Well This Can't Be Good: AMD Axes Carrell Killebrew & Other Employees

[Red Hawk]

Cancelling contracts with console makers would be incredibly stupid. In fact, I wouldn't be surprised if the requisite research for console development is the biggest thing keeping GPU development at AMD afloat, since the research can be applied to desktop GPU designs as well.

 

[BFG10K]

As far as I remember it was a MASSIVE improvement over the HD 2900, enough for it to be very competitive with the 8800 series. In fact I think its the reason that Nvidia brought out the 8800GT at the price they did.

Nope, in terms of performance the 3870 was basically the same as the 2900XT (i.e. not even close to being competitive with the 8000 series). It still had the same broken hardware AA resolve too.

The main advantage was the large die shrink which lowered manufacturing costs and caused a huge reduction in TDP.

 

[boxleitnerb]

Nope, in terms of performance the 3870 was basically the same as the 2900XT (i.e. not even close to being competitive with the 8000 series). It still had the same broken hardware AA resolve too.

The main advantage was the large die shrink which lowered manufacturing costs and caused a huge reduction in TDP.

This. It never came close to the 8800GTX and was often behind the 8800GT.

 

[Dribble]

Cancelling contracts with console makers would be incredibly stupid. In fact, I wouldn't be surprised if the requisite research for console development is the biggest thing keeping GPU development at AMD afloat, since the research can be applied to desktop GPU designs as well.

Console gpu's have very tight margins - Sony and MS drive a very hard bargain. The question is that where you want to spend you limited R&D engineers time - dealing with all the little quirks and tight deadlines a console has? Obviously not - you want them focusing on things that will make you lots of money.

 

[Ancalagon44]

Nope, in terms of performance the 3870 was basically the same as the 2900XT (i.e. not even close to being competitive with the 8000 series). It still had the same broken hardware AA resolve too.

The main advantage was the large die shrink which lowered manufacturing costs and caused a huge reduction in TDP.

I meant massive improvement by way of slightly improved AA, smaller die, massively reduced power consumption and heat production. Yeah it wasnt faster - in fact I think it had a slower memory bus - but it was a much better card all the same.

EDIT: See here. It was competitive with the 8800GT on price.

 

[sontin]

Console gpu's have very tight margins - Sony and MS drive a very hard bargain. The question is that where you want to spend you limited R&D engineers time - dealing with all the little quirks and tight deadlines a console has? Obviously not - you want them focusing on things that will make you lots of money.

But the revenue from the royalities are nearly 100% profit (=~100% Marge). That's the reason why AMD has a very good Q4 every year. It's the best sellingt quarter for Microsoft and Nintendo.

 

[BenSkywalker]

$5 Billion dollars, financed at 3% generates $150 million dollars per year in interest costs.

5 years of $150 million dollars per year is $750 million.

If the GPU business of AMD started making a billion dollars in profit per quarter it would take roughly a year and a half from today to break even for AMD concerning their ATi acquisition. They have not made a billion dollars in total profit in the last five years. Obviously at this point AMD has significantly reduced the value of what was ATi, they sold off the division that became Snapdragon for a pittance, they got into a price war with a company that has a lock down on the highest margin segment of the industry, and they did everything they could to avoid entering the highly profitable HPC space.

There is a certain level of stupid you can allow for things not going as planned, but AMD has been shockingly great at getting everything wrong.

Eyefinity? Tegra and the 'failure' Fermi architecture(non gaming use only) brought in $377 Million in revenue for nV last quarter, AMD's GPU division as a whole barely edged that out. Fanboys may fap-fap all day over Eyefinity, it isn't doing a whole lot for AMD's bottom line while the 'failure' that the fanboys bashed Fermi for being is bringing in truck loads of cash for nV(in terms of profit- more then nV's and AMD's gaming GPUs *combined*) Fermi, the non gaming elements of it that made it so huge and such a 'bad design' has made more profit by itself then all of AMD's GPU business for the last five years combined.

