tnl engine...
source: http://www.powervr.com/Release.asp?ID=36
Joe,
simply the fact that it has AGPx4 SBA vs AGPx2 SBA in 4000/4500 would mark a limiting factor in hardware.
Ah, thanks, Ailuros. That's the bit I was talking about regarding "SE core feature."
In no way did I want to imply that the actual enT&L implementation is hardware--but it might very well need things only the "improved" SE core supports.
ta,
.rb
________
buy vapor genie
________
green crack pictures
In no way did I want to imply that the actual enT&L implementation is hardware--but it might very well need things only the "improved" SE core supports.
ta,
.rb
________
buy vapor genie
________
green crack pictures
Last edited by a moderator:
I stand corrected on behalf of the initial KYRO. Who knows maybe the 4000 could eventually benefit from eTAL too???
As for AGPx2 on the 4500: afaik x4 was claimed to being disabled due to compatibility reasons with some motherboards (if my memory serves me well).
As for AGPx2 on the 4500: afaik x4 was claimed to being disabled due to compatibility reasons with some motherboards (if my memory serves me well).
Dave B(TotalVR)
Regular
The KYRO family has always been capable of AGP 4x, its just on the KYRO II it is limited to AGP 2x for stability, limited at the *board* level.
Now EnTnL makes out like it requires AGP 4x and implicit guard band clipping, the first isn't available on KYRO II cards and the second, well who knows?
Dave
Now EnTnL makes out like it requires AGP 4x and implicit guard band clipping, the first isn't available on KYRO II cards and the second, well who knows?
Dave
The AGP4X blurb is ridiculus.
My mainboard does not even support AGP4X, and even if it would, with my SDR memory it can't benefit of the higher AGP bandwith anyway. There is no further technical difference between AGP2X and AGP4X.
It is just marketing, very disappointing that I will have to do without EnT&L because of this lame PR tactis.
My mainboard does not even support AGP4X, and even if it would, with my SDR memory it can't benefit of the higher AGP bandwith anyway. There is no further technical difference between AGP2X and AGP4X.
It is just marketing, very disappointing that I will have to do without EnT&L because of this lame PR tactis.
Mephisto,
It's a combination between software and hardware enhancements. I wouldn't guess that the difference between a 4500 and a 4800 would be soooooo big after all, besides the higher clock speed.
You can always wait a couple of months and get the next batch with a real T&L unit. No one after all promised from either company that you'll get an upgrade via drivers for free.
Teasy,
Isn't binning space being subtracted from the total amount of onboard memory? Programs report f.e. 62mb total where 55mb Vram and 7mb AGP. Increasing binning space to 12mb (well you have to add the strings with the latest drivers) gives me 51mb Vram and 11mb for AGP.
Not to speak that AGP texturing has been enabled since 8xxx series drivers in combination with 8.1 if my memory serves me well
It's a combination between software and hardware enhancements. I wouldn't guess that the difference between a 4500 and a 4800 would be soooooo big after all, besides the higher clock speed.
You can always wait a couple of months and get the next batch with a real T&L unit. No one after all promised from either company that you'll get an upgrade via drivers for free.
Teasy,
Isn't binning space being subtracted from the total amount of onboard memory? Programs report f.e. 62mb total where 55mb Vram and 7mb AGP. Increasing binning space to 12mb (well you have to add the strings with the latest drivers) gives me 51mb Vram and 11mb for AGP.
Not to speak that AGP texturing has been enabled since 8xxx series drivers in combination with 8.1 if my memory serves me well
Yes it is, what I'm saying is that Kyro can address more then 64mb of video ram, it can only address 64mb of texture/framebuffer ram but the ram used for poly binning can be retrieved in AGP texturing. So the total amount of ram that Kyro is able to address is dependant on how much ram is used for poly binning. 6mb is used at defualt which leaves a total of 64mb on-board ram (58mb usable for anything other then poly binning) and 6mb AGP = 70mb of usable ram. I wouldn't take those reported MB numbers as exact, its probably more like 58mb Vram (64mb - 6mb binning space) and 6mb AGP rather then 55mb Vram and 7mb AGP.
Although I should mention that this wasn't always the case, originally Kyro could only address a total of 64mb of ram including binning space. But IMGTEC managed to make it so they could retreive whatever amount of ram was being used for poly binning in AGP texturing in a new driver (well its not new now, it was quite a few months ago now).
