[Groo The Wanderer]

Originally Posted by Razor1

if you want me to be specific, nV's and AMD's wafer purchasing isn't the same, I know that nV's purchasing model isn't based on buying wafers, which if you didn't get the hint in my second post to you, well......

I'm going to be blatent there is no truth to your article on profit's per chip/card.

BTW I don't know why you pointed to your bump gate article, (self promotion )

Chipset article, since you wrote that article its been 1 year before anouncement that they are still in legal proceedings, they aren't jumping off the boat, Intel and nV havn't gotten to an agreement, quite different then what your article has stated.

The wafers are probably sold at a different price, mainly due to processing steps required, not volume. That said, it isn't enough to change the price substantially, and not enough to move the price of a good die by much more than rounding error. What's your point?

I know SOMETIMES NV's model isn't based on wafer cost, which is why I PRINTED THAT AND THE YIELD BREAKPOINT here.
http://www.semiaccurate.com/2009/08/...kes-huge-risk/
Then again, you can't troll as well if you read the articles. Facts have this nasty way of scattering the roaches in forums.

To sum up, you agree with most of my point, most of my numbers, but jump up and down saying I am wrong. Then you refuse to provide any numbers of your own. Wow, wonderful argument there!

To quote you from the above, "I'm going to be blatent [SIC] there is no truth to your article on profit's per chip/card." OK, there are two ways to solve this issue, one involves numbers, the other does not. They both involve the word 'up'. I will repeat my earlier statement, put up or shut up. You have only posted one vague number that agreed with me.

As for the bumpgate article, I was trying to point out to you that I do actually get the science behind this. You posted a long list of things that affect wafer cost instead of answer the questions posed, so I pointed out that I do in fact understand those. The link was to point you to an example. I am starting to think you don't read very much.

To belabor that point, you seem to be one of those people that wait for PR announcements before believing something. I keep forgetting that you don't actually go out and talk to the people doing the work and making the products. My bad.

Let me explain it a little more slowly. The article was written in August 2008. The suit was filed on Feb 18, 2009. I will assume you have read it fully before commenting on it like I have, otherwise you would look pretty stupid and be trolling.

To refresh your memory, in the filing on P8 para 18, it says, "In early 2007, Intel informed NVIDIA that it planned to introduce Nehalem architecture processors in 2008." It goes on on para 19 to say, "A series of discussions ensued between Intel and NVIDIA as to whether Disputed NVIDIA MCPs are licensed under the Agreements. It was, and is, Intel's position that the Disputed NVIDIA MCPs are not licensed under the CLA because they cannot provide an interface between an Intel processor and system memory."

There you have it, it started in early 2007, and was very public a year later. Maybe in your world disputes start after the court filing, but not in mine. In my world, court filings only happen after a dispute can not be solved through negotiation or other means.

In this case, it was very clear, and very public over a year before you noticed. It was one of the points, along with Nvidia canceling the development of future products and reassigning engineers, that clued me in. Actually, you might say that the licensing issue was minor in comparison to them ending the programs.

Then again, you seem to be the type that listens to EVERY word that PR says, and takes it as gospel, so keep on believing that they JUST made that decision. And they are, giggle, using those, heh heh, resources to.... heh... no.... pain in side from laughter.... to.... to.... work on the SLI licensing program for Intel chipsets. Bwahahahaha! I wish I had that link handy, it will give you another talking point to explain how this could have happened. Bwahahahaha!

-Charlie

[Razor1]

lol now we are talking about risk wafers? BTW I wasn't taking about Fermi or risk wafers.

Oh and that article has a lot of holes too. Hot lots aren't strickly used for yield, not to mention according to that article if they did get a return of such low chips and they were doing a yield analysis, they wouldn't even have done it. BTW do you know the equation's used when doing hot lots for yield analysis? (you do realize even if nV got X number of chips from the hot lots, as you stated they would have to calculate risk by using using a formula) And yes I do have yield numbers for Fermi from about the time you wrote that article its more like 15% for fully funtional chips and it goes up to around double that with salvage parts. Keep pulling up your old articles to make yourself look good, just doesn't work its easy to cut them down like swiss cheese because the factual information you have is castrated by your fanatical ramblings which make no sense.

