6800 / AMD / IBM Speculation?

[jrphoenix]

I was attempting to step back and make something out of all of the new technology that has come to light / will come to light. Here is some of the latest information:

AMD's new 90 nm process is coming with the assistance of IBM.
There has been news that all of IBM capacity for 90 nm chips is being used to assist AMD?
IBM is fabing the new Nvidia chips.
Nvidia make motherboards for AMD none for Intel.
Nvidia's new chips take video encoding away from the CPU to the GPU (or is it VPU).

Intel is struggling with their 90 nm process and really only comes out in front on the processing wars in video and audio encoding / decoding. With the encoding / decoding going to the graphics card.... AMD could beat or at least be even with Intel in almost every aspect.

AMD did great in its latest financial reports but has bled quite a bit in the past. Do you see this as a possible "proxy" war between Intel and these others? Hmmm? If IBM ever wanted to get back into the chip wars they could buy AMD and buy a big stake in Nvidia...

Anyway, just some thoughts.

 

[Shyatic]

Me likey AMD.

 

[Regs]

The current CEO of AMD is too passionate about his technology to sell off his company. But an hostile take over could be a possibility if the stock owners wanted to jeopardize their own pipe line.

 

[MDE]

If you can't beat Intel by yourself, get a couple of buddies to help.

 

[JBT]

Originally posted by: MDE
If you can't beat Intel by yourself, get a couple of buddies to help.

sounds good.

 

[jrphoenix]

I agree this "triumphrant" is shaping up to be quite an interesting battle.... I'm pulling for AMD, they're our only hope of keeping prices low.

 

[AIWGuru]

There has been news that all of IBM capacity for 90 nm chips is being used to assist AMD?

Well, first of all, that's wrong.
MOST of IBM .09u fab is directed towards the PPC 970 (G5) which they are already late in delivering to Apple.

IBM also worked with AMD on SOI.

IBM has its fingers in everything lately including the cell processor (Playstation 3 and other things) nintendo's console processor, and the processor for the Xbox.

 

[CTho9305]

Originally posted by: jrphoenix
I was attempting to step back and make something out of all of the new technology that has come to light / will come to light. Here is some of the latest information:

AMD's new 90 nm process is coming with the assistance of IBM.
There has been news that all of IBM capacity for 90 nm chips is being used to assist AMD?

I thought AMD fabbed 100% of their CPUs in Dresden, and 100% of their flash in Austin, and don't make much other high volume stuff. What's your source for that?

 

[Cerb]

IBM and AMD have had a manufacturing relationship for awhile now, but now AMD is doing things right, and needs a bit more help.
Last I heard, NVidia went back to TSMC, keeping IBM as an open option.
NVidia could make them for Intel CPUs, but I don't think they want the hassle, or Intel doesn't want them, since they could actually compete very well w/ IEG, and offer pretty good sound for big OEM boxes. I haven't seen anything concrete about why they don't do P4 ones.
Encoding/decoding...they have an edge because they have the market. W/o P4-specific enhancements, they wouldn't do better than the Athlons, and even with that, the Athlon FXs are giving a great showing. Not bad considering Intel also does the compiler .
Video encoding isn't going to the GPU too fast. Decoding, yes, but encoding will take a long time, unless somebody manages some DirectX features that allow hybrid work, going from full GPU encoding--including filters and such--to full software encoding, in the same executables. It may happen (it will get most of the way there, of that you can be certain), but not quickly.

AMD has had some executive changes in the last 3-4 years. My guess is that AMD doing well is probably because of that. CPUs really help, but that's not their main thing, and even losing money on CPUs they could do well w/ flash and get by just fine.

With the current situation, IBM and AMD have a mutually beneficial relationship, and I doubt AMD would allow themselves to be bought any time soon. 30 years, most in debt, in a tech company, and plodding along...nah. They must be able to attract people with a certain mentality to keep up like they have. And with the A64s, I think AMD has had somebody, or bodies, with a good mind for getting things done well and timely...it has been awhile since they were able to keep so dead on to their roadmaps for CPU speeds, and doing so with relatively low OCing capability (so the CPUs are being released when they can get enough at that speed).