If you listened to the fanboys, right up to the bitter end, 3dfx was about to crush nVidia. Charlie and his flock of lemmings were talking about nV going bankrupt last year because of two product line choices that have already had more positive cash flow then all of AMD for the last *decade*.

The people responsible for these choices at AMD had a job to do, and they failed at it horribly. Their job was not to make Charlie and his kin have nerdgasms, it was to make money. The last several generations of products from AMD's GPU division have failed at pushing into new markets(moving integrated chipset to integrated on die is *not* a new segment), they have failed at making a profit remotely close to their business costs(not ATi's fault, but this is the real world and that does not matter), they nigh gave away the UP market right before it became a multi billion dollar profit printing segement.

None of this is even a tiny bit surprising, it was extremely obvious years ago(as clearly AMD noticed and hence the shift in their upcoming direction). The people in charge of what the loyalists considrered the great 4xxx-6xxx product lines did a very poor job. That isn't the fault of the lower level engineers who clearly delivered what they were asked to in a very solid state, but the people that made the choice to go in that direction were very wrong. Try and argue what you would like to see all day, at the end of the day it is numbers in a spreadsheet and what is popular on this forum clearly isn't the best way to run a business.

 

[SirPauly]

I find it odd that one only looks at only the GPU division numbers. To even compare 3dfx to AMD is odd.

 

[LOL_Wut_Axel]

The problem was the price war at the time of the HD4000 series. It reverberated all the way through to today. Nvidia had to cut prices for their GT200 line and then AMD could not sell the HD5000 for so much more (like 500 bucks). And so forth. If you raise prices but your competitor does not, no one will buy your product. If you don't raise prices, you make to little money. Quite a pickle to break free of that cycle.

I have no problem paying a bit more for a Geforce because of features and compatibility/support (better AA compatibility, early good SLI profiles etc.). I feel that AMD has treated its enthusias segment quite stepmotherly ever since the HD2000 series came out. It's the image, that's a problem. AMD stands for value offerings (aka cheap) while Nvidia, retaining the GPU crown (single card) ever since 2006, has kind of a leader image. AMD could turn things around and raise prices if they actually beat Nvidia again, and care more for the enthusiast community. Do that 2 generations in a row and you can raise prices also on lower tier products.

Just my 2 cents.

Edit:
But I wonder, is it just image or is there something else? Nvidias Geforce business is very profitable compared to AMDs GPU segment, despite their large dies (and accordingly higher prices). If the die/price thing is breaking AMDs profitability, I would expect Nvidia doing worse than they actually are for the same reasons.

Alright, then explain to me the HD 5970 and the HD 6990. The HD 5870 was unmatched by anything for six months until the GTX 480 came out, and AMD still had the HD 5970 and that was vastly faster (albeit at a much higher cost). Then the HD 6990 came before the GTX 590, and it's slightly faster than it. AMD has had the fastest graphics card crown since 2009. NVIDIA has had the fastest single GPU graphics card since 2010, but not the fastest graphics card.

 

[SirPauly]

Would you expect anything different coming from AMD. "IPC increases!" lol. These APUs are very cheap. Always expect the sugar coating. Its a double edge sword. Their fusion touting has also caused intel to step up to the plate. PPL are gona be surprised at the quick advancing intel will make in this field. They will push this route to the max. In future generations AMD will have to out do intels cpu/gpu combos or face the embarrassment. As AMD makes more capable APUs, this will take more and more chunks out of their GPU division. It cannot happen any other way. Fusion just doesnt seem like the greatest thing to me at all. AMDs big dream just took them way to long to exicute. So long that intel beat them to it even after they wasted yrs tring create their own high powered dGPU larrabee. AMD took soo long with fusion. This is why ppl believe it was a mistake to buy ATI. Why did they take so long, funds? In this light, ATI cost them a lot more than they could afford. How can you be sure the BD result wasnt because of lack of funds? Many many different possible directions and they took this one. Currently, its far from a success story.

Its way to early to claim much of anything still. AMD may not be in as bad of shape as some ppl think, but you must realize things arent just rosy. To me its sort of like a long silent moment. It remains to be seen how any of this plays out. AMD has been stumbling, i am not so sure they are on their feet yet. Lots has changed and real quick like. Maybe its all for the better, i can be hopeful but i cannot say it for sure is!