The bump gate article, I did give you credit for that in the past, but I don't believe you wrote that yourself. You would have needed alot of help and alot is an understatement.

The license agreement between Intel and nV which was done 5 years ago, Intel wanted to renegotiate the deal with nV, nV didn't want to, and thats when Intel filed for breach. TO REFRESH YOUR MEMORY if you didn't know that would happen before Feb of 2008 which would have been impossible to know outside of said people, or are you just omniscient, possibly you knew it was going to happen 7 months before it actually happened. The negotiations didn't go sour till oh end of 2008 early 2009. Negotiations were know publicly not the course of action by either company till Feb of 2009.

[DegustatoR]

Originally Posted by aaronspink

A linking to BSOBS as a credible site while at the same time disparaging any other author or side is humorous as best.

Have you actually read that link or is it again the same thing as with all of NV's bashers -- just assume a bunch of stuff and you're done?
Hint: Theo's providing a link to Develop magazine. I guess it's "humorous as best" too.

[Demirug]

Originally Posted by DegustatoR

But if Battleforge is any indication they'll probably do it in such a way that only DX11 GPUs will benefit from added quality while most of what will be added is probably going to be possible to implement on DX11 feature level 10 h/w (that's mostly NV's today's h/w). I wonder what the general reaction to such an event will be. Battleforge ported to DX11 ignoring all the DX11 stuff for DX10 h/w went largely unnoticed somehow but NV's AA implementation in BAA turned into a shitstorm of stupidity.

There are only two new DX11 features for DX10 hardware.
1. Multithread render context support. In BattleForge there was no use for this at all. The engine already has an API independent multithread render support.
2. Compute Shader. It was my personal decision to not use CS 4.xright now. There were two primary reasons for this:
a. The SDK sucks when it comes to CS 4.x. No documentation at all. As our AO shader goes on the limits of CS 5.0 it is not funny at all to find all the CS 4.x limitations that block us from doing a CS 4.x version by playing trial and error with the shader compiler.
b. CS 4.x support was not enabled by default in the nvidia drivers at the time we had to decide what to do. This was a clear sign for us that the CS 4.x support isn’t ready for prime time.

[DegustatoR]

Originally Posted by Demirug

There are only two new DX11 features for DX10 hardware.

Yes and there is only one beyond that for DX11 hardware -- tesselation (I know about new BCs and other stuff but they are smaller and less important features). And you're not using tesselation in BF DX11 at all. That's why I find it very strange that you've decided not to use FL10 h/w since you're basically not using FL11 h/w's features.

Originally Posted by Demirug

1. Multithread render context support. In BattleForge there was no use for this at all. The engine already has an API independent multithread render support.

How can it be API independent? Are you sure that you're not mixing two different things here -- engine multithreading and renderer multithreading? AFAIK the second thing depends on API and isn't possible in <DX11.

Originally Posted by Demirug

2. Compute Shader. It was my personal decision to not use CS 4.xright now. There were two primary reasons for this:
a. The SDK sucks when it comes to CS 4.x. No documentation at all. As our AO shader goes on the limits of CS 5.0 it is not funny at all to find all the CS 4.x limitations that block us from doing a CS 4.x version by playing trial and error with the shader compiler.
b. CS 4.x support was not enabled by default in the nvidia drivers at the time we had to decide what to do. This was a clear sign for us that the CS 4.x support isn’t ready for prime time.

Why are we talking about NVIDIA here? AMD has a big bunch of FL10(.1) h/w itself. If they care about their current customers and not about selling their new cards to those customers they should've probably push for CS4(.1) before CS5. I mean, isn't it clear that CS4 will be more widely used in the forseeable future -- even after the whole EG line will be on shelves and NV will release Fermi cards?
As for the SDK -- i'm sure that MS and NV would provide whatever documents you may have needed to implement CS4 support. It's not like this was a feature that couldn't wait another month or two anyway.
Are you planning on implementing CS4 support after Win7 launch, NV and AMD enabling them in their drivers by default and CS4 SDK matures?

[Demirug]

Originally Posted by DegustatoR

How can it be API independent? Are you sure that you're not mixing two different things here -- engine multithreading and renderer multithreading? AFAIK the second thing depends on API and isn't possible in <DX11.