Also, IIRC, IBM hasn't done any manufacturing for AMD in a along time (but didn't they in the 386 days?), and are having problems with their strained silicon 90nm process for the PPC...trying to make it to 3GHz by year end.

 

[AIWGuru]

NVidia could make them for Intel CPUs, but I don't think they want the hassle, or Intel doesn't want them

It's a dispute over licencing the P4 bus.

Decoding, yes, but encoding will take a long time, unless somebody manages some DirectX features that allow hybrid work,

Live under a rock? NV40 (released yesterday) supports hardware mpeg 2, 4, wmv encoding. It intercepts directx calls and does all the work itself. Software support is not required. The driver handles the intercepts.

 

[Cerb]

Originally posted by: AIWGuru

NVidia could make them for Intel CPUs, but I don't think they want the hassle, or Intel doesn't want them

It's a dispute over licencing the P4 bus.

So it's the hassle.

Decoding, yes, but encoding will take a long time, unless somebody manages some DirectX features that allow hybrid work,

Live under a rock? NV40 (released yesterday) supports hardware mpeg 2, 4, wmv encoding. It intercepts directx calls and does all the work itself. Software support is not required. The driver handles the intercepts.

Aparently so, I never saw anything about it just doing the encoding on its lonesome already. Any MPEG creation tests to prove it anywhere? And how much software uses DirectX to get most of the job done, instead of say, actually telling it to do specific work (IE, if they do x * y - 2 / 4, where does the video card come in?)

 

[AIWGuru]

Originally posted by: Cerb

Originally posted by: AIWGuru

NVidia could make them for Intel CPUs, but I don't think they want the hassle, or Intel doesn't want them

It's a dispute over licencing the P4 bus.

Decoding, yes, but encoding will take a long time, unless somebody manages some DirectX features that allow hybrid work,

Live under a rock? NV40 (released yesterday) supports hardware mpeg 2, 4, wmv encoding. It intercepts directx calls and does all the work itself. Software support is not required. The driver handles the intercepts.

Aparently so, I never saw anything about it just doing the encoding on its lonesome already. Any MPEG creation tests to prove it anywhere?

The current 60.x BETA drivers haven't added support yet but it's implemented in hardware.
What you're saying is unlikely to happen for years is here already. The drivers should have the functionality enabled by shipping.
Also, you can just buy a hardware mpeg encoder and pop it into a PCI slot. Of course, this requires software support.
NV40 (and possibly R420) are transparent.
If you want to know the details, read a few reviews.

 

[AIWGuru]

Looks like the WMV9 support is decode only but mpg is encode, my bad:
"High Quality Video
Motion Adaptive De-interlacing
High quality scaling & filtering
Video de-blocking
Integrated TV-encoder
Complete HDTV Solution
Transport stream handling
HDTV Output (720p, 1080i,
480p, CGMS)
Complete PVR Solution
Hardware Audio/Video
Synchronization
MPEG 1/2/4 encode/decode
WMV9 decode acceleration
HDCP Support "

No changes to any program or video files are necessary to take advantage of the acceleration decoding features, as the driver intercepts all DirectX calls and forwards them directly to the video processor.

 

[miniMUNCH]

Originally posted by: Regs
The current CEO of AMD is too passionate about his technology to sell off his company. But an hostile take over could be a possibility if the stock owners wanted to jeopardize their own pipe line.

A CEO really can't do that much to stop their company from being sold...if IBM or someone else wants to pay enough for AMD, the board directors has the responsibility to do whats best for stock-holders and take the deal.

Also, the CEO has plenty to profit from the deal...he too gets the benefit of the poison pill which is often really tough to pass up.

 

[Cerb]

Originally posted by: AIWGuru

Originally posted by: Cerb

Originally posted by: AIWGuru

NVidia could make them for Intel CPUs, but I don't think they want the hassle, or Intel doesn't want them

It's a dispute over licencing the P4 bus.

