So it looks like it will come in in 2005 instead of the original 2H 2004. It'll have 24k L1 instead of 8k or 16k like current and prescott have. When it is made at 65nm insteadof 90nm it'll have 2megs L2 instead of 1meg.
It should start eventually run as high as 5Ghz. Maybe that is on the 65nm process years from now? Bus speed should be 1066Mhz (266*4) or 1200.
It should have some new instructions in order to make life harder for AMD.
Fortunately for AMD Prescott was already supposed to be shipping at 3.8Ghz, but Intel is a bit behind on their road map too:)
"AMD did also verify that the "new" core is 193mm2 in die size. This is the same size as specified on the original 1MB L2 Athlon 64 cores as noted here back in September of this year. "
I remember this back in the PentiumII/Celeron days, but does this happen anymore? Has anybody looked at the chip to see if it is any smaller or the same size as its bigger brother?
I don't know when the deltachrome will be on the market, but it looks like ATI and nvidia will have some new cards on the market possibly by April which will push the price of the 5900 and 9800 way down, which will in turn push the price of the 5700 and 9600 down which is going to put some serious pressure on everybody else.
I see XGI's Volari as the biggest compitition to S3's DeltaChrome.
It is just as easy to cheat... Err... include application specific optimizations in applications like Quack3. http://www.hardocp.com/article.html?art=M TEx
Re:Good, there needs to be some more competition
on
The Return of S3
·
· Score: 1
ATI's drivers aren't completely mature, but they're pretty good IMO. And ATI has been very good about fixing problems *that are reported to them*.
Please get the new 3.10 drivers released this week and see if you have any problems at all. If you find any problems head to http://apps.ati.com/driverfeedback/index.asp and give them feedback. I'd be surprised if the problems you report aren't all totally fixed by 3.11 or 3.12.
Re:Future support? Driver updates?
on
The Return of S3
·
· Score: 1
You made the correct choice. The kyro is missing T&L hardware so your poor Athlon would have to process every single triangle instead of simply handing entire objects off to the video card for processing. Also the Kyro is missing cube map support and is stuck at OpenGL 1.2 while your old GF2 supports OpenGL 1.4 (although shadows are emulated and not real time) and will probably soon support OpenGL 1.5.
I would never take a 1st generation challanger brand myself.
If I were to buy a card today it'd probably be a 5700 or maybe a 9600.
Re:I'm A Little Disappointed
on
The Return of S3
·
· Score: 2, Insightful
The question is how long will it take? How about all the 3dfx owners that never got decent drivers? Did Matrox ever get around to decent OpenGL drivers for the G400 or did they just have their Quake OpenGL->d3d wrapper? Did SiS ever give good drivers for their Xabre? Did Trident ever release good drivers for its products? Are the Kyro drivers good enough to run all applications? Only recently has ATI started producing stable drivers, and even then 7x00 users seem to be experiencing problems sometimes still.
What if you get card X with bad drivers, but it takes 11/2 years for good drivers? By the time you finally get decent drivers your card is obsolete. Why not get a card you know has good drivers and then when brand Y produces good drivers buy their card?
IMO if you buy a card with bad drivers because it gives you a slightly better price/performance ratio and its bad drivers *might* get fixed you are an idiot:P
I hope S3 gets their drivers fixed soon and they become the 3rd most popular 3d company, but until this happens I have to tell everybody to stick with nvidia or ATI hardware.
It was intended to be a great 3D card: -Four textures sampled per cycle. The Radeon 8500/GeForce4Ti can apply four texture per pass like the Parhelia, but they require two cycles instead of one. The Radeon 9000/9200 require four cycles instead of just one. -Giga colour mode: This reduces banding. Check out Homeworld2 on Giga colour vs normal colour. -Pixel Shader 1.3 just like the GeForce4. -Displacement mapping. I don't even know if any other cards on the market support this wonderful feature. -256bit memory bus. -up to 256megs of memory. -IIRC support for DX9 level vertex shaders was planned making this card partially DX9 compatible.
The old Matrox Millenium was a great 2D card. The Parhelia is a decent 3D card. It was supposed to be a GeForce3/4Ti/Radeon8500 killer and in many ways it was.
The Parhelia was intended to be a high end gamer's dream card. It just fell a bit short. Hopefully they get to try again.
Opps, that was supposed to be in POT, not HTML. This is what was missing:
BTW currently no graphics hardware exists which can emulate the standard T&L pipeline with eight lights! What would be a simple loop like this: for( i = 0; i 8; ++i) {
if(!light[i].enabled)
continue;
switch( light[i].type)
{
case directional:...
break;
case point:...
break;
case spotlight:...
break;
} } is impossible. One program needs to be written for every permutation of different lighting settings which is actually 65536 (4^8) different programs! Only many of those programs are impossible to write because you only have 128 intructions (or 256 on Radeon 9500+ and GeForceFX). By the time you calculate specular and colour for a point light (or worse, spot light) you've already used a ton of instructions. You simply can not have eight point lights on current cards. Of course to make vertex programs useful you really need to use other instructions for things other than calculating lighting such skinning of characters.
