AMD: "[Developers use PhysX only] because they’re paid to do it"

XMAN26 said:
I'd argue Havok and teh others are far moer "useless' Than PhysX is. Atleast with PhysX it adds a more realistoc feal to a game rather than crap prerendered/scripted that never changes.
Click to expand...
That isn't Havok; you're confusing things.

XMAN26 said:
Right, being able to interact with the environment around you in a multitude of different ways with different results depending on the former is non-interactive.
Click to expand...
Actually, you've got it exactly backwards. Up till now, GPU-accelerated physics has absolutely no impact on the worldspace in the game. None. I don't think you understand this concept well enough to speak about it...
 
aaronspink said:
nvidia has done plenty of positive things, I've owned several of their cards, unfortunately not a whole lot of positive things lately and I've never been a fan of physx stretching all the way back to the ageia days. Proprietary lock in software is proprietary lock in software.
Click to expand...
It just struck me...so x86 is free now...or? :p
 
Lonbjerg said:
Most I have tried was 8 people...canno remember the limit.



Well, welcome the previous posts? :oops:
you know, 56K needs to die...SHDSL (2ms to gateway) ect...are you trolling...or just short term memory problems?
Click to expand...

whats a gateway? you mean LNS? I would expect much higher latency then 2ms to the LNS around 20ms is far more likely.
 
itsmydamnation said:
whats a gateway? you mean LNS? I would expect much higher latency then 2ms to the LNS around 20ms is far more likely.
Click to expand...

You should look up SHDSL Gbond.
I don't have a router either.
I have a DSLAM.
I run DSLAM to DSLAM to our core network...and I have 2ms ping in DK...3-4ms to sweeden...~100ms to New York (damn you speed of light!!!)
20 MS sounds like a lantency above ADSL (~10ms) infact it sounds like ADSL2+ (20-35ms)...but no where near SHDSL lantency.

EDIT:
From my SHDSL Gbond

Code: 
Pinger [URL="http://www.tm.se/"]www.tm.se[/URL] [84.243.62.13] med 32 byte data:
Svar fra 84.243.62.13: byte=32 tid=4ms TTL=55
Svar fra 84.243.62.13: byte=32 tid=4ms TTL=55
Svar fra 84.243.62.13: byte=32 tid=4ms TTL=55
Svar fra 84.243.62.13: byte=32 tid=4ms TTL=55

From my ADSL2+ line

Code: 
Pinger [URL="http://www.tm.se/"]www.tm.se[/URL] [84.243.62.13] med 32 byte data:
Svar fra 84.243.62.13: byte=32 tid=29ms TTL=54
Svar fra 84.243.62.13: byte=32 tid=29ms TTL=54
Svar fra 84.243.62.13: byte=32 tid=29ms TTL=54
Svar fra 84.243.62.13: byte=32 tid=29ms TTL=54
 
Last edited by a moderator:
What does any of that have to do with this conversation? Are we talking about sending physics data over the wire? Because if so, it doesn't matter if you have an OC-192 piped directly into your box, because you aren't the general masses.

The general masses might aggregate 10mbps on a relatively latent network, as a 'norm'. Even that's questionable, but since you pinned it down to 'gamers' (however you think you can realistically define that) then so be it.
 
Lonbjerg said:
You should look up SHDSL Gbond.
I don't have a router either.
I have a DSLAM.
I run DSLAM to DSLAM to our core network...and I have 2ms ping in DK...3-4ms to sweeden...~100ms to New York (damn you speed of light!!!)
20 MS sounds like a lantency above ADSL (~10ms) infact it sounds like ADSL2+ (20-35ms)...but no where near SHDSL lantency.

EDIT:
From my SHDSL Gbond

Code: 
Pinger [URL="http://www.tm.se/"]www.tm.se[/URL] [84.243.62.13] med 32 byte data:
Svar fra 84.243.62.13: byte=32 tid=4ms TTL=55
Svar fra 84.243.62.13: byte=32 tid=4ms TTL=55
Svar fra 84.243.62.13: byte=32 tid=4ms TTL=55
Svar fra 84.243.62.13: byte=32 tid=4ms TTL=55

From my ADSL2+ line

Code: 
Pinger [URL="http://www.tm.se/"]www.tm.se[/URL] [84.243.62.13] med 32 byte data:
Svar fra 84.243.62.13: byte=32 tid=29ms TTL=54
Svar fra 84.243.62.13: byte=32 tid=29ms TTL=54
Svar fra 84.243.62.13: byte=32 tid=29ms TTL=54
Svar fra 84.243.62.13: byte=32 tid=29ms TTL=54
Click to expand...

from my DSL line

Code: 
Reply from 203.16.215.189: bytes=32 time=23ms TTL=254
Reply from 203.16.215.189: bytes=32 time=22ms TTL=254
Reply from 203.16.215.189: bytes=32 time=23ms TTL=254
Reply from 203.16.215.189: bytes=32 time=23ms TTL=254