I am fearing the next core GPU too. AMD would say its the bomb. I hope it is cause i imagine nvidia will step up to the plate for a space blasting kepler. Core sounded very interesting a while ago, but i have heard some concerning things now. Its something that interest me. I am hopeful they can make something wonderful but i heard the power cap and it will be very limiting is f it holds true.

Again, its too early for me to conclude anything really. But from what i do gather, it twists my gut.

I don't think anyone would think things are easy and rosy and there is a lot of work to do in my mind. It's not easy to compete against Intel, nVidia and others -- yet they're growing slowly and making profits. Curious to see how much of a predator and aggressor AMD can be with the new CEO's leadership. AMD could easily fall to becoming irrelevant with future mistakes, but I believe most of the growth mistakes was placed on management --- not reacting quick enough to emerging markets.

I really like nVidia's management on how quickly they adapt and innovate.

 

[boxleitnerb]

Alright, then explain to me the HD 5970 and the HD 6990. The HD 5870 was unmatched by anything for six months until the GTX 480 came out, and AMD still had the HD 5970 and that was vastly faster (albeit at a much higher cost). Then the HD 6990 came before the GTX 590, and it's slightly faster than it. AMD has had the fastest graphics card crown since 2009. NVIDIA has had the fastest single GPU graphics card since 2010, but not the fastest graphics card.

I prefer a single GPU card every day over CF/SLI as a substitute for the enthusiast segment. Only when the fastest GPU doesn't cut it anymore, I'm in for one or two more.
You cannot put two performance chips together with limited memory, microstuttering, profile dependency, bad scaling and call that a legitimate highend product. AMD is still way behind with their CF-tech compared to SLI (don't only look at mainstream or newer games - alienbabletech did an extensive test where CF often failed miserably). They don't combat microstuttering as well as Nvidia and are often late with their profiles. The 5970 and 6990 are good on paper, but in reality it's a different story.

I dislike this strategy and those dual-gpu cards don't really count for me - no matter if they are from NV or AMD. They are just a cost effective compromise solution so they don't have to develop and make an even bigger chip.

 

[Ancalagon44]

I prefer a single GPU card every day over CF/SLI as a substitute for the enthusiast segment. Only when the fastest GPU doesn't cut it anymore, I'm in for one or two more.
You cannot put two performance chips together with limited memory, microstuttering, profile dependency, bad scaling and call that a legitimate highend product. AMD is still way behind with their CF-tech compared to SLI (don't only look at mainstream or newer games - alienbabletech did an extensive test where CF often failed miserably). They don't combat microstuttering as well as Nvidia and are often late with their profiles. The 5970 and 6990 are good on paper, but in reality it's a different story.

I dislike this strategy and those dual-gpu cards don't really count for me - especially the ones from AMD.

But thats only your opinion. As far as AMD is concerned, these are their best cards, whatever processors they actually pack. Yes, they have all of the issues you mentioned, but that doesnt change the fact that they are still single card solutions.

Solutions from both companies are dependent on driver optimization anyway. Shader compilers etc.

 

[LOL_Wut_Axel]

I prefer a single GPU card every day over CF/SLI as a substitute for the enthusiast segment. Only when the fastest GPU doesn't cut it anymore, I'm in for one or two more.
You cannot put two performance chips together with limited memory, microstuttering, profile dependency, bad scaling and call that a legitimate highend product. AMD is still way behind with their CF-tech compared to SLI (don't only look at mainstream or newer games - alienbabletech did an extensive test where CF often failed miserably). They don't combat microstuttering as well as Nvidia and are often late with their profiles. The 5970 and 6990 are good on paper, but in reality it's a different story.

I dislike this strategy and those dual-gpu cards don't really count for me - no matter if they are from NV or AMD. They are just a cost effective compromise solution so they don't have to develop and make an even bigger chip.

Wat. Dual-GPU scaling is better with the HD 6000 series than the 500 series.