Decoding, yes, but encoding will take a long time, unless somebody manages some DirectX features that allow hybrid work,

Live under a rock? NV40 (released yesterday) supports hardware mpeg 2, 4, wmv encoding. It intercepts directx calls and does all the work itself. Software support is not required. The driver handles the intercepts.

Aparently so, I never saw anything about it just doing the encoding on its lonesome already. Any MPEG creation tests to prove it anywhere?

The current 60.x BETA drivers haven't added support yet but it's implemented in hardware.
What you're saying is unlikely to happen for years is here already. The drivers should have the functionality enabled by shipping.
Also, you can just buy a hardware mpeg encoder and pop it into a PCI slot. Of course, this requires software support.
NV40 (and possibly R420) are transparent.
If you want to know the details, read a few reviews.

All I saw in any reviews was that it had the hardware support and would have driver support. The GF3 had all the great DX 8 stuff, and look at how long it took for that to really get used.
Every tuner and encoder card I've used has had special software to do the job, and haven't seen anything to make this any different. If I get a VIVO 6800, and use the S-video in, what software (either out right this instant or on the drivers CD) will use these 'transparent' features to give me a MPEG2 video of it?

 

[AIWGuru]

Originally posted by: Cerb

Originally posted by: AIWGuru

Originally posted by: Cerb

Originally posted by: AIWGuru

NVidia could make them for Intel CPUs, but I don't think they want the hassle, or Intel doesn't want them

It's a dispute over licencing the P4 bus.

Decoding, yes, but encoding will take a long time, unless somebody manages some DirectX features that allow hybrid work,

Live under a rock? NV40 (released yesterday) supports hardware mpeg 2, 4, wmv encoding. It intercepts directx calls and does all the work itself. Software support is not required. The driver handles the intercepts.

Aparently so, I never saw anything about it just doing the encoding on its lonesome already. Any MPEG creation tests to prove it anywhere?

The current 60.x BETA drivers haven't added support yet but it's implemented in hardware.
What you're saying is unlikely to happen for years is here already. The drivers should have the functionality enabled by shipping.
Also, you can just buy a hardware mpeg encoder and pop it into a PCI slot. Of course, this requires software support.
NV40 (and possibly R420) are transparent.
If you want to know the details, read a few reviews.

All I saw in any reviews was that it had the hardware support and would have driver support. The GF3 had all the great DX 8 stuff, and look at how long it took for that to really get used.
Every tuner and encoder card I've used has had special software to do the job, and haven't seen anything to make this any different. If I get a VIVO 6800, and use the S-video in, what software will use these 'transparent' features to give me a MPEG2 video of it?

ALL SOFTWARE. All software the do encoding anyway. That's what transparent means. As I just quoted from that review (the first random one I went to) "No changes to any program or video files are necessary to take advantage of the acceleration decoding features, as the driver intercepts all DirectX calls and forwards them directly to the video processor.
What this means is that as far as any program doing any encoding is concerned, the processor is doing the work. That's what transparent means. No program support is necessary, unlike your example of GF3 shader support where each program must be written for it.

 

[Cerb]

Originally posted by: AIWGuru

Originally posted by: Cerb

Originally posted by: AIWGuru

Originally posted by: Cerb

Originally posted by: AIWGuru

NVidia could make them for Intel CPUs, but I don't think they want the hassle, or Intel doesn't want them

It's a dispute over licencing the P4 bus.

Decoding, yes, but encoding will take a long time, unless somebody manages some DirectX features that allow hybrid work,

Live under a rock? NV40 (released yesterday) supports hardware mpeg 2, 4, wmv encoding. It intercepts directx calls and does all the work itself. Software support is not required. The driver handles the intercepts.

Aparently so, I never saw anything about it just doing the encoding on its lonesome already. Any MPEG creation tests to prove it anywhere?

The current 60.x BETA drivers haven't added support yet but it's implemented in hardware.
What you're saying is unlikely to happen for years is here already. The drivers should have the functionality enabled by shipping.
Also, you can just buy a hardware mpeg encoder and pop it into a PCI slot. Of course, this requires software support.
NV40 (and possibly R420) are transparent.
If you want to know the details, read a few reviews.