Code: 
home#show dsl int atm 0
ATM0
Alcatel 20190 chipset information
                ATU-R (DS)                      ATU-C (US)
Modem Status:    Showtime (DMTDSL_SHOWTIME)
DSL Mode:        ITU G.992.1 (G.DMT) Annex A
ITU STD NUM:     0x01                            0x1
Vendor ID:       'STMI'                          'BDCM'
Vendor Specific: 0x0000                          0x619F
Vendor Country:  0x0F                            0xB5
Chip ID:         C196 (0)
DFE BOM:         DFE3.0 Annex A (1)
Capacity Used:   99%                             97%
Noise Margin:     9.5 dB                         11.0 dB
Output Power:    19.5 dBm                        12.0 dBm
Attenuation:     51.0 dB                         31.5 dB

I CBF logging into work to look at SHDSL latencies but the tail isn't anywhere near 2ms.
im at a distance from the DSLAM where i get 0 benifit from ADSL2 and all the extra interleave latency.

is that SHDSL over some kind of FTTN/FTTC style network? seems much faster latency wise then what we get over POTS.
 
Albuquerque said:
What does any of that have to do with this conversation? Are we talking about sending physics data over the wire? Because if so, it doesn't matter if you have an OC-192 piped directly into your box, because you aren't the general masses.

The general masses might aggregate 10mbps on a relatively latent network, as a 'norm'. Even that's questionable, but since you pinned it down to 'gamers' (however you think you can realistically define that) then so be it.
Click to expand...

It was more of a argument brought up by eastman, that interactive physics over WAN was impossible...perhaps in the US, but not in DK.

itsmydamnation said:
from my DSL line

Code: 
Reply from 203.16.215.189: bytes=32 time=23ms TTL=254
Reply from 203.16.215.189: bytes=32 time=22ms TTL=254
Reply from 203.16.215.189: bytes=32 time=23ms TTL=254
Reply from 203.16.215.189: bytes=32 time=23ms TTL=254

Code: 
home#show dsl int atm 0
ATM0
Alcatel 20190 chipset information
                ATU-R (DS)                      ATU-C (US)
Modem Status:    Showtime (DMTDSL_SHOWTIME)
DSL Mode:        ITU G.992.1 (G.DMT) Annex A
ITU STD NUM:     0x01                            0x1
Vendor ID:       'STMI'                          'BDCM'
Vendor Specific: 0x0000                          0x619F
Vendor Country:  0x0F                            0xB5
Chip ID:         C196 (0)
DFE BOM:         DFE3.0 Annex A (1)
Capacity Used:   99%                             97%
Noise Margin:     9.5 dB                         11.0 dB
Output Power:    19.5 dBm                        12.0 dBm
Attenuation:     51.0 dB                         31.5 dB

I CBF logging into work to look at SHDSL latencies but the tail isn't anywhere near 2ms.
im at a distance from the DSLAM where i get 0 benifit from ADSL2 and all the extra interleave latency.

is that SHDSL over some kind of FTTN/FTTC style network? seems much faster latency wise then what we get over POTS.
Click to expand...

It this:
http://en.wikipedia.org/wiki/Single-pair_high-speed_digital_subscriber_line

The extended version and then in Gbond mode (line bundling) so i have (3x)5696/5696 kbit/s:

Code: 
ds1-op> show linerate 10-6
slot-port=10-6, DSL line rate
link              = up_gbond
min rate          = 4096 kbps
max rate          = 5696 kbps
actual rate       = 5696 kbps
noise margin      = 0 dB
attenuation       = 11 dB
linkup count      = 1
 
ds1-op> show linerate 10-7
slot-port=10-7, DSL line rate
link              = up_gbond
min rate          = 4096 kbps
max rate          = 5696 kbps
actual rate       = 5696 kbps
noise margin      = 6 dB
attenuation       = 12 dB
linkup count      = 1
 
ds1-op> show linerate 10-8
slot-port=10-8, DSL line rate
link              = up_gbond
min rate          = 4096 kbps
max rate          = 5696 kbps
actual rate       = 5696 kbps
noise margin      = 0 dB
attenuation       = 11 dB
linkup count      = 1

(current DSLAM firmware has issuse reporting accurate SNR/attenuation values when in Gbond)

So it is symmetric xDSL(ATM based), with 3 lines bundled via this:
http://www.zyxel.com/web/product_fa...316E-3E79-4CE6-98B6-0BAC34EB4C57&display=6243

I was considering going 100/100 Mbit/s fiber...but I actually don't need the extra bandwith as i don't do warez :D
But SHDSL is perfect for gamers due to the low lantency.
 
aaronspink said:
x86 is the standard, has been, will be. Its available in multiple implementations from multiple companies. Its pretty much a definition of non-proprietary.
Click to expand...