The rest, all I have to say it's wrong, misinformed, and because of that I don't care about your opinion on it. When you have an opinion that's based on facts, reply. AMD still has the fastest graphics card, and scaling as well as compatibility have been in general excellent. You can try to skew things in your favor, but I was replying to a comment saying NVIDIA has the fastest graphics card, and that's simply false. The HD 6990 is a very small amount faster than the GTX 590, but still faster.

 

[boxleitnerb]

But thats only your opinion. As far as AMD is concerned, these are their best cards, whatever processors they actually pack. Yes, they have all of the issues you mentioned, but that doesnt change the fact that they are still single card solutions.

Solutions from both companies are dependent on driver optimization anyway. Shader compilers etc.

I was talking about performance, true. But as I said, it's also about image. And in my eyes, Nvidia is an enthusiast company while AMD is not so much. Leadership always commands a premium. And this leadership is not only about performance but also dev support, features (SSAO, 3DVision, advanced image quality features), listening to the community and attitude in general. At least in my opinion, their attitude puts them into 2nd place. All the fuss about their AF optimizations, no PhysX competition, saying microstuttering doesn't matter (read it in an AMD interview), Batmangate...I don't deny that Nvidia are all holy - they sure aren't. But at least they are doers while AMD are talkers. That is my impression.

To get around to my original point:
Be innovative, do more than you have to, support your enthusiast products properly and then you can raise prices and enjoy synergetic effects throughout your portfolio. There is a reason that the average Joe first points to Nvidia and then to AMD. And it's not that green looks better than red

Wat. Dual-GPU scaling is better with the HD 6000 series than the 500 series.

The rest, all I have to say it's wrong, misinformed, and because of that I don't care about your opinion on it. When you have an opinion that's based on facts, reply. AMD still has the fastest graphics card, and scaling as well as compatibility have been in general excellent. You can try to skew things in your favor, but I was replying to a comment saying NVIDIA has the fastest graphics card, and that's simply false. The HD 6990 is a very small amount faster than the GTX 590, but still faster.

I said it has the performance leadership in single-GPU and that is true. I didn't say anything about performance of dual gpu cards. Take a look here:
http://alienbabeltech.com/main/sli-vs-crossfire-part-2-high-end-multi-gpu-scaling
To this day, Crysis 2 is not fixed, and that is an AAA title. CF support was disabled for several games just recently.

 

[SirPauly]

It's simple: A pro-active nature and doing more allows customers to enjoy more, which hopefully rewards the company with a modest premium.

 

[Ancalagon44]

I was talking about performance, true. But as I said, it's also about image. And in my eyes, Nvidia is an enthusiast company while AMD is not so much. Leadership always commands a premium. And this leadership is not only about performance, dev support, features, listening to the community and attitude in general. At least in opinion, their attitude puts them into 2nd place. All the fuss about their AF optimizations, no PhysX competition, saying microstuttering doesn't matter (read it in an AMD interview), Batmangate...I don't deny that Nvidia are all holy - they sure aren't. But at least they are doers while AMD are talkers. That is my impression.

To get around to my original point:
Be innovative, do more than you have to, support your enthusiast products properly and then you can raise prices and enjoy synergetic effects throughout your portfolio. There is a reason that the average Joe first points to Nvidia and then to AMD. And it's not that green looks better than red

More opinion!

Where is your proof the average Joe points to Nvidia? I'm just not seeing it.

The only thing that Nvidia have really outperformed AMD in - badly - is GPGPU. Their implementations are far more robust as far as I know.

 

[SirPauly]

Desktop discrete market share.

 

[Genx87]

I called the ATI acquisition a turd back in the day and it still hasnt been shined enough to make it look pretty. Fusion is not worth the 5 billion price tag for ATI. For that kind of cash, AMD could have built something home grown and not stunted their CPU development.

Looking at the market cap. AMD has mismanaged the merger to the point where it has evaporated from the books.


But Bens post is spot on. AMD has managed to hit a grand slam when it comes to making wrong decisions.