All I saw in any reviews was that it had the hardware support and would have driver support. The GF3 had all the great DX 8 stuff, and look at how long it took for that to really get used.
Every tuner and encoder card I've used has had special software to do the job, and haven't seen anything to make this any different. If I get a VIVO 6800, and use the S-video in, what software will use these 'transparent' features to give me a MPEG2 video of it?

ALL SOFTWARE. All software the do encoding anyway. That's what transparent means. As I just quoted from that review (the first random one I went to) "No changes to any program or video files are necessary to take advantage of the acceleration decoding features, as the driver intercepts all DirectX calls and forwards them directly to the video processor.
What this means is that as far as any program doing any encoding is concerned, the processor is doing the work. That's what transparent means. No program support is necessary, unlike your example of GF3 shader support where each program must be written for it.

So every single video recording application uses directx calls for all the MPEG encoding? So if I get the 6800, DVDshrink uses significantly less CPU? Ulead's stuff as well? etc.?
All software doesn't do everything the same way, or use the same features. If it uses those DirectX calls, the software has to call them first. If 90% or more do, then great. The question is, do they?
Certainly they will, but do they right now, with 6-12 month old software (IE, what people already have or what is on store shelves)/

 

[AIWGuru]

Originally posted by: Cerb
So every single video recording application uses directx calls for all the MPEG encoding? So if I get the 6800, DVDshrink uses significantly less CPU? Ulead's stuff as well? etc.?

Yes, all programs that do encoding would have to make a directx call. Directx covers everything from mouse/keyboard input to what's drawn on your monitor, to sound to devices talking to one another. Hence the X.
Direct sound/direct draw/direct 3d etc etc.
We'll have to see how it pans out in the end since not many people actually own one of these cards but one of its features is transparent hardware encoding.
This is a natural evolution since, with 4x the transistors of a P4 and unlimited branch instructions (up from 65k) (PS3.0)modern GPUs are basically co-processors.
I think I read that the pipes being used for this encoding aren't dedicated and can be programmed for any old task.
This is why PCI-express will be big.
AGP 8x may have tons of bandwidth going to the GPU which is good for gaming but it has only 233mb/sec upstream (and only when nothing is going downstream) which is lame for applications such as this.

edit: re your question about what percentage: the language of the quote does say "no changes to any program or video file..."

Take it for what you will...

 

[Cerb]

Originally posted by: AIWGuru

Originally posted by: Cerb
So every single video recording application uses directx calls for all the MPEG encoding? So if I get the 6800, DVDshrink uses significantly less CPU? Ulead's stuff as well? etc.?

Yes, all programs that do encoding would have to make a directx call. Directx covers everything from mouse/keyboard input to what's drawn on your monitor, to sound to devices talking to one another. Hence the X.
Direct sound/direct draw/direct 3d etc etc.
We'll have to see how it pans out in the end since not many people actually own one of these cards but one of its features is transparent hardware encoding.

If the include it on the $150 cards, better yet the $70 cards, and in some degree on the IGP version (say, 2 years from now?), anyone who wants it can easily have it.

This is a natural evolution since, with 4x the transistors of a P4 and unlimited branch instructions (up from 65k) modern GPUs are basically co-processors.

You mean 76% more (unless I missed their 500m transistor GPU somewhere...).

I think I read that the pipes being used for this encoding aren't dedicated and can be programmed for any old task.

With the shaders going the way they have been, this was pretty certain. NVidia is making up for the FX 5x00 with a veangence.

This is why PCI-express will be big.
AGP 8x may have tons of bandwidth going to the GPU which is good for gaming but it has only 233mb/sec upstream which is lame for applications such as this.

 

[AIWGuru]

NV40 is basically one huge-ass SIMD processor. With its 16 pipelines and special SIMD instructions, it should whoop the holy hell outta any general purpose CPU at video encoding (which is strictly SIMD.)

 

[AIWGuru]

You mean 76% more (unless I missed their 500m transistor GPU somewhere...).