In view of this

http://www.agner.org/optimize/blog/read.php?i=25

your claim of x86 being non-proprietary seems hard to digest.

IMHO, x86 is a proprietary ISA, grudgingly licensed to others, all of whom are busy making super-proprietary forks with occasional convergence.
 
rpg.314 said:
IMHO, x86 is a proprietary ISA, grudgingly licensed to others, all of whom are busy making super-proprietary forks with occasional convergence.
Click to expand...

x86 has been made by roughly 13 different companies over the years. Most weren't competitive enough to stay in the market place. If anything, its the least proprietary of the mainstream ISAs outside of Sparc.
 
Lonbjerg said:
Oh yeah?
http://www.tomshardware.com/news/amd-intel-x86-cpu,7285.html
Click to expand...

If a company is using patents without a cross license, then they are in trouble. This has been a pretty basic concept for oh say 200+ years. You want to try to make a modern high performance processor without violating at least one patent of a company the size of Intel? Hell, if you are going to complain about this, you must hate IBM which basically requires everyone in the electronics industry to have a license from them.

Because, I quote: "Proprietary lock in software is proprietary lock in software." :p
Click to expand...

you mean C and C++ are owned by Intel or that code produced by Intel's compilers don't run on industry standard x86? Come on, if you want to have a comparison at least make it in the same ballpark. I never said anything nasty about nvidia making their physx backend only run on OpenCL or CS. The front end, middle, AND backend for physx is entirely proprietary. The only thing proprietary about an Intel compiler is the the middle which is pretty normal in a compiler outside of GCC. The compilers take industry standard code and output industry standard machine code.

When Nvidia decides to transfer the APIs to an ISO and have their stack output to openCL/CS then you can talk.
 
Looks like Huddy's epic troll worked. Only at B3D could a bunch of the world's smartest fanboys prostitute themselves this much in public :love:
 
aaronspink said:
If a company is using patents without a cross license, then they are in trouble. This has been a pretty basic concept for oh say 200+ years. You want to try to make a modern high performance processor without violating at least one patent of a company the size of Intel? Hell, if you are going to complain about this, you must hate IBM which basically requires everyone in the electronics industry to have a license from them.



you mean C and C++ are owned by Intel or that code produced by Intel's compilers don't run on industry standard x86? Come on, if you want to have a comparison at least make it in the same ballpark. I never said anything nasty about nvidia making their physx backend only run on OpenCL or CS. The front end, middle, AND backend for physx is entirely proprietary. The only thing proprietary about an Intel compiler is the the middle which is pretty normal in a compiler outside of GCC. The compilers take industry standard code and output industry standard machine code.

When Nvidia decides to transfer the APIs to an ISO and have their stack output to openCL/CS then you can talk.
Click to expand...

I guess we see things differently, no biggie...I still do not consider x86 for free and open *shrugs*

Rys said:
Looks like Huddy's epic troll worked. Only at B3D could a bunch of the world's smartest fanboys prostitute themselves this much in public :love:
Click to expand...

Of course Huddy is trolling...what else can you do when you promised GPU-physics in 2006..and still fail to deliver? ;)
 
Lonbjerg said:
Of course Huddy is trolling...what else can you do when you promised GPU-physics in 2006..and still fail to deliver? ;)
Click to expand...
Blame Intel on that, they bought off Havok and cancelled Havok FX :D
And last I checked, Bullet+DMM OpenCL accelerated are out already?
 
Rys said:
Looks like Huddy's epic troll worked. Only at B3D could a bunch of the world's smartest fanboys prostitute themselves this much in public :love:
Click to expand...

What I find disturbing is regular members spending time trying to educate the latest batch of trolls that registered since November 09 as if they were posting here to learn something and not just spouting their biases and nothing esle :devilish:
 
Actually the meta discussions Rys, you and now I are engaging in add nothing except our biases ... as repetitive as the rest of the thread is, it generally adds more even if by a minimal amount.

PS. x86 is proprietary, but it's ISA is fairly static, openly documented and there is competition in it's acceleration. None of that is true for the PhysX API.
 
PatrickL said:
What I find disturbing is regular members spending time trying to educate the latest batch of trolls that registered since November 09 as if they were posting here to learn something and not just spouting their biases and nothing esle :devilish:
Click to expand...

Ahh so I am not the only one who noticed that as well... I suppose just as we had to wade through the Red Tide before, we now have to make our way through the Green Fog now
 
The problem with "fanboyism waves" on every board is IMO pretty much the same, the underdog gets always a lot of supporters who go in far too deep into the fanaticism, and then there's always the few who are if possible even more fanatical about the other brand
Lately, AMD/ATI has been the underdog, thus had a lot more forum support

(yes, I know I'm blamed for even being paid by AMD/ATI (I wish :LOL: ) on some other board(s), but it's just the dislike towards nVidia due 3dfx "incidents", not that I'd like AMD/ATI more)
 
You must log in or register to reply here.
Back
Top