Opps, that was supposed to be in POT, not HTML. This is what was missing:
BTW currently no graphics hardware exists which can emulate the standard T&L pipeline with eight lights! What would be a simple loop like this: for( i = 0; i
ATI can execute up to 96 ALU instructions in a fragment progam on their top of the line video card. That is actually 64 native instructions. The GeForceFX can handle 1024.
Think about that for a second. Lets say you want to compute 1,000,000 things in parallel so you create a 1000x1000 floating point texture to render to. You have to fit your algorithm into just 64 instructions.
And you can't make function calls (Yes, DX9 HLSL has functions, but they are all "inline").
And your algorithm has to have no loops.
And your algorithm has to have essentially no branches. Well it can have "branches" that are of the form a = x >= y ? b : c; No real branches.
And you can't have static or global variables. Each pixel is executed by the same 64 instruction program. If you want a static variable to save between different programs you have to store it to another 1000x1000 texture. Which is what this system does.
If you want an if(whatever){ bunch of instructions } you have to compute all those instructions and then multiply by 0 or 1 at the end in order to emulate branching.. Terribly ineffecient.
BTW currently no graphics hardware exists which can emulate the standard T&L pipeline with eight lights! What would be a simple loop like this:
for( i = 0; i
We need far more flexible video cards before this is useful.
ATI's ASHLI is sort of like this. It allows you to compile long Renderman or OpenGL shaders into multiple passes so they can be execute on our current hardware like the crappy;) 9800XT
http://www.ati.com/developer/ashli.html
This will be a lot more interesting when we get pixel shaders that can do emulation of the standard T&L pipeline.
Slashdot Mirror
http://www.aceshardware.com/forum?read=105063294
http://www.x86-secret.com/
Apparently anandtech is showing a prescott?
2/27/2003:, 3973,900185, 00.asp/ cpu/display/200310110 84615.htmlt ag=Tejas
:)
http://www.extremetech.com/article2/0
10/11/2003:
http://www.xbitlabs.com/news
misc:
http://endian.net/details.asp?
So it looks like it will come in in 2005 instead of the original 2H 2004. It'll have 24k L1 instead of 8k or 16k like current and prescott have. When it is made at 65nm insteadof 90nm it'll have 2megs L2 instead of 1meg.
It should start eventually run as high as 5Ghz. Maybe that is on the 65nm process years from now? Bus speed should be 1066Mhz (266*4) or 1200.
It should have some new instructions in order to make life harder for AMD.
Fortunately for AMD Prescott was already supposed to be shipping at 3.8Ghz, but Intel is a bit behind on their road map too
GNU/Linux gets dynamic shader compilers!h tml
http://www.nvidia.com/object/IO_9292.
Do these drivers export all the same extensions as their windows counter parts?
I hope Microsoft wins this lawsuit.
This is stupid. No other company can make a RPG that features Norse gods?? Its not like they were invented by that company.
Also at http://www.firingsquad.com/hardware/s3_deltachrome _s8_preview/
Tech Report's was the best of the previews I've read.
Maybe it's because of the generally grumpy tone found on the main news page of amdzone?
:(
I didn't know HP was shipping Mandrake. I wish that had made it as a story.
Opps. :)
http://www.hardocp.com/article.html?art=NTYw
"AMD did also verify that the "new" core is 193mm2 in die size. This is the same size as specified on the original 1MB L2 Athlon 64 cores as noted here back in September of this year. "
I remember this back in the PentiumII/Celeron days, but does this happen anymore? Has anybody looked at the chip to see if it is any smaller or the same size as its bigger brother?
PCI Express will help, but it is still slow. It only is faster if you can send a large batch for processing to the VPU and then later read it back.
http://www.digitimes.com/NewsShow/Article.asp?date Publish=2003/12/19&pages=A7&seq=47
I don't know when the deltachrome will be on the market, but it looks like ATI and nvidia will have some new cards on the market possibly by April which will push the price of the 5900 and 9800 way down, which will in turn push the price of the 5700 and 9600 down which is going to put some serious pressure on everybody else.
I see XGI's Volari as the biggest compitition to S3's DeltaChrome.
It is just as easy to cheat... Err... include application specific optimizations in applications like Quack3.M TEx
http://www.hardocp.com/article.html?art=
ATI's drivers aren't completely mature, but they're pretty good IMO. And ATI has been very good about fixing problems *that are reported to them*.
Please get the new 3.10 drivers released this week and see if you have any problems at all. If you find any problems head to http://apps.ati.com/driverfeedback/index.asp and give them feedback. I'd be surprised if the problems you report aren't all totally fixed by 3.11 or 3.12.
You made the correct choice. The kyro is missing T&L hardware so your poor Athlon would have to process every single triangle instead of simply handing entire objects off to the video card for processing. Also the Kyro is missing cube map support and is stuck at OpenGL 1.2 while your old GF2 supports OpenGL 1.4 (although shadows are emulated and not real time) and will probably soon support OpenGL 1.5.
I would never take a 1st generation challanger brand myself.
If I were to buy a card today it'd probably be a 5700 or maybe a 9600.
The question is how long will it take? How about all the 3dfx owners that never got decent drivers? Did Matrox ever get around to decent OpenGL drivers for the G400 or did they just have their Quake OpenGL->d3d wrapper? Did SiS ever give good drivers for their Xabre? Did Trident ever release good drivers for its products? Are the Kyro drivers good enough to run all applications? Only recently has ATI started producing stable drivers, and even then 7x00 users seem to be experiencing problems sometimes still.
:P
What if you get card X with bad drivers, but it takes 11/2 years for good drivers? By the time you finally get decent drivers your card is obsolete. Why not get a card you know has good drivers and then when brand Y produces good drivers buy their card?
IMO if you buy a card with bad drivers because it gives you a slightly better price/performance ratio and its bad drivers *might* get fixed you are an idiot
I hope S3 gets their drivers fixed soon and they become the 3rd most popular 3d company, but until this happens I have to tell everybody to stick with nvidia or ATI hardware.
It was intended to be a great 3D card:
-Four textures sampled per cycle. The Radeon 8500/GeForce4Ti can apply four texture per pass like the Parhelia, but they require two cycles instead of one. The Radeon 9000/9200 require four cycles instead of just one.
-Giga colour mode: This reduces banding. Check out Homeworld2 on Giga colour vs normal colour.
-Pixel Shader 1.3 just like the GeForce4.
-Displacement mapping. I don't even know if any other cards on the market support this wonderful feature.
-256bit memory bus.
-up to 256megs of memory.
-IIRC support for DX9 level vertex shaders was planned making this card partially DX9 compatible.
The old Matrox Millenium was a great 2D card. The Parhelia is a decent 3D card. It was supposed to be a GeForce3/4Ti/Radeon8500 killer and in many ways it was.
The Parhelia was intended to be a high end gamer's dream card. It just fell a bit short. Hopefully they get to try again.
Opps, that was supposed to be in POT, not HTML.
... ... ...
This is what was missing:
BTW currently no graphics hardware exists which can emulate the standard T&L pipeline with eight lights! What would be a simple loop like this:
for( i = 0; i 8; ++i)
{
if(!light[i].enabled)
continue;
switch( light[i].type)
{
case directional:
break;
case point:
break;
case spotlight:
break;
}
}
is impossible. One program needs to be written for every permutation of different lighting settings which is actually 65536 (4^8) different programs! Only many of those programs are impossible to write because you only have 128 intructions (or 256 on Radeon 9500+ and GeForceFX). By the time you calculate specular and colour for a point light (or worse, spot light) you've already used a ton of instructions. You simply can not have eight point lights on current cards. Of course to make vertex programs useful you really need to use other instructions for things other than calculating lighting such skinning of characters.
Opps, that was supposed to be in POT, not HTML.
This is what was missing:
BTW currently no graphics hardware exists which can emulate the standard T&L pipeline with eight lights! What would be a simple loop like this:
for( i = 0; i
ATI can execute up to 96 ALU instructions in a fragment progam on their top of the line video card. That is actually 64 native instructions. The GeForceFX can handle 1024. Think about that for a second. Lets say you want to compute 1,000,000 things in parallel so you create a 1000x1000 floating point texture to render to. You have to fit your algorithm into just 64 instructions. And you can't make function calls (Yes, DX9 HLSL has functions, but they are all "inline"). And your algorithm has to have no loops. And your algorithm has to have essentially no branches. Well it can have "branches" that are of the form a = x >= y ? b : c; No real branches. And you can't have static or global variables. Each pixel is executed by the same 64 instruction program. If you want a static variable to save between different programs you have to store it to another 1000x1000 texture. Which is what this system does. If you want an if(whatever){ bunch of instructions } you have to compute all those instructions and then multiply by 0 or 1 at the end in order to emulate branching.. Terribly ineffecient. BTW currently no graphics hardware exists which can emulate the standard T&L pipeline with eight lights! What would be a simple loop like this: for( i = 0; i We need far more flexible video cards before this is useful. ATI's ASHLI is sort of like this. It allows you to compile long Renderman or OpenGL shaders into multiple passes so they can be execute on our current hardware like the crappy ;) 9800XT
http://www.ati.com/developer/ashli.html
This will be a lot more interesting when we get pixel shaders that can do emulation of the standard T&L pipeline.