No...
P4northwood - 55million transistors
NV40 - 220million transistors
That's 4 times.
Me thinks you need to go back to math class

 

[CTho9305]

Originally posted by: AIWGuru

Originally posted by: Cerb
So every single video recording application uses directx calls for all the MPEG encoding? So if I get the 6800, DVDshrink uses significantly less CPU? Ulead's stuff as well? etc.?

Yes, all programs that do encoding would have to make a directx call. Directx covers everything from mouse/keyboard input to what's drawn on your monitor, to sound to devices talking to one another. Hence the X.
Direct sound/direct draw/direct 3d etc etc.
We'll have to see how it pans out in the end since not many people actually own one of these cards but one of its features is transparent hardware encoding.
This is a natural evolution since, with 4x the transistors of a P4 and unlimited branch instructions (up from 65k) (PS3.0)modern GPUs are basically co-processors.
I think I read that the pipes being used for this encoding aren't dedicated and can be programmed for any old task.
This is why PCI-express will be big.
AGP 8x may have tons of bandwidth going to the GPU which is good for gaming but it has only 233mb/sec upstream (and only when nothing is going downstream) which is lame for applications such as this.

edit: re your question about what percentage: the language of the quote does say "no changes to any program or video file..."

Take it for what you will...

Unless DirectX actually provides a "encodeDivxVideo()" and "encodeXvidVideo()" and "encodeQuicktimeVideo()" and ____ and _____ (i.e. every possible codec), directx WOULDN'T be involved in the actual encoding. Accelerating the functions you listed (talking to IO devicess) won't help for pure number crunching.

edit: What I'm saying is, if I hand wrote an MPEG4 encoder without calling any directx functions, there's nothing the driver could do to offload work to the video card.
edit: I just looked through the xvid source... no mention of "directx" anywhere. the only mention of "windows" was an include of <windows.h>. There were plently of references to dx, but ignoring the references to the x86 "EDX" register, there didn't seem to be anything directx-releated. Thus directx would not affect the performance of the encoder (or even decoder, I'm thinking)

 

[AIWGuru]

Originally posted by: CTho9305

Originally posted by: AIWGuru

Originally posted by: Cerb
So every single video recording application uses directx calls for all the MPEG encoding? So if I get the 6800, DVDshrink uses significantly less CPU? Ulead's stuff as well? etc.?

Yes, all programs that do encoding would have to make a directx call. Directx covers everything from mouse/keyboard input to what's drawn on your monitor, to sound to devices talking to one another. Hence the X.
Direct sound/direct draw/direct 3d etc etc.
We'll have to see how it pans out in the end since not many people actually own one of these cards but one of its features is transparent hardware encoding.
This is a natural evolution since, with 4x the transistors of a P4 and unlimited branch instructions (up from 65k) (PS3.0)modern GPUs are basically co-processors.
I think I read that the pipes being used for this encoding aren't dedicated and can be programmed for any old task.
This is why PCI-express will be big.
AGP 8x may have tons of bandwidth going to the GPU which is good for gaming but it has only 233mb/sec upstream (and only when nothing is going downstream) which is lame for applications such as this.

edit: re your question about what percentage: the language of the quote does say "no changes to any program or video file..."

Take it for what you will...

Unless DirectX actually provides a "encodeDivxVideo()" and "encodeXvidVideo()" and "encodeQuicktimeVideo()" and ____ and _____ (i.e. every possible codec), directx WOULDN'T be involved in the actual encoding. Accelerating the functions you listed (talking to IO devicess) won't help for pure number crunching.

edit: What I'm saying is, if I hand wrote an MPEG4 encoder without calling any directx functions, there's nothing the driver could do to offload work to the video card.

none of those codec that you listed are supported, I believe.
but your edit is correct. We're assuming (which is all we can do at this point) that commerical programs will not have this problem, however and that they will use a plugin to do their encoding.

 

[ingenuiti]

IBM has positioned its future in services not hardware.

 

[Cerb]

Originally posted by: AIWGuru

You mean 76% more (unless I missed their 500m transistor GPU somewhere...).

No...
P4northwood-55million transistors
NV40-220million transistors
That's 4 times.
Me thinks you need to go back to math class

P4 Prescott: 125m transistors. +76%
A64: 106m transistors. +108%