The main job of the new DX11 multithread feature is moving driver overhead to another thread. We have an own solution for this on a higher level. I talked about this with Microsoft and they agree that DX11 could not do improve the performances at this point for us.

Originally Posted by DegustatoR

Why are we talking about NVIDIA here? AMD has a big bunch of FL10(.1) h/w itself. If they care about their current customers and not about selling their new cards to those customers they should've probably push for CS4(.1) before CS5. I mean, isn't it clear that CS4 will be more widely used in the forseeable future -- even after the whole EG line will be on shelves and NV will release Fermi cards?
As for the SDK -- i'm sure that MS and NV would provide whatever documents you may have needed to implement CS4 support. It's not like this was a feature that couldn't wait another month or two anyway.
Are you planning on implementing CS4 support after Win7 launch, NV and AMD enabling them in their drivers by default and CS4 SDK matures?

CS 4.x support on older hardware has no priority for AMD. This makes CS 4.x an nvidia only feature for quite some time.

As a DirectX MVP I have access to documentation beyond the SDK. But this doesn’t help either. The whole CS 4.x stuff was a late feature that wasn’t in the original feature set planned for DX11. Therefore it is natural that the documentation is not ready yet.

I am not sure if CS 4.x will be widely used. CS 5.0 is significant better and in most cases you need a pixelshader fallback. CS 4.x is the middle child that get’s no love.

We might add CS 4.x support later but we already know that the performances would be worse than 5.0 because of the 4.x limitations.

[aaronspink]

Originally Posted by DegustatoR

Have you actually read that link or is it again the same thing as with all of NV's bashers -- just assume a bunch of stuff and you're done?
Hint: Theo's providing a link to Develop magazine. I guess it's "humorous as best" too.

Yeah, I read the link. Its a random non-scientific survey. Woopity doo da. As far as shear number of different games, Physx is probably in the lead., as far as number of games sold, Havoc is probably fairly comfortably in the lead.

[DegustatoR]

Originally Posted by aaronspink

Yeah, I read the link. Its a random non-scientific survey. Woopity doo da. As far as shear number of different games, Physx is probably in the lead., as far as number of games sold, Havoc is probably fairly comfortably in the lead.

Havok is exactly the same thing as PhysX. LRB just not out yet.

[Jaaanosik]

Originally Posted by Demirug

The main job of the new DX11 multithread feature is moving driver overhead to another thread. We have an own solution for this on a higher level. I talked about this with Microsoft and they agree that DX11 could not do improve the performances at this point for us.

...

I am just curious. How much is the driver overhead when compared to the whole CPU load of a game?
Thanks...

[Groo The Wanderer]

Originally Posted by Razor1

Oh and that article has a lot of holes too. Hot lots aren't strickly used for yield, not to mention according to that article if they did get a return of such low chips and they were doing a yield analysis, they wouldn't even have done it. BTW do you know the equation's used when doing hot lots for yield analysis? (you do realize even if nV got X number of chips from the hot lots, as you stated they would have to calculate risk by using using a formula) And yes I do have yield numbers for Fermi from about the time you wrote that article its more like 15% for fully funtional chips and it goes up to around double that with salvage parts. Keep pulling up your old articles to make yourself look good, just doesn't work its easy to cut them down like swiss cheese because the factual information you have is castrated by your fanatical ramblings which make no sense.

Do I know the equations? No. Do I do that for a living? No. Do I have at least 3 people who do in my address book that don't mind if I call and ask questions? Yes.

Originally Posted by Razor1

The bump gate article, I did give you credit for that in the past, but I don't believe you wrote that yourself. You would have needed alot of help and alot is an understatement.

I started out in college in chemical engineering, then went to chemistry, then biology, and ended up in genetics before I dropped out (ADHD makes it hard to study when you are bored with a topic). I realized that I was making more in computers then than I would ever in biology, so why bother? That said, I took a fair number of matsci courses, tons of csci, and a lot of related science courses. There isn't much difference between a reading a gas/liquid chromatograph and xray materials analysis.

Did I do the actual sampling and reading? No, I don't have the equipment. Are there any of 20 teardown houses within an hour of your HQ that do? Sure. Are some of them bored? Sure. Do the math, pun intended.

I wrote it 100% myself, researched it over 2 months or so, and had to up my cell time from 1500 to unlimited after burning through it 3 months straight on the topic. I did have help for things I don't have the lab for. I do have a lot of friends in the valley though, and they have all the toys I need.

Originally Posted by Razor1

The license agreement between Intel and nV which was done 5 years ago, Intel wanted to renegotiate the deal with nV, nV didn't want to, and thats when Intel filed for breach. TO REFRESH YOUR MEMORY if you didn't know that would happen before Feb of 2008 which would have been impossible to know outside of said people, or are you just omniscient, possibly you knew it was going to happen 7 months before it actually happened. The negotiations didn't go sour till oh end of 2008 early 2009. Negotiations were know publicly not the course of action by either company till Feb of 2009.

I am pretty sure (but I didn't go dig up my notes), that NV was bitching about chipsets and licensing during the 2007 fall IDF, which would have been a year before I wrote. If I am a year off on this, sorry, but there were public hints long long before it was an open war, you just had to have your ear to the ground. I did.

They were shouting to the press about how they were licensed, and how teh aswum their stuff would be. They were breaking NDAs left and right telling every analyst who would listen, and the occasional tame journalist, what Intel chips were coming, and what busses they used.

They were mentioning Bloomfield and Lynnfield by name, and saying CSI and DMI before the NDAs went up, long before. This was done, I think it was Evendon FWIW, to try and sway analysts to pump up the stock. The analysts promptly went to Intel and asked them the questions, and Intel responded.

I can guarantee you that Intel was NOT pleased about the leaks. If anything, the problem became acute, and the negotiations soured because of Nvidia's moronic insistence to try and publicly up the stakes. Intel doesn't take well to being threatened, nor do they take well to people attempting to hurt their image for monetary gain. Think "whoop-ass".

NV shot themselves in the foot, and pissed Intel off. Intel had every right to be annoyed at NV's childish behavior. Then again, you can't accuse Nvidia of being overly bright, trying to be a bully from a position of weakness isn't the smartest thing in the world.

Nvidia has one tool for relationships, be they internal, external, public or private, a hammer. If they don't like you, they hit you and tell you to step into line or you will be hit again. This didn't work with Intel, and the bright 'strategerists' NV likely sunk the company over this. FWIW, it is more or less the same thing they tried with me, do what we say, or we will go after you. I didn't. Intel didn't. Nvidia doesn't have another tool to use.

-Charlie

[rpg.314]

The chipset business is officially shut down now.

http://arstechnica.com/hardware/news...ng-arrives.ars

The funny bit is their bailing from AMD chipset business. I guess, they told them that there will be no HT3.2 goodness for them.

[Razor1]

Originally Posted by Groo The Wanderer

Do I know the equations? No. Do I do that for a living? No. Do I have at least 3 people who do in my address book that don't mind if I call and ask questions? Yes.

Then why didn't you ask them before you did that article, or add an addedum to your article after all of the sh*t hitting the fan?

I started out in college in chemical engineering, then went to chemistry, then biology, and ended up in genetics before I dropped out (ADHD makes it hard to study when you are bored with a topic). I realized that I was making more in computers then than I would ever in biology, so why bother? That said, I took a fair number of matsci courses, tons of csci, and a lot of related science courses. There isn't much difference between a reading a gas/liquid chromatograph and xray materials analysis.

Cool, never doubted your over all intellegence, just the way your articles read......

Did I do the actual sampling and reading? No, I don't have the equipment. Are there any of 20 teardown houses within an hour of your HQ that do? Sure. Are some of them bored? Sure. Do the math, pun intended.

I wrote it 100% myself, researched it over 2 months or so, and had to up my cell time from 1500 to unlimited after burning through it 3 months straight on the topic. I did have help for things I don't have the lab for. I do have a lot of friends in the valley though, and they have all the toys I need.

No, but as you I do have a lot of connections to information from my past job, which was using GPGPU computing to create an AI for financial models and previous jobs in the game industry.

I am pretty sure (but I didn't go dig up my notes), that NV was bitching about chipsets and licensing during the 2007 fall IDF, which would have been a year before I wrote. If I am a year off on this, sorry, but there were public hints long long before it was an open war, you just had to have your ear to the ground. I did.

They were shouting to the press about how they were licensed, and how teh aswum their stuff would be. They were breaking NDAs left and right telling every analyst who would listen, and the occasional tame journalist, what Intel chips were coming, and what busses they used.

They were mentioning Bloomfield and Lynnfield by name, and saying CSI and DMI before the NDAs went up, long before. This was done, I think it was Evendon FWIW, to try and sway analysts to pump up the stock. The analysts promptly went to Intel and asked them the questions, and Intel responded.

I can guarantee you that Intel was NOT pleased about the leaks. If anything, the problem became acute, and the negotiations soured because of Nvidia's moronic insistence to try and publicly up the stakes. Intel doesn't take well to being threatened, nor do they take well to people attempting to hurt their image for monetary gain. Think "whoop-ass".

NV shot themselves in the foot, and pissed Intel off. Intel had every right to be annoyed at NV's childish behavior. Then again, you can't accuse Nvidia of being overly bright, trying to be a bully from a position of weakness isn't the smartest thing in the world.

Nvidia has one tool for relationships, be they internal, external, public or private, a hammer. If they don't like you, they hit you and tell you to step into line or you will be hit again. This didn't work with Intel, and the bright 'strategerists' NV likely sunk the company over this. FWIW, it is more or less the same thing they tried with me, do what we say, or we will go after you. I didn't. Intel didn't. Nvidia doesn't have another tool to use.

-Charlie

I don't agree with what nV did, personally I don't think it was a smart move to play hardball with Intel, but again its a lot of speculation which I see it as a culmination of events that Intel went down the path they did, not one or two occassions.

[trinibwoy]

Originally Posted by rpg.314

The funny bit is their bailing from AMD chipset business. I guess, they told them that there will be no HT3.2 goodness for them.

I doubt AMD would withhold a license. After all, they're the good guys right and HT is supposedly an open standard unlike Quickpath or DMI. Most likely it's that they realized they had no competitive advantage in that space. AMD's IGPs are pretty good and they have the platform advantage. Nforce was only ever good when it was the best thing out there but fared much worse with real competition. Same thing is happening with ION right now, it's doing well because the alternative is shit.

[mmendez]

Originally Posted by rpg.314

The chipset business is officially shut down now.

http://arstechnica.com/hardware/news...ng-arrives.ars

The funny bit is their bailing from AMD chipset business. I guess, they told them that there will be no HT3.2 goodness for them.

Another one with some revenue numbers:

http://www.internetnews.com/bus-news...0791_3843026_1

It's not like we didn't see it coming.

[Groo The Wanderer]

Originally Posted by Razor1

Then why didn't you ask them before you did that article, or add an addedum to your article after all of the sh*t hitting the fan?

If I needed to for the article, why do you assume I didn't?

-Charlie

[ANova]

Originally Posted by Groo The Wanderer

FWIW, it is more or less the same thing they tried with me, do what we say, or we will go after you. I didn't. Intel didn't. Nvidia doesn't have another tool to use.

I'd like to hear more about this. Nv's escapades are always entertaining.

[Razor1]

Originally Posted by Groo The Wanderer

If I needed to for the article, why do you assume I didn't?

-Charlie

because you put hot lots = yields in your article, which is not the case if they got 7 chips back from 1 hot lot (4 wafers), they could have been working with yields, they could have been tweaking the chip for performance, they could have been tweaking the chip for power consumption, etc. Without knowing exactly what nV was going for, we don't know anything about those hot lots. Each wafer used in the hot lot is a test, and then they change something else to see how that test effects the chip in a certain expected fashion.

[Davros]

Originally Posted by ANova

I'd like to hear more about this. Nv's escapades are always entertaining.

Seconded

[Sxotty]

Originally Posted by Groo The Wanderer

There isn't much difference between a reading a gas/liquid chromatograph and xray materials analysis.

Except the TGA was the important thing anyway...

Originally Posted by rpg.314

The chipset business is officially shut down now.

http://arstechnica.com/hardware/news...ng-arrives.ars

The funny bit is their bailing from AMD chipset business. I guess, they told them that there will be no HT3.2 goodness for them.

From your link

On AMD platforms, we continue to sell a higher quantity of chipsets than AMD itself. MCP61-based platforms continue to be extremely well positioned in the entry CPU segments where AMD CPUs are most competitive vs. Intel

I don't understand why you would link to an article that doesn't really support what you said. It just links to another article that supports what you said, then goes on to say Nvidia denies it. Though I am with Triniboy and don't see how Nvidia will have a competitive advantage in the AMD market.

[rpg.314]

That bs is from nv pr, from the ars article

PC Magazine also reported as an unexplained aside that work on AMD chipsets has also been halted, despite the fact that the Intel DMI issue should have no bearing on NVIDIA's AMD platform plans.

and

NVIDIA has clearly ducked the question about AMD chipset R&D that we included in our initial request for a statement from them, so we followed up again with a point-blank question about the company's R&D plans for AMD chipsets, asking if it was true that the AMD work was also halted. NVIDIA has yet to respond, so we're assuming that the PC Mag report is accurate, and that NVIDIA's chipset "postponement" also applies to the AMD chipsets.

[Groo The Wanderer]

Originally Posted by Razor1

because you put hot lots = yields in your article, which is not the case if they got 7 chips back from 1 hot lot (4 wafers), they could have been working with yields, they could have been tweaking the chip for performance, they could have been tweaking the chip for power consumption, etc. Without knowing exactly what nV was going for, we don't know anything about those hot lots. Each wafer used in the hot lot is a test, and then they change something else to see how that test effects the chip in a certain expected fashion.

Fair enough, but I was pretty specific in saying that it was only this lot, and there was another one due later.

Several people have complained that I used yield in a way that wasn't the same as industry people did, and that is true, but I don't have a better term to describe it to the 99.9% of people reading the article that don't work in the industry.

To me, yield = good dies/candidates. I didn't say that the whole line, and every one after that was going to be the same. So, for what I was doing, I think it was a valid use of the term. Also, I don't care what they were doing, if 7 came out, 7 came out. I didn't speculate about what they were doing because I don't know, didn't ask, and only marginally care.

To me, the problem is that there are certain things you need to do with the initial silicon, especially if your ass is to the wall like it is at NV now. 7 isn't enough to do that, and given their 'puppy' piss-poor showing, as opposed to their normal whoop-ass showing, at GTC, it says they don't have enough for the engineering they need to do. From what I understand, the count of real cards shown off at GTC was zero, but there may have been some behind closed doors things that I wasn't clued in on. The ones I was didn't have real silicon.

Another piece of the puzzle is that they haven't put the respin in yet, or at least they hadn't when I last checked a few days ago. That tells me that the chips is in deep trouble (very likely) and/or they don't have enough silicon to test what they need to test (fairly likely).

The lack of silicon is the problem that I was pointing to. If it was based on a decision, you have to pat management on the back for being that ballsy, I would never have sent my CEO out on stage to look like a total idiot in front of his most loyal patsies. Then again, I don't get the whole NV culture thing either. If a person I employed did that more than once, I would fire them.

-Charlie

[Groo The Wanderer]

Originally Posted by ANova

I'd like to hear more about this. Nv's escapades are always entertaining.

To explain it, I would have to publish a long thread of emails, or at least one summary email. I don't publish emails without permission of the person on the other end, or without extraordinary circumstances. If I can get those permissions, I would be happy to publish them, nothing better than hanging a company with their own words. So far, the main protagonist has said no.

If you are familiar with the company, there isn't much that will surprise you though, just their behaving like they are known to do, and me not playing their games.

-Charlie

[Groo The Wanderer]

Originally Posted by Sxotty

Except the TGA was the important thing anyway...

And that is such a hard concept to grasp. I think I covered the basics of that in about 6th grade science. The main part to grasp is the curve and how it relates to the real world. That part isn't hard, it was much harder to get the material sheets on it.

-Charlie

[AlexV]

The Charlie love-train goest through another station. The hate one too. This one is only for nVidia's strained or unstrained status. Gracias.

[trinibwoy]

Partners bail on GT200. Due to poor availability. Fudo.