 

[Genx87]

More opinion!

Where is your proof the average Joe points to Nvidia? I'm just not seeing it.

The only thing that Nvidia have really outperformed AMD in - badly - is GPGPU. Their implementations are far more robust as far as I know.

That and marketshare and making money.

 

[LOL_Wut_Axel]

I said it has the performance leadership in single-GPU and that is true. I didn't say anything about performance of dual gpu cards. Take a look here:
http://alienbabeltech.com/main/sli-vs-crossfire-part-2-high-end-multi-gpu-scaling
To this day, Crysis 2 is not fixed, and that is an AAA title. CF support was disabled for several games just recently.

No, you said SINGLE CARD. Here's your comment:

AMD stands for value offerings (aka cheap) while Nvidia, retaining the GPU crown (single card) ever since 2006, has kind of a leader image.

AMD has the fastest single card ever since 2009.

Using "to this day" is inaccurate as well. That review uses older drivers, and newer Catalyst drivers were released 8 days ago. You're obviously trying to nitpick. In general, for both AMD and NVIDIA CF/SLI support has been excellent. You can find one out of thirty recent games that has problems, but it doesn't change the general picture. As far as scaling itself goes, again, the HD 6000 series does better than the 500 series. That's why you see a smaller gap in comparison to single card when comparing the HD 6970 and GTX 580 in CF/SLI.

 

[Idontcare]

There is a certain level of stupid you can allow for things not going as planned, but AMD has been shockingly great at getting everything wrong.

So true.

Bobcat appears to be one of the few things they've done right in quite a while.

And the irony is not lost on me that the key enabler (40nm process tech) was avoiding the use of their own internal fabs (the decision to go with TSMC for production was made when AMD still owned, and planned to continue to own, their own fabs).

Not that today's AMD can survive on Bobcat's future alone, but in a much smaller footprint they can recede the same as Via while avoiding bankruptcy and ruin.

As for the glory days...not likely to happen again.

 

[boxleitnerb]

No, you said SINGLE CARD. Here's your comment:



AMD has the fastest single card ever since 2009.

Using "to this day" is inaccurate as well. That review uses older drivers, and newer Catalyst drivers were released 8 days ago. You're obviously trying to nitpick. In general, for both AMD and NVIDIA CF/SLI support has been excellent. You can find one out of thirty recent games that has problems, but it doesn't change the general picture. As far as scaling itself goes, again, the HD 6000 series does better than the 500 series. That's why you see a smaller gap in comparison to single card when comparing the HD 6970 and GTX 580 in CF/SLI.

Sorry, I meant single GPU. Honest mistake.

Two weeks ago:
http://forums.guru3d.com/showthread.php?t=352774
What you call nitpicking, I call an example.

Concerning scaling, read the article in the link. Especially with 3 cards and up, you often get into trouble with CF, the minimum fps can also be an issue with just 2 cards. Furthermore, a single avg fps number is not enough to describe gameplay experience. fps consistency and frame distribution are important too. I know enough people who had SLI as well as CF lately (since 2009) who can confirm that Nvidia is often doing better there. Anyway, this is getting a bit too off topic. I think I made my opinion on AMD's enthusiast status clear.

 

[LOL_Wut_Axel]

True.

Also, Krishna and Wichita are gonna be in the 28nm process node, so AMD is being quite aggressive there.

 

[SirPauly]

AMD could have built something home grown

How? And what would be the resources here? How much time needed? Intel is a very capable company and they can't make a dent in the discrete market and the graphics aspect of AMD's APU's, there is differentiation.

 

[Seero]

So true.

Bobcat appears to be one of the few things they've done right in quite a while.

And the irony is not lost on me that the key enabler (40nm process tech) was avoiding the use of their own internal fabs (the decision to go with TSMC for production was made when AMD still owned, and planned to continue to own, their own fabs).

Not that today's AMD can survive on Bobcat's future alone, but in a much smaller footprint they can recede the same as Via while avoiding bankruptcy and ruin.

As for the glory days...not likely to happen again.

I thought Apple